Date   
Re: Searching for IMD

Henning Weddig
 

Glenn,

please compare the schematics of the Wes Hayward BiDI amp and Ashars version.

As Wes states his version is designed for 15 dB gain, resulting that he has a series feedback (680 ohms + cap) from collector to base. The biasing is done with two extra resistors.

Ashar uses only the biasing and feedback with the two resistors.

The result is that Ashars amp has about 20 dB gain, worse freqeuncy response (due to the limited ft of the 2N3904) and worse input reflection (S11). Due to the higher gain IMD must be worse at the same input level. If the textbook curve applies (3 dB for every 1 dB of input power change) the IMD will be 15 dB worse.

73

Henning Weddig

DK5LV



Am 15.09.2018 um 06:44 schrieb Glenn:

Post 59439 is partly incorrect.  Just after I posted mentioning the visible clipping I realised, stupid me, was measuring gain during clipping which of course compresses the result.

I deleted that post. And have attached the new plots where you can see the  gain difference between uBITX and Hayward version. This obviously partly accounts for the difference in input levels before output clipping occurs. The biasing arrangement gives different Iq also.  Gain is reduced about  4dB in the Hayward version I built over uBITX. It now accords more closely also with  his test results of 15.5dB I get 16dB at 30MHz.

Haywards paper gives some values for varying the gain to other values also.







deleted this:-
On Sat, Sep 15, 2018 at 12:16 PM, Glenn wrote:
Hi Ashhur,
I just built up a proto of both versions on vero board.

As a rough check of IP3, since I am not set up to do proper measurement of IP3, I found that the 'visible distortion' of the Hayward version occurs about 10dB higher than the uBITX version.  ie the point of clipping (negative side in this case)  
~-22dBm for uBITX
~ -13dBm for Hayward.   This would then imply IMD is better.

In my set-up I found gain to be similar and pretty flat. I used junk-box parts of mixed tolerances. I built only one side of the Bi-Amps.
uBITX:-   ~ 12dB gain, flat over the range 3-45MHz flat within 0.8dB
Hayward:- ~ 10db gain, flat over same range within 0.6db

glenn

Re: Searching for IMD

Ashhar Farhan
 

glen, you can do this,

find out, accurately, the gain of the amps at very low input levels.
Now, keep increasing the drive level until you see the gain dip by 1 db.
that is the compression point. For bipolar device amplifiers, the IIP3 will be about 15 above that point.

- f

On Sat, Sep 15, 2018 at 10:14 AM, Glenn <glennp@...> wrote:
Post 59439 is partly incorrect.  Just after I posted mentioning the visible clipping I realised, stupid me, was measuring gain during clipping which of course compresses the result.

I deleted that post. And have attached the new plots where you can see the  gain difference between uBITX and Hayward version. This obviously partly accounts for the difference in input levels before output clipping occurs. The biasing arrangement gives different Iq also.  Gain is reduced about  4dB in the Hayward version I built over uBITX. It now accords more closely also with  his test results of 15.5dB I get 16dB at 30MHz.

Haywards paper gives some values for varying the gain to other values also.







deleted this:-
On Sat, Sep 15, 2018 at 12:16 PM, Glenn wrote:
Hi Ashhur,
I just built up a proto of both versions on vero board.

As a rough check of IP3, since I am not set up to do proper measurement of IP3, I found that the 'visible distortion' of the Hayward version occurs about 10dB higher than the uBITX version.  ie the point of clipping (negative side in this case)  
~-22dBm for uBITX
~ -13dBm for Hayward.   This would then imply IMD is better.

In my set-up I found gain to be similar and pretty flat. I used junk-box parts of mixed tolerances. I built only one side of the Bi-Amps.
uBITX:-   ~ 12dB gain, flat over the range 3-45MHz flat within 0.8dB
Hayward:- ~ 10db gain, flat over same range within 0.6db

glenn


Re: Searching for IMD

Glenn
 

Post 59439 is partly incorrect.  Just after I posted mentioning the visible clipping I realised, stupid me, was measuring gain during clipping which of course compresses the result.

I deleted that post. And have attached the new plots where you can see the  gain difference between uBITX and Hayward version. This obviously partly accounts for the difference in input levels before output clipping occurs. The biasing arrangement gives different Iq also.  Gain is reduced about  4dB in the Hayward version I built over uBITX. It now accords more closely also with  his test results of 15.5dB I get 16dB at 30MHz.

Haywards paper gives some values for varying the gain to other values also.







deleted this:-

On Sat, Sep 15, 2018 at 12:16 PM, Glenn wrote:
Hi Ashhur,
I just built up a proto of both versions on vero board.

