Changing the servo motor encoders from 256 to 512 or 1024. Any reason not to do this?


 

The full gory details of how the position counter value is determined are in Microchip document 70208C.pdf (“Quadrature Encoder Interface”):

 

https://www.microchip.com/content/dam/mchp/documents/OTH/ProductDocuments/ReferenceManuals/70208C.pdf

 

David

 

From: Losmandy_users@groups.io <Losmandy_users@groups.io> On Behalf Of Paul Kanevsky
Sent: 27 July 2021 20:42
To: Losmandy_users@groups.io
Subject: Re: [Losmandy_users_io] Changing the servo motor encoders from 256 to 512 or 1024. Any reason not to do this?

 

Peter,

Entirely possible and even likely. I'm not arguing with your results. All I was trying to do was to correct some information that was posted earlier on the encoder precision. Gemini PIC controller reads encoders in quadrature, and so the effective resolution is 4x smaller than the (Gemini-addressable) step size.

Regards,

      -Paul

On Tue, Jul 27, 2021 at 03:32 PM, <pcboreland@...> wrote:

Paul,

This is where people are getting confused I think.  The positioning step accuracy is 0.5625". Therefore if you what to make a Dec correction it will be that large. You can not move the motor a 1/4 step.   If you install a 1024 encoder you can make a 0.1405" step move. The results are night and day as far as Dec guiding performance. In the chart just posted Dec rms was 0.15", about the same size as the step resolution. There is no point fighting the facts, a higher res encoder will greatly improve your Dec guiding. 

Peter

PS. Before this change by Dec guided at about 0.74" rms.


Peter Boreland
 

Paul,

I appreciate that, but it's kind of irrelevant, and confuses most people into thinking you are controlling the motor at that resolution. Frankly, I can not recommend enough that Losmandy owners upgrade their Dec encoders and see what results they get. It's such a simple mod. Moreover, people should zero out their backlash. The Allen wrench method is balls on accurate. Zeroing backlash is what 10 micron mount do in their software.  If you change the slew setting to 1200 you can also move just as fast. Sticking with the 25:1 gearboxes make that possible.
 
Upgrading the Ra encoder was a bust. Folks should focus instead on elimination the large period errors. (80s (OPW), 240s and 2600s.  I've shone a light onto those problems and how I dealt with them. If you follow the steps I took you will have no use for PEC. The gearbox upgrade and thrust bearing upgrade are a must if you want a calm Ra guiding curve.   

All the best,

Peter


Paul Kanevsky
 

Peter,

Entirely possible and even likely. I'm not arguing with your results. All I was trying to do was to correct some information that was posted earlier on the encoder precision. Gemini PIC controller reads encoders in quadrature, and so the effective resolution is 4x smaller than the (Gemini-addressable) step size.

Regards,

      -Paul


On Tue, Jul 27, 2021 at 03:32 PM, <pcboreland@...> wrote:
Paul,

This is where people are getting confused I think.  The positioning step accuracy is 0.5625". Therefore if you what to make a Dec correction it will be that large. You can not move the motor a 1/4 step.   If you install a 1024 encoder you can make a 0.1405" step move. The results are night and day as far as Dec guiding performance. In the chart just posted Dec rms was 0.15", about the same size as the step resolution. There is no point fighting the facts, a higher res encoder will greatly improve your Dec guiding. 

Peter

PS. Before this change by Dec guided at about 0.74" rms.


Peter Boreland
 
Edited

Paul,

This is where people are getting confused I think.  The positioning step accuracy is 0.5625". Therefore if you what to make a Dec correction it will be that large. You can not move the motor a 1/4 step.   If you install a 1024 encoder you can make a 0.1405" step move. The results are night and day as far as Dec guiding performance. In the chart just posted Dec rms was 0.15", about the same size as the step resolution. There is no point fighting the facts, a higher res encoder will greatly improve your Dec guiding. 

Peter

PS. Before this change by Dec guided at about 0.74" rms.


Paul Kanevsky
 

Here is my math: (360*3600)/(256*25*360) = 0.5625 arc sec per step.This seems rather coarse. 
This is a single step addressable by the Gemini controller. The actual accuracy of the encoder reading is 4x the step or 0.1406" and even smaller for Titan.

Regards,

    -Paul


Keith
 

Peter:  since you're doing a deep dive, might be interesting to check the orthogonality of the axis shafts to their top plate. I had an issue in mine leading to some stiff spots when rotating the axis by hand (similar to what one might feel with a wonky needle bearing).  See this thread if interested:  https://groups.io/g/Losmandy_users/message/59944

Keith


Peter Boreland
 
Edited

I was able to run further tests last night, but smoke moving into our area cut the session short.

I put back the 25:1 McLennan gearbox on the Ra drive and ran a guiding test with the 256 tick optical encoder. Slew rates are now fully restored. I also changed out the Ra thrust bearings to those sourced from McMaster. The net result saw a very significant reduction of 2600s error I previously identified and attributed to the top thrust bearing. As you can see from the attached PHD2 screen shot by guiding rms error was 0.28 arc-sec. Later in the evening I changed out the Ra motor encoder to 1024. I saw no real improved in Ra performance which leaves to with the following overall conclusions:

1. The stock Losmandy thrust bearings are a major source of problems. I attribute stiffness of rotation, worm misalignment, and significantly large 2600s error to them. My suggestion with anyone who buys an older or new mount is to change these to German made bearings sourced from McMaster.

2. The Dec axis greatly benefits from a smaller step size. This can be achieved by switching out the 256 tick encoder to a 1024 tick encoder. Cost is around $65 sourced from US digital. A smaller step size make sense as it leads to a more subtle control, and it most noticed when one's polar alignment is excellent.  

3. Replace the Oldham couplers with rigid couplers. This facilitates alignment of the motor to the worm. This mod it difficult to pull off as the couplers have to be made from two parts, previously listed. Cost is around $20 for both Dec and Ra. Since there are two pivot points that have to be aligned, it requires loosening of the coupler screws and gearbox mount screws and alignment the assembly while rotating the motor. When aligned correctly the motor will not bob up and down, Done right the native 240s error in Ra will be very much reduced and there will be no requirement to used PEC to achieved a 240s guided error of under 0.2 - 0.15 arc-sec.

4. Reduction of the 80s bearing error in the OPW remained stubbornly large. After repeated tests I found the best thing was to used a single Bevell washer arranged to put pressure on the ball race. In my case I reduced the 80s error to under 0.3 arc-sec.

5. Replace the Ra gearbox to the McLennan gearbox (not necessary for Dec) . I saw no real benefit to reducing the step size. This makes sense because the Ra worm is continuously rotating. The reason to replace the gearbox is to eliminate the large 32s error of the Losmandy gearbox. 

6. Spring load the Ra worm ( did not test the locked down cased), and add a 2 lb bias weight to load the worm on the east side. The keeps the worm permanently engaged on the right face of the worm gear both before and after a meridian flip. 

7. Use TVC correctly to eliminate Dec backlash. The ideal arrangement  is to have the worm mash on both sides to the ring gear, but not have the pressure be too great so as not to be able to turn with one's fingers. I found a spring loaded arrange in Dec to be detrimental to achieving a low backlash value. TVC is there to cancel out any backlash. The thrust bearings play a big role here because you want the ring gear to turn easily when the clutch is tightened. If using the 1024 tick encoder then TVC will be 4x larger in value than in the 256 tick encoder case (my setting is 45). 

That it folks! Happy imaging,

Pete


alan137
 

Please start another thread to discuss this


Peter Boreland
 
Edited

I moved this discussion to its own thread:

https://groups.io/g/Losmandy_users/topic/needle_bearings_and_axial/84141331?p=,,,20,0,0,0::recentpostdate%2Fsticky,,,20,2,0,84141331


alan137
 

On Fri, Jul 9, 2021 at 06:10 AM, Michael Herman wrote:
Alan Fang has that GM811 version and he is also investigating its behavior (or he is happy and his system is stable and in use). 
 
 
Alan, while visiting his parents in southern California, stopped by the humble Losmandy building and swapped out his G8 DEC axis for a G11.  They were very nice to give me a credit towards purchase of the new axis.  So now I am a G11 owner.
In the end, I had the G8 axis (while using a 40 lbs 10" F4 Newtonian) working "ok, but not trouble free".  Which means that on a day with good seeing, I could crank out many good frames in a row, but will occasionally have to cancel a frame before completion due to excessive DEC movement 30 seconds after a dither move, etc.  Given that it has a "resolution" of 1.x arc-sec, I would consider that it is good only up to 1000mm focal length and about 40 lbs.  The g11 has a worm wheel twice as big, so I expect all errors to be half as much, and a much better baseline to work from.


Peter Boreland
 

On Fri, Jul 9, 2021 at 06:10 AM, Michael Herman wrote:
Peter,
 
You have initiated many good experiments that have not been described earlier.  Like your alternative gearboxes over 50:1.  So you are pioneering new ground.  We all benefit from your research. 
 
If you have the tucked motors and spring loaded worm systems then your mount is past my versions of G11 and GM8.  Alan Fang has that GM811 version and he is also investigating its behavior (or he is happy and his system is stable and in use). 
 
 I think this is the end result of all our systems: when we image reliably with good results we stop, lock everything down, and just drive the mount.  It's like a cake recipe.  The work goes into the recipe then the recipe is "frozen".  
 
One last thought: having the R4 spring in place only works if the bearing above it can slide in and out by the spring force.  If the bearing gets stuck, the spring can't level the (far) bearing. What about the bearing nearest the gearbox? I've never fiddled with that one.  Could that one have a tight spot in its race?  Would undersizing that and lubing that do any improvement?   I don't think that area has been tried or considered.
 
I'm still awaiting some bearings with ceramic balls to arrive.  Hugo in Tehachapi told me he got some amazingly smooth tracking from full ceramic bearings (they were like $50 each) but I thought there would be no way to undersize their OD, so I kept the idea of using stainless steel bearing parts.  Maybe those ceramics he got are so perfect they don't need undersizing.  Anyway ... more experiments have been tried with further progress.  
 
Each of us uses different scopes, different cameras... as well as different latitudes, making our comparisons somewhat fuzzy.  We can compare our unguided PE peak to peak and RMS values.  We can show our exposure subs and final images.  
 
On other topics: The delta variant of Covid is being seen here in the US and is now in many places the dominant strain. Hospitalizations are going up, though still under ICU capacity (except in parts of Missouri?). And the initial vaccines are starting to wear off.  And the use of masks is waning too.  
 
Stay well, all...
 
 
Michael
 
Michael,

Thank you for your help this. There are still a number of issues needing to be locked down to achieve the level of performance one expects from expensive premium mounts (accuracy, no transients moves, very low noise for a clean accurate PEC curve). Through this path of discovery I have learned several important things that will greatly improve the performance of a mount like the G11 to get you in the range of around 0.6 arcsec or better plus or minus with a 35 to 40 lb payload (summertime). Here is my current summery of the important steps to take to improve mount performance:

1. Change the Ra coupler a one-piece rigid one to significantly reduce the 240s error. This unfortunately requires some engineering work. https://www.ruland.com/clx-4-3-a.html Since it has a dia of 5/8" it will not rotate without the screws touching the sides in the OPW. I replaced the 4-40 screw with tapered ones, drilled out the coupler screw seating's to seat the tapered screws, and parred down the heads to as small as possible. They still will not clear the base plate so a hollow will need to be ground. I did this with a file. There maybe a 1/2" dia rigid coupler out there but I could not find one. My belief is if people have very large values it is due to misalignment of the worm and the motor drive shaft. The Oldham set screw coupler is the real problem here. 

2. Increase the number of steps per Dec worm revolution. Going to a 50:1 gearbox might be enough to overcome erratic PHD2 behavior in this axis. My suggestion however is not to change out the gearbox but rather the encoder as I originally planned to do before going off down the gearbox route. You have two choices 512 or 1024. I would opt for 1024. The nice thing about going this route is slewing speed should be preserved and you  not need to change the stock Losmandy gearbox.

3. Increase the number of steps for the Ra axis by 4x. I did this using a McLennan 125:1 gearbox (5x). This gearbox has really nice high frequency characteristics, and a large pinion gear (same as 25:1 McLennan gearbox), but the slew rate will have to be dropped down to 160 - 200. Again, I will in the future explore changing over to a 1024 encoder and use the McLennan 25:1 gearbox. This gearbox does not have the large 32s error component of the stock Losmandy gearbox. The other important thing is increasing the number of steps per worm revolution allows one to create a high resolution PEC curve. I've demonstrated that the Gemini II controller will accept a curve with 32000 steps, contrary to what was previously thought. Elimination of the higher frequency noise components does allow a very precise PEC curve to be created, which I've demonstrated to myself using PECPrep. However PEMPro is required in  order to sync and download a PEC curve. 

With that said, I want to reintegrate that the easiest way I think to to make these improvements is simply to change the encoders. I have not done this but will be doing this in the near future. I feel this is important to do for completeness. I have an 80s error problem as previously noted in earlier  (possibly from the gearbox), which should be removable via PEC. Just in case anyone is wondering, these mods have not caused me to loose any imaging time. As an engineer I want to explore the boundaries of what is possible with this mount. I will be loading it up with 70 - 75lb payload soon, and the way it performed previously would have forced me to give in up and lay out for a premium mount.

Work planned:

1. Take measurements with PEMPro generated PEC curve,
2. Take measurements with and without spring loading the Ra worm.
3. Find source of the 80s error.
4. Take measurements using higher res encoders with 25:1 gearbox so slew speeds are preserved. 

Peter








Michael Herman
 

Peter,

You have initiated many good experiments that have not been described earlier.  Like your alternative gearboxes over 50:1.  So you are pioneering new ground.  We all benefit from your research. 

If you have the tucked motors and spring loaded worm systems then your mount is past my versions of G11 and GM8.  Alan Fang has that GM811 version and he is also investigating its behavior (or he is happy and his system is stable and in use). 

 I think this is the end result of all our systems: when we image reliably with good results we stop, lock everything down, and just drive the mount.  It's like a cake recipe.  The work goes into the recipe then the recipe is "frozen".  

