1. SoftwareControlledHamRadio
  2. Topics

New filter (50 ohms) on AD985x module to normalize output over the whole frequency range.

Edwin Houwertjes <ehouwertjes@...>
 


I sold my first AA because I was more fond of my other analyzer. I did still have parts that could be used to make a second W8TEE/K2ZIA AA.

When I tested it, it seemed that the output power differed quite a lot. 

On the other AD9851 board I already had big variations in signal strength that looked like this?

AA output power:

The lowpass filter on de the AD9850 and AB9851 boards was not designed for 50 ohms, but more like around 150 ohms.

The signals I had now, with this new AD9851 board, had very nasty dips around 80 meter. I took off all the filter parts and build a new filter.
Now my output power looks like this:
(sorry for the lager frequency range in this picture, I ran the AA up to 75Mhz)

Yes, I know, it is still dropping a little, probably because of the quality of the parts I used. I did at first use some inductors I had in some cheep kit, but these had such band Q factor so I got me some new rf inductors. The small drop could also be caused by the ad chip itself but I suspect the quality of the capacitors.
I also ran it over its specs, op to 75Mhz so a lower output there should be expected.


This is the filter I used. It uses al lot of the same value components, the capacitors on the in and output already should be 22pf but I replaced them just to be sure.


The design program (Elsie) shows a curve lik this:



This is what the schematic looks like:


Here are the parts located.



I already had a quick and dirty fix by removing R7 from the DDS board and putting a 82 ohm resistor in series to improve the output signal.

With this filter we still do not want R7 mounted, but we do not want to put the resistor in series with C7 on the mainboard.
With this new filter and the fact that we do not need an 82 ohms resistor in series with C7 we should expect a little more output signal.
I do not know if that could cause us any problems. I do not expect any, but I did not yet test it.

In the lines above, i was talking about part the PE1PWF mods for the W8TEE / K2ZIA antenna analyzer. You could of course also use the same board/filter in other equipment that use a 50 ohm circuit and are not matched to the original approx. 150 Ohm filter on the DDS board. 


73'

Edwin PE1PWF











John P
 

Very nice Edwin! Thanks for adding more work to my to-do list! LOL!
--
John P.
WA2FZW

dick faust
 

You might be interested in some work done by Rudolf Reuter at
http://www.rudiswiki.de/wiki/AmateurRadioDDSgenerator
He presents some convincing evidence that the original filter is 200 ohms and relatively flat, along with several other designs.
You did not mention the outputs for either filter, but if you were to use a minimum loss pad from 200 ohms to 50 oms it would be just under 12dB and may correct the bumps in the response [I believe most Spectrum Analyzers are 50 ohm input].  Perhaps someone could run the LTspice file with a 50 ohm termination and see what that response would yield.  Pass band ripple is very dependent on termination impedance.
Independent of the results of the synthesis, the results of your work looks like a very suitable solution.
Dick K9IVB

Edwin Houwertjes <ehouwertjes@...>
 

Thank Dick, I did see this info before but forgot where to find it.

I do not have ltspice, I use Elsie.
The images in the link use a large scale so a 3db loss would only be a few pixels in the image.




At 200 ohms it is reasonably flat but does not look that perfect.



I think it looks even better around 130 ohms....










If you use the output of the AD985x module into a 50 ohm load, it will look like this.



A 50 ohms filter like the one I made should look like this.

In the DDS60 and Sark100 is a simpler smaller filter, it gives these results according to Elsie



I did measure the output signal of the AA, and not the DDS output. The DDS output is amplified by a MMIC which has 50 ohm in- and output.
This output is pushed into the measuring bridge. Terminating the AA output with a 50 Ohm analyzer will mean the load on the MMIC also is 50 ohms. 

In my measurements you can see the differences between the original filter and my design mountned in the AA.
So it is not just the DDS board output although in both situations the graph should look similar because the output is "terminated" by the input of the MMIC. 

Why we see such a big drop-off after 55Mhz is a bit puzzeling. The graph shows a dip at 65.5 Mhz but it is not that huge.

Anyway, the mod used in the PE1PWF mods already kind of helps to equalize the output power of the AA, but making a new lowpass filter just is the best solution.

I hope to post a new image soon using some better caps.

73'

Edwin
















2018-06-09 20:27 GMT+02:00 dick faust via Groups.Io <dick_faust@...>:

You might be interested in some work done by Rudolf Reuter at
http://www.rudiswiki.de/wiki/AmateurRadioDDSgenerator
He presents some convincing evidence that the original filter is 200 ohms and relatively flat, along with several other designs.
You did not mention the outputs for either filter, but if you were to use a minimum loss pad from 200 ohms to 50 oms it would be just under 12dB and may correct the bumps in the response [I believe most Spectrum Analyzers are 50 ohm input].  Perhaps someone could run the LTspice file with a 50 ohm termination and see what that response would yield.  Pass band ripple is very dependent on termination impedance.
Independent of the results of the synthesis, the results of your work looks like a very suitable solution.
Dick K9IVB


Edwin Houwertjes <ehouwertjes@...>
 

Sorry John ;-) 

Did not mean to add work to your todo list.

For the inductor I used a coilcraft 0805CS-101XJLC

For caps, I might change to these.
251R15S820JV4E 
251R15S220JV4E

It seems like overkill, but a hi q cap would probably work better than the standard caps I use now.
Maybe I can find some other RF caps in my collection. I think better caps should improve the filter characteristics.


73'

Edwin






2018-06-09 14:29 GMT+02:00 John P <j.m.price@...>:

Very nice Edwin! Thanks for adding more work to my to-do list! LOL!
--
John P.
WA2FZW


Edwin Houwertjes <ehouwertjes@...>
 

I just realized the graphs I posted do not show the suppression, only the passband.

The filter was optimized to have the least amount of loss, cutt off close to 70Mhz, and the least amount of different parts.
This will mean the suppression at around 250Mhz is only about 43dB.

If you would change the in and output both caps to 33pf the passband improves a tiny bit but the suppression also is better with about -43db.







I think this gives a very nice 50 ohm filter with a passband loss of less than 0.5dB

If you want a better rejection, you will see the losses below 70Mhz will increase.
If you look at the original filter, you will see the rejection is 76dB or better but you also see 3 dips in the passband of 1.1 , 1.4 and 2.6dB.  (terminated at 200 Ohm)

The filter will always be a compromise, I think for the AA it would be nice to have the most constant power output. SInce we only have a small signal, I don't think we need to aim for a bigger rejection.

About the drop in signal we did see on the analyzer output, that could also have to do with the MMIC. The MMIC does not have the same gain over the oprtational frequency range.

Maybe an AD8007 or AD8008 would have neen nice as an amplifier.

Well, I am not going to mess with that, I will however get me some nice 22,33 and 82 pF caps to see if those perform better than the ones I had in my sample-book.

73'

Edwin



2018-06-09 22:54 GMT+02:00 Edwin Houwertjes <ehouwertjes@...>:

Sorry John ;-) 

Did not mean to add work to your todo list.

For the inductor I used a coilcraft 0805CS-101XJLC

For caps, I might change to these.
251R15S820JV4E 
251R15S220JV4E

It seems like overkill, but a hi q cap would probably work better than the standard caps I use now.
Maybe I can find some other RF caps in my collection. I think better caps should improve the filter characteristics.


73'

Edwin






2018-06-09 14:29 GMT+02:00 John P <j.m.price@...>:
Very nice Edwin! Thanks for adding more work to my to-do list! LOL!
--
John P.
WA2FZW



Edwin Houwertjes <ehouwertjes@...>
 

I promised to run a test with some capacitors I expected to be better.

Unfortunately the output signal is just about the same.



I still see the same drop in the signal....

After examinging just the filter it seems it does kind off looks the same...
The output Elsie showed was a very flat line within about 0.5db, the real results show the signal differs 1.5db up to about 55Mhz, 
after that it drops another 2dB an rapidly drops near 70Mhz. 

I am not sure why that happens, but at least we have some results and I know I don't need expensive hi q caps for a simple filter like this. 
The caps in my sample book seem good enough.






2018-06-08 23:23 GMT+02:00 Edwin Houwertjes <ehouwertjes@...>:


I sold my first AA because I was more fond of my other analyzer. I did still have parts that could be used to make a second W8TEE/K2ZIA AA.

When I tested it, it seemed that the output power differed quite a lot. 

