Author Topic: What is a good and cheap way to do frequency response testing on a scope?  (Read 5528 times)

0 Members and 1 Guest are viewing this topic.

Offline Amazing

  • Regular Contributor
  • *
  • Posts: 59
  • Country: us
I know about using a fast rise time test, but that requires that you already know if a scope has a gaussian or maximally flat response.

My two (working) scopes are a Rigol DS1202CA, 200MHz, and LeCroy WavePro 960, with a 2GHz bandwidth.

I'd like to know more about the frequency response curve (bode plot) of the 960 in particular.

Right now the only signal generator I have is a Siglent that goes up to 40MHz.   I also have an HP 213B fast risetime pulse generator, but I don't think that will be useful, nor am I certain of its actual rise time.

I'm thinking of trying to find something ancient, used on eBay for cheap.  But how do I know if it is flat across it's output range?

Which brings me to my next question, is there a wide-band RF power meter that is cheap?  Something that can be used to characterize a signal generator, which can then be used to characterize a scope?

I'd love to hear your thoughts!

 

Online tautech

  • Super Contributor
  • ***
  • Posts: 16668
  • Country: nz
  • Taupaki Technologies Ltd. NZ Siglent Distributor
    • Taupaki Technologies Ltd.
I know about using a fast rise time test, but that requires that you already know if a scope has a gaussian or maximally flat response.

My two (working) scopes are a Rigol DS1202CA, 200MHz, and LeCroy WavePro 960, with a 2GHz bandwidth.

I'd like to know more about the frequency response curve (bode plot) of the 960 in particular.
Why ?
Is the manufacturers datasheet/s not adequate ?  :-//
Avid Rabid Hobbyist
 

Offline Amazing

  • Regular Contributor
  • *
  • Posts: 59
  • Country: us
Why ?
Is the manufacturers datasheet/s not adequate ?  :-//

Curiosity mostly.

1. The manual and datasheet on say "-3dB at 2GHz" and don't specify the curve

2. I was wondering how much point there is to sampling at 16GHz with 2GHz front end.  I've read the old threads on this topic... Just curious how they apply to my scope.

3. I'm curious about what the actual rise time is of my HP213B -- or at least get a maximum bound on it.

4. I thought it would provoke an interesting discussion about testing and calibrating on a budget.
 

Offline jonovid

  • Frequent Contributor
  • **
  • Posts: 814
  • Country: au
    • JONOVID
given the level of white RF noise in most cities nowadays. a pace of wire  will work as a signal generator
add a tuned circuit ?  :-//
Hobby of evil genius      basic knowledge of electronics
 
The following users thanked this post: Amazing

Offline Amazing

  • Regular Contributor
  • *
  • Posts: 59
  • Country: us
given the level of white RF noise in most cities nowadays. a pace of wire  will work as a signal generator
add a tuned circuit ?  :-//

Not sure if you are serious or joking, but environmental noise is unlikely to be flat.  Also any tuned circuit or filter will have it's own frequency response curve that needs to be characterized. 

Also it's probably too low level to give good results on an oscilloscope.

Although this reminds that I also have on hand a Tekbox TBWA2 wide band amplifier, 3MHz to 3GHz, which may be useful as part of a test rig.  This amp comes with a Bode plot showing a variation of about 1.4dB from 10MHz to 2.5GHz.  I don't know if I can trust that data, but at least it is a starting place.
 

Offline jonovid

  • Frequent Contributor
  • **
  • Posts: 814
  • Country: au
    • JONOVID
i stand corrected  :-[
Hobby of evil genius      basic knowledge of electronics
 

Online tautech

  • Super Contributor
  • ***
  • Posts: 16668
  • Country: nz
  • Taupaki Technologies Ltd. NZ Siglent Distributor
    • Taupaki Technologies Ltd.
Why ?
Is the manufacturers datasheet/s not adequate ?  :-//

Curiosity mostly.

1. The manual and datasheet on say "-3dB at 2GHz" and don't specify the curve

2. I was wondering how much point there is to sampling at 16GHz with 2GHz front end.  I've read the old threads on this topic... Just curious how they apply to my scope.