One last thought: having the R4 spring in place only works if the bearing above it can slide in and out by the spring force.  If the bearing gets stuck, the spring can't level the (far) bearing. What about the bearing nearest the gearbox? I've never fiddled with that one.  Could that one have a tight spot in its race?  Would undersizing that and lubing that do any improvement?   I don't think that area has been tried or considered.

I'm still awaiting some bearings with ceramic balls to arrive.  Hugo in Tehachapi told me he got some amazingly smooth tracking from full ceramic bearings (they were like $50 each) but I thought there would be no way to undersize their OD, so I kept the idea of using stainless steel bearing parts.  Maybe those ceramics he got are so perfect they don't need undersizing.  Anyway ... more experiments have been tried with further progress.  

Each of us uses different scopes, different cameras... as well as different latitudes, making our comparisons somewhat fuzzy.  We can compare our unguided PE peak to peak and RMS values.  We can show our exposure subs and final images.  

On other topics: The delta variant of Covid is being seen here in the US and is now in many places the dominant strain. Hospitalizations are going up, though still under ICU capacity (except in parts of Missouri?). And the initial vaccines are starting to wear off.  And the use of masks is waning too.  

Stay well, all...


Michael


On Thu, Jul 8, 2021, 4:29 PM pcboreland via groups.io <pcboreland=yahoo.com@groups.io> wrote:
On Thu, Jul 8, 2021 at 03:46 PM, Michael Herman wrote:
All good notes and another good conversation.
 
The 80 sec oscillation has been attributed (by Michael Sinescalchi in his www.helixgate.net) to a pinched bearing ball.  He saw either 78 or 80 seconds from that.  I appreciate that this 80 sec oscillation could arise from a 25:1 McLennan gearbox....that's new to me.  
 
If Peter has not installed a R4 Belleville spring behind his (far) RA and DEC worm bearing, he will get:
 
1. Longer Hysteresis time lag adversely affecting autoguiding
 
2. A possible pinched R4ZZ bearing ball at 78 to 80 seconds period affecting PE.
 
The solution on how to add in the R4 Belleville spring, and better quality R4ZZ bearings, and make the bearing slide and swivel in the mounting block, is written up in the attached PDF.  
 
Hope these ideas help you get your mount to it's best operation.
 
Michael
I do have the bevel washer(s) installed per your document. It is funny but this never used to the be a problem! Only when I changed the gearboxes. But I have made so many alterations. I'm looking into it and will report back.

Peter


Peter Boreland
 

On Thu, Jul 8, 2021 at 03:46 PM, Michael Herman wrote:
All good notes and another good conversation.
 
The 80 sec oscillation has been attributed (by Michael Sinescalchi in his www.helixgate.net) to a pinched bearing ball.  He saw either 78 or 80 seconds from that.  I appreciate that this 80 sec oscillation could arise from a 25:1 McLennan gearbox....that's new to me.  
 
If Peter has not installed a R4 Belleville spring behind his (far) RA and DEC worm bearing, he will get:
 
1. Longer Hysteresis time lag adversely affecting autoguiding
 
2. A possible pinched R4ZZ bearing ball at 78 to 80 seconds period affecting PE.
 
The solution on how to add in the R4 Belleville spring, and better quality R4ZZ bearings, and make the bearing slide and swivel in the mounting block, is written up in the attached PDF.  
 
Hope these ideas help you get your mount to it's best operation.
 
Michael
I do have the bevel washer(s) installed per your document. It is funny but this never used to the be a problem! Only when I changed the gearboxes. But I have made so many alterations. I'm looking into it and will report back.

Peter


Peter Boreland
 

On Thu, Jul 8, 2021 at 03:26 PM, Keith wrote:
What are the internal gear ratios?  In the 25:1 gearbox (P5-G11), the third stage gear ends up being 3:1 to the worm period (so 79.8s for the G11), so perhaps that's what you're seeing here.
Keith,

I'm reluctant to disassemble at this point to figure that out. I too though that the 80s period might be from the gearbox. Both the 80:1 and 125:1 gearboxes do have the same gear size on the final output stage, so when I swapped them over it could have been a shared problem. If I can not find another source I'll dig into this further and get the gear ratios. I'm hoping I can neutralize it via other means. Its a good thought!

Peter


Peter Boreland
 

On Thu, Jul 8, 2021 at 02:55 PM, Ray Gralak wrote:
Which build of PEMPro are you using? I'm guessing you were trying to use the ZWO ASCOM driver?

I have a ASI294MM so I am sure I can get you a version that works with that camera. Please post a ticket to this forum:

https://siriusimaging.zendesk.com

-Ray
Ray, super thanks!  I'll do as you say.

Peter


Michael Herman
 

All good notes and another good conversation.

The 80 sec oscillation has been attributed (by Michael Sinescalchi in his www.helixgate.net) to a pinched bearing ball.  He saw either 78 or 80 seconds from that.  I appreciate that this 80 sec oscillation could arise from a 25:1 McLennan gearbox....that's new to me.  

If Peter has not installed a R4 Belleville spring behind his (far) RA and DEC worm bearing, he will get:

1. Longer Hysteresis time lag adversely affecting autoguiding

2. A possible pinched R4ZZ bearing ball at 78 to 80 seconds period affecting PE.

The solution on how to add in the R4 Belleville spring, and better quality R4ZZ bearings, and make the bearing slide and swivel in the mounting block, is written up in the attached PDF.  

Hope these ideas help you get your mount to it's best operation.

Michael


On Thu, Jul 8, 2021, 3:26 PM Keith <keithdnak@...> wrote:
On Thu, Jul 8, 2021 at 05:29 PM, <pcboreland@...> wrote:
Use of the 125:1 gear box has removed the higher frequency Ra tracking errors. The one remaining problem is the 80s error which I have had not impact on removing, and must be related to something that I had not yet touched. I'm at a loss right now! Not sure what even causes it?
What are the internal gear ratios?  In the 25:1 gearbox (P5-G11), the third stage gear ends up being 3:1 to the worm period (so 79.8s for the G11), so perhaps that's what you're seeing here.

Keith


Keith
 

On Thu, Jul 8, 2021 at 05:29 PM, <pcboreland@...> wrote:
Use of the 125:1 gear box has removed the higher frequency Ra tracking errors. The one remaining problem is the 80s error which I have had not impact on removing, and must be related to something that I had not yet touched. I'm at a loss right now! Not sure what even causes it?
What are the internal gear ratios?  In the 25:1 gearbox (P5-G11), the third stage gear ends up being 3:1 to the worm period (so 79.8s for the G11), so perhaps that's what you're seeing here.

Keith


Ray Gralak
 

One big improvement was installing a rigid coupler in the RA axis. This net effect was to significantly reduced the
240s error. My raw peak-peak error is now under 3s, which corrected translated to p-p of 0.4s (0.1s rms). This is
without PEC. Sadly I have not been able to get PEMPro to work with either of my two cameras (ASI294 or ASI178),
so no way to download a PEC curve. Searching for a way to get my money back?
Which build of PEMPro are you using? I'm guessing you were trying to use the ZWO ASCOM driver?

I have a ASI294MM so I am sure I can get you a version that works with that camera. Please post a ticket to this forum:

https://siriusimaging.zendesk.com

-Ray

-----Original Message-----
From: Losmandy_users@groups.io [mailto:Losmandy_users@groups.io] On Behalf Of pcboreland via groups.io
Sent: Thursday, July 8, 2021 2:30 PM
To: Losmandy_users@groups.io
Subject: Re: [Losmandy_users_io] Changing the servo motor encoders from 256 to 512 or 1024. Any reason not to
do this?

Thumbs up to Michael for the grease recommendation: CRC Brake and Caliper grease has MolyD, Graphite, and
Teflon. I bought it at Napa Auto Parts locally, but they had to go into the back to find it, as was not on the shelf.
Also for the hanging weight idea. I had trouble not hitting the pier, but with some creativity implemented it. I took
careful note on how Michael attached it.

I've made the upgrades I intended, but ran into some trying problems with my Dec axis. Bottom line I changed out
the coupler to one I had previously noted to find it introduced too much backlash, so went back the set screw
Oldham coupler. Backlash is at 2500ms. I think is just horrible, but likely the best that can be done with the current
worm design. Yes, it has been better, but totally hit and miss. If you take it apart again who knows what it will be?
This is the design flaw one has to live with with this mount unfortunately. I did install the 80:1 gearbox which did
make a noticeable rms tracking improvement (smaller step sizes). Dropped from about 0.8 rms to 0.4 rms.
Interestedly this is about the same level of improvement I saw with the Ra axis going to the 125:1 gearbox.

One big improvement was installing a rigid coupler in the RA axis. This net effect was to significantly reduced the
240s error. My raw peak-peak error is now under 3s, which corrected translated to p-p of 0.4s (0.1s rms). This is
without PEC. Sadly I have not been able to get PEMPro to work with either of my two cameras (ASI294 or ASI178),
so no way to download a PEC curve. Searching for a way to get my money back?

Use of the 125:1 gear box has removed the higher frequency Ra tracking errors. The one remaining problem is the
80s error which I have had not impact on removing, and must be related to something that I had not yet touched. I'm
at a loss right now! Not sure what even causes it?

Here are the raw FFT and corrected FFT for the RA axis as things currently stand. The 80s peak is 0.45s on the
second image if it is not readable. Overall my guiding with good seeing but poor transparency last night was around
0.65 rms. Previous to these modes it would have been all over the place frame to frame from 0.9ms to 2.5s,
averaging around 1.4s. The really big improvement, which has the most impact on imaging really, is how nicely the
mount performs. Smooth, no sudden transient jumps in either axis. I feel this is directly attributable to having a
higher step resolution (0.1125" in RA and 0.176" in Dec).









Peter Boreland
 

Thumbs up to Michael for the grease recommendation: CRC Brake and Caliper grease has MolyD, Graphite, and Teflon.  I bought it at Napa Auto Parts locally, but they had to go into the back to find it, as was not on the shelf. Also for the hanging weight idea. I had trouble not hitting the pier, but with some creativity implemented it. I took careful note on how Michael attached it.

I've made the upgrades I intended, but ran into some trying problems with my Dec axis. Bottom line I changed out the coupler to one I had previously noted to find it introduced too much backlash, so went back the set screw Oldham coupler. Backlash is at 2500ms. I think is just horrible, but likely the best that can be done with the current worm design. Yes, it has been better, but totally hit and miss. If you take it apart again who knows what it will be? This is the design flaw one has to live with with this mount unfortunately. I did install the 80:1 gearbox which did make a noticeable rms tracking improvement (smaller step sizes).  Dropped from about 0.8 rms to 0.4  rms.  Interestedly this is about the same level of improvement I saw with the Ra axis going to the 125:1 gearbox. 

One big improvement was installing a rigid coupler in the RA axis. This net effect was to significantly reduced the 240s error.  My raw peak-peak error is now under 3s, which corrected translated to p-p of 0.4s (0.1s rms). This is without PEC. Sadly I have not been able to get PEMPro to work with either of my two cameras (ASI294 or ASI178), so no way to download a PEC curve. Searching for a way to get my money back? 

Use of the 125:1 gear box has removed the higher frequency Ra tracking errors. The one remaining problem is the 80s error which I have had not impact on removing, and must be related to something that I had not yet touched. I'm at a loss right now! Not sure what even causes it?    

Here are the raw FFT and corrected FFT for the RA axis as things currently stand. The 80s peak is 0.45s on the second image if it is not readable. Overall my guiding with good seeing but poor transparency last night was around 0.65 rms.  Previous to these modes it would have been all over the place frame to frame from 0.9ms to 2.5s, averaging around 1.4s. The really big improvement, which has the most impact on imaging really, is how nicely the mount performs. Smooth, no sudden transient jumps in either axis. I feel this is directly attributable to having a higher step resolution (0.1125" in RA and 0.176" in Dec). 






 


Michael Herman
 

Peter,

I had heard about the hanging weight many years ago but I avoided using that for silly cosmetic reasons (my own vanity...thinking about the beauty of the mount and not on ultimate performance) ...and I never seemed to need it on my G11 systems.  I always got very good results using the Belleville R4 springs, McLennan gearbox, ABEC-7 bearings, etc.) and my old straight out drives worms are in light but full contact...lucky me.  Here are photos of the C14EdgeHD on the G11 (not G11T) mount.  Note the very long 33 inch long counterweight bar made for me by my nice neighbor.  

On my G11T, I heard from many Titan owners that the hanging cord was essential.  And it proved that it is essential any time your worm has a gap to the ring gear.  You must assure the worm and ring gear contact point remains consistent.  If the contact point flips over you'd get a big glitch (near CWD) from that and also you could get vibration (chatter) if the system has its upper weight to the West not East.  

That effect has to do with the attack angle the worm is making with the ring gear teeth.  If that contact is "dragging" like pulling a chalk across a chalkboard like this:
_____ /____-> 
with the chalk moving to the right, you'd get a smooth rubbing with no chatter.  But if you'd push the chalk to the right like this: 
_____\ ______-> 
then you'd get chatter.  Try it if you have some chalk and a chalkboard!

I had good results with up to 12 inch Meade (even 14 inch Celestron EdgeHD) SCTs on my G11s.  I did use an extended counterweight bar to balance the C14 on the G11.  

However as Brian suggested to me, the G11T is a more massive heavy mount with the massive Titan worm system, more suitable to the C14EdgeHD scope.  That scope alone is about 62 lbs.  Here is a report about my PE tested on the smaller, older direct drive G11 system (non-tucked, non spring loaded).  

I did find one of my G11 tripod legs slipping (contracting,) under the extreme weight of the scope and heavy counterweights.  I think the total weight on the mount head was like ~125 lbs, and add the weight if the mount and that's like 170 lbs on the tripod.  
So be extra careful in looking for the telescoping tripod legs to be securely clamped.  

(A guess: perhaps one reason for slippage is that the mount leg surfaces are anodized and very hard surfaces. The anodized surfaces won't deform much so they may slip more easily than bare aluminum surfaces. )

As for the CRC grease:
I generously painted that all around the massive Titan RA ring gear and worm.  The cap of the small CRC grease jar has a wire brush for the application.  As I've mentioned before, that grease had no reaction on test strip of brass, and it's soec sheet gives the same rating (bright) for the copper corrosion test as Superlube and ArcticJetlube.  That is: no reaction after 24 hours at 100C temperature.  So the worry from the AP group about some MolyD lubricants in extreme machine uses was not found in this brass worm application.  The highest pressure, and highest friction in the mount, is the worm to ring gear contact point.  That area is very small and all the torque on the axis is countered right there.  It needs the lowest friction lubricant that can stay put and handle the high pressure there.  

