Frequency Response Analyzer with Bode Plots

Post discussions on applications you are writing
Post Reply
hexamer
Advanced User
Advanced User
Posts: 5
Joined: Tue Aug 12, 2014 10:09 pm

Re: Frequency Response Analyzer with Bode Plots

Post by hexamer »

Sorry for taking so long to get back with you.

I don't know that my TPS5420 plots are necessarily any better than those from your Rohm buck converter. I think part of that may be perception due to my using less steps/decade and the plotting routines smoothing things out. But above about 200-300 kHz, it does appear pretty noisy. It looks like the upper end (500kHz) is still pretty far away from the switching frequency (1MHz min for that part). So I'm not sure the noise is due to the stimulation disturbing the converter, filter BW being too wide, or both. Both of the enhancements I mentioned earlier might help that.

Just curious, does the Rohm dev board have any published Bode plots to compare to?

On the PFC, I'm having trouble envisioning the way you have it setup. Is it possible to point to a figure showing the overall topology and measurement points.

Your work has been very helpful as I was able to find and fix bugs in the time domain plots. Thank you! I don't want to bother you to take more diagnostic plots of your PFC with the fixed version. So, it occurred to me that it might be helpful to look at time domain in the PicoScope application. Since the stimulus might be obscured there, I also wonder if it's possible to use maths channel filters in a way that helps see the stimulus (bandpass) or background noise (bandstop). Perhaps that could help determine if the stimulus is disturbing the converter.

While waiting on an adaptive amplitude implementation, an alternative would be to run multiple segments, each having a specific stimulus amplitude. I guess in many converters it's safe to use a higher stimulus at lower frequencies, which is convenient because that's where the feedback loop gain is strong, and thus the input signal goes very small.

ibdami21
Newbie
Posts: 0
Joined: Sat Mar 26, 2016 3:11 am

Re: Frequency Response Analyzer with Bode Plots

Post by ibdami21 »

Hi Aaron,

No worries.

I actually don't see any published data on the Rohm switcher for stability analysis, so nothing to really compare that to. It does seem to be measuring correctly with how the converter behaves though.

The active PFC converter is similar to what can be found in this ST document:
http://www.st.com/resource/zh/applicati ... 004041.pdf

Figure 8 shows a similar circuit as I've got. I'm using a cousin of this part with some enhancements, however similar in nature. This doc shows a lot more in depth information here vs the vendor I am using, so this would be better to reference for the general idea.

The nature of these converters makes them pretty noisy. Over the sine wave cycle of mains, the frequency of the converter moves around. This is great for minimising EMI/EMC, but terrible when we're trying to see the loop response since I have many frequencies now in the output spectrum. That along with the strong 120Hz (or 100Hz for 50Hz mains) signal make this a bit more tricky.

Measurement is made by breaking the loop on the secondary side as I have shown. I was able to use a 22 ohm resistor and drew out the points where I inject from the transformer and where the scope is measuring.

I should have captured some time domain plots using the picoscope, but a notch filter as I built plus putting a digital filter at ~10kHz on both channels made this very easy to measure. I'll have to repeat this at the correct signal amplitude since that was again at the high 1V drive. I have a hunch that would allow me to measure in the mV stimulus range and get correct crossover data.

It will be a little time until I can get back to measuring that, but I'll check hopefully in the next week.

Thanks,
Mitch
Attachments
PFC_setup.png

Hitesh

Re: Frequency Response Analyzer with Bode Plots

Post by Hitesh »

Hi Aaron,

I've just been testing a PicoScope 3203D with FRA 0.5.4b and it returns a status code of 286 (PICO_USB3_0_DEVICE_NON_USB3_0_PORT) when connected to a USB 2.0 port when you try to start the data collection.

When using the ps3000a driver, If the status codes of 282 (PICO_POWER_SUPPLY_NOT_CONNECTED) or 286 are returned by ps3000aOpenUnit(), please ensure that the ps3000aChangePowerSource() function is called passing in the actual status code that was returned by ps3000aOpenUnit().

The status code of 286 is also returned by the ps4000a driver when the PicoScope 4824 is connected to a USB 2.0 port.

Many thanks,

hexamer
Advanced User
Advanced User
Posts: 5
Joined: Tue Aug 12, 2014 10:09 pm

Re: Frequency Response Analyzer with Bode Plots

Post by hexamer »

Thank you for testing that Hitesh. When I coded flexible power for the PS3000A, I wondered what was the right value to pass to ChangePowerSource. If I recall correctly, the programmers guide only lists PICO_POWER_SUPPLY_NOT_CONNECTED and PICO_POWER_SUPPLY_CONNECTED as valid parameters for ps3000AChangePowerSource and that seemed suspicious to me. Could you please check that and update if necessary? I'll fix the 3000A code and also add the code for 4000A.

Hitesh

Re: Frequency Response Analyzer with Bode Plots

Post by Hitesh »

Hi Aaron,

The Programmer's Guide will require an update to the ps3000aOpenUnit() function (our Technical Publications team have already been notified about this).

Regards,

hexamer
Advanced User
Advanced User
Posts: 5
Joined: Tue Aug 12, 2014 10:09 pm

Re: Frequency Response Analyzer with Bode Plots

Post by hexamer »

Hitesh,

In reading the PS 4000A programmers guide it seems that rather than doing the intuitive/symmetric thing and being able to return PICO_USB3_0_DEVICE_NON_USB3_0_PORT, ps4000aCurrentPowerSource only ever returns PICO_OK. Is that right? If so, I'll have to code things a little differently there.

Edit: OK, I guess I should also ask whether ps3000aCurrentPowerSource can return PICO_USB3_0_DEVICE_NON_USB3_0_PORT. The existing code in the FRA application assumes it can.

Thanks,

Aaron.

Hitesh

Re: Frequency Response Analyzer with Bode Plots

Post by Hitesh »

Hi Aaron,

The psXXXXCurrentPowerSource functions are for use with 4-channel oscilloscopes to indicate whether a DC power supply is connected or not.

In the case of the ps4000aCurrentPowerSource() function, this has been reserved for future use, and in the case of the PicoScope 4824 returns PICO_OK if called.

For the ps3000aCurrentPowerSource() function, the function will only return PICO_POWER_SUPPLY_NOT_CONNECTED for 4-channel PicoScope 3000 Series models where the DC power supply is not connected. For 2-channel models, it will return PICO_OK, as the driver has already been informed that USB power will be used via the ps3000aChangePowerSource() function.

I hope this helps,

hexamer
Advanced User
Advanced User
Posts: 5
Joined: Tue Aug 12, 2014 10:09 pm

Re: Frequency Response Analyzer with Bode Plots

Post by hexamer »

Hitesh,

This was helpful, but as I began to implement, it generated even more questions.

First, could you please directly confirm that CurrentPowerSource cannot return PICO_USB3_0_DEVICE_NON_USB3_0_PORT.

Something I hope the documentation updates will clarify ... which condition takes precedence on opening a 4 channel 3000D device - PICO_USB3_0_DEVICE_NON_USB3_0_PORT or PICO_POWER_SUPPLY_NOT_CONNECTED? Also, I assume that when a 3000D device is on a USB3 port, but with no aux dc power connected it returns PICO_OK at Open. But what about on calling CurrentPowerSource?

These questions are relevant because of the way I'm coding PicoScope::GetNumChannels, and I hope I'm not over complicating things. But this asymmetry between Open and CurrentPowerSource in the API seems problematic. I had considered using GetUnitInfo to detect what kind of USB port the device is connected to, but it's not clear to me if that's referring to the scope's port or the PC's port. I'll publish my code shortly - perhaps you can tell me if it looks correct, or point out a simpler way.

Thanks,

Aaron

Hitesh

Re: Frequency Response Analyzer with Bode Plots

Post by Hitesh »

Hi Aaron,

I ran some tests to check the ps3000aCurrentPowerSource() function after your previous post and can confirm that the ps3000aCurrentPowerSource function will return either PICO_OK or PICO_POWER_SUPPLY_NOT_CONNECTED.

When opening a 4-channel PicoScope 3000 Series device with the DC power supply connected, ps3000aOpenUnit() will return PICO_OK. If the power supply is not connected to a 4-channel device, then PICO_POWER_SUPPLY_NOT_CONNECTED is returned.

The ps3000aCurrentPowerSource() is more geared towards determining whether the DC power supply is connected or not for 4-channel devices as opposed to determining the type of USB port that you are on.

The ps3000aUnitInfo() function will return 2.0 for the USB version if the device is connected on a USB 2.0 port or connected via USB 2.0 cable to a USB 3.0 port.

Regards,

Hitesh

hexamer
Advanced User
Advanced User
Posts: 5
Joined: Tue Aug 12, 2014 10:09 pm

Re: Frequency Response Analyzer with Bode Plots

Post by hexamer »

Hitesh,

I've made some changes based on the info you supplied. If you'd like to give it a try, I made a new binary: https://bitbucket.org/hexamer/fra4picos ... 0.5.5b.msi

Thanks,

Aaron.

dantooth
Newbie
Posts: 0
Joined: Wed Jun 29, 2016 4:56 pm

Re: Frequency Response Analyzer with Bode Plots

Post by dantooth »

Hi

I'm using a 3206B picoscope. I've downloaded the PicoSDK and FRA software and it starts-up ok. So far so good!
My goal is to measure dc/dc converter Bode plot responses.