3. I'm curious about what the actual rise time is of my HP213B -- or at least get a maximum bound on it.

4. I thought it would provoke an interesting discussion about testing and calibrating on a budget.
1. Does it really matter if the -3dB point IS NOT exceeded before 2 GHz ?
2. LeCroys are famous for their considerable signal analysis capabilities.....you need lots of datapoints to do this accurately.
3. It's probably still very close to the specs in the datasheet, check it.
4. With this level of equipment forget calibration on a budget, it's a highly specialised field with very expensive calibrated gear and elaborate methodology to get the job done properly. Shortcuts can not normally be taken and procedures need to be followed to the letter and completed in full.

Use your gear and gain familiarity with it and you'll know when something's not right however it's good to have a few simple shop references as sanity checks. (frequencies, voltages, etc.)
Avid Rabid Hobbyist
 
The following users thanked this post: Amazing

Offline Amazing

  • Regular Contributor
  • *
  • Posts: 59
  • Country: us
i stand corrected  :-[

You do make an interesting point about using some kind of noise as a signal source instead of pure sine wave. 
 

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 10284
  • Country: us
  • DavidH
Which brings me to my next question, is there a wide-band RF power meter that is cheap?  Something that can be used to characterize a signal generator, which can then be used to characterize a scope?

We discussed this extensively in connection with verifying the performance of replacement leveling heads for the Tektronix SG504 leveled generator up to 1 GHz on the TekScopes@yahoogroups.com email list.  My suggestion and what I have used for this is a sampling oscilloscope.  Sampling heads have a very predictable sin(x)/x frequency response which can be characterized if necessary by measuring the first null using an RF signal generator.  Some RF sampling voltmeters might be suitable although I am not familiar with any which operate significantly above 1 GHz.

I would rather know the transition time and pulse response of an oscilloscope since it is primarily a time domain instrument but a sampling oscilloscope is also suitable for calibrating reference level pulse generators.
 
The following users thanked this post: Amazing

Offline Amazing

  • Regular Contributor
  • *
  • Posts: 59
  • Country: us
Correction: I found that my tests were invalid in this post because I was using a sub-par BNC cable that was not quite 50 Ohms.   Also my scope seems to distort a tiny bit if the signal goes into the top division on the screen.  But I'm leaving this here anyway because I think the process was interesting.  Actual final test results are down below.

David:  That's an interesting option.  I did some reading this morning on sampling scopes like you suggested.  Seems like a whole other rabbit hole to go down.   Thanks for the pointer to the tek scopes forum; I have a 321A that is awaiting repair...

What do you all think of this method of checking scope bandwidth:

https://community.keysight.com/community/keysight-blogs/oscilloscopes/blog/2016/09/01/how-to-measure-your-oscilloscope-and-probe-s-bandwidth-yourself

I ran through it this today on the WavePro 960.  Here are my results.  Equipment:

HP 213B pulse generator, which is supposed to have a less than 100ps rise time.  But it's also like 50 years old, so who knows.   

Lecroy WavePro 960, calibrated by LeCroy almost a year ago. I expect it to be working pretty well.

First, the rise time measurement of the HP213B.  The "A" trace on this screenshot is the derivative of the actual signal. 



Next, I have a wide view to get more FFT points.  The red "A" trace is the derivative of the signal.  The pink "B" trace is the FFT of the derivative.  And the lavender "C" trace is an average of the FFT.



there is a strange spur at 2GHz.  The response seems to be down about 6dB leading up to 2 GHz, which should not be correct for this scope.  There is also a small step after the initial rise on the signal.  I think this is due to a reflection on the 2 meter BNC cable.

Here is another version of the previous image, zoomed in a bit on the horizontal to get rid of the step.  The spur a 2GHz disappears.  But the response at 2GHz is even lower.



Note that all of the above is plotting the POWER spectrum.  If I change the display to MAGNITUDE,  all of my FFT values are suddenly in MV/s, which is appropriate given the that the source data is a derivative of voltage, but not particularly helpful in this case.  However, the max value at 100MHz is 27.4 MV/s and the min value at about 1.5GHz is 14 MV/s, and that corresponds to a drop of 6dB.

 Here is the view with the FFT set to magnitude:



What do you folks think of all this?
« Last Edit: January 03, 2017, 04:45:04 am by Amazing »
 

Offline Amazing

  • Regular Contributor
  • *
  • Posts: 59
  • Country: us
Re: What is a good and cheap way to do frequency response testing on a scope?
« Reply #10 on: January 02, 2017, 11:12:16 pm »
Ok, after thinking about that for a bit, I'm starting to think that my 2GHz oscilloscope only has  a bandwidth of 1.5 GHz.

That's using the 10% to 90% rise time of 317ps, and using a multiplier of 0.45.
 