Best,
Michael

On Mon, Jun 28, 2021, 9:00 AM pcboreland via groups.io <pcboreland=yahoo.com@groups.io> wrote:
Michael,

Thanks for this info. I assume the grease is for the worm? Your rig looks quite heavy. How much weight are you sporting? I have a 12" truss Newtonian that I want to mount in the fall. It seems from looking at your setup my G11 should be able to handle it? I estimate the weight of the gear will be around 65 - 70lbs.

I'll defiantly give the loose hanging weight a go. One thing to note, with the worm blocks removed you have very free rotation so easy to get a precise balance. 

Thanks for all the support guys,
  
Peter


Michael Herman
 

Jason:

They are nice handset holsters,...cleverly fabricated to be flexible by 3D printer of plastic.  

To get them contact:

Warren Schultz
136 Torrey Heights Ln
Durham, NC 27703

His email is

handset holsters were at the time:
 ~$30 each

I had him make me one for Gemini-2,
And 2 more for Gemini-1 units.  They are quite useful.

Very best,
Michael





On Mon, Jun 28, 2021, 6:30 PM Jason Close <jason@...> wrote:
Hey Michael,

How can I get one of those hand controller mounts that attaches to HD tripod??

Jason


Jason Close
 

Hey Michael,

How can I get one of those hand controller mounts that attaches to HD tripod??

Jason


 

Yes I have seen many pictures of Michael’s set up. I have already reported him for crimes against nature :-)

But as I recall he primarily does planetary imaging which is a whole different kind of tracking requirement.

He just bought a G 11 T as well

On Mon, Jun 28, 2021 at 2:35 PM pcboreland via groups.io <pcboreland=yahoo.com@groups.io> wrote:
On Mon, Jun 28, 2021 at 12:12 PM, Brian Valente wrote:
i wasn't sure if your dob was for imaging or visual, if visual it would be fine
 
Brian,

Strictly imaging. Did you see the pictures Michael posted earlier of him setup? Look at those weights!

Peter

--
Brian 



Brian Valente


Peter Boreland
 

On Mon, Jun 28, 2021 at 12:12 PM, Brian Valente wrote:
i wasn't sure if your dob was for imaging or visual, if visual it would be fine
 
Brian,

Strictly imaging. Did you see the pictures Michael posted earlier of him setup? Look at those weights!

Peter


 

>>> I know. But you and other are carrying more than that. So not very helpful there!

definitely not my setup - it's around 40lb

i wasn't sure if your dob was for imaging or visual, if visual it would be fine



On Mon, Jun 28, 2021 at 11:57 AM pcboreland via groups.io <pcboreland=yahoo.com@groups.io> wrote:
On Mon, Jun 28, 2021 at 09:02 AM, Brian Valente wrote:
the spec of the G11 is 60lbs imaging, 75lbs visual
I know. But you and other are carrying more than that. So not very helpful there!



--
Brian 



Brian Valente


Peter Boreland
 

On Mon, Jun 28, 2021 at 09:02 AM, Brian Valente wrote:
the spec of the G11 is 60lbs imaging, 75lbs visual
I know. But you and other are carrying more than that. So not very helpful there!


 

>>> I estimate the weight of the gear will be around 65 - 70lbs.

the spec of the G11 is 60lbs imaging, 75lbs visual

On Mon, Jun 28, 2021 at 9:00 AM pcboreland via groups.io <pcboreland=yahoo.com@groups.io> wrote:
Michael,

Thanks for this info. I assume the grease is for the worm? Your rig looks quite heavy. How much weight are you sporting? I have a 12" truss Newtonian that I want to mount in the fall. It seems from looking at your setup my G11 should be able to handle it? I estimate the weight of the gear will be around 65 - 70lbs.

I'll defiantly give the loose hanging weight a go. One thing to note, with the worm blocks removed you have very free rotation so easy to get a precise balance. 

Thanks for all the support guys,
  
Peter



--
Brian 



Brian Valente


Peter Boreland
 

Michael,

Thanks for this info. I assume the grease is for the worm? Your rig looks quite heavy. How much weight are you sporting? I have a 12" truss Newtonian that I want to mount in the fall. It seems from looking at your setup my G11 should be able to handle it? I estimate the weight of the gear will be around 65 - 70lbs.

I'll defiantly give the loose hanging weight a go. One thing to note, with the worm blocks removed you have very free rotation so easy to get a precise balance. 

Thanks for all the support guys,
  
Peter


Michael Herman
 

Peter...

You have revealed a clue to your RA drive issues: your 240 sec error (worm fundamental period) depends on which side of the meridian you are targeting.

Try this:

1. Fully balance your RA by testing the motor (total Gemini) current with the counterweight bar horizontal.  Adjust your DEC balance first, then balance your RA to get to minimum total Gemini power current.

2. Wrap a length of flexible cord, such as clothesline, around the RA axis do that the cord hangs off the  East side of the RA.  This direction minimizes ring gear chatter.

3. Tie a one or two pound weight to the cord, so that when the RA rotates during imaging, the weight does not bump the tripod or other part of the mount.

That torque from the hanging weight will provide a constant torque to the RA worm and ring gear and always "preload" the contact surface.  In this way the contact region will not flip over as you image through the meridian. 

Then take your PE data on both sides of meridian, and see what your results look like.  I'll hope your flippin' problem is eliminated.

Here is how this appears on my G11T.  

I suggest use of the CRC Brake and Caliper grease to minimize worm to ring gear "chatter" that was observed by me and others....if you put the wrapped cord hanging off the West side by mistake.  This particular grease has 3 lubricants: MolyD, Graphite, and Teflon in a heavy very wide temp range (-30C to +200C?) anticorrosion grease.  It eliminated chatter observed using other greases like Super lube.  

Best of luck,

Michael


On Mon, Jun 28, 2021, 4:29 AM pcboreland via groups.io <pcboreland=yahoo.com@groups.io> wrote:
Alan,

I will try a rigid one piece coupler. My RA has the OPW. I've found it's possible to feel it the gearbox is aligned or not with the coupler by rotating worm with your fingers. The nice thing about the Mclennan gearboxes is they have a metal plate.  I have removed the motor cover as is is really cosmetic, so will then be able to use a hex head screw to tighten down the motor box more sensitively to hopefully get a good alignment. My planned testing on this issue is as follows.

1. Install a new worm undamaged by the use of the Oldham coupler set screws. Use a dial caliper to measure any wobble in the worm for a baseline measurement.

2. Install the gearbox using two different clamp style couplers from Ruland. I have already posted the part number of one and will get the part number for a rigid coupler if there is one with the correct O.D. Repeat the dial measurement.

3. Run an imaging session and get new FFT data.

The reason for these tests is mostly to better understand to magnitude of errors caused by different elements in the drive chain. I've already demonstrated how much better the FFT looks with the 125:1 gearbox in terms of higher frequency components. I'll be testing the 50:1 and 80:1 gearboxes in near future too. 

Currently, there is an 80s and a 240s error as show early in the FFT plot I posted. These have a corrected magnitude of 0.35 arcsec using the hysteresis PHD2 algorithm. I've not found the Predictive PEC algorithm works. These numbers change depending on which site of the meridian I'm on. Usually, after the flip the 240s goes away completely, but since my current project is the Cygnes Wall no flip has been needed so the error is a problem for the resolution I'm at.

Peter


Peter Boreland
 

Alan,

I will try a rigid one piece coupler. My RA has the OPW. I've found it's possible to feel it the gearbox is aligned or not with the coupler by rotating worm with your fingers. The nice thing about the Mclennan gearboxes is they have a metal plate.  I have removed the motor cover as is is really cosmetic, so will then be able to use a hex head screw to tighten down the motor box more sensitively to hopefully get a good alignment. My planned testing on this issue is as follows.

1. Install a new worm undamaged by the use of the Oldham coupler set screws. Use a dial caliper to measure any wobble in the worm for a baseline measurement.

2. Install the gearbox using two different clamp style couplers from Ruland. I have already posted the part number of one and will get the part number for a rigid coupler if there is one with the correct O.D. Repeat the dial measurement.

3. Run an imaging session and get new FFT data.

The reason for these tests is mostly to better understand to magnitude of errors caused by different elements in the drive chain. I've already demonstrated how much better the FFT looks with the 125:1 gearbox in terms of higher frequency components. I'll be testing the 50:1 and 80:1 gearboxes in near future too. 

Currently, there is an 80s and a 240s error as show early in the FFT plot I posted. These have a corrected magnitude of 0.35 arcsec using the hysteresis PHD2 algorithm. I've not found the Predictive PEC algorithm works. These numbers change depending on which site of the meridian I'm on. Usually, after the flip the 240s goes away completely, but since my current project is the Cygnes Wall no flip has been needed so the error is a problem for the resolution I'm at.

Peter


alan137
 

I'm just saying, that from a theoretical perspective, a one-piece coupler maintains a kind of symmetry where it's indifferent to whatever the shaft angle is, and thus it will rotate at a constant speed.  This is as opposed to the oldham coupler where the center plastic piece slides back and forth, and thus the "gear ratio" changes a little bit for different orientations.  Remember those Shimano Biopace elliptical chainrings?
Another nice thing about the ruland type couplings is that if the colinear alignment is off, it will be rather obvious since it won't fit.

With that being said, Michael is right that there are many different causes for the 240s error.  Maybe try purposely mis-aligning the oldham coupler and see if the problem gets worse.
I forgot what system you have, but is there some way to pre-align the worm and gear shafts before putting on the coupler?


Michael Herman
 

I'm not Alan...but I'll pitch in some ideas... 

240 sec is the worm period of a G11 (and half the worm period of a GM8).  {Actually 239.344 sec}

The term "confounded" is used in statistics to mean when several variables can have the same effect on an outcome...so you can't tell what variable is making the effect.  

The 239 sec worm period is confounded with everything that returns to it's same position at that time:
1. The worm itself...can be bad.  Try exchanging the RA and DEC worms to see which gives the lowest PE...then keep the lower one on RA.
2. The Oldham coupler.  Ideally you want the Oldham coupler input and output shafts to be exactly in the same line.  Watch for the center white plastic part movement during worm driving...if that white plastic part is moving up and down or in and out, you should try to adjust your gearbox location (or inner drive shaft location if you have a tucked motor design)  (you may have to enlarge or slot the mounting ear holes if the older direct drive design) to minimize the movement of that center white part.
3. Transfer gears can wobble if you have a tucked motor system. Don't mesh these too tightly... have no tight point.
4. Inner transfer gear drive shaft can wobble if you have a tucked motor system.  
5. The gearbox drive shaft final gear also rotates at the same rate as the worm. Try a different gearbox.

As Alan has said, the worm period of all these mounts is very long (239 sec on a G11) and can autoguide out very nicely most of the time.  It is the faster oscillation from a bad gearbox (1/7.5 of the worm period or 32 sec on a G11), or a bad RA worm bearing ball getting pinched (~78-80 sec on a G11) that really challenge autoguiding efforts.

Have fun,
Michael






On Sun, Jun 27, 2021, 12:16 PM pcboreland via groups.io <pcboreland=yahoo.com@groups.io> wrote:
Alan,

Are saying that a piece coupler might be best overall in minimizing co-linear problem of the two shafts and provide an overall lower  240s error? Do you have any recommendations?

Peter


Peter Boreland
 

Alan,

Are saying that a piece coupler might be best overall in minimizing co-linear problem of the two shafts and provide an overall lower  240s error? Do you have any recommendations?

Peter


alan137
 

The set screw is supposed to dig into the shaft of whatever it screws into.  If you like, you can file or mill a small flat for it.  It is the bearing that sets the concentricity of the worm, not the set screw.  In theory, the coupler is supposed to take into account that the worm shaft is pushed over on one side of its bore.  OF course, you are supposed to keep jamming the set screw into the same spot, otherwise all bets are off.  If you have a 0.0001" micrometer, you can check what is the actual shaft diameter of the worm where it passes thru the bearing.
The problem is not that of worm run-out, the problem is if the worm and [whatever shaft drives it on the other side of the coupler] are not co-linear.  One piece couplers will not give a periodic error even in this case.  The oldham might - I don't know how much.  Just re-adjust the input shaft, then.
The 240s error is not much to worry about because it is SLOW and will easily guide out.  The 80 / 76 s error is 3x worse, even though it has the same amplitude, it has 3x the slew rate.


Peter Boreland
 
Edited

Eric,

Isolating the cause of these errors is a difficult process I think. Really I need to test different adjustments, configurations, etc during daylight hours.  I do have a 0.0001" dial indicator, so I'm thinking of drilling an access hole in the worm block to get access to the worm shaft to make proper measures of the wobble in realtime. We are looking at 1/3rd and 1 whole period errors, these would be visible to read on the dial indicator I think. I've made the assumption that the 80s error is either a harmonic of the worm or the bearings. I do not think it is from the gearbox. I have assumed any coupler misalignment contributes to the 240s error, but may be it is part of creating a harmonic of the worm period.

There are other factors at play too, particularly RA mount balance, leading to bearing and worm loading. I'm not sure there has been a realtime study as such.

Peter


Cyclone
 

Peter,
I believe the 76 and 80 second PE components are caused by different mechanisms.

The 76 second component is the bearing BPFO (that period can be calculated from the bearing specifications), which does not appear to be visible on the FFT plot you posted, but difficult to say on a low resolution image. It should be fairly obvious on a chart in the time domain, as the 76 second component is not in phase with the worm.

With the 25:1 McLennan gearbox on my mount, the last stage has a period of 80 sec, so it is difficult to tell if the 80 sec component comes from the gearbox or between the gearbox and the worm.

