Admittance Measurements with DSO & AWG with Bode Function

[mawyatt]

This is an extension to the Impedance measurements discussed earlier, see:

https://www.eevblog.com/forum/testgear/capacitive-impedance-plots-with-sds2104x-plus-bode-function/msg4342009/#msg4342009

However decided to start another thread since some very interesting and unexpected results have shown up, actually started to add to the above until we made some measurements, you'll understand why in a few movements.

Here's an interesting configuration that many folks may have at hand if they have an LCR Bench meter.

Using a popular fixture for SMD that often are used with Bench LCR meters, add a AWG and DSO and you have an Impedance/Admittance Measuring device for SMD capable of generating plots!!

In the simplest form you just need a couple BNC adapters and cables. Connect the AWG output to the Fixture Force (H or L cur) and with the DSO CH1 to Fixture Sense (H or L pot), (use a BNC cable or Scope probe). Now set DSO Ch1 & Ch2 to 50 ohms and Ch2 to the other side of the SMD fixture (H or L) as shown. Set the Bode Plot parameters as required and the plot will show the voltage across the 50 Ch2 resistor divided by the voltage sensed on Ch1.

A little analysis that assumes the voltage on Ch2 is much smaller than Ch1 at the measurement of interest, Ch1 is the AWG output in this case.

I of DUT = VCh2/R, where R is Ch2 termination 50 ohms

DUT Z is ~ VCh1/I or (VCh1/VCh2)/R

VCh1/VCh2 which is inverse of dB version of screen Bode Plot Result (VCh2/VCh1) in dB, so

VCh1/VCh2 = 1/[10^(dB/20)], where dB is Bode Reading in dB

Then Z ~ R/[10^(dB/20) ]

and  C ~ 1/(|Z|*2pi*F) where R is 50 ohms in this case (Ch2 termination)

C ~ [10^(dB/20)]/(R*2pi*F), where F is obviously the frequency of the measurement.

Note: We could have swapped Ch1 and Ch2 so that the Bode Plot was inverted but wanted to keep the display relative to Admittance.

Also note when making these measurement the "Quality" of the measurement for Hi Q capacitors is when the phase isn't too far from +90 degrees (remember this is admittance)!!

We ran these for starters:

1nF COG cap, which measured 992.8pF @ 10KHz (plot 115) and 993.2pF on Hioki IM3536 Lab LCR Meter
100pF COG measured 102.1pF @ 1MHz (plot 116) and 101.0 on IM3536
10pF COG measured 10.12pF @ 1MHz (117) and 9.943pF on IM3536
1pF COG measured 1.07pF @ 1MHz (118) and 1.015pF on IM3536
0.3pF COG measured 0.345pF @ 1MHz (120) and 0.3134pF on IM3536

I know this seems crazy to be able to measure these smaller caps with a DSO, but to actually get this performance we didn't expect and this was with no calibration or measurement correction. We did measure the short thur and got less than 0.1dB loss and for an open we got less than 0.05pF equivalent parasitic capacitance

Will admit that some time ago we did modify this fixture to improve the isolation and internal grounding!

And who said that accurate DSO Channel 50 ohm termination isn't useful, and low noise DSO front ends aren't necessary and DSO dynamic range doesn't matter 

Anyway more to come later, and here's the really simply setup with only DSO, AWG, SMD Fixture and 3 BNC cables and thru adapters.

Best,

[mawyatt]

Also works well with the lead component LCR Meter test fixture.

Best,

[nctnico]

I did a similar experiment several years ago with my GW Instek DSO using free-form math. With amplitude and phase measurements (both available from a math function in that DSO), you can obtain the values from a math trace directly.

[mawyatt]

You can add a lower value resistor to the Force terminal of the fixture as shown for larger value capacitors. Here a 1 ohm shunt to ground is added, and Ch2 is set to 1Meg Input impedance. Just requires another BNC thru and BNC Terminal adapter where the shunt resistor is installed.

Best,

[TopQuark]

Very interesting as usual!   

I can only dream for the day when bode plots can be launched, controlled and retrieved through SCPI. With the right PC software, we can turn the scope bode plot function into a kickass LCR machine and much more!

We can beat the crap out of the Zurich Instrument Impedance Analyser that's always being advertised on top of the forum page. 

[Mechatrommer]

Very interesting as usual!   

I can only dream for the day when bode plots can be launched, controlled and retrieved through SCPI. With the right PC software, we can turn the scope bode plot function into a kickass LCR machine and much more!