Offline joeqsmith

  • Super Contributor
  • ***
  • Posts: 6009
  • Country: us
Re: What is a good and cheap way to do frequency response testing on a scope?
« Reply #11 on: January 03, 2017, 12:57:42 am »
0.45 assumes a certain roll off that modern DSOs may not have. 
How electrically robust is your meter?? https://www.youtube.com/channel/UCsK99WXk9VhcghnAauTBsbg
 
The following users thanked this post: Amazing

Offline Amazing

  • Regular Contributor
  • *
  • Posts: 59
  • Country: us
Re: What is a good and cheap way to do frequency response testing on a scope?
« Reply #12 on: January 03, 2017, 01:00:07 am »
0.45 assumes a certain roll off that modern DSOs may not have.

That's actually how I started down this path, trying to determine the rolloff for this scope.
 

Offline Amazing

  • Regular Contributor
  • *
  • Posts: 59
  • Country: us
Re: What is a good and cheap way to do frequency response testing on a scope?
« Reply #13 on: January 03, 2017, 02:11:33 am »
Back from dinner and now I'm getting the expected results with a -3dB point a 2GHz.

I came back intending to check that the other 3 channels had the same response as channel 1, and they did.

Going back to channel 1, the only things I changed (I think) were:

  • the sensitivity of channel 1 (was at 33mV/div, now at 40mV/div)
  • move the trigger point close to the vertical center -- might be a bit more stable
  • changed to a shorter BNC cable (1 meter) that is giving a slightly better rise time value

Playing with it, the FFT seems to get weird as the signal waveform gets closer to the top of the display, even if the waveform is not obviously being clipped.

Here's the latest result, showing 3dB down at 2GHz, just like it should be.



Again, the lavender "C" trace is an average of the FFT, scaled at 1Ghz/div horizontal, 5.9dBm vertical (oops, must have bumped the encoder, should have been 6dBm)

So, questions remaining:
  • Is this technique for measuring scope bandwidth valid?
  • Given a new measured rise time value of 305ps, that implies that the correct bandwidth multiplier for determining bandwidth from rise time, for this scope, is 0.61.  Does that seem like a plausible number?
Lastly, it looks like it will be pointless to do any equivalent time sampling at 16GHz, due to the sharp rolloff above 2GHz.  However that sample rate gives a lot of extra data points which seems to be useful for generating detailed FFTs, as tautech mentioned.

Thanks!
« Last Edit: January 03, 2017, 04:36:22 am by Amazing »
 

Offline macboy

  • Super Contributor
  • ***
  • Posts: 1990
  • Country: ca
Re: What is a good and cheap way to do frequency response testing on a scope?
« Reply #14 on: January 03, 2017, 04:02:43 am »
Lastly, it looks like it will be pointless to do any equivalent time sampling at 16GHz, due to the sharp rolloff above 2GHz.  However that sample rate gives a lot of extra data points which seems to be useful for generating detailed FFTs, as tautech mentioned.

Thanks!
Don't forget that the calculated frequency response is of the system, not of the scope. The system consists of your step/pulse generator (which has a finite, non-ideal rise time and other inherent irregularities), the cable, and the scope. You cable is certainly not ideal; it has a limited frequency response of its own, and may not be a perfect match to the source and/or scope input impedances. Speaking of which, the response shows an obvious mismatch evidenced by the second small step. It looks to me that the cable has a lower impedance that the source has, so it loads the output of the source a little too much, until the reflection of the pulse reaches the source and the system is basically at steady state (DC). To truly measure the response of only the scope this way, you need to provide an ideal step (zero rise time) to its input, which is impossible. You can do quite well with an input 10x faster than the scope but you don't quite have that either.

You said equivalent time sampling at 16 GS/s, but this scope does real-time or single shot sampling at 16 GS/s. It can do equivalent time sampling at 50 GS/s.

I just picked up a WavePro 960 about a week ago. Mine has 16 MS memory, WaveAnalyzer Pro, DDM/PRML, LAN, PCCard, and more... essentially fully loaded. I am far less equipped to test its bandwidth than you are. The jitter analysis (part of WaveAnalyzer Pro) has been fun to play with so far. I have an HP 33120A signal gen, which is based on 40 MHz DDS. It is interesting, but not surprising, to see just how bad the jitter on the square waves is.
 