On the other AD9851 board I already had big variations in signal strength that looked like this?

AA output power:

The lowpass filter on de the AD9850 and AB9851 boards was not designed for 50 ohms, but more like around 150 ohms.

The signals I had now, with this new AD9851 board, had very nasty dips around 80 meter. I took off all the filter parts and build a new filter.
Now my output power looks like this:
(sorry for the lager frequency range in this picture, I ran the AA up to 75Mhz)

Yes, I know, it is still dropping a little, probably because of the quality of the parts I used. I did at first use some inductors I had in some cheep kit, but these had such band Q factor so I got me some new rf inductors. The small drop could also be caused by the ad chip itself but I suspect the quality of the capacitors.
I also ran it over its specs, op to 75Mhz so a lower output there should be expected.


This is the filter I used. It uses al lot of the same value components, the capacitors on the in and output already should be 22pf but I replaced them just to be sure.


The design program (Elsie) shows a curve lik this:



This is what the schematic looks like:


Here are the parts located.



I already had a quick and dirty fix by removing R7 from the DDS board and putting a 82 ohm resistor in series to improve the output signal.

With this filter we still do not want R7 mounted, but we do not want to put the resistor in series with C7 on the mainboard.
With this new filter and the fact that we do not need an 82 ohms resistor in series with C7 we should expect a little more output signal.
I do not know if that could cause us any problems. I do not expect any, but I did not yet test it.

In the lines above, i was talking about part the PE1PWF mods for the W8TEE / K2ZIA antenna analyzer. You could of course also use the same board/filter in other equipment that use a 50 ohm circuit and are not matched to the original approx. 150 Ohm filter on the DDS board. 


73'

Edwin PE1PWF












Brian Minnis
 

Dear Edwin et al

 

Unless I am mistaken, the roll-off in the output signal from the DDS (and therefore the signal on the top of the bridge) is caused by the sample-and-hold function.  The sample-and hold function incorporated in the DAC of the DDS produces an inherent Sin(x)/x filter function as shown by the picture below.

 

I have assumed a clock frequency of 125 MHz which forces the first zero of transmission at 125 MHz.  The attenuation at 50 MHz caused by this filtering function is about 2.5 dB which I think is consistent with your measurements.  I hope this sheds some light on your mystery.  I don’t think the roll-off has much to do with the ohmic losses in the filter.

 

Kind regards,

 

Brian

M0KGW

 

From: SoftwareControlledHamRadio@groups.io [mailto:SoftwareControlledHamRadio@groups.io] On Behalf Of Edwin Houwertjes
Sent: 13 June 2018 17:43
To: SoftwareControlledHamRadio@groups.io
Subject: Re: [SoftwareControlledHamRadio] New filter (50 ohms) on AD985x module to normalize output over the whole frequency range.

 

I promised to run a test with some capacitors I expected to be better.

 

Unfortunately the output signal is just about the same.

 

 

I still see the same drop in the signal....

 

After examinging just the filter it seems it does kind off looks the same...

The output Elsie showed was a very flat line within about 0.5db, the real results show the signal differs 1.5db up to about 55Mhz, 

after that it drops another 2dB an rapidly drops near 70Mhz. 

 

I am not sure why that happens, but at least we have some results and I know I don't need expensive hi q caps for a simple filter like this. 

The caps in my sample book seem good enough.

 

 

 

 

 

2018-06-08 23:23 GMT+02:00 Edwin Houwertjes <ehouwertjes@...>:

 

I sold my first AA because I was more fond of my other analyzer. I did still have parts that could be used to make a second W8TEE/K2ZIA AA.

 

When I tested it, it seemed that the output power differed quite a lot. 

 

On the other AD9851 board I already had big variations in signal strength that looked like this?

 

AA output power:

The lowpass filter on de the AD9850 and AB9851 boards was not designed for 50 ohms, but more like around 150 ohms.

 

The signals I had now, with this new AD9851 board, had very nasty dips around 80 meter. I took off all the filter parts and build a new filter.

Now my output power looks like this:

(sorry for the lager frequency range in this picture, I ran the AA up to 75Mhz)

Yes, I know, it is still dropping a little, probably because of the quality of the parts I used. I did at first use some inductors I had in some cheep kit, but these had such band Q factor so I got me some new rf inductors. The small drop could also be caused by the ad chip itself but I suspect the quality of the capacitors.

I also ran it over its specs, op to 75Mhz so a lower output there should be expected.

 

 

This is the filter I used. It uses al lot of the same value components, the capacitors on the in and output already should be 22pf but I replaced them just to be sure.

 

The design program (Elsie) shows a curve lik this:

 

 

This is what the schematic looks like:

 

Here are the parts located.

 

 

I already had a quick and dirty fix by removing R7 from the DDS board and putting a 82 ohm resistor in series to improve the output signal.

 

With this filter we still do not want R7 mounted, but we do not want to put the resistor in series with C7 on the mainboard.

With this new filter and the fact that we do not need an 82 ohms resistor in series with C7 we should expect a little more output signal.

I do not know if that could cause us any problems. I do not expect any, but I did not yet test it.

 

In the lines above, i was talking about part the PE1PWF mods for the W8TEE / K2ZIA antenna analyzer. You could of course also use the same board/filter in other equipment that use a 50 ohm circuit and are not matched to the original approx. 150 Ohm filter on the DDS board. 

 

 

73'

 

Edwin PE1PWF

 

 

 

 

 

 

 

 

 

 

 

 

DuWayne Schmidlkofer
 

I think it would be interesting if you could compare the response of the unfiltered output of the DDS to see if the drop off is caused by the filter or the DDS itself.
DuWayne KV4QB

On 6/15/2018 6:38 AM, Brian Minnis via Groups.Io wrote:
Dear Edwin et al
Unless I am mistaken, the roll-off in the output signal from the DDS (and therefore the signal on the top of the bridge) is caused by the sample-and-hold function.  The sample-and hold function incorporated in the DAC of the DDS produces an inherent Sin(x)/x filter function as shown by the picture below.
I have assumed a clock frequency of 125 MHz which forces the first zero of transmission at 125 MHz.  The attenuation at 50 MHz caused by this filtering function is about 2.5 dB which I think is consistent with your measurements.  I hope this sheds some light on your mystery.  I don’t think the roll-off has much to do with the ohmic losses in the filter.
Kind regards,
Brian
M0KGW
*From:*SoftwareControlledHamRadio@groups.io [mailto:SoftwareControlledHamRadio@groups.io] *On Behalf Of *Edwin Houwertjes
*Sent:* 13 June 2018 17:43
*To:* SoftwareControlledHamRadio@groups.io
*Subject:* Re: [SoftwareControlledHamRadio] New filter (50 ohms) on AD985x module to normalize output over the whole frequency range.
I promised to run a test with some capacitors I expected to be better.
Unfortunately the output signal is just about the same.
I still see the same drop in the signal....
After examinging just the filter it seems it does kind off looks the same...
The output Elsie showed was a very flat line within about 0.5db, the real results show the signal differs 1.5db up to about 55Mhz,
after that it drops another 2dB an rapidly drops near 70Mhz.
I am not sure why that happens, but at least we have some results and I know I don't need expensive hi q caps for a simple filter like this.
The caps in my sample book seem good enough.
2018-06-08 23:23 GMT+02:00 Edwin Houwertjes <ehouwertjes@... <mailto:ehouwertjes@...>>:
I sold my first AA because I was more fond of my other analyzer. I did still have parts that could be used to make a second W8TEE/K2ZIA AA.
When I tested it, it seemed that the output power differed quite a lot.
On the other AD9851 board I already had big variations in signal strength that looked like this?
AA output power:
The lowpass filter on de the AD9850 and AB9851 boards was not designed for 50 ohms, but more like around 150 ohms.
The signals I had now, with this new AD9851 board, had very nasty dips around 80 meter. I took off all the filter parts and build a new filter.
Now my output power looks like this:
(sorry for the lager frequency range in this picture, I ran the AA up to 75Mhz)
Yes, I know, it is still dropping a little, probably because of the quality of the parts I used. I did at first use some inductors I had in some cheep kit, but these had such band Q factor so I got me some new rf inductors. The small drop could also be caused by the ad chip itself but I suspect the quality of the capacitors.
I also ran it over its specs, op to 75Mhz so a lower output there should be expected.
This is the filter I used. It uses al lot of the same value components, the capacitors on the in and output already should be 22pf but I replaced them just to be sure.
The design program (Elsie) shows a curve lik this:
This is what the schematic looks like:
Here are the parts located.
I already had a quick and dirty fix by removing R7 from the DDS board and putting a 82 ohm resistor in series to improve the output signal.
With this filter we still do not want R7 mounted, but we do not want to put the resistor in series with C7 on the mainboard.
With this new filter and the fact that we do not need an 82 ohms resistor in series with C7 we should expect a little more output signal.
I do not know if that could cause us any problems. I do not expect any, but I did not yet test it.
In the lines above, i was talking about part the PE1PWF mods for the W8TEE / K2ZIA antenna analyzer. You could of course also use the same board/filter in other equipment that use a 50 ohm circuit and are not matched to the original approx. 150 Ohm filter on the DDS board.
73'
Edwin PE1PWF