We can beat the crap out of the Zurich Instrument Impedance Analyser that's always being advertised on top of the forum page. 

you can bode plot with just rigol ds1054z + uni-t utg962 (or any other brand that supports) sweep function. download data to PC and use PC software to plot mag and phase... but the key is... "With the right PC software" ... btw those bk precision's test fixtures posted by mawyatt are eye watering, now i got idea for 3d printed project...

[TopQuark]

Very interesting as usual!   

I can only dream for the day when bode plots can be launched, controlled and retrieved through SCPI. With the right PC software, we can turn the scope bode plot function into a kickass LCR machine and much more!

We can beat the crap out of the Zurich Instrument Impedance Analyser that's always being advertised on top of the forum page. 

you can bode plot with just rigol ds1054z + uni-t utg962 (or any other brand that supports) sweep function. download data to PC and use PC software to plot mag and phase... but the key is... "With the right PC software" ... btw those bk precision's test fixtures posted by mawyatt are eye watering, now i got idea for 3d printed project...

I mean currently we can save the bode plot raw gain/phase data to a usb stick and transfer the data to a PC for further processing, just wishing for more SCPI automation to be possible.

[Mechatrommer]

i only know programmed rigol through visa and usb port, automation is possible. because thats only what i have. other brand you need to make yourself. i recently purchased owon scope meter, i saw some api doc but i dont have time for it.

[2N3055]

i only know programmed rigol through visa and usb port, automation is possible. because thats only what i have. other brand you need to make yourself. i recently purchased owon scope meter, i saw some api doc but i dont have time for it.

Difference is that you pulled raw data and did all math yourself. TopQuark is talking about fact that those Siglent scopes have magnitude/phase measurement (Bode plot) built in, and he would like to automate that and get that data..

[mawyatt]

While on this subject, does anyone know/measured the actual time to complete the Bode Plot for the SDS2000X+, vs HD, SDS6000? Obviously need the same Bode parameters for each to compare times.

One would "think" the SDS2000HD and SDS6000 should be faster with the higher performance FPGA and such.

Best,

[nctnico]

i only know programmed rigol through visa and usb port, automation is possible. because thats only what i have. other brand you need to make yourself. i recently purchased owon scope meter, i saw some api doc but i dont have time for it.

Difference is that you pulled raw data and did all math yourself. TopQuark is talking about fact that those Siglent scopes have magnitude/phase measurement (Bode plot) built in, and he would like to automate that and get that data..

How is that any different than using magnitude and phase measurement functions? Since you'll only need these measurements for a single or limited number of frequencies, that route will be much quicker. I don't see the need for using bode plot at all.

[TopQuark]

While on this subject, does anyone know/measured the actual time to complete the Bode Plot for the SDS2000X+, vs HD, SDS6000? Obviously need the same Bode parameters for each to compare times.

One would "think" the SDS2000HD and SDS6000 should be faster with the higher performance FPGA and such.

Best,

Haven't timed it myself, but I am pretty sure the HD is just as slow as the non HD. I think the time required is dominated by the sampling time at lower frequencies, and the multiple samples per frequency points that the scope takes when sweeping the frequency range.

i only know programmed rigol through visa and usb port, automation is possible. because thats only what i have. other brand you need to make yourself. i recently purchased owon scope meter, i saw some api doc but i dont have time for it.

Difference is that you pulled raw data and did all math yourself. TopQuark is talking about fact that those Siglent scopes have magnitude/phase measurement (Bode plot) built in, and he would like to automate that and get that data..

How is that any different than using magnitude and phase measurement functions? Since you'll only need these measurements for a single or limited number of frequencies, that route will be much quicker. I don't see the need for using bode plot at all.

I have thought of rolling my own bode plot program that controls the instruments through SCPI, and reading the amplitude and phase measurements with the scope built-in math functions. Main problem with using the scope's amplitude and phase readout is, the math functions are not frequency selective, so in noisy measurements (e.g. measure loop response of a fast switching smps), the measurement might be inaccurate.

[mawyatt]

... btw those bk precision's test fixtures posted by mawyatt are eye watering, now i got idea for 3d printed project...

The B&K is a nice fixture and comes with various fixtures for Axial and Radial leaded components. We made some BNC to Bananna cables to use with Bode Capability or with the Tek 577 Curve Tracer as shown, works very well. These cables also allow the SMD Fixture to be used with both measurement methods.

Inexpensive and very functional fixture the B&K TL89F1

Anyway, hope this helps!!