As a rough check of IP3, since I am not set up to do proper measurement of IP3, I found that the 'visible distortion' of the Hayward version occurs about 10dB higher than the uBITX version.  ie the point of clipping (negative side in this case)  
~-22dBm for uBITX
~ -13dBm for Hayward.   This would then imply IMD is better.

In my set-up I found gain to be similar and pretty flat. I used junk-box parts of mixed tolerances. I built only one side of the Bi-Amps.
uBITX:-   ~ 12dB gain, flat over the range 3-45MHz flat within 0.8dB
Hayward:- ~ 10db gain, flat over same range within 0.6db

glenn

Re: Which xmit transistor is blown?

Sean W7SKD
 

Excellent point (cw key). One of those situations where you go down a road and don’t think to step back and rethink things. I convinced myself that it was a transmit issue related to amplification and didn’t challenge tha notion. 

Re: 2N2222A vs 2N3904 sourcing

ajparent1/kb1gmx
 

Hi Jim,...High gain in small pieces with input close to output...around and around it goes,,,

12:05 today I said...
So I know its doable, but you can not fold the amp on itself and have antennas (heatsinks) with RF and hope for stability.

Lumps of gain can be stable strung together in a cramped pace is much more of a challenge.  

I have amps that work very well.  Hans has done it as well.  All of the have one thing in common the
input is as far as possible from the output. Its not magic but trying to wring out the last drop of gain
is an unsuccessful way to go as its not stable even with good layout.  The next batch may or many
not work due to nominal variations.

Allison

Re: Searching for IMD

ajparent1/kb1gmx
 

Farhan,

>>>crucial thing is bias current and feedback. The biasing method itself is irrelevant.<<<

Generally yes but its easier to have the bias independent of the feedback as one forces
possibly less optimum values.

The other is that only the gain needed is employed not a case of what the circuit can do.

Allison

Re: Searching for IMD

Glenn
 

Hi Ashhur,
I just built up a proto of both versions on vero board.

As a rough check of IP3, since I am not set up to do proper measurement of IP3, I found that the 'visible distortion' of the Hayward version occurs about 10dB higher than the uBITX version.  ie the point of clipping (negative side in this case)  
~-22dBm for uBITX
~ -13dBm for Hayward.   This would then imply IMD is better.

In my set-up I found gain to be similar and pretty flat. I used junk-box parts of mixed tolerances. I built only one side of the Bi-Amps.
uBITX:-   ~ 12dB gain, flat over the range 3-45MHz flat within 0.8dB
Hayward:- ~ 10db gain, flat over same range within 0.6db

glenn

Re: 2N2222A vs 2N3904 sourcing

jim
 

Like most devices intended to oscillate, output getting back into input ...High gain in small pieces with input close to output...around and around it goes

Jim


On Friday, September 14, 2018, 12:32:38 PM PDT, ajparent1/KB1GMX <kb1gmx@...> wrote:


Allison, why would the MMICs oscillate and not a 2N2219A? Might screening of the mmic help?

Iz oos,

Simple the board layout is not suitable.  Screening is not effective.

Allison

Re: 2N2222A vs 2N3904 sourcing

Timothy Fidler
 

Re the bitch about RF power transistors from PRC being substandard or fake . If they are popular old CB radio PA transistors and are cheap they will most likely be fake .  " Cecil the parts place " in the US has a few RF or pushed into RF service transistors and mosfets. Most are from Fairchild  mafr ..  one hopes the real Fairchild.  Some have rather nice specs and he has links to small amps built with them.

NO I am not his brother  or a schiller for him as the Yanks say. Just noticed the site which is linked to a Steve Webber  design site.

Re: Searching for IMD

Ashhar Farhan
 

The article does mention the oip3 which is a measure of IMD. The crucial thing is bias current and feedback. The biasing method itself is irrelevant.
- f

On Sat, 15 Sep 2018, 04:21 Glenn, <glennp@...> wrote:
Perhaps it's possible to improve the Bi-Amps?   The article by Wes Hayward and Bob Kopski shows a slightly different biasing arrangement and feedback in the first stage compared to that used in uBITX. As shown, gain is 15dB and flat to 100MHz within 1dB.

More conventional resistive voltage divider is used and the feedback is AC only. Two extra parts are used. The article doesn't mention IMD though specifically, as a performance target or measure it.

glenn

Re: 2N2222A vs 2N3904 sourcing

Tim Gorman
 

Back in March, 2018 I tried building up an amp section using an HP mmic
I have in the junk box. It worked very well on the desktop. Flat from
3Mhz to 30Mhz. No oscillation. I simply could find no way to add it
into the ubitx around Q90 (i.e. before and after) where it didn't break
into oscillation as soon as I tried to transmit.