Also, my mount came with the one piece worm block and Ruland disk coupling, as per your picture. I checked the coupling runout using a dial indicator and it has a significant jump; one of the disks may be damaged as this type of coupling does not allow much misalignment between the gearbox and the worm. Below is the FFT for my mount:


Eric


Peter Boreland
 

With a little digging one quickly realizes that problems with the design of Losmandy's worm block goes back 15 years or more. Their solution is the OPW and more recently to spring load it. I feel they could have done a so much better job in the intervening years. Their dimensional choice for the blocks really limits the choice of available couplers. The use of set screws that dig into the shaft of the gearbox and the worm I find totally unacceptable. If you need to remove the worm one will need to file it to get it to pass through the bearings. The worm end will longer be round and will not line up with the coupler without play, guaranteeing a larger 240s error. I just had to buy a second one in less than 2 months due to this. If you want to use a clamping coupler you only have two choices as far as I can tell. These are both from Ruland. Here is the coupler I'm currently using https://www.ruland.com/servo-couplings/jaw-couplings.html    It does not do the job I feel as it does not remove any misalignment, Ruland has Oldham style couplers, but nothing that is clamp style under 3/4" O.D. There is the seriously more expensive disc coupler https://www.ruland.com/dcd10-1-4-4mm-a.html  This version has 1/4" I.D. at one end and 4mm at the other. It's $182! If anyone has experience with this part and feels it work better than what I'm presently using please advise. 

This image of the OPW from 2010 seems show show the use of this disc coupler, since replaced by a horrid $20 part with set screws. 



This article talks in detail about the earlier 76s error and the 80s error with the OPW and its couases http://helixgate.net/G11opw8.html

Here is my FFT plot from the other night. 




Right now PEC is not working for me for reasons discussed below. I'm optimistic this can be resolved. With the 125:1 McLennan gearbox there very little gear noise. This is great and what I was hoping for. The 80s (bearings) and the 240s (worm + coupler) are the real problem to be focused on. I'm wondering if the axial load on the bearings causing the 80s error can be reduced with bevel washers on the worm side at both ends? I'm also wondering if the 240s error could be reduced with the use of the $182 disc coupler? Your kind input would be appreciate, and sorry for ranting!

Peter


Brendan Smith
 

Ahhh thankyou Peter

I'd doubt the slight divisor "point error" would cause it to lag that much

Cheers
--
Brendan


Peter Boreland
 

On Fri, Jun 25, 2021 at 07:28 PM, Ray Gralak wrote:
This slight difference in RA tracking divisor should not cause much drift. Something else is likely wrong.
Ray,

Any thoughts what it could be?

Peter


Peter Boreland
 

On Fri, Jun 25, 2021 at 04:51 PM, Brendan Smith wrote:
Ummm My calc's for the tracking divisor are different:
RA divisor = (1,500,000/prescaler)/(steps per rev/worm period)

Steps per rev = 256 x 125 = 32,000

Its a G11 isn't it?   Thus worm period is the same 239.344658
Pre-scaler = 1
Steps per worm revolution = 32,000

Thus Ra divisor = ((1,500,000/1)(/32,000/239.344658)) = 11,219.2    This is my figure?

Even if its a Titian I get 14959......

I'm confused about how you got a Ra divisor of 89754
Brenden,

You had me going there for a while. That number I believe is for Gemini 1. For Gemini II with the gearbox set at 25:1 the divisor is 448771, which I then divided by 5 for the 125:1 gearbox, which give you 89754.2. The controller automatically puts in these numbers. 

Peter


Ray Gralak
 

My only though right now is the RA tracking divisor of 89754 ought to be 89754.2. If you let
the mount track for a couple of hours the mount does fall behind quite a bit.
This slight difference in RA tracking divisor should not cause much drift. Something else is likely wrong.

-Ray

-----Original Message-----
From: Losmandy_users@groups.io [mailto:Losmandy_users@groups.io] On Behalf Of pcboreland via groups.io
Sent: Friday, June 25, 2021 3:14 PM
To: Losmandy_users@groups.io
Subject: Re: [Losmandy_users_io] Changing the servo motor encoders from 256 to 512 or 1024. Any reason not
to do this?

On Thu, Jun 24, 2021 at 05:34 AM, Ray Gralak wrote:
Ray,



1. Either your graph is incorrect, or PEC is introducing a lot of drift. By the end of the worm cycle the
accumulated PEC value should be 0 (but it's not).

I tried numerous times last night to create a PEC curve that starts and ends at zero or the same point. If you try to
use any curve that does not you will of coarse get a sudden jump in guiding so they are not usable and one has to
keep PEC turned off. My only though right now is the RA tracking divisor of 89754 ought to be 89754.2. If you let
the mount track for a couple of hours the mount does fall behind quite a bit. This problem will apply to any gearbox
ratio other than 25:1. The only solution that I can see is to download the PEC curve and not use Gemini II to
calculate it.

Peter


Brendan Smith
 

Ummm My calc's for the tracking divisor are different:
RA divisor = (1,500,000/prescaler)/(steps per rev/worm period)

Steps per rev = 256 x 125 = 32,000

Its a G11 isn't it?   Thus worm period is the same 239.344658
Pre-scaler = 1
Steps per worm revolution = 32,000

Thus Ra divisor = ((1,500,000/1)(/32,000/239.344658)) = 11,219.2    This is my figure?

Even if its a Titian I get 14959......

I'm confused about how you got a Ra divisor of 89754

--
Brendan


Peter Boreland
 

On Thu, Jun 24, 2021 at 05:34 AM, Ray Gralak wrote:
Ray,

1. Either your graph is incorrect, or PEC is introducing a lot of drift. By the end of the worm cycle the accumulated PEC value should be 0 (but it's not).
I tried numerous times last night to create a PEC curve that starts and ends at zero or the same point. If you try to use any curve that does not you will of coarse get a sudden jump in guiding so they are not usable and one has to keep PEC turned off.  My only though right now is the RA tracking divisor of 89754 ought to be 89754.2. If you let the mount track for a couple of hours the mount does fall behind quite a bit. This problem will apply to any gearbox ratio other than 25:1. The only solution that I can see is  to download the PEC curve and not use Gemini II to calculate it.

Peter


Peter Boreland
 

Paul,

Yes indeed. Through Michael Herman I have communicated my thoughts and specifically what might be a bug. I'm hoping Rene will in time clarify things. It is important to know the true limits of the system and have an accurate spec. 

Peter


Paul Kanevsky
 

That's good to hear, Peter. Since the document I referenced is directly from Rene, the author of Gemini, I'd like to hear from him before making any corrections to it.

Regards,

      -Paul


On Thu, Jun 24, 2021 at 06:22 PM, <pcboreland@...> wrote:
Paul, 

I've confirmed that PECmax can be set to 32000 and everything works just fine. I monitored the PEC index through the web interface and it counts up to 32000. It's clearly misinformation. I'm glade I decided to test the limit for myself.


Peter


Peter Boreland
 

Paul, 

I've confirmed that PECmax can be set to 32000 and everything works just fine. I monitored the PEC index through the web interface and it counts up to 32000. It's clearly misinformation. I'm glade I decided to test the limit for myself.


Peter


Paul Kanevsky
 

By serial command.


On Thu, Jun 24, 2021 at 11:53 AM, crocco1250 wrote:
Paul

Is the pecmax value set by serial cmd or by the webpage?


crocco1250
 

Paul

Is the pecmax value set by serial cmd or by the webpage?

Chuck

On Wednesday, June 23, 2021, 09:02:07 PM PDT, Paul Kanevsky <yh@...> wrote:


Hi Peter,

PECMax value can be set by a command to Gemini. This can even be multiples of the worm cycle. I'm sorry if you already know this, I'm not up to speed as to what you're trying to test.

Regards,

     -Paul

On Wed, Jun 23, 2021 at 10:12 PM, <pcboreland@...> wrote:
Tonight I was able to test  if the PECmax limit  was indeed 25600 as some have suggested. As this image below shows the the counter went up to 3200 and rolled over. I run the test using a synthetic PEC file I created. The fact that PEC would not record the PEC curve correctly with the steps per worm revolution at 3200, is I think a separate issue. Using PEMProp I should be able to create a PEC curve and download it to the mount.  Keeping fingers crossed.




Paul Kanevsky
 

Maybe this can help. From https://gemini-2.com/web/L5V1serial.html



Regards,

      -Paul


Ray Gralak
 

I have something exciting to report. I was successful in recording a PEC curve with 32000 steps. It seems there is
a bug with Gemini II PEC recording. When I first attempted to record a curve it stopped after 6400 steps. Since
uploading the dummy test file where the PECmax variable set to 32000, it now seems to work just fine. Go figure.
Here is the curve I recorded. I like its fine detail.
1. Either your graph is incorrect, or PEC is introducing a lot of drift. By the end of the worm cycle the accumulated PEC value should be 0 (but it's not).

2. Whether PEC is using 32000 should be confirmed by sampling the PEC index from the mount.

-Ray

-----Original Message-----
From: Losmandy_users@groups.io [mailto:Losmandy_users@groups.io] On Behalf Of pcboreland via groups.io
Sent: Wednesday, June 23, 2021 8:47 PM
To: Losmandy_users@groups.io
Subject: Re: [Losmandy_users_io] Changing the servo motor encoders from 256 to 512 or 1024. Any reason not
to do this?

I have something exciting to report. I was successful in recording a PEC curve with 32000 steps. It seems there is
a bug with Gemini II PEC recording. When I first attempted to record a curve it stopped after 6400 steps. Since
uploading the dummy test file where the PECmax variable set to 32000, it now seems to work just fine. Go figure.
Here is the curve I recorded. I like its fine detail.



Paul Kanevsky
 

No, it’s currently only available using  a serial command. A simple script could be used to change it.

Regards,

   -Paul

On Thu, Jun 24, 2021 at 12:14 AM, <pcboreland@...> wrote:

Paul,

Is this from the hand controller? I do not see this feature in the web interface or the screen based controller.

Peter


Michael Herman
 

Hi Paul,

As Peter tried, I also could not find PECmax in any of my G-2 web pages or the Gemini.net sub panels. 

 Can you guide us to where that setting is available?

Thanks,
Michael

On Wed, Jun 23, 2021, 9:14 PM pcboreland via groups.io <pcboreland=yahoo.com@groups.io> wrote:
Paul,

Is this from the hand controller? I do not see this feature in the web interface or the screen based controller.

Peter


alan137
 

I would have to completely clean off the old grease from the worm and ring before adding the new grease.
I think you should wait for a day when seeing is good enough that you don't even need to apply guide pulses to DEC, and then you can really tell if the RA axis adds noise or not.  Or use the PHD guiding assistant which will temporarily turn guiding off.


Peter Boreland
 

Paul,

Is this from the hand controller? I do not see this feature in the web interface or the screen based controller.

Peter


Paul Kanevsky
 

Hi Peter,

PECMax value can be set by a command to Gemini. This can even be multiples of the worm cycle. I'm sorry if you already know this, I'm not up to speed as to what you're trying to test.

Regards,

     -Paul


On Wed, Jun 23, 2021 at 10:12 PM, <pcboreland@...> wrote:
Tonight I was able to test  if the PECmax limit  was indeed 25600 as some have suggested. As this image below shows the the counter went up to 3200 and rolled over. I run the test using a synthetic PEC file I created. The fact that PEC would not record the PEC curve correctly with the steps per worm revolution at 3200, is I think a separate issue. Using PEMProp I should be able to create a PEC curve and download it to the mount.  Keeping fingers crossed.




Peter Boreland
 

I have something exciting to report. I was successful in recording a PEC curve with 32000 steps. It seems there is a bug with Gemini II PEC recording. When I first attempted to record a curve it stopped after 6400 steps. Since uploading the dummy test file where the PECmax variable set to 32000, it now seems to work just fine. Go figure. Here is the curve I recorded. I like its fine detail.


Peter Boreland
 

Tonight I was able to test  if the PECmax limit  was indeed 25600 as some have suggested. As this image below shows the the counter went up to 3200 and rolled over. I run the test using a synthetic PEC file I created. The fact that PEC would not record the PEC curve correctly with the steps per worm revolution at 3200, is I think a separate issue. Using PEMProp I should be able to create a PEC curve and download it to the mount.  Keeping fingers crossed.




Peter Boreland
 

Hi Alan,

I've looked quite a few peoples guiding stats and that is always seems to the case. Bold statement I know! Does anyone disagree? The fact that increasing the Ra gear ratio inverted the Ra/Dec relationship in my testing thus far is interesting, but it's early days.  I've not yet worked on the Dec axis at all, waiting on parts. I'm optimistic  that I can get similar guiding results during the summer as I had during the winters months when seeing was so much better. It's not worm chatter?  Not sure why you have to mix greases?

Peter


alan137
 

My RA guiding is also not as good as DEC.  Sometimes I wonder if it really is due to the worm chatter, which may be fixed with the CRC brake and caliper fix like Michael Herman suggested.  I haven't tried it myself since I don't like to mix greases.


Peter Boreland
 

Eric, Yes, and I assume 20480 with a 80:1 gearbox. The count value at which it rolls over is defined by PECmax.
My test is with a PEC file that has PECmax set to 32000. In that case it should roll over at 25600? Or do I have this wrong?

Peter


Cyclone
 

Peter, the PEC index will roll over at 12800 with 50:1 gearbox numbers entered in the Gemini, not 25600. And you don’t need to load a PEC file, the PEC index is always running.

Eric


Peter Boreland
 

Eric,

I see the pec pointer on the sate page. Thanks! I'll run my hand made pec file and see if it rolls over at 25600.

Peter


Peter Boreland
 

Eric,

I just looked at the document. This has to be a variable length issue possibly.  Will move the 125:1 gearbox to Dec and go with 50:1 for Ra. I just bought a second one😬 If anyone what's to buy it from me let me know, I will not open it. 

There is also an 80:1 ratio gearbox made by RS Pro, RS part number 440537 which I just found. This would provide a 0.175 arc sec resolution. Not many details on it though. Looks like one would have to locate, drill and tap the motor mounting holes. Really 0.11 arcsec was over kill. Would be nice not to have to slew so slowly. 

Oh well I'll just keep spending money so you guys don't have to😀

Bottom line is a higher ratio gearbox, I.e. more steps per worm cycle, does have a significant effect on guiding performance.  This past week my guiding on consecutive nights when from RA/Dec 1.02/0.79 arcsec (25:1) to 0.78/0.88 arcsec (125:1). RA was better than Dec over three 90 min imaging sessions by same margin. Looking at the Dec number seeing was not as good second night and has generally being getting much worse as June progresses.  For comparison purposes: May 20th was 0.79/0.54 arcsec. May 6th was 0.64/0.45 arcsec. Always RA not as good as Dec axis.