Best,

Edit: Lined thur incorrect information above, Bode function just uses standard BNC cables, custom cables not required.

[mawyatt]

While on this subject, does anyone know/measured the actual time to complete the Bode Plot for the SDS2000X+, vs HD, SDS6000? Obviously need the same Bode parameters for each to compare times.

One would "think" the SDS2000HD and SDS6000 should be faster with the higher performance FPGA and such.

Best,

Haven't timed it myself, but I am pretty sure the HD is just as slow as the non HD. I think the time required is dominated by the sampling time at lower frequencies, and the multiple samples per frequency points that the scope takes when sweeping the frequency range.

Was thinking more in the range of say 1KHz to 1MHz where acquiring the samples shouldn't take too long.

I have thought of rolling my own bode plot program that controls the instruments through SCPI, and reading the amplitude and phase measurements with the scope built-in math functions. Main problem with using the scope's amplitude and phase readout is, the math functions are not frequency selective, so in noisy measurements (e.g. measure loop response of a fast switching smps), the measurement might be inaccurate.

The Siglent Bode Function appears to utilize some form of frequency selective sampling, maybe a derivative of Synchronous Sampling. Say this because of the ability to effectively utilize such high DR in the presence of noise and interference, where conventional sampling just wouldn't cut it!! This observation is with actual "use" rather than the "speculation" as some other anti-Siglent types often respond. Recall rf-loop having shown some remarkable plots utilizing a high DR step attenuator. Duplicating this kind of performance may be much more difficult than just using the simple amplitude and phase captured and read by SCPI to remotely display a plot, as some serious DSP is apparently going on behind the scene!!

Edit: Now if they allowed the DSP post processed data accessible by remote means, then one might have "the cake and eat it too"

Best,

[nctnico]

i only know programmed rigol through visa and usb port, automation is possible. because thats only what i have. other brand you need to make yourself. i recently purchased owon scope meter, i saw some api doc but i dont have time for it.

Difference is that you pulled raw data and did all math yourself. TopQuark is talking about fact that those Siglent scopes have magnitude/phase measurement (Bode plot) built in, and he would like to automate that and get that data..

How is that any different than using magnitude and phase measurement functions? Since you'll only need these measurements for a single or limited number of frequencies, that route will be much quicker. I don't see the need for using bode plot at all.

I have thought of rolling my own bode plot program that controls the instruments through SCPI, and reading the amplitude and phase measurements with the scope built-in math functions. Main problem with using the scope's amplitude and phase readout is, the math functions are not frequency selective, so in noisy measurements (e.g. measure loop response of a fast switching smps), the measurement might be inaccurate.

For bode plotting yes (even though waveform averaging could help improve things). For measuring components with a known (and controlled) sine wave source you should be able to get rid of noise using high-res and/or averaging modes. With software at the PC side, you could even average several readings and still have a reasonably quick update rate. The higher end LCR meters typically have a selectable update speed and also allow extra averaging for further noise reduction.

[TopQuark]

i only know programmed rigol through visa and usb port, automation is possible. because thats only what i have. other brand you need to make yourself. i recently purchased owon scope meter, i saw some api doc but i dont have time for it.

Difference is that you pulled raw data and did all math yourself. TopQuark is talking about fact that those Siglent scopes have magnitude/phase measurement (Bode plot) built in, and he would like to automate that and get that data..

How is that any different than using magnitude and phase measurement functions? Since you'll only need these measurements for a single or limited number of frequencies, that route will be much quicker. I don't see the need for using bode plot at all.

I have thought of rolling my own bode plot program that controls the instruments through SCPI, and reading the amplitude and phase measurements with the scope built-in math functions. Main problem with using the scope's amplitude and phase readout is, the math functions are not frequency selective, so in noisy measurements (e.g. measure loop response of a fast switching smps), the measurement might be inaccurate.

For bode plotting yes (even though waveform averaging could help improve things). For measuring components with a known (and controlled) sine wave source you should be able to get rid of noise using high-res and/or averaging modes. With software at the PC side, you could even average several readings and still have a reasonably quick update rate. The higher end LCR meters typically have a selectable update speed and also allow extra averaging for further noise reduction.

Don't think averaging will do much good for isolating signals from a specific frequency. Think a better way would be to measure the gain and phase shift in the frequency domain with FFT, so that you can compare in and out at a specific frequency. Sounds like a lot of work though.