At this point it would be easier to make up a new PA driver chain on a
replacement circuit board. Mount it on standoffs above the existing
board using coax in from Q90 and coax out to IRF510's. It's not a
project I want to take on. I would rather set the power out to 1w or
2w, have a clean signal, and use an external amplifier.

tim ab0wr

On Fri, 14 Sep 2018 15:23:49 -0700
"RCBoatGuy via Groups.Io" <ijnfan-HamRadio=yahoo.com@groups.io> wrote:

In response to Warren's question about why the MMIC's would be
sensitive to oscillating..  

According to the datasheets, they are very sensitive to ground
loops.  The datasheet for the HP/Aligent MMIC on the 30dB board
available on Amazon was quite clear on this, and recommended using an
ultra-thin PCB (0.032 rather than 0.0625) to keep the inductance of
the ground vias down where it needed to be to avoid oscillation. 

Given that, it would be very hard to retro-fit one of these MMIC's
into an existing board designed for transistors rather than MMIC's
and not have it oscillate.

73,

Carl, K0MWC

Re: Searching for IMD

Glenn
 

Perhaps it's possible to improve the Bi-Amps?   The article by Wes Hayward and Bob Kopski shows a slightly different biasing arrangement and feedback in the first stage compared to that used in uBITX. As shown, gain is 15dB and flat to 100MHz within 1dB.

More conventional resistive voltage divider is used and the feedback is AC only. Two extra parts are used. The article doesn't mention IMD though specifically, as a performance target or measure it.

glenn

Re: si5351 crosstalk #radiuno

MadRadioModder
 

Glenn… Now that is a sweet solution.

 

 

 


Virus-free. www.avg.com

--

…_. _._

Re: 2N2222A vs 2N3904 sourcing

RCBoatGuy
 

In response to Warren's question about why the MMIC's would be sensitive to oscillating..  

According to the datasheets, they are very sensitive to ground loops.  The datasheet for the HP/Aligent MMIC on the 30dB board available on Amazon was quite clear on this, and recommended using an ultra-thin PCB (0.032 rather than 0.0625) to keep the inductance of the ground vias down where it needed to be to avoid oscillation. 

Given that, it would be very hard to retro-fit one of these MMIC's into an existing board designed for transistors rather than MMIC's and not have it oscillate.

73,

Carl, K0MWC

Bandscan and Spectrum using I2C dsp meter

N5KBP
 

Has anyone gotten the bandscan and spectrum pages to work on KD8CEC's firmware using the standalone  I2C dspmeter? The only way I can get it to work kinda sorta is to compile the firmware to not use the dsp meter then attach the purple wire (a7) to the junction of the 2 resistors and .01 cap on the standalone dsp uno. This works but the readings on the band scan are very low and pretty much useless. What am I not understanding about this feature?
--
N5KBP

Re: 2N2222A vs 2N3904 sourcing

Trystan G0KAY
 

..and at Farnell

Re: 2N2222A vs 2N3904 sourcing

Trystan G0KAY
 

I've just had a look, and 2n222A are available readily enough at Cricklewood. 

Re: 2N2222A vs 2N3904 sourcing

ajparent1/kb1gmx
 

On Fri, Sep 14, 2018 at 11:07 AM, Tom, wb6b wrote:
Assuming the MMC amps can work without the issues that are being pointed out, the idea of using an assembled amplifier board, for us retrofitters (like the assembled filter boards folks are creating) might make mods easer. Here is an amplifier I bought on Amazon. The Mini-Circuits amplifier looks like it has better spec for use in the TX chain. But, this board cost $10.99 (and I have it here). If only 10DBm output drive is usable, it has a little more gain (oscillation caveat noted). I saw similar boards for less money, elsewhere. 

I have  handful of those.  Handy but a lot of gain. The gain has conditional stability as well.
I've had them take off when using poo input or output match.  The part shown is MAR8SM+,
I'll attach a datasheet.   FYI they prefer regulated voltages.
 
There are better parts in the Gali and other series I have them too and protoboards like that one.
I do a lot of VHF and UHF stuff so I even have connectorized block amps with 20db of gain and 
26dbm (-1db) output.  Consider this replacieng Q90, pre-drivers, Drivers with 2n5109s or other
devices that can deliver the gain based on feedback values ends up with the whole mess
oscillating as now with enough gain the isolation due to layout fails.