Peter


Ray Gralak
 

I thought this too, but the mount (worn turn) does have integer multiples.
There are other sources for non-integer multiples. For example, the 76-second cycle found in some mounts:

http://helixgate.net/G11opw8.html

-Ray

-----Original Message-----
From: Losmandy_users@groups.io [mailto:Losmandy_users@groups.io] On Behalf Of pcboreland via groups.io
Sent: Friday, June 18, 2021 2:04 PM
To: Losmandy_users@groups.io
Subject: Re: [Losmandy_users_io] Changing the servo motor encoders from 256 to 512 or 1024. Any reason not
to do this?

[Edited Message Follows]

On Fri, Jun 18, 2021 at 11:38 AM, Brendan Smith wrote:


And, worse, if the mount has any non-integer fundamentals, the recorded PEC curve will not match
every worm cycle because each worm cycle will vary

Thankyou Ray. Explains quite the opposite effect to many are striving for

I thought this too, but the mount (worn turn) does have integer multiples. It is defined by the steps per revolution
(6400 for the 25:1 gearbox). I noticed in "configure PEC" a Set PEC Period selection and a drop down box. There
are multiple options 1x6400, 2x6400 up to I think 4x6400. I had expected this to change to being indexed to
32000, but the pec recording only indexed up to 6400. I wonder what would happen if one created the PEC curve
using third party software. How is that curve indexed?

I have ordered. 50:1 gearbox. I can confirm the the small gear can be reamed out to 1/8th. I'm evermore
convinced that upping the hear ratio is defiantly worth doing. There is the option of changing out the encoder too,
but these PEC issues would still remain. I confident we can solve this as a group.


Cyclone
 

Peter,
The Gemini 2 serial commands document says:
27     2000..25600 Number of RA steps for one worm revolution (since is a product of spur ratio and motor encoder ratio,this command can only be used for reading out the maximum step count, not for setting it).
Note: This product must not exceed 25600. If higher values are reported the combination of RA spur gear ratio and motor encoder is invalid.

Therefore, the limit is 25600, and Gemini probably defaults to 6400 if the gear ratio and encoder combination is invalid. The 50:1 gearbox will result in a valid combination, so PEC should work. You can do the tests I suggested previously to confirm the PEC index will roll over at 12800 while you wait for your new gearbox.

Eric


Peter Boreland
 

I took a look at the PEC file and ran a quick test to see if a curve with 3200 steps is possible and it seems to be so. The PEC process is very simple. I only changed out one gearbox, I wonder if PEC is getting the step information from the Dec axis by mistake which has 6400 steps?




Peter Boreland
 
Edited

On Fri, Jun 18, 2021 at 11:38 AM, Brendan Smith wrote:

And, worse, if the mount has any non-integer fundamentals, the recorded PEC curve will not match every worm cycle because each worm cycle will vary

Thankyou Ray.  Explains quite the opposite effect to many are striving  for 

I thought this too, but the mount (worn turn) does have integer multiples. It is defined by the steps per revolution (6400 for the 25:1 gearbox). I noticed in "configure PEC" a Set PEC Period selection and a drop down box. There are multiple options 1x6400, 2x6400 up to I think 4x6400. I had expected this to change to being indexed to 32000, but the pec recording only indexed up to 6400. I wonder what would happen if one created the PEC curve using third party software. How is that curve indexed?

I have ordered. 50:1 gearbox. I can confirm the the small gear can be reamed out to 1/8th. I'm evermore convinced that upping the hear ratio is defiantly worth doing. There is the option of changing out the encoder too, but these PEC issues would still remain. I confident we can solve this as a group.


Michael Herman
 

Good question, about pulse widths!, but ..I do not know the answer.   The pulse widths should be varied and capable of being quite short by the uP, but I have not scoped out the actual pulses to see how short they do go.
  
If Rene were listening in, he could tell you the answer immediately.  I emailed him this morning about the PEC riddle.  

Best,
Michael





On Fri, Jun 18, 2021, 1:16 PM crocco1250 via groups.io <crocco1250=aol.com@groups.io> wrote:
Michael

Is the pulse width to the motor fixed in duration such that it results in 256 tics per rev to match the optical encoder or can the pulse width vary. Since the quadriture can determine where the motor is at (0, pi/2, pi, 3pi/2, 2pi) with much finer resolution, it could vary but does it? When you see phd graphs of the RA axis trying to "catch up" on poorly aligned mounts, you see multiple pulses being sent, one after the other, resulting in overshoot, this appears to be constant duration pulses. Or am I missing something?

Chuck
On Friday, June 18, 2021, 08:18:43 AM PDT, Michael Herman <mherman346@...> wrote:


Nice catch, Ray.  That could explain the puzzle... and also my puzzle on my 1:2 belt drive G11T.  

(Please see my PEMpro question at the bottom of this email...)

Some thoughts... 

The number of data points recorded in the PEC run is very small.  For my 1:2 belt drive recorded at 1 sec intervals, the number of points is only 2*319 = 638 points.  Even recording at 0.1 sec intervals this is 6,380 points.   (There is a CPU RAM storage limit of 96 kbytes but there is also large external Flash available.)

Ray suggests that the optical encoder tick counts per worm rotation has to do with the puzzle.  

A worm rotation value in raw optical encoder ticks is:
  256 ticks per motor rev
x  25 stock gearbox (or 50 for me, 125 for PCBoreland) gearbox rotation reduction.

At 25 you get 6400
At 50 you get 12800
At 125 you get 32000 (over 25,600 Ray mentions as a limit)

The stock 25:1 and 50:1 gearbox values are at or below 25,600 Ray mentions as a limit for encoder nominal ticks.  The 125:1 gearbox goes higher.  

However if the system is counting "quadruture"  ticks (and we know it can)  then the values become
At 25 you get 6400 -> 25,600 
At 50 you get 12800 -> 51,200 (>>25,600 or 32,767)
At 125 you get 32000->128,000 (>>25,600 or 32,767)

I mention +/- 32,767 as the size of a C "int" 2 byte number.  You can double the "int" value using "uint" to go from 0 to 65535 still taking only 2 bytes. 

The question is: why is the tick count limit 25,600 or 32,000?
 Isn't there sufficient RAM or Flash memory? Spec sheet says there is 98kb of RAM on the LPC2388 ARM CPU chip.  Maybe using only on-CPU RAM is the limiting problem.  There should be huge room on the Flash drive for the data.  

An int size 2 bytes can go to +/- 32767
A uint size 2 bytes can go 0 to 65535 
A long size 4 bytes can go to +/-  (huge)

This now seems like a firmware code or memory storage issue. It would need the memory to allocate for a larger tick number, or more PEC data collection numbers, or both.   If anyone can address it, it will be Rene. 

Ray:  will PEMpro generates a useable PEC curve for gearboxes above 25:1.....or ....will the PEMpro PEC curve also fail if the problem is with the G-2 tick size data storage in the present firmware?  

All the best,
Michael

  




On Fri, Jun 18, 2021, 7:24 AM Ray Gralak <iogroups@...> wrote:
> PEC training quits after 48s, 1/5th period of the worm. I think this has to do Pec working with 6400 ticks instead of
> 32000.

32000 is above the maximum number of ticks, which I think is 25600.

The tick count probably wraps at 25,600 so 32000-25600 = 6400, which would explain the 1/5 cycle.

-Ray

> -----Original Message-----
> From: Losmandy_users@groups.io [mailto:Losmandy_users@groups.io] On Behalf Of pcboreland via groups.io
> Sent: Friday, June 18, 2021 6:22 AM
> To: Losmandy_users@groups.io
> Subject: Re: [Losmandy_users_io] Changing the servo motor encoders from 256 to 512 or 1024. Any reason not
> to do this?
>
> Ray,
>
> PEC training quits after 48s, 1/5th period of the worm. I think this has to do Pec working with 6400 ticks instead of
> 32000.
>
> Peter
>







crocco1250
 

Michael

Is the pulse width to the motor fixed in duration such that it results in 256 tics per rev to match the optical encoder or can the pulse width vary. Since the quadriture can determine where the motor is at (0, pi/2, pi, 3pi/2, 2pi) with much finer resolution, it could vary but does it? When you see phd graphs of the RA axis trying to "catch up" on poorly aligned mounts, you see multiple pulses being sent, one after the other, resulting in overshoot, this appears to be constant duration pulses. Or am I missing something?

Chuck

On Friday, June 18, 2021, 08:18:43 AM PDT, Michael Herman <mherman346@...> wrote:


Nice catch, Ray.  That could explain the puzzle... and also my puzzle on my 1:2 belt drive G11T.  

(Please see my PEMpro question at the bottom of this email...)

Some thoughts... 

The number of data points recorded in the PEC run is very small.  For my 1:2 belt drive recorded at 1 sec intervals, the number of points is only 2*319 = 638 points.  Even recording at 0.1 sec intervals this is 6,380 points.   (There is a CPU RAM storage limit of 96 kbytes but there is also large external Flash available.)

Ray suggests that the optical encoder tick counts per worm rotation has to do with the puzzle.  

A worm rotation value in raw optical encoder ticks is:
  256 ticks per motor rev
x  25 stock gearbox (or 50 for me, 125 for PCBoreland) gearbox rotation reduction.

At 25 you get 6400
At 50 you get 12800
At 125 you get 32000 (over 25,600 Ray mentions as a limit)

The stock 25:1 and 50:1 gearbox values are at or below 25,600 Ray mentions as a limit for encoder nominal ticks.  The 125:1 gearbox goes higher.  

However if the system is counting "quadruture"  ticks (and we know it can)  then the values become
At 25 you get 6400 -> 25,600 
At 50 you get 12800 -> 51,200 (>>25,600 or 32,767)
At 125 you get 32000->128,000 (>>25,600 or 32,767)

I mention +/- 32,767 as the size of a C "int" 2 byte number.  You can double the "int" value using "uint" to go from 0 to 65535 still taking only 2 bytes. 

The question is: why is the tick count limit 25,600 or 32,000?
 Isn't there sufficient RAM or Flash memory? Spec sheet says there is 98kb of RAM on the LPC2388 ARM CPU chip.  Maybe using only on-CPU RAM is the limiting problem.  There should be huge room on the Flash drive for the data.  

An int size 2 bytes can go to +/- 32767
A uint size 2 bytes can go 0 to 65535 
A long size 4 bytes can go to +/-  (huge)

This now seems like a firmware code or memory storage issue. It would need the memory to allocate for a larger tick number, or more PEC data collection numbers, or both.   If anyone can address it, it will be Rene. 

Ray:  will PEMpro generates a useable PEC curve for gearboxes above 25:1.....or ....will the PEMpro PEC curve also fail if the problem is with the G-2 tick size data storage in the present firmware?  

All the best,
Michael

  




On Fri, Jun 18, 2021, 7:24 AM Ray Gralak <iogroups@...> wrote:
> PEC training quits after 48s, 1/5th period of the worm. I think this has to do Pec working with 6400 ticks instead of
> 32000.

32000 is above the maximum number of ticks, which I think is 25600.

The tick count probably wraps at 25,600 so 32000-25600 = 6400, which would explain the 1/5 cycle.

-Ray

> -----Original Message-----
> From: Losmandy_users@groups.io [mailto:Losmandy_users@groups.io] On Behalf Of pcboreland via groups.io
> Sent: Friday, June 18, 2021 6:22 AM
> To: Losmandy_users@groups.io
> Subject: Re: [Losmandy_users_io] Changing the servo motor encoders from 256 to 512 or 1024. Any reason not
> to do this?
>
> Ray,
>
> PEC training quits after 48s, 1/5th period of the worm. I think this has to do Pec working with 6400 ticks instead of
> 32000.
>
> Peter
>







Brendan Smith
 

And, worse, if the mount has any non-integer fundamentals, the recorded PEC curve will not match every worm cycle because each worm cycle will vary

Thankyou Ray.  Explains quite the opposite effect to many are striving  for 

cheers
--
Brendan


Michael Herman
 

My Goto's with my G11T 1:2 belt drive and 25:1 McLennan gearbox are spot on.  That works fine.

And when I used a 50:1 McLennan and the stock 3 transfer gears on tucked Titan RA the GoTos also worked fine.  And PCBoreland's report on the 125:1 gearbox also said that worked fine.  Only PEC showed trouble. 

Those drive systems and gotos seem to have all been tested and debugged correctly.

It's the PEC that ... needs debugging and a solution for the non-stock gear ratios above 25:1.  

I'll hope Rene sees these notes and enters the fray!

It is also good that these issues are seeing the light of day so they can get resolved.  Any one person who observed a problem was considered an anomaly.  Nowit is clear that there is a consistent PEC issue and it comes about when the gear ratios are over 25:1.   

Best regards,
Michael

On Fri, Jun 18, 2021, 9:10 AM Ray Gralak <iogroups@...> wrote:
Hi Michael,

> Ray:  will PEMpro generates a useable PEC curve for gearboxes above 25:1.....or ....will the PEMpro PEC curve
> also fail if the problem is with the G-2 tick size data storage in the present firmware?

PEMPro will retrieve the count from the Gemini and use that to calculate the PEC curve. In this case, the number would be 6400, which would be wrong, so I suspect the resulting curve would be bad.

Depending on the firmware implementation, other things might not work as expected, like GOTO's and autoguiding. Have you tried either?

-Ray

> -----Original Message-----
> From: Losmandy_users@groups.io [mailto:Losmandy_users@groups.io] On Behalf Of Michael Herman
> Sent: Friday, June 18, 2021 8:18 AM
> To: Losmandy_users@groups.io
> Subject: Re: [Losmandy_users_io] Changing the servo motor encoders from 256 to 512 or 1024. Any reason not
> to do this?
>
> Nice catch, Ray.  That could explain the puzzle... and also my puzzle on my 1:2 belt drive G11T.
>
> (Please see my PEMpro question at the bottom of this email...)
>
> Some thoughts...
>
> The number of data points recorded in the PEC run is very small.  For my 1:2 belt drive recorded at 1 sec intervals,
> the number of points is only 2*319 = 638 points.  Even recording at 0.1 sec intervals this is 6,380 points.   (There is
> a CPU RAM storage limit of 96 kbytes but there is also large external Flash available.)
>
> Ray suggests that the optical encoder tick counts per worm rotation has to do with the puzzle.
>
> A worm rotation value in raw optical encoder ticks is:
>   256 ticks per motor rev
> x  25 stock gearbox (or 50 for me, 125 for PCBoreland) gearbox rotation reduction.
>
> At 25 you get 6400
> At 50 you get 12800
> At 125 you get 32000 (over 25,600 Ray mentions as a limit)
>
> The stock 25:1 and 50:1 gearbox values are at or below 25,600 Ray mentions as a limit for encoder nominal ticks.
> The 125:1 gearbox goes higher.
>
> However if the system is counting "quadruture"  ticks (and we know it can)  then the values become
> At 25 you get 6400 -> 25,600
> At 50 you get 12800 -> 51,200 (>>25,600 or 32,767)
> At 125 you get 32000->128,000 (>>25,600 or 32,767)
>
> I mention +/- 32,767 as the size of a C "int" 2 byte number.  You can double the "int" value using "uint" to go from 0
> to 65535 still taking only 2 bytes.
>
> The question is: why is the tick count limit 25,600 or 32,000?
>  Isn't there sufficient RAM or Flash memory? Spec sheet says there is 98kb of RAM on the LPC2388 ARM CPU
> chip.  Maybe using only on-CPU RAM is the limiting problem.  There should be huge room on the Flash drive for
> the data.
>
> An int size 2 bytes can go to +/- 32767
> A uint size 2 bytes can go 0 to 65535
> A long size 4 bytes can go to +/-  (huge)
>
> This now seems like a firmware code or memory storage issue. It would need the memory to allocate for a larger
> tick number, or more PEC data collection numbers, or both.   If anyone can address it, it will be Rene.
>
> Ray:  will PEMpro generates a useable PEC curve for gearboxes above 25:1.....or ....will the PEMpro PEC curve
> also fail if the problem is with the G-2 tick size data storage in the present firmware?
>
> All the best,
> Michael
>
>
>
>
>
>
> On Fri, Jun 18, 2021, 7:24 AM Ray Gralak <iogroups@...> wrote:
>
>
>       > PEC training quits after 48s, 1/5th period of the worm. I think this has to do Pec working with 6400 ticks
> instead of
>       > 32000.
>
>       32000 is above the maximum number of ticks, which I think is 25600.
>
>       The tick count probably wraps at 25,600 so 32000-25600 = 6400, which would explain the 1/5 cycle.
>
>       -Ray
>
>       > -----Original Message-----
>       > From: Losmandy_users@groups.io [mailto:Losmandy_users@groups.io] On Behalf Of pcboreland via
> groups.io
>       > Sent: Friday, June 18, 2021 6:22 AM
>       > To: Losmandy_users@groups.io
>       > Subject: Re: [Losmandy_users_io] Changing the servo motor encoders from 256 to 512 or 1024. Any
> reason not
>       > to do this?
>       >
>       > Ray,
>       >
>       > PEC training quits after 48s, 1/5th period of the worm. I think this has to do Pec working with 6400 ticks
> instead of
>       > 32000.
>       >
>       > Peter
>       >
>
>
>
>
>
>
>
>
>







Ray Gralak
 

Hi Michael,

Ray: will PEMpro generates a useable PEC curve for gearboxes above 25:1.....or ....will the PEMpro PEC curve
also fail if the problem is with the G-2 tick size data storage in the present firmware?
PEMPro will retrieve the count from the Gemini and use that to calculate the PEC curve. In this case, the number would be 6400, which would be wrong, so I suspect the resulting curve would be bad.

Depending on the firmware implementation, other things might not work as expected, like GOTO's and autoguiding. Have you tried either?

-Ray

-----Original Message-----
From: Losmandy_users@groups.io [mailto:Losmandy_users@groups.io] On Behalf Of Michael Herman
Sent: Friday, June 18, 2021 8:18 AM
To: Losmandy_users@groups.io
Subject: Re: [Losmandy_users_io] Changing the servo motor encoders from 256 to 512 or 1024. Any reason not
to do this?

Nice catch, Ray. That could explain the puzzle... and also my puzzle on my 1:2 belt drive G11T.

(Please see my PEMpro question at the bottom of this email...)

Some thoughts...

The number of data points recorded in the PEC run is very small. For my 1:2 belt drive recorded at 1 sec intervals,
the number of points is only 2*319 = 638 points. Even recording at 0.1 sec intervals this is 6,380 points. (There is
a CPU RAM storage limit of 96 kbytes but there is also large external Flash available.)

Ray suggests that the optical encoder tick counts per worm rotation has to do with the puzzle.

A worm rotation value in raw optical encoder ticks is:
256 ticks per motor rev
x 25 stock gearbox (or 50 for me, 125 for PCBoreland) gearbox rotation reduction.

At 25 you get 6400
At 50 you get 12800
At 125 you get 32000 (over 25,600 Ray mentions as a limit)

The stock 25:1 and 50:1 gearbox values are at or below 25,600 Ray mentions as a limit for encoder nominal ticks.
The 125:1 gearbox goes higher.

However if the system is counting "quadruture" ticks (and we know it can) then the values become
At 25 you get 6400 -> 25,600
At 50 you get 12800 -> 51,200 (>>25,600 or 32,767)
At 125 you get 32000->128,000 (>>25,600 or 32,767)

I mention +/- 32,767 as the size of a C "int" 2 byte number. You can double the "int" value using "uint" to go from 0
to 65535 still taking only 2 bytes.

The question is: why is the tick count limit 25,600 or 32,000?
Isn't there sufficient RAM or Flash memory? Spec sheet says there is 98kb of RAM on the LPC2388 ARM CPU
chip. Maybe using only on-CPU RAM is the limiting problem. There should be huge room on the Flash drive for
the data.

An int size 2 bytes can go to +/- 32767
A uint size 2 bytes can go 0 to 65535
A long size 4 bytes can go to +/- (huge)

This now seems like a firmware code or memory storage issue. It would need the memory to allocate for a larger
tick number, or more PEC data collection numbers, or both. If anyone can address it, it will be Rene.

Ray: will PEMpro generates a useable PEC curve for gearboxes above 25:1.....or ....will the PEMpro PEC curve
also fail if the problem is with the G-2 tick size data storage in the present firmware?

All the best,
Michael






On Fri, Jun 18, 2021, 7:24 AM Ray Gralak <iogroups@siriusimaging.com> wrote:


> PEC training quits after 48s, 1/5th period of the worm. I think this has to do Pec working with 6400 ticks
instead of
> 32000.

32000 is above the maximum number of ticks, which I think is 25600.

The tick count probably wraps at 25,600 so 32000-25600 = 6400, which would explain the 1/5 cycle.

-Ray

> -----Original Message-----
> From: Losmandy_users@groups.io [mailto:Losmandy_users@groups.io] On Behalf Of pcboreland via
groups.io
> Sent: Friday, June 18, 2021 6:22 AM
> To: Losmandy_users@groups.io
> Subject: Re: [Losmandy_users_io] Changing the servo motor encoders from 256 to 512 or 1024. Any
reason not
> to do this?
>
> Ray,
>
> PEC training quits after 48s, 1/5th period of the worm. I think this has to do Pec working with 6400 ticks
instead of
> 32000.
>
> Peter
>









Michael Herman
 

Nice catch, Ray.  That could explain the puzzle... and also my puzzle on my 1:2 belt drive G11T.  

(Please see my PEMpro question at the bottom of this email...)

Some thoughts... 

The number of data points recorded in the PEC run is very small.  For my 1:2 belt drive recorded at 1 sec intervals, the number of points is only 2*319 = 638 points.  Even recording at 0.1 sec intervals this is 6,380 points.   (There is a CPU RAM storage limit of 96 kbytes but there is also large external Flash available.)

Ray suggests that the optical encoder tick counts per worm rotation has to do with the puzzle.  

A worm rotation value in raw optical encoder ticks is:
  256 ticks per motor rev
x  25 stock gearbox (or 50 for me, 125 for PCBoreland) gearbox rotation reduction.

At 25 you get 6400
At 50 you get 12800
At 125 you get 32000 (over 25,600 Ray mentions as a limit)

The stock 25:1 and 50:1 gearbox values are at or below 25,600 Ray mentions as a limit for encoder nominal ticks.  The 125:1 gearbox goes higher.  

However if the system is counting "quadruture"  ticks (and we know it can)  then the values become
At 25 you get 6400 -> 25,600 
At 50 you get 12800 -> 51,200 (>>25,600 or 32,767)
At 125 you get 32000->128,000 (>>25,600 or 32,767)

I mention +/- 32,767 as the size of a C "int" 2 byte number.  You can double the "int" value using "uint" to go from 0 to 65535 still taking only 2 bytes. 

The question is: why is the tick count limit 25,600 or 32,000?
 Isn't there sufficient RAM or Flash memory? Spec sheet says there is 98kb of RAM on the LPC2388 ARM CPU chip.  Maybe using only on-CPU RAM is the limiting problem.  There should be huge room on the Flash drive for the data.  

An int size 2 bytes can go to +/- 32767
A uint size 2 bytes can go 0 to 65535 
A long size 4 bytes can go to +/-  (huge)

This now seems like a firmware code or memory storage issue. It would need the memory to allocate for a larger tick number, or more PEC data collection numbers, or both.   If anyone can address it, it will be Rene. 

Ray:  will PEMpro generates a useable PEC curve for gearboxes above 25:1.....or ....will the PEMpro PEC curve also fail if the problem is with the G-2 tick size data storage in the present firmware?  

All the best,
Michael

  




On Fri, Jun 18, 2021, 7:24 AM Ray Gralak <iogroups@...> wrote:
> PEC training quits after 48s, 1/5th period of the worm. I think this has to do Pec working with 6400 ticks instead of
> 32000.

32000 is above the maximum number of ticks, which I think is 25600.

The tick count probably wraps at 25,600 so 32000-25600 = 6400, which would explain the 1/5 cycle.

-Ray

> -----Original Message-----
> From: Losmandy_users@groups.io [mailto:Losmandy_users@groups.io] On Behalf Of pcboreland via groups.io
> Sent: Friday, June 18, 2021 6:22 AM
> To: Losmandy_users@groups.io
> Subject: Re: [Losmandy_users_io] Changing the servo motor encoders from 256 to 512 or 1024. Any reason not
> to do this?
>
> Ray,
>
> PEC training quits after 48s, 1/5th period of the worm. I think this has to do Pec working with 6400 ticks instead of
> 32000.
>
> Peter
>







Ray Gralak
 

PEC training quits after 48s, 1/5th period of the worm. I think this has to do Pec working with 6400 ticks instead of
32000.
32000 is above the maximum number of ticks, which I think is 25600.

The tick count probably wraps at 25,600 so 32000-25600 = 6400, which would explain the 1/5 cycle.

-Ray

-----Original Message-----
From: Losmandy_users@groups.io [mailto:Losmandy_users@groups.io] On Behalf Of pcboreland via groups.io
Sent: Friday, June 18, 2021 6:22 AM
To: Losmandy_users@groups.io
Subject: Re: [Losmandy_users_io] Changing the servo motor encoders from 256 to 512 or 1024. Any reason not
to do this?

Ray,

PEC training quits after 48s, 1/5th period of the worm. I think this has to do Pec working with 6400 ticks instead of
32000.

Peter


Peter Boreland
 

Ray,

PEC training quits after 48s, 1/5th period of the worm. I think this has to do Pec working with 6400 ticks instead of 32000. 

Peter


Ray Gralak
 

Hi Michael,

I am speaking about trying to use the built in PEC recorder feature advertised as being internal to Gemini-2. It
does not work for the G11T. It appears it does not work for the 125:1 gearbox either. In both cases the G-2 quits
taking data before a complete worm cycle period is complete. Without using a complete worm cycle it cannot
generate a correct PEC correction curve.
How are you determining that it is quitting early? And, how much earlier does is it quitting for you?

And again, using an autoguider to create a PEC curve is not going to give the best result. Not only are the recorded values phase delayed, but the data is subject to random star scintillation. And, worse, if the mount has any non-integer fundamentals, the recorded PEC curve will not match every worm cycle because each worm cycle will vary. In this case, the only way to produce a useful PEC curve is to remove the non-integer frequencies, which PEMPro can do.

-Ray

-----Original Message-----
From: Losmandy_users@groups.io [mailto:Losmandy_users@groups.io] On Behalf Of Michael Herman
Sent: Thursday, June 17, 2021 10:55 PM
To: Losmandy_users@groups.io
Subject: Re: [Losmandy_users_io] Changing the servo motor encoders from 256 to 512 or 1024. Any reason not
to do this?

Ray,

I am speaking about trying to use the built in PEC recorder feature advertised as being internal to Gemini-2. It
does not work for the G11T. It appears it does not work for the 125:1 gearbox either. In both cases the G-2 quits
taking data before a complete worm cycle period is complete. Without using a complete worm cycle it cannot
generate a correct PEC correction curve.

You are right....I have not used PEMpro. I have used Gemini-1 to generate it's internal PEC with good results. I
have used PECprep for PE analysis, not generate PEC for the Gemini system. I only use non-guided star
tracking to acquire PE data for analysis.

Best regards,
Michael






On Thu, Jun 17, 2021, 10:37 PM Ray Gralak <iogroups@siriusimaging.com> wrote:


> I encountered the same problem as you report: the Gemini-2 PEC training acquisition time is broken. The
internal
> PEC that is advertisef to work in G-2 does not work. The only way to get a working PEC at this time is by
> purchasing PEMpro, then use a PHD2 unguided log file fed into PEMpro to have it create the correct PEC
file.

Michael, you obviously haven't used PEMPro, because you don't feed a PHD2 unguided log file into
PEMPro to create a PEC curve. Maybe you are thinking of the PEMPro Log Viewer, which does accept PHD2 logs
for analysis?

> I reported the G-2 problem to Renew Goerlich who asked me to download certain files from the G-2, but I
did not
> have a CAT5 cable connection so I did not deliver the files requested. Bottom line: G-2 PEC does not yet
work.