[nctnico]

For using a known source, I assume there is a good signal to noise ratio without too much external interference. Averaging should make things better though because interference is likely to be random (noise) compared to the source's frequency. If your oscilloscope has the ability to filter, this would also help.

Not sure whether FFT is the perfect approach; I'd probably go for an algorithm that filters a single frequency to reduce the amount of processing needed.

Synchronous sampling -as mentioned by mawyatt already- would be another option but for most accurate results you'd need to have the generator (DAC) clock locked to the sampler (ADC) clock. Still, it takes processing quite a few cycles to get to accurate results. In the past I have made an LCR meter-ish device to measure micro-Ohm level AC impedances on battery cells. The circuit itself is pretty crude using a relatively crappy DAC and a 10 or 12 bit ADC in a microcontroller.

[TopQuark]

For using a known source, I assume there is a good signal to noise ratio without too much external interference. Averaging should make things better though because interference is likely to be random (noise) compared to the source's frequency. If your oscilloscope has the ability to filter, this would also help.

Not sure whether FFT is the perfect approach; I'd probably go for an algorithm that filters a single frequency to reduce the amount of processing needed.

Synchronous sampling -as mentioned by mawyatt already- would be another option but for most accurate results you'd need to have the generator (DAC) clock locked to the sampler (ADC) clock. Still, it takes processing quite a few cycles to get to accurate results. In the past I have made an LCR meter-ish device to measure micro-Ohm level AC impedances on battery cells. The circuit itself is pretty crude using a relatively crappy DAC and a 10 or 12 bit ADC in a microcontroller.

In an ideal world I'd give all the options above a try and see what works best. But given my limited time, I think I'll put my time developing my DSA project, which in theory could give a bode plot every few milliseconds (if it works) 

[mawyatt]

Don't think averaging will do much good for isolating signals from a specific frequency. Think a better way would be to measure the gain and phase shift in the frequency domain with FFT, so that you can compare in and out at a specific frequency. Sounds like a lot of work though.

Agree, simple averaging won't help much with strong interferers, especially ones that are not totally random during the sampling intervals. We ran into this problem long ago with a special test capability, no reasonable amount of signal averaging helped much and we decided to utilize synchronous sampling and place a low frequency integrator right on top of the sampling effect which translates into a narrow bandpass filter centered at the sampling frequency. This worked very well indeed! This is also the similar technique to what quality LCR meters utilize, Synchronous Sampling shines in this type interference environment.

In a way the FFT is doing something similar, placing narrow Frequency "bins" to isolate energy within those frequency bins span and reject outside signals.

A really good demonstration of just how good the Siglent implementation of the Bode Function is in this threads about the Peltz Oscillator Close Loop Response and Injection Locking of such. Still amazed that these kind of plots could be pulled off with a simple DSO & AWG

https://www.eevblog.com/forum/projects/things-coming-together-bode-plot-diy-isolation-transformer-peltz-oscillator/msg4288363/#msg4288363

https://www.eevblog.com/forum/projects/injection-locked-peltz-oscillator-with-bode-analysis/msg4424434/#msg4424434

Best,

[gf]

Not sure whether FFT is the perfect approach; I'd probably go for an algorithm that filters a single frequency to reduce the amount of processing needed.

Then just calculate the DFT for a single frequency only. This will do both, bandpass filtering, and act as vector detector. The shape of the filter is determined by the chosen window function. A full FFT does the same for N frequencies simultaneously with O(N*log(N)) complexity, but there are constraints on the frequencies.

[2N3055]

Not sure whether FFT is the perfect approach; I'd probably go for an algorithm that filters a single frequency to reduce the amount of processing needed.

Then just calculate the DFT for a single frequency only. This will do both, bandpass filtering, and act as vector detector. The shape of the filter is determined by the chosen window function. A full FFT does the same for N frequencies simultaneously with O(N*log(N)) complexity, but there are constraints on the frequencies.

Goertzel could be used for this, for instance..

[Mechatrommer]

Don't think averaging will do much good for isolating signals from a specific frequency. Think a better way would be to measure the gain and phase shift in the frequency domain with FFT, so that you can compare in and out at a specific frequency. Sounds like a lot of work though.