Actually for several tried replacing the 3904s with BFR106 in the q90 slot and dialing down the bias.
Replacing the 3904s in teh pre-driver and driver and both cases change up the emitter resistors
to 11ohms or 8.2 and then rest idling bias to under 20ma with series RLC feedback rather than RC
was easily done good bang for the money and reproducable(,-- magical) and in most cases set
to about 12w at 80m (q90 emitter 70/680/820pf) cap mod it was around 4-5W at 10M. 
Good enough for reasonable effort.

Allison 
one pill wishful thinking.

Re: 2N2222A vs 2N3904 sourcing

ajparent1/kb1gmx
 

On Fri, Sep 14, 2018 at 11:13 AM, Warren Allgyer wrote:
an see how that could happen if trying to fit the MMICs into the existing board layout. One critical thing has changed however: we have moved a PA filter section that was laid out extremely poorly and leaking like a sieve, off the board along with it's high power leakage. The MMICs in question are designed to be laid out on 50 ohm PCB transmission lines and, if done that way, are nearly impervious to external unwanted feedback. Fitting them into the existing board is a non-starter.... agreed. But, as an overlay card with properly laid out transmission line inputs and outputs, replacing each of the transmit bi-directional amps and the driver stages up to the -7 dBm level or so would have a dramatic effect on IMD and it is hard for me to accept they would not be far less prone to oscillation than existing components in the existing layout.

Warren,

I've used MMICs for many years.  Generally the design was built with them in mind and not as a bandaid.
The application for the ubitx was to make up for low gain IE: spool the stage gains down so 2222A or 2219
could be used at 11-12db gain with feedback and make up the need with a 10db MMIC.    Stability in a word
was poor.  I had the same issue trying devices that would make the same gain at 10M as 80m (5109s or 6661Vmos)
and the thing was unstable as well.    Last pass before I moved to bigger issues was make the IRF510s an offboard
output element using a board I had with seriously better layout and that did work with the 5109 drivers and gave
me less than 2db gain tilt. However it was again unstable when I tried to route the power back through the low pass
filter array as it was.  Spurs and carrier leak and the infamous low pass filter were bigger fish that required frying.

A project for my self is a resurrected a Atlas210x ( really old radio with DBMs for mixers) and I've built a requisite
60db 10W power amp for it with minimal difficulty that is flat (+-1db) to 30mhz using IRF510 PP.  Its all about layout. 
So I know its doable, but you can not fold the amp on itself and have antennas (heatsinks) with RF and hope for stability.
I assume you know that but its also for the audience.

Allison

Re: Which xmit transistor is blown?

iz oos
 

I mean into a dummy load


Il 14/set/2018 21:44, "iz oos" <and2oosiz2@...> ha scritto:

Just to see whether it puts out the correct amount of power why not just keying down CW?


Il 14/set/2018 20:11, "Sean W7SKD" <sean.jrdalys@...> ha scritto:
Ok, I think I have this solved.

Really short version:
It appears that my major problem was the laptop that I was using for FT-8.  The laptop that I have been using is not putting out consistent volume/drive on wsjt-x with ft8 (even with a different sound card).  A different computer was tried and everything now works fine.

Longer version:
Ok, I'm a bit at a loss for 'why' this has happened, but here's what happened.  I spent a LOT of time tracing the transmit signal, going back through the mixer stages and into the mic amp (q6).  What I started to notice was that everything looked just great at DC levels, but when under drive ('tune' signal from wsjt-x), the signal levels and frequencies didnt seem to make sense.  I connected a commercial hf transceiver up to the same PC (which I hadnt done, as I didnt have access to one) and discovered that it was doing the SAME thing - effectively no output.   I tried a different sound card (usb connected) with the same problem.  When I switched to a different PC with the commercial transceiver, TWO things were discovered:  1) suddenly I had signal coming out and 2) the transceiver was warning me of high swr.
so...high swr.  It appears that my tuner was set up incorrectly (mfj-971 - has 2 internal jumpers that have to be set for qrp operation, and the swr movement was incorrectly set), and I was likely running high swr for quite a long time with my ubitx and not realizing it. 
Once I got the tuner situation dealt with, the commercial rig works great with the NEW laptop, and still does not work well with the old laptop.  The ubitx ALSO works well with the NEW laptop now.  Whats really odd is that the OLD pc's soundcard still works fine with earbuds, just not driving either radio.

so...what I'm left with appears to be an odd pc situation (inability to drive the radio, but no other noted malfunction), but functioning radio gear.  If anyone has ideas on that, I'm open to them, but I'm not going to expend much effort in that area.

THANK YOU to those who posted and emailed me things to check.  I got into radio in order to learn more about electronics, and that is absolutely happening

73

Sean