You must mean recording PE via an autoguider doesn't work, right? PEMPro does work with the Gemini-2
PEC.

BTW, recording via autoguider is likely to provide inferior results. In fact, this method doesn't work very well
at all if there are any non-integer fundamental frequencies, which is not uncommon. I can explain why this is so, if
this isn't obvious?

-Ray

> -----Original Message-----
> From: Losmandy_users@groups.io [mailto:Losmandy_users@groups.io] On Behalf Of Michael Herman
> Sent: Thursday, June 17, 2021 8:56 PM
> To: Losmandy_users@groups.io
> Subject: Re: [Losmandy_users_io] Changing the servo motor encoders from 256 to 512 or 1024. Any
reason not
> to do this?
>
> PCBoreland,
>
> (What is your first name?)
>
> I encountered the same problem as you report: the Gemini-2 PEC training acquisition time is broken. The
internal
> PEC that is advertisef to work in G-2 does not work. The only way to get a working PEC at this time is by
> purchasing PEMpro, then use a PHD2 unguided log file fed into PEMpro to have it create the correct PEC
file.
>
> The PEC training in Gemini-1 does work. You don't need PEMpro if you have a G-1. You can use
PEMpro if you
> want of course.
>
> I reported the G-2 problem to Renew Goerlich who asked me to download certain files from the G-2, but I
did not
> have a CAT5 cable connection so I did not deliver the files requested. Bottom line: G-2 PEC does not yet
work.
>
> Meanwhile I am happy to read your detailed reports on the 125:1 gearbox. I agree that PHD2 autoguiding
will
> correct the DEC and RA. It would be the same if tracking the sun or moon or a planet...you must use
autoguiding.
>
> And yes. The 125:1 vs 25:1 means that slewing will be somewhere like 5x slower. Yet it is the tracking
perfection
> that you are focussing on.
>
> Best,
> Michael
>
>
>
>
>
> On Thu, Jun 17, 2021, 8:10 PM pcboreland via groups.io <pcboreland=yahoo.com@groups.io> wrote:
>
>
> As was pointed out by David Partridge, it is possible to change out the gear box on just RA if one uses
the
> web interface to select the parameters. I installed the 125:1 gearbox earlier in the day, and have been
able to
> gather some initial data. As might be expected there is good and bad.
>
> The Good. Unmoded, my Ra guided on average 30% worse than the Dec axis looking back over 6
months
> of logs. With the limited short run tests done earlier this evening RA is now guiding 10 to 20% better than
Dec. A
> huge shift in performance. A seeming improvement of greater than 0.6 arcsec on 10 min runs. I'll be
imaging in 90
> minute blocks later tonight so will hopefully get data worth sharing, and this improvement holds
>
> The Bad. 1. Ra runs slow. Over a 3hr period it fell behind the object a noticable bit. This I assume is
> because the Ra tracking divisor is set to 89754 when it needs to be set 89754.2. I go not think this is a big
deal as
> guiding ought to handle this problem. 2. Pec training runs for just 48s. 1/5th of a worm cycle. Not sure
how to fix
> this? I do not see any parameters to lengthen the record time. Pec is turned off and I'm using Predictive
Pec in
> PHD2, but hoping this can be fixed?
>
> First Ha image of the Cygnus Wall just came in!
>
>









Cyclone
 

Peter,
The difference between 89754 and 89754.2 over 3 hours is about 0.4 arcsec; it is not likely the cause of the mount running slow.

Did you do the test I suggested in a previous post? Did the PEC index roll over at 32000? Is you RA index 5760000 at CWD position?

Michael,
The built in PEC training always runs the full 4 minutes on my G11 Gemini 2; I used it last time a few weeks ago for some testing without any issue. Which files did Rene ask you to download from the Gemini?

Eric


Michael Herman
 

Ray,

I am speaking about trying to use the built in PEC recorder feature advertised as being internal to Gemini-2.  It does not work for the G11T.  It appears it does not work for the 125:1 gearbox either.  In both cases the G-2 quits taking data before a complete worm cycle period is complete.  Without using a complete worm cycle it cannot generate a correct PEC correction curve. 

You are right....I have not used PEMpro.  I have used Gemini-1 to generate it's internal PEC with good results.    I have used PECprep for PE analysis, not generate PEC for the Gemini system.   I only use non-guided star tracking to acquire PE data for analysis.    

Best regards,
Michael 






On Thu, Jun 17, 2021, 10:37 PM Ray Gralak <iogroups@...> wrote:
> I encountered the same problem as you report: the Gemini-2 PEC training acquisition time is broken.  The internal
> PEC that is advertisef to work in G-2 does not work. The only way to get a working PEC at this time is by
> purchasing PEMpro, then use a PHD2 unguided log file fed into PEMpro to have it create the correct PEC file.

Michael, you obviously haven't used PEMPro, because you don't feed a PHD2 unguided log file into PEMPro to create a PEC curve. Maybe you are thinking of the PEMPro Log Viewer, which does accept PHD2 logs for analysis?

> I reported the G-2 problem to Renew Goerlich who asked me to download certain files from the G-2, but I did not
> have a CAT5 cable connection so I did not deliver the files requested.  Bottom line: G-2 PEC does not yet work.

You must mean recording PE via an autoguider doesn't work, right? PEMPro does work with the Gemini-2 PEC.

BTW, recording via autoguider is likely to provide inferior results. In fact, this method doesn't work very well at all if there are any non-integer fundamental frequencies, which is not uncommon. I can explain why this is so, if this isn't obvious?

-Ray

> -----Original Message-----
> From: Losmandy_users@groups.io [mailto:Losmandy_users@groups.io] On Behalf Of Michael Herman
> Sent: Thursday, June 17, 2021 8:56 PM
> To: Losmandy_users@groups.io
> Subject: Re: [Losmandy_users_io] Changing the servo motor encoders from 256 to 512 or 1024. Any reason not
> to do this?
>
> PCBoreland,
>
> (What is your first name?)
>
> I encountered the same problem as you report: the Gemini-2 PEC training acquisition time is broken.  The internal
> PEC that is advertisef to work in G-2 does not work. The only way to get a working PEC at this time is by
> purchasing PEMpro, then use a PHD2 unguided log file fed into PEMpro to have it create the correct PEC file.
>
> The PEC training in Gemini-1 does work.  You don't need PEMpro if you have a G-1. You can use PEMpro if you
> want of course.
>
> I reported the G-2 problem to Renew Goerlich who asked me to download certain files from the G-2, but I did not
> have a CAT5 cable connection so I did not deliver the files requested.  Bottom line: G-2 PEC does not yet work.
>
> Meanwhile I am happy to read your detailed reports on the 125:1 gearbox.  I agree that PHD2 autoguiding will
> correct the DEC and RA.  It would be the same if tracking the sun or moon or a planet...you must use autoguiding.
>
> And yes. The 125:1 vs 25:1 means that slewing will be somewhere like 5x slower.  Yet it is the tracking perfection
> that you are focussing on.
>
> Best,
> Michael
>
>
>
>
>
> On Thu, Jun 17, 2021, 8:10 PM pcboreland via groups.io <pcboreland=yahoo.com@groups.io> wrote:
>
>
>       As was pointed out by David Partridge, it is possible to change out the gear box on just RA if one uses the
> web interface to select the parameters. I installed the 125:1 gearbox earlier in the day, and have been able to
> gather some initial data. As might be expected there is good and bad.
>
>       The Good. Unmoded, my Ra guided on average 30% worse than the Dec axis looking back over 6 months
> of logs. With the limited short run tests done earlier this evening RA is now guiding 10 to 20% better than Dec. A
> huge shift in performance. A seeming improvement of greater than 0.6 arcsec on 10 min runs. I'll be imaging in 90
> minute blocks later tonight so will hopefully get data worth sharing, and this improvement holds
>
>       The Bad. 1. Ra runs slow. Over a 3hr period it fell behind the object a noticable bit. This I assume is
> because the Ra tracking divisor is set to 89754 when it needs to be set 89754.2. I go not think this is a big deal as
> guiding ought to handle this problem. 2. Pec training runs for just 48s. 1/5th of a worm cycle. Not sure how to fix
> this? I do not see any parameters to lengthen the record time. Pec is turned off and I'm using Predictive Pec in
> PHD2, but hoping this can be fixed?
>
>       First Ha image of the Cygnus Wall just came in!
>
>







Ray Gralak
 

I encountered the same problem as you report: the Gemini-2 PEC training acquisition time is broken. The internal
PEC that is advertisef to work in G-2 does not work. The only way to get a working PEC at this time is by
purchasing PEMpro, then use a PHD2 unguided log file fed into PEMpro to have it create the correct PEC file.
Michael, you obviously haven't used PEMPro, because you don't feed a PHD2 unguided log file into PEMPro to create a PEC curve. Maybe you are thinking of the PEMPro Log Viewer, which does accept PHD2 logs for analysis?

I reported the G-2 problem to Renew Goerlich who asked me to download certain files from the G-2, but I did not
have a CAT5 cable connection so I did not deliver the files requested. Bottom line: G-2 PEC does not yet work.
You must mean recording PE via an autoguider doesn't work, right? PEMPro does work with the Gemini-2 PEC.

BTW, recording via autoguider is likely to provide inferior results. In fact, this method doesn't work very well at all if there are any non-integer fundamental frequencies, which is not uncommon. I can explain why this is so, if this isn't obvious?

-Ray

-----Original Message-----
From: Losmandy_users@groups.io [mailto:Losmandy_users@groups.io] On Behalf Of Michael Herman
Sent: Thursday, June 17, 2021 8:56 PM
To: Losmandy_users@groups.io
Subject: Re: [Losmandy_users_io] Changing the servo motor encoders from 256 to 512 or 1024. Any reason not
to do this?

PCBoreland,

(What is your first name?)

I encountered the same problem as you report: the Gemini-2 PEC training acquisition time is broken. The internal
PEC that is advertisef to work in G-2 does not work. The only way to get a working PEC at this time is by
purchasing PEMpro, then use a PHD2 unguided log file fed into PEMpro to have it create the correct PEC file.

The PEC training in Gemini-1 does work. You don't need PEMpro if you have a G-1. You can use PEMpro if you
want of course.

I reported the G-2 problem to Renew Goerlich who asked me to download certain files from the G-2, but I did not
have a CAT5 cable connection so I did not deliver the files requested. Bottom line: G-2 PEC does not yet work.

Meanwhile I am happy to read your detailed reports on the 125:1 gearbox. I agree that PHD2 autoguiding will
correct the DEC and RA. It would be the same if tracking the sun or moon or a planet...you must use autoguiding.

And yes. The 125:1 vs 25:1 means that slewing will be somewhere like 5x slower. Yet it is the tracking perfection
that you are focussing on.

Best,
Michael





On Thu, Jun 17, 2021, 8:10 PM pcboreland via groups.io <pcboreland=yahoo.com@groups.io> wrote:


As was pointed out by David Partridge, it is possible to change out the gear box on just RA if one uses the
web interface to select the parameters. I installed the 125:1 gearbox earlier in the day, and have been able to
gather some initial data. As might be expected there is good and bad.

The Good. Unmoded, my Ra guided on average 30% worse than the Dec axis looking back over 6 months
of logs. With the limited short run tests done earlier this evening RA is now guiding 10 to 20% better than Dec. A
huge shift in performance. A seeming improvement of greater than 0.6 arcsec on 10 min runs. I'll be imaging in 90
minute blocks later tonight so will hopefully get data worth sharing, and this improvement holds

The Bad. 1. Ra runs slow. Over a 3hr period it fell behind the object a noticable bit. This I assume is
because the Ra tracking divisor is set to 89754 when it needs to be set 89754.2. I go not think this is a big deal as
guiding ought to handle this problem. 2. Pec training runs for just 48s. 1/5th of a worm cycle. Not sure how to fix
this? I do not see any parameters to lengthen the record time. Pec is turned off and I'm using Predictive Pec in
PHD2, but hoping this can be fixed?

First Ha image of the Cygnus Wall just came in!


Michael Herman
 

PCBoreland,

(What is your first name?)

I encountered the same problem as you report: the Gemini-2 PEC training acquisition time is broken.  The internal PEC that is advertisef to work in G-2 does not work. The only way to get a working PEC at this time is by purchasing PEMpro, then use a PHD2 unguided log file fed into PEMpro to have it create the correct PEC file. 

The PEC training in Gemini-1 does work.  You don't need PEMpro if you have a G-1. You can use PEMpro if you want of course. 

I reported the G-2 problem to Renew Goerlich who asked me to download certain files from the G-2, but I did not have a CAT5 cable connection so I did not deliver the files requested.  Bottom line: G-2 PEC does not yet work.  

Meanwhile I am happy to read your detailed reports on the 125:1 gearbox.  I agree that PHD2 autoguiding will correct the DEC and RA.  It would be the same if tracking the sun or moon or a planet...you must use autoguiding.  

And yes. The 125:1 vs 25:1 means that slewing will be somewhere like 5x slower.  Yet it is the tracking perfection  that you are focussing on.  

Best,
Michael





On Thu, Jun 17, 2021, 8:10 PM pcboreland via groups.io <pcboreland=yahoo.com@groups.io> wrote:
As was pointed out by David Partridge, it is possible to change out the gear box on just RA if one uses the web interface to select the parameters. I installed the 125:1 gearbox earlier in the day, and have been able to gather some initial data. As might be expected there is good and bad.

The Good. Unmoded, my Ra guided on average 30% worse than the Dec axis looking back over 6 months of logs. With the limited short run tests done earlier this evening RA is now guiding 10 to 20% better than Dec. A huge shift in performance. A seeming improvement of greater than 0.6 arcsec on 10 min runs. I'll be imaging in 90 minute blocks later tonight so will hopefully get data worth sharing, and this improvement holds

The Bad. 1. Ra runs slow. Over a 3hr period it fell behind the object a noticable bit. This I assume is because the Ra tracking divisor is set to 89754 when it needs to be set 89754.2. I go not think this is a big deal as guiding ought to handle this problem. 2. Pec training runs for just 48s. 1/5th of a worm cycle. Not sure how to fix this? I do not see any parameters to lengthen the record time. Pec is turned off and I'm using Predictive Pec in PHD2, but hoping this can be fixed?

First Ha image of the Cygnus Wall just came in!


