20dB 'RF' attenuator - seeking feedback to improve

[joeqsmith]

To be honest, it's way outside of my area.    Most of my hobby experience is below 2GHz.   It would be hard for me to say where to begin.   Maybe use something better than FR4 to start and some better connectors.   When I first saw the previous attenuator design, I was concerned to see the single plane with those vias.  All that added inductance can't be good.   It wasn't. 

These are fairly inexpensive if you don't mind the wait.   Maybe someone will verify my findings for starts. 

I bought two of each type and planned to see if I could improve them.  The first one, I doubt I will do anything with it but maybe I will try some different configurations with these new ones.   

[joeqsmith]

I swapped out all the parts for 2X0603s and trimmed off some of the excess around the launch connectors.  In the end, I absolutely destroyed the flatness.      

The width of the 2X0603s could be the problem.  I may put the original ones back on the board and see how that behaves.   

I have gained an appreciation for our friends in China who made this little board work so well at such a low cost.

[enut11]

Thanks @joeqsmith. I just bought the green 0/10/20/30dB set (again) now that I know it has good performance for my frequency range (currently <1GHz) and will use them to compare the performance of my home made attenuators.

Staying with those 'horrible' metal film resistors because I have some, I built two new 50 ohm 20dB attenuators with SMA connectors and had a friend check them with the following results:

All I can conclude at this stage is that the H-pad is better than the T-pad, but I don't know why. The resistor junctions are floating very close to the ground plane.
enut11

[ogden]

AFAIK real world Pi attenuators (shunt first) have better return loss compared to T. If possible - ask your friend to measure not only S21 but S11, S22 as well. BTW "green 0/10/20/30dB set" also uses Pi, not T type. Could be nice to see measurements of your first attenuator shown in the first post.

[joeqsmith]

It appears now that you have seen some higher frequency data off your friends system, the hook has been set.      

Both of those look pretty good at 250MHz.   Your friend may be able to help you out with your experiments and learning how to use your Nano.

[enut11]

Yes, he has some nice equipment but lives 250Km away.

I came across this VNA demo board on eBay. I has SMA screw SMB connectors and demonstrates the proper response of various component s. A bit expensive but it gives me some ideas.

[enut11]

First tests with my NanoVNA, these span 1-300MHz which is the basic range for the Nano.

Comparing:
1) the original boat-anchor attenuator with BNC connectors from the start of this thread (BA1).
2) slightly improved boat-anchor attenuator with BNC connectors (BA2)
3) improved Pi attenuator with SMA connectors
4) improved T attenuator with SMA connectors
5) commercial 6GHz in-line attenuator for reference (6G)

[enut11]

Attaching 'BA1' 1-300MHz

My conclusions so far:
1) all attenuators are OK from 1-300MHz (blue line) with small dB differences due to resistor tolerances
2) The yellow line (CH0 LOGMAG reflected signal?) shows significant improvement with each successive build. 
Using the commercial 6GHz as reference, it appears that the yellow line should be as low as possible across the frequency sweep.

BA1, my first BNC 20dB attenuator is the worst as frequency goes up, even with all that shielding!
BA2, also BNC equipped, is a little better but still poor.
SMA equipped T atten and Pi atten are better but no match for the commercial unit (6G)

For me it is good to see some progress in performance due to improved layout and connectors.
More to come...
enut11

[joeqsmith]

Yes, he has some nice equipment but lives 250Km away.

I came across this VNA demo board on eBay. I has SMA screw connectors and demonstrates the proper response of various component s. A bit expensive but it gives me some ideas.

In the picture, it looks like SMBs.   SMA would be much better for this.   

2) The yellow line (CH0 LOGMAG reflected signal?) shows significant improvement with each successive build. 
Using the commercial 6GHz as reference, it appears that the yellow line should be as low as possible across the frequency sweep.

Try leaving the CH0 cable open on one end, then try adding the 50 ohm terminator.  Then try the short.  You will find all of signal is reflected back with the short and open, giving a reading near 0.  With the 50 ohm terminator, nothing will be returned and the reading will be -infinity.   Well, obviously it's not a perfect world but you will get the point.

For the really low cost attenuators, they publish the SWR which is giving you some figure of merit for the match.  I think the Nano can read that directly.  For a perfect match (like the 50 ohm terminator) it will read 1.0.   Try it with the open and short as well.    The low cost attenuators were advertised with an SWR of 1.2.  They don't show a different SWR for each attenuator.   

I have been avoiding showing you the match (reflection) to avoid adding confusion.  Up to 2GHz, I can provide some reasonable numbers but moving beyond that requires a higher grade cal kit than I currently have.  Hopefully I will have that sorted out soon.   When I made the video for the Nano, I measured some resistors to check the SWR.  You will find that section here:

https://youtu.be/mKi6s3WvBAM?t=2167

You can try these same experiments to help cement some of these concepts. 

Too bad your friend is so far away.  They could have been a very good resource.   

[enut11]

" 2) The yellow line (CH0 LOGMAG reflected signal?) shows significant improvement with each successive build. 
    Using the commercial 6GHz as reference, it appears that the yellow line should be as low as possible across the frequency sweep.

Quote from @joeqsmith
Try leaving the CH0 cable open on one end, then try adding the 50 ohm terminator.  Then try the short.  You will find all of signal is reflected back with the short and open, giving a reading near 0.  With the 50 ohm terminator, nothing will be returned and the reading will be -infinity.   Well, obviously it's not a perfect world but you will get the point."

Pix of CH0 Open, CH0 50ohm and CH0 Short.
Is the display for CH0 50ohm OK?

[enut11]

Second tests with my NanoVNA, these span 1-900MHz which is the extended range for the Nano.
I believe beyond 300MHz the Nano relies on harmonics with reduced performance.