Agree, simple averaging won't help much with strong interferers, especially ones that are not totally random during the sampling intervals. We ran into this problem long ago with a special test capability, no reasonable amount of signal averaging helped much and we decided to utilize synchronous sampling and place a low frequency integrator right on top of the sampling effect which translates into a narrow bandpass filter centered at the sampling frequency. This worked very well indeed! This is also the similar technique to what quality LCR meters utilize, Synchronous Sampling shines in this type interference environment.

i'm not sure whats a synchronous sampling is but i did use FFT to get bode plot out of DS1052E and Hantek DDS3x25 automated for each sine sweep frequency back then.. but i never tested more complicated stuffs other than a simple capacitors and inductors. iirc do the full FFT, take the largest component's magnitude (assume its the fundamental, ignore the rest of bins) and compare the phase, take another sweep frequency and redo and plot, its been a while so i could be mistaken in the process, but attached was my plot, iirc no averaging needed.

but both my DS1052E and Hantek are now inoperational, and i dont have an urgent need anymore. i even experimented with square wave, its like feed one single square wave frequency say 1 or 10MHz, do full FFT, compare magnitudes and phases for all FFT components/bins and plot, very quick! but the result is noisy and less trustworthy (2nd attachment), i guess averaging or very good DSO/AWG front end to get perfect and stable square wave is needed to improve result. fwiw.

[nctnico]

@Mechatrommer: AFAIK synchronous sampling is adding all samples from the incoming signal but flipping the sign at the place where you expect a zero crossing for the real part and flipping the sign at the peak for the imaginary part of the signal (if you are interested in that). At least, that is home I implemented it.

Brain fart: An alternative for implementing an LCR meter / bode plot function could be to use 2 IQ demodulators. One for the stimulus and one for the measured signal across the DUT. Adjust the modulation frequency & phase to have the imaginary part of the stimulus at 0, Low-pass filter the IQ signals from the DUT and you should have the real and imaginary parts. This should eliminate errors due to differences in frequency between the generator's clock and the sampler. From my experience with digital function generators I've seen that there are many that have a frequency offset due to limited DDS resolution. For example: 1Hz isn't 1Hz but 1.005Hz. Since the function generator inside a DSO is relatively simple, it would not surprise me if many suffer the same issue. If you measure over a lot of cycles, then this frequency offset could accumulate to a significant error when doing synchronous sampling.

[gf]

AFAIK synchronous sampling is adding all samples from the incoming signal but flipping the sign at the place where you expect a zero crossing for the real part and flipping the sign at the peak for the imaginary part of the signal (if you are interested in that). At least, that is home I implemented it.

What you describe, I would rather call a synchronous detector.

Synchronous sampling IMO just means that the sample rate is an exact integral multiple (>2x) of the signal frequency, and that the measurement interval is an exact integral multiple of the signal period. This can be helpful for building a synchronous detector in the digital domain. DFT is eventually yet another synchronous detector too (for all bin frequencies), and so is Goertzel.

[mawyatt]

AFAIK synchronous sampling is adding all samples from the incoming signal but flipping the sign at the place where you expect a zero crossing for the real part and flipping the sign at the peak for the imaginary part of the signal (if you are interested in that). At least, that is home I implemented it.

What you describe, I would rather call a synchronous detector.

Synchronous sampling IMO just means that the sample rate is an exact integral multiple (>2x) of the signal frequency, and that the measurement interval is an exact integral multiple of the signal period. This can be helpful for building a synchronous detector in the digital domain. DFT is eventually yet another synchronous detector too (for all bin frequencies), and so is Goertzel.

Agree, that "Synchronous Detection" is the technically proper term, however most of the folks we've known use Synchronous Sampling in the same context.

Another way of thinking about this technique is that the input signal is BiPhase Modulated with a squarewave of +-1 amplitude at the sampling frequency.

A recent discovery (~2008) we called the Polyphase Mixer, others prefer N-Path Mixer, is a multi-phase version of Synchronous Detection (think of this as Synchronous Detection with multiple N samplers equally distributed across the LO phase of +-pi). This technique has proved very useful in RF/MW/MMW use as it provided direct downconversion with I and Q baseband output and because it's a bilateral function allowed creating RF/MW/MMW complex impedance matching at the input without inductors or varactors, that tracked the LO!! Also it violates (measured below 2 dB) convention Biphase mixer minimum Noise Figure theory!! If interested, be sure to thoroughly read the IEEE articles mentioned, and follow all the related threads and such.

Absolutely fascinating circuit discovery, that has proven extremely useful. Believe Apple has employed such for some time now, and maybe this new technique will show up in some new LCR meters, or FRA, understand it's in a few high performance SDRs too 

https://www.eevblog.com/forum/rf-microwave/polyphase-or-n-path-mixer/msg3381802/#msg3381802

Best,