Peter Boreland
 

As was pointed out by David Partridge, it is possible to change out the gear box on just RA if one uses the web interface to select the parameters. I installed the 125:1 gearbox earlier in the day, and have been able to gather some initial data. As might be expected there is good and bad.

The Good. Unmoded, my Ra guided on average 30% worse than the Dec axis looking back over 6 months of logs. With the limited short run tests done earlier this evening RA is now guiding 10 to 20% better than Dec. A huge shift in performance. A seeming improvement of greater than 0.6 arcsec on 10 min runs. I'll be imaging in 90 minute blocks later tonight so will hopefully get data worth sharing, and this improvement holds

The Bad. 1. Ra runs slow. Over a 3hr period it fell behind the object a noticable bit. This I assume is because the Ra tracking divisor is set to 89754 when it needs to be set 89754.2. I go not think this is a big deal as guiding ought to handle this problem. 2. Pec training runs for just 48s. 1/5th of a worm cycle. Not sure how to fix this? I do not see any parameters to lengthen the record time. Pec is turned off and I'm using Predictive Pec in PHD2, but hoping this can be fixed?

First Ha image of the Cygnus Wall just came in!


Cyclone
 

On Wed, Jun 16, 2021 at 04:52 PM, <pcboreland@...> wrote:
Eric,

Thank you for the information you have provided. I have a far better understanding of what is going on under the hood. A final question, which I'm sure has been asked here many times, how is the PE curve get synced with the worm?

Peter
The following description is my understanding for the regular gearbox, and will likely change for your 125:1 gearbox. The PEC index is not absolute and rolls over at every worm cycle (6400 steps), so a method is used to sync the PEC index to the RA axis index. The RA index is relative to the CWD position, where the RA index is 1152000. As the RA axis rotates, the RA index and PEC index decrement/increment by 1 for each encoder tick. At RA index 1152000 (CWD), the PEC index is 0 when you first boot up the Gemini. When training PEC, Gemini resets the PEC index to 0 and starts recording guiding commands for a full worm cycle (6400 PEC steps for the regular gearbox). Resetting the PEC index at an arbitrary RA index location would cause the PEC index to be out of sync with the RA index, so Gemini also calculates an offset as it resets the PEC index, aka PEC Worm Offset, which you can find in the currpec.pec header. You can calculate the PEC worm offset for any arbitrary RA index in Excel as follows: 6400 - MOD(PEC index + RA index - 1152000, 6400). Then, at the next boot up, at CWD the RA index is 1152000 and the PEC index is non zero as calculated by Gemini based on the PEC worm offset save in memory (PEC index + PEC worm offset = 6400 at RA index 1152000).

Hopefully this all makes sense and you can figure out how to translate this for your 125:1 gearbox.

Eric


Cyclone
 

To check PEC, you could first confirm that the PEC index rolls over at 32000. You can verify that by going to the web interface, State page, and monitor the RA PEC Pointer while the mount tracks at sidereal rate. The RA PEC Pointer should increment and roll over at 32000 with the 125:1 gearbox (or 6400 with the regular gearbox).

You could also create a dummy PEC file during the day by simply training PEC with the hand controller, and send a guiding command periodically with the hand controller (both East and West commands). Then open the currpec.pec file in the PEC folder on the SD card and check that the header information matches your setup and the PEC data contains the correct divisors for the entire worm cycle. 

Eric


Brendan Smith
 

Its funny..Mclennan lists its 25:1 gearbox as P5-G11

Had a laugh.......
--
Brendan


Peter Boreland
 

Eric,

Thank you for the information you have provided. I have a far better understanding of what is going on under the hood. A final question, which I'm sure has been asked here many times, how is the PE curve get synced with the worm?

Peter


Cyclone
 

Peter,
The guide rate you selected (0.8x) has very coarse steps in RA. Selecting 0.5x or slower would result in much smaller steps. However step size is constant in Dec.

The 0.11 arcsec step size with the new gearbox will allow for a smaller minimum guide move if your seeing is good enough and you can somehow manage Dec backlash to take advantage of the smaller step size. In other words, the limiting factor should not be the step sizes anymore, but the other factors impacting guiding will still be there, which I believe is your objective in implementing this mod.

Eric


Peter Boreland
 

Eric, 

Your thoughts on how a 0.11arcsec step resolution will affect PEC programming and operation would also be most appreciated. We now have 32000 ticks per worm revolution.

Peter


Peter Boreland
 

Eric,

Your terrific document does indeed answer my question. How do you think having a 0.11 arcsec step resolution will affect this statement:

"Therefore, when a small move in the East direction is required, the mount may have to first take a large step East and overshoot the desired position, and then make a series of West steps. Such direction reversals are programmable in PEC as described in the next section. However, it is unknown if Gemini implements such reversals when autoguiding, or if it simply ignores guiding commands that are less than half the East step size. "

This seems to be a big problem and is possibly what I was seeing that drove me to want to do this mod as far as overshoot (extream movement) that I do see affect the quality of my images.  You see it in Dec. 0.56 arcsec is a big jump. Getting round stars is one thing. Getting hi res dust lanes is another. I'm even considering lucky imaging as a solution if this mod fails. Seems easier to focus on getting better performance ffrom the mount first than tackling seeing.

I will get you the gear ratios.


 

You should be able to replace just one gearbox and set the custom spur gear value for just that axis.

 

From: Losmandy_users@groups.io <Losmandy_users@groups.io> On Behalf Of pcboreland via groups.io
Sent: 16 June 2021 00:19
To: Losmandy_users@groups.io
Subject: Re: [Losmandy_users_io] Changing the servo motor encoders from 256 to 512 or 1024. Any reason not to do this?

 

First impression of the McLennan 125:1 gear box. The build quality is better than McLennan 25:1, which had noticeable slop in the drive shaft. This did add an additional frequency component to the 240s error. While the shaft measured 4mm the bushings where oversize. This is not the case with the 125:1 gearbox. The bushings used are not the same as in the 25:1. There is no perceivable movement in the output shaft lrelative to the bushings. 

It looks like I can not operate the mount with only one gearbox changed out so have to order another one. 


Cyclone
 

On Tue, Jun 15, 2021 at 04:48 PM, <pcboreland@...> wrote:
Cyclone,

Thanks for thinking this over. Just to check I did this a few times. At 25:1 gear ratio RA divisor is as you said reads 448771.

1. Set gear ratio to 125 and steps per worm rev to 32000.
2. Reboot.  

RA divisor now reads 89754. But the calculated value is 89754.2, so mount is going to run a bit slow! Speed values stay the same, except may have to reduce the slew rate. What is the calculation behind the guide rate please?

I would reduce the slew rate to be within the gearbox spec, which is 5000 rpm, or 83.333 motor revolutions per sec, or 0.667 worm revolutions per sec, or 2400 arcsec per sec, which is 160 times sidereal rate. The default slew multiplier is 800 I believe, which is 5 times too fast for the gearbox.

 

Can you clarify your question regarding the guide rates? Are you looking for the step size calculation? If so, the explanation is in the PDF document attached in a previous post.

Eric


Peter Boreland
 

First impression of the McLennan 125:1 gear box. The build quality is better than McLennan 25:1, which had noticeable slop in the drive shaft. This did add an additional frequency component to the 240s error. While the shaft measured 4mm the bushings where oversize. This is not the case with the 125:1 gearbox. The bushings used are not the same as in the 25:1. There is no perceivable movement in the output shaft lrelative to the bushings. 

It looks like I can not operate the mount with only one gearbox changed out so have to order another one. 


Peter Boreland
 

Cyclone,

Thanks for thinking this over. Just to check I did this a few times. At 25:1 gear ratio RA divisor is as you said reads 448771.

1. Set gear ratio to 125 and steps per worm rev to 32000.
2. Reboot.  

RA divisor now reads 89754. But the calculated value is 89754.2, so mount is going to run a bit slow! Speed values stay the same, except may have to reduce the slew rate. What is the calculation behind the guide rate please?


Cyclone
 

Is there a reason why you use a 0.8x guide rate? It does not seem to match your stated intent to use small steps for guiding. If you read my document, you will see that the RA  step sizes at 0.8x are 2.25 arcsec East and 0.25 arcsec West. If you want small step sizes, you need to select 0.2x guide rate. The step size in Dec is constant at 0.5625 arcsec.

Eric


Cyclone
 
Edited

I don’t see why the speed settings would have to change as they are a multiple of sidereal rate, unless the 125:1 gearbox supports less torque.
edit: I just saw the maximum gearbox input speed is 5000 rpm, so the slew speed would have to be divided by 5, but the other speeds should not have to change.

Also, you will have to change the RA divisor to 89755 to ensure tracking at sidereal rate, currently set at 448771.

It would be interesting to have the data on the 4 gearbox stages so the PE frequencies can be calculated. The motor will rotate with a 1.92 sec period, so hopefully the corresponding PE amplitude is low as it will be difficult to correct.

Eric


Michael Herman
 

Very very good conversation going on here.  The topic is: what alternative gearboxes work better than the stock gearbox.  Pros and cons...

The candidates seem to be the McLennan series of gearboxes...you can get 2mm or 3mm prebored pinion gears for a motor shaft:

  25:1. Easy install proven good to get rid of the known 1/7.5 worm period oscillation from stock gearboxes.  The 3mm pinion can be easily enlarged to 1/8 inch to fit the HiTorque motor.  But 25:1 does not reduce the 0.56 arcsec step size of the optical encoder.

  50:1.   Problem: very tiny pinion gear might leave little material in fitting the 1/8th inch HiTorque motor shaft.  Proponent is Marc Aragnou in Australia for older motors that had a 2mm shaft ... says gives better results than 25:1 McLennan gearbox...can be a perfect fit for the 2mm bore pinion available.  50:1 reduces optical encoder step size to 0.28 arcsec.  Maybe reduces slew speed to about half standard.  
 
  125:1.  New proposal by PCBoreland. Reduces optical encoder step size to 0.11 arcsec.  Nice idea!  Probably limits slew speed but...we can accept that if RA sidereal PE goes low enough!

Waiting for PE test data from PCBoreland!

All the best,...nice work gang!

Michael




On Mon, Jun 14, 2021, 4:47 PM Paul Kanevsky <yh@...> wrote:
On Mon, Jun 14, 2021 at 07:16 PM, <pcboreland@...> wrote:
Paul,

But can one instruct the motor to move a quarter step? My understanding is the minimum step size is still 0.56 arc sec. My experience with guiding at 0.8 arc sec rms and looking to improve upon that is considerable instability is injected by the coarse step size. This is particulary true of the Dec axis. If you turn off guiding it's pretty flat most of the time. Turn it on and it can starts flapping around +/- 0.5 arc sec or worse. 
No, externally "addressable" positions are in steps, so you can't tell the mount to go to a quarter step position. For tracking and maintaining/achieving the correct desired position, the precision is 1/4 of a step.

Regards,

     -Paul


Paul Kanevsky
 

On Mon, Jun 14, 2021 at 07:16 PM, <pcboreland@...> wrote:
Paul,

But can one instruct the motor to move a quarter step? My understanding is the minimum step size is still 0.56 arc sec. My experience with guiding at 0.8 arc sec rms and looking to improve upon that is considerable instability is injected by the coarse step size. This is particulary true of the Dec axis. If you turn off guiding it's pretty flat most of the time. Turn it on and it can starts flapping around +/- 0.5 arc sec or worse. 
No, externally "addressable" positions are in steps, so you can't tell the mount to go to a quarter step position. For tracking and maintaining/achieving the correct desired position, the precision is 1/4 of a step.

Regards,

     -Paul


Peter Boreland
 

Paul,

But can one instruct the motor to move a quarter step? My understanding is the minimum step size is still 0.56 arc sec. My experience with guiding at 0.8 arc sec rms and looking to improve upon that is considerable instability is injected by the coarse step size. This is particulary true of the Dec axis. If you turn off guiding it's pretty flat most of the time. Turn it on and it can starts flapping around +/- 0.5 arc sec or worse. 


Peter Boreland
 

Michael,

RS Components: 336-422. I ordered through the US site, not the UK site. I looked at the 50:1,  but as previously mentioned it has a very small pinion gear that might prove difficult to enlarge to fit a 1/8th shaft. I already have the 25:1. Yes you need to change the sign to reverse direct.

I'll report back on the results for good or bad. 


Michael Herman
 

Well...if you really did buy 125:1 not 25:1,
 That's a new experiment!

Can you share with us the part number and place you ordered from?  

Thanks,
Michael


On Mon, Jun 14, 2021, 12:58 PM Michael Herman via groups.io <mherman346=gmail.com@groups.io> wrote:
Surely you mean 25:1, not 125:1

The stock gearbox is 25:1 

The McLennan 25:1 gearbox rotates the opposite direction, do all you need to do is change the sign of the ring gear teeth (for G11 change 360 to -360).  You leave the gearbox setting at 25.

Best,
Michael

On Mon, Jun 14, 2021, 12:38 PM pcboreland via groups.io <pcboreland=yahoo.com@groups.io> wrote:
I've opted to change out the gear boxes with McLennan 125:1 ratio.  I have one unit arriving tomorrow and will be ordering a second. I was able to successfully change the controller settings for the gear box values to 125 and set the steps per worm revolution to 32000. I guess the speed settings need to be divided by 5, except for Guiding speed which I've kept at 0.8. Everything ran OK. So 0.112 arc sec per step. Less expensive than changing the encoders.


Michael Herman
 

Surely you mean 25:1, not 125:1

The stock gearbox is 25:1 

The McLennan 25:1 gearbox rotates the opposite direction, do all you need to do is change the sign of the ring gear teeth (for G11 change 360 to -360).  You leave the gearbox setting at 25.

Best,
Michael

On Mon, Jun 14, 2021, 12:38 PM pcboreland via groups.io <pcboreland=yahoo.com@groups.io> wrote:
I've opted to change out the gear boxes with McLennan 125:1 ratio.  I have one unit arriving tomorrow and will be ordering a second. I was able to successfully change the controller settings for the gear box values to 125 and set the steps per worm revolution to 32000. I guess the speed settings need to be divided by 5, except for Guiding speed which I've kept at 0.8. Everything ran OK. So 0.112 arc sec per step. Less expensive than changing the encoders.