I'm thinking I need an isolation transformer. Before I invest in one, can you say how effective the 3206B is for this kind of work?

One thing I note is that the output impedance of the signal generator on 3206B is 600-Ohms, so I'm thinking I need to buffer that before I try and drive the isolation transformer, like the VB-1BB.

Searching this forum I can find various tips on what else I need to do - use the high noise mode. I see you got a good result using 40mV signal, 10-steps/decade over 15Hz-200kHz range. Do you have a document that summarizes all this and shows the exact test set-up?

My background is h/w rather than s/w.

Regards Dan

hexamer
Advanced User
Advanced User
Posts: 5
Joined: Tue Aug 12, 2014 10:09 pm

Re: Frequency Response Analyzer with Bode Plots

Post by hexamer »

Hi Dan,

Thank you for your interest. I can't make any firm assessment of the 3206B since I don't have one to test with. It does have some specs that would seem less favorable (vs PS5000A). But in general it does not seem to have any specs that immediately disqualify it.

A real challenge with SMPS is the need for good dynamic range, especially on the low end. I think 3206B is 8 bit +/- 50 mV, whereas the 5000A is 15 bit +/- 10 mV (not sure how the ENOB compares to PS5000A). To predict how well the 3206B would do, you may be able to construct some sort of analogous circuit which does not require an isolation transformer and use it to assess performance. To do this you probably will want to estimate the low frequency gain of the SMPS. Then make a filter with similar attenuation and measure it with low input Vpp. To make it complete, you may need to also inject some noise simulating ripple. If the output is really noisy in the high attenuation region, then the SMPS may be similarly difficult.

I'm not sure whether the high output impedance would be a significant issue because I don't remember the transformer impedance. Without a buffer, it could limit the upper voltage output. This would typically be more important at lower frequencies where the SMPS gain is the greatest.

Thanks,

Aaron.

Hitesh

Re: Frequency Response Analyzer with Bode Plots

Post by Hitesh »

Hi Aaron,

I have tested FRA 0.5.5b with the following USB 3.0 devices on USB 2.0 ports:
  • PicoScope 3203D
  • PicoScope 3206D MSO
  • PicoScope 4824
These work fine :D

There are a couple of other things that I noticed but I will send those off-line from here.

Thanks,

dantooth
Newbie
Posts: 0
Joined: Wed Jun 29, 2016 4:56 pm

Re: Frequency Response Analyzer with Bode Plots

Post by dantooth »

Hi hexamer

I did look at the 5000A series and indeed it has variable resolution, up to 16 bits. Very good Picoscope.
I did note this comment at the bottom of its spec table, seen in the link below:-

"** Maximum effective resolution is limited on the lowest voltage ranges: ±10 mV = 8 bits • ±20 mV = 12 bits. All other ranges can use full resolution."
https://www.picotech.com/oscilloscope/5 ... ifications

My question is are you using the +/-10mV range? If so then I understand you are limited to 8 bits.

The 3206B can go down to 10mV/div and it is 8-bit device.

Regards Dan

hexamer
Advanced User
Advanced User
Posts: 5
Joined: Tue Aug 12, 2014 10:09 pm

Re: Frequency Response Analyzer with Bode Plots

Post by hexamer »

dantooth wrote:Hi hexamer

I did look at the 5000A series and indeed it has variable resolution, up to 16 bits. Very good Picoscope.
I did note this comment at the bottom of its spec table, seen in the link below:-

"** Maximum effective resolution is limited on the lowest voltage ranges: ±10 mV = 8 bits • ±20 mV = 12 bits. All other ranges can use full resolution."
https://www.picotech.com/oscilloscope/5 ... ifications

My question is are you using the +/-10mV range? If so then I understand you are limited to 8 bits.

The 3206B can go down to 10mV/div and it is 8-bit device.

Regards Dan
Hi Dan,

Thanks for pointing that out. I was aware that the ENOB dropped off for these lower ranges. For some reason I was thinking it was 10 bits at +/- 10 mV.

If I'm interpreting it correctly, the 10 mV per division on the 3206B means +/- 50 mV (10 vertical divisions).

We can do some rough calculations to estimate performance. If your ripple and input happened to be small enough to allow use of +/- 50 mV (best case), then the resolution would be about 391 uV. If you had a fixed input of 10 mV and it was attenuated by 40 dB (typical at low frequency), the it would be about 100 uV. So with this example, it seems that it would not work so well.

However I don't think this rules it out. It may just mean that you'll need to limit the lower frequency. After all, for the phase margin, you're looking for the 0 dB point.

Thanks,

Aaron.

Post Reply