Edwin Houwertjes <ehouwertjes@...>
 

Hi Brian,

I have to admit I do not quite understand what you mean. All electronics knowledge is just what I learned myself I am a mechanical engineer and not a RF electronics engineer. The AD9851 is the board I use (180Mhz clock)
As far as I can remember I have seen much flatter output signal levels on the Sark100 analyzer which allso uses teh AD9851 chip but an other type off filter.

About the tests, in my last message before this one, I used two pictures, one shows the output signal of my AA, the second image is the filter on a tracking generator.
So the AD is not the signal source the signal source is the tracking generator in a Marconi 2945 radiotestset/spectrum analyser. Yes I know this is not quite typical mechanical engineering equipment :-)


I now have the feeling my inductors might not be what they were sold for....I will see if I can hook them up to my LC meter.

Edwin




Brian Minnis
 

Dear Edwin

 

What I’m suggesting in short is that it is the DDS itself that is largely responsible for the roll-off of signal with frequency and not the ohmic losses in the filter on the board.  The filter losses will increase with frequency but I don’t think this effect is dominant.  But you can check this by measurement quite easily…….

 

The DDS has two outputs, one which is used to drive the LP filter and another inverting one which just terminates in a 100ohm resistor (R10) on the DDS board. R10 is shunted to ground by the jumper J4. If you take off the J4 jumper, you can then measure the DDS output signal directly on pin 2 of the jumper.  Apart from the phase inversion, this is exactly what you would get if you measured the main DDS output on pin 21 of the chip with the filter shorted from end to end and the main SWR bridge disconnected.  You should see that the signal on R10 rolls off gently as per your own observations.

 

The DAC on the output of the DDS chip is followed by a sample-and-hold block which aims to make the sinusoidal output as smooth as possible without any returns to zero in between clock samples.  As I explained in my previous note, this S&H block, by its very nature, introduces a filtering function that has a Sin(x)/x characteristic.  If your clock frequency is 180MHz then the first zero of this filter function will be at 180 MHz.  This derives from the basic theory of DSP.

 

So I don’t think you need to search so hard for excess losses in your filter components.  They are unlikely to be to blame.

 

But I suppose I could be missing something………….. J

 

Kind regards,

 

Brian

M0KGW

 

From: SoftwareControlledHamRadio@groups.io [mailto:SoftwareControlledHamRadio@groups.io] On Behalf Of Edwin Houwertjes
Sent: 15 June 2018 16:59
To: SoftwareControlledHamRadio@groups.io
Subject: Re: [SoftwareControlledHamRadio] New filter (50 ohms) on AD985x module to normalize output over the whole frequency range.

 

Hi Brian,

I have to admit I do not quite understand what you mean. All electronics knowledge is just what I learned myself I am a mechanical engineer and not a RF electronics engineer. The AD9851 is the board I use (180Mhz clock)
As far as I can remember I have seen much flatter output signal levels on the Sark100 analyzer which allso uses teh AD9851 chip but an other type off filter.

About the tests, in my last message before this one, I used two pictures, one shows the output signal of my AA, the second image is the filter on a tracking generator.
So the AD is not the signal source the signal source is the tracking generator in a Marconi 2945 radiotestset/spectrum analyser. Yes I know this is not quite typical mechanical engineering equipment :-)


I now have the feeling my inductors might not be what they were sold for....I will see if I can hook them up to my LC meter.

Edwin



1 - 11 of 11