Comparing:
1) the original boat-anchor attenuator with BNC connectors from the start of this thread (BA1).
2) slightly improved boat-anchor attenuator with BNC connectors (BA2)
3) SMA improved Pi attenuator
4) SMA improved T attenuator
5) commercial 6GHz in-line attenuator for reference (6G)

Again, up to 250MHz, the 6GHz commercial attenuator shows good performance. Beyond that, performance deteriorates and this may be a limitation of the Nano.
For my home made attenuators, ranking is as before.

There is room for improvement however...

[enut11]

Looking at BA1, the gain (loss) blue line goes crazy at the top end.

[joeqsmith]

Looking at BA1, the gain (loss) blue line goes crazy at the top end.

Of course it does.  It's $50, not $50,000.     

Your SOL data looks what I would expect.  For a reference, RCLoad was taken from my Nano. 
(added)

Changing to display, trace, trace0, format, SWR
Showing SWR of the 50 ohm load and a 100 ohm load. 

Measuring the SWR of one of the home made 20dB attenuators.
(should add that you would want to terminate the unused port)

Again, up to 250MHz, the 6GHz commercial attenuator shows good performance. Beyond that, performance deteriorates and this may be a limitation of the Nano.
For my home made attenuators, ranking is as before.

 

Go back and look at some of the data I have provided comparing the Nano with my other VNAs.  Keep it below 400MHz or so.   I would trust a high grade attenuator long before I would trust the $50 Nano once you get above 300MHz. 

[enut11]

Correct, the Nano seems to work well up to ~300MHz.

For my four home made 20dB attenuators from above, the SWR at 300MHz is:
1) BNC BA1 - 2.00 (ie very poor)
2) BNC BA2 - 1.44
3) SMA Pi - 1.13 (not too bad?)
4) SMA T - 1.16 (ditto?)
5) Commercial 6GHz - 1.05 (very good)
enut11

[joeqsmith]

After you receive your low cost attenuators, it would be interesting to have your friend characterize them with their system.   

...
3) SMA Pi - 1.13 (not too bad?)
4) SMA T - 1.16 (ditto?)
...

The following article is on SWR. It may help you interpret your results.
http://www.arrl.org/files/file/Technology/tis/info/pdf/q1106037.pdf

[joeqsmith]

Graphs showing the reflection coefficient using the Nano, sweeping from 1 to 100MHz, using a from 0 to 30dB step attenuator with 1dB/step.   You can see it's fairly linear to 16dB or so.   20dB even at a MHz looks like it is asking a bit much for your $50. 

[joeqsmith]

If I step from 0-20dB but sweep from 1-300MHz which is in the region you are operating,  you can clearly see how nonlinear the Nano is.  Ideally you would want to constrain all of your measurements to avoid these errors.   

You may find you get to a point where you want something better than the Nano.  My advice is first, learn as much as you can with  the Nano.  Then if you still decide to upgrade, consider even my old HP8754A will put the Nano to shame when we start looking at noise, flatness of the signal path and linearity.   

Maybe by then the Nano V2 will be readily available and we will have a better idea on if it would be a suitable upgrade.   

[ogden]

Then if you still decide to upgrade, consider even my old HP8754A will put the Nano to shame when we start looking at noise, flatness of the signal path and linearity.   

Why don't you demonstrate it by running linearity test of 0 to 30dB step attenuator?

[enut11]

Graphs showing the reflection coefficient using the Nano, sweeping from 1 to 100MHz, using a from 0 to 30dB step attenuator with 1dB/step.   You can see it's fairly linear to 16dB or so.   20dB even at a MHz looks like it is asking a bit much for your $50.

Thanks @joeqsmith. A very interesting way to present data and show the limitations of the Nano

I have just acquired some SMD resistors and looking to build a very small 20dB attenuator.

Considering this method which includes a small double sided PCB for ground and soldering the resistors between and across the SMA pins.
Shown next to my boat-anchor attenuator to lend scale. Metal film resistors will not be used in this attenuator.
Hardest part will be how to hold the SMD resistors in place while soldering.
enut11

[joeqsmith]

As you continue to improve your construction practices, just keep in mind my early posts where I've shown some of the problems with the Nano.    If you find your new design is not as flat as you may expect, you may want to make the drive to your friends place or ship the parts to them to.   With the one you are now making, depending on the frequency range, I expect you will see a pretty big difference between the two VNAs.   

https://www.eevblog.com/forum/projects/20db-rf-attenuator-seeking-feedback-to-improve/msg2902996/#msg2902996

[enut11]

Well working with tiny SMD resistors is a challenge for me. The ones I have are only 2mm long.
Already lost 2 just by picking them up with tweezers! Popped out never to be seen again.
I did manage to measure resistance with a home-made clip. This one was stamped 5110 meaning 511 ohms.
enut11

[ogden]

The ones I have are only 2mm long.

Those are 0805, considered "big" from modern (automatic) SMT point of view. Note that your SMA end launch (edge) connectors have little over 2mm distance between bottom of the central pin and bottom of the bottom legs. Hint: you can use same "dead bug" technique as for your boat anchor attenuator by putting whole SMA edge-launch connectors on top of solid ground plane. Start with single connector, then solder resistors, when done with resistors - only then add 2nd SMA at the end of resistor chain.

[enut11]

Hi @ogden. I had already soldered the SMD resistors to the SMA connectors before I read your post. I hope I got it right. I tested it at DC but have yet to test it at VHF.
enut11

[ogden]

No, you did not place connectors as I told - when no parts of connector touch bottom side of PCB. Is it so that you use double sided PCB with top copper layer still in place, very, very close to central pins? - Not good.

[enut11]

@ogden, thanks. Now I understand. I will also make your idea and it will be interesting to compare them.
enut11