The following users thanked this post: tautech, Amazing

Online tautech

  • Super Contributor
  • ***
  • Posts: 16668
  • Country: nz
  • Taupaki Technologies Ltd. NZ Siglent Distributor
    • Taupaki Technologies Ltd.
Re: What is a good and cheap way to do frequency response testing on a scope?
« Reply #15 on: January 03, 2017, 04:21:48 am »
macboy nailed it.  :clap:

And this:
Quote
I think this is due to a reflection on the 2 meter BNC cable.
:scared:
Avid Rabid Hobbyist
 
The following users thanked this post: Amazing

Offline Amazing

  • Regular Contributor
  • *
  • Posts: 59
  • Country: us
Re: What is a good and cheap way to do frequency response testing on a scope?
« Reply #16 on: January 03, 2017, 04:34:15 am »
Don't forget that the calculated frequency response is of the system, not of the scope...

Quite correct.  When I changed to a shorter and better BNC cable, the step disappeared and the response started to look more like it should.

You can do quite well with an input 10x faster than the scope but you don't quite have that either.

It's nice to know that  the performance is probably better than I measured.

It can do equivalent time sampling at 50 GS/s.

Right again.  I was actually just playing with this to see if the FFT resolution would improve for short time captures.  Indeed it does.

One thing I learned to tonight is to keep the signal out of the top division of the screen.  It has noticable effects on the FFT if the signal protrudes into the top division.

Have you tried the scope explorer software?  It's a bit clunky but useful for doing screen grabs.  It connects over ethernet and works with Windows 7.  not sure about newer versions of windows.

http://teledynelecroy.com/support/softwaredownload/scopeexplorer.aspx

I just picked up a WavePro 960 about a week ago

Have fun with it!
 

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 10284
  • Country: us
  • DavidH
Re: What is a good and cheap way to do frequency response testing on a scope?
« Reply #17 on: January 03, 2017, 06:03:20 am »
David:  That's an interesting option.  I did some reading this morning on sampling scopes like you suggested.  Seems like a whole other rabbit hole to go down.   Thanks for the pointer to the tek scopes forum; I have a 321A that is awaiting repair...

It was mostly a problem of economics.  There seem to be more used but working high bandwidth sampling oscilloscopes available than precision high bandwidth power meters and many of the oscilloscopes are repairable while the power meters are not.  All of these considerations seem to make the oscilloscopes less expensive than the power meters.

Quote
What do you all think of this method of checking scope bandwidth:

...

Correction: I found that my tests were invalid in this post because I was using a sub-par BNC cable that was not quite 50 Ohms.   Also my scope seems to distort a tiny bit if the signal goes into the top division on the screen.  But I'm leaving this here anyway because I think the process was interesting.  Actual final test results are down below.

The other posts and your new results basically covered the problem.  Cables become a significant factor as the transition time decreases and the bandwidth increases.

Either the pulse generator has to be connected directly to the oscilloscope (1) or a short but good transmission line needs to be used.

(1) This becomes a problem with tunnel diode or avalanche pulse generators and sampling oscilloscopes.  If the tunnel diode or avalanche transistor is too close to the sampling gate, then the kick-out from the sampling gate strobe can trigger the pulse generator and the sampling oscilloscope ends up "seeing" its own sampling strobe.  The solution is to add a short section of transmission line between the pulse generator and oscilloscope.
 
The following users thanked this post: Amazing

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 10284
  • Country: us
  • DavidH
Re: What is a good and cheap way to do frequency response testing on a scope?
« Reply #18 on: January 03, 2017, 06:06:33 am »
Don't forget that the calculated frequency response is of the system, not of the scope. The system consists of your step/pulse generator (which has a finite, non-ideal rise time and other inherent irregularities), the cable, and the scope. You cable is certainly not ideal; it has a limited frequency response of its own, and may not be a perfect match to the source and/or scope input impedances. Speaking of which, the response shows an obvious mismatch evidenced by the second small step. It looks to me that the cable has a lower impedance that the source has, so it loads the output of the source a little too much, until the reflection of the pulse reaches the source and the system is basically at steady state (DC). To truly measure the response of only the scope this way, you need to provide an ideal step (zero rise time) to its input, which is impossible. You can do quite well with an input 10x faster than the scope but you don't quite have that either.

Tektronix had this problem when they designed the 1 GHz 7A29 vertical amplifier and discovered that 50 ohm BNC connectors were generally not 50 ohms even though they were marketed that way; they ended up making their own 50 ohm BNC connectors.
 
The following users thanked this post: Amazing


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf