Control to Output gain-phase measurement

[Magnethicc]

Hi everyone.
I want to know if it is possible to measure the control to output transfer function in real life and if yes then how?
In LTSpice I use AC analysis of the small signal model of the DC-DC and then take the ratio of output voltage to error amplifier output.
Lets assume I have a this buck converter (in reality I will have a PSFB converter):

1) In real life I think to inject a signal at point 4 (I have access to this point) and hook the Frequency Response Analyzer (FRA) at Vb and Vout. Is this a good way to measure the control to output transfer function?

2) I would also like to measure the compensator transfer function so I though this time to hook the FRA at FB and Va. again is this a good way to measure the compensator gain?

3) a question about the frequency of the injected signal - can I inject a signal with a frequency of 1/2switching frequency and above and get results of meaning? or after 1/2 switching frequency I will get garbage information?

Any responses and further information is highly appreciated!

 

[cupoftea]

In this course
SMPS course_little folders

https://drive.google.com/file/d/1WvElhDNfHI0EBei4rouSNEdWWTO1oCtJ/view?usp=sharing

...there is a folder called switchingpowermagazine.com.....in the top three PDFs it tells how to do it.

--- Updated May 9, 2022 ---

SMPS Feedback loop measurements​

The Ridley engineering group make a gain phase analyser ( AP300) which is good for measuring the gain and phase margin of an SMPS…
http://www.ridleyengineering.com/analyzer.html

The following 4 videos go through use of the AP300 frequency analyser, which you can use to measure gain and phase margin of the SMPS feedback loop.

--- Updated May 9, 2022 ---

Attached also is a Buck in LTspice with the injection source going in......you can check the phase of the signals either side of the 50R injection resistor, with respect to gnd, to get the phase margin.....the amplitude difference tells the gain margin....sorry its a bit messy....you dont see it until after 20ms.

 

[Magnethicc]

In this course
SMPS course_little folders

https://drive.google.com/file/d/1WvElhDNfHI0EBei4rouSNEdWWTO1oCtJ/view?usp=sharing

...there is a folder called switchingpowermagazine.com.....in the top three PDFs it tells how to do it.

--- Updated May 9, 2022 ---

SMPS Feedback loop measurements​

The Ridley engineering group make a gain phase analyser ( AP300) which is good for measuring the gain and phase margin of an SMPS…
http://www.ridleyengineering.com/analyzer.html

The following 4 videos go through use of the AP300 frequency analyser, which you can use to measure gain and phase margin of the SMPS feedback loop.

--- Updated May 9, 2022 ---

Attached also is a Buck in LTspice with the injection source going in......you can check the phase of the signals either side of the 50R injection resistor, with respect to gnd, to get the phase margin.....the amplitude difference tells the gain margin....sorry its a bit messy....you dont see it until after 20ms.

Thank you for your response.
The PDFs and videos does not say really answer the questions though they do provide good information

 

[cupoftea]

2) I would also like to measure the compensator transfer function so I though this time to hook the FRA at FB and Va. again is this a good way to measure the compensator gain?

...You usually calc this with complex numbers..only if your loop bandwidth is really high do you need to measure compensator......and if you did, you do it in isolation..no need to do it in PSU.

Opamp speed is not usually relevant to error amplifier compensator....its not usually set that high in bandwidth...so the opamp pole isnt usually relevant.

I think all your qus are answered in the videos i gave you....including the one about 1/2 fsw

1) In real life I think to inject a signal at point 4 (I have access to this point) and hook the Frequency Response Analyzer (FRA) at Vb and Vout. Is this a good way to measure the control to output transfer function?

No, you measure the open loop gain & phase margins by the injection technique....to work out Con to output..you would subtract out the error amplifier from that . The vids deffo tell this.

 

[KlausST]

Hi,

I guess this is configured as push-pull driver.

Then the output voltage is: V_in x duty_cycle
And duty_cycle depends on sawtooth_min and sawtooth_max and Vb.

Duty_cycle is:
* 0 when Vb < sawtooth_min
* 100% when Vb > sawtooth_max
* linear from 0 to 100% when Vb moves from sawtooth_min to sawtooth_max.
--> dc = 100% * (Vb - sawtooth_min) / (sawtooth_max - sawtooth_min).

Klaus

 

[Magnethicc]

Hi cupoftea, I must have missed it when he talked about it.. I will watch the videos again.
I agree with you, it does makes sense to subtract the bode of the compensator.

the question still remains - is it possible to measure the control to output by just changing the position of the probes of the FRA.
from this Texas Instruments Application Notes it is possible (page 7, sub-title 9).
When I will receive the PSFB module I will test both your way and Texas Instrument way and will report back on if they produce the same results or not.


KlausST - I'm sorry but not really sure what you mean

 

[cupoftea]

the question still remains - is it possible to measure the control to output by just changing the position of the probes of the FRA.

...that pg 7 is basically saying that the error amp isnt affecting that measurement, but it is...so i dont agree with that.....you cant JUST measure power_stage_and_modulator Transfer function if the error amp is still there.

 

[Magnethicc]

...that pg 7 is basically saying that the error amp isnt affecting that measurement, but it is...so i dont agree with that.....you cant JUST measure power_stage_and_modulator Transfer function if the error amp is still there.

How come? correct me if I'm wrong but you measure a sine wave the at the error amplifier output and measure also the sine wave after the modulator + plant then divide the amplitudes and measure the phase difference - you have the control-output bode.

For example:
You inject the signal right above Rtop. The sine wave propagates through the compensator, the compensator changes the sine wave gain and phase with regards to the injected signal.
At the point after the compensator you measure the sine wave (input 1 of FRA)
Sine wave continue to propagate through the modulator and plant, changing it's gain and phase again.
You measure again the sine wave this time at the output of the plant (input 2 of FRA).
You now have 2 sine waves with different amplitudes and phases which you can construct a bode from.
This bode will tell you how a sine from the error amplifier output is affecting the output of the plant which is exactly the control to output bode.

The FRA does not know where do you inject the signal. It will only measure sine waves amplitudes and phase difference.

Please, correct me if I am wrong - I would gladly learn if and how my way of thinking is wrong.

 

[KlausST]

Hi,

Give example values.

This is just a practical approach of
Vb to Vout
(via comparator and half bridge)

Klaus

 

[FvM]

the question still remains - is it possible to measure the control to output by just changing the position of the probes of the FRA.

It's possible. A full featured FRA like AP300 has a source output and two probe inputs, so it's just the question how you connect it.

 

[Warpspeed]

Yes its possible.
Can it be done, probably not, unless you have some very expensive and exotic high end test equipment.

The idea that you can just feed a function generator into a signal insertion transformer and view two sine waves on an oscilloscope, and manually plot a Bode curve sounds pretty straightforward, until you actually try to do it.

The problems of noise and dynamic range will quickly defeat all your efforts.

It sounds really nice +80db gain here, -40db loss there to plot your gain/phase curves, but how can you resolve a signal of a few microvolts in the presence of perhaps hundreds of millivolts of switching supply noise spikes and broad band harmonic crap ?

Only possible way is with an extremely narrow band tracking filter with a bandwidth of perhaps 1Hz.
If you are injecting 2.403 Khz, you need to reject everything else down below the noise floor, and have some kind of auto ranging for gain as well. Its not really possible to make an accurate phase measurement of a noisy jittery signal you can hardly see.

So the old analog method typically uses hetrodyning up in frequency into a very narrow band 100 kHz quartz crystal filter to achieve the very narrow bandwidths required. That is all now obsolete technology, its all done digitally with high speed data processing, and out comes a Bode plot with at least 100db of dynamic range, and nice smooth phase curve.

A wide band network analyser does not come cheap though, its not the sort of thing hobbyists or even small businesses can afford. If you have access to some top end gear, its all pretty straightforward.
If not, its going to be a real challenge to make meaningful measurements.

 

[Akanimo]

Hi,
See this.

 

[KlausST]

Hi,

Only possible way is with an extremely narrow band tracking filter with a bandwidth of perhaps 1Hz.

The problem with a filter is that it introduces new ... and maybe unknown ... phaseshift.

A lock in amplifier could be another option to improve signal to noise while getting amplitude and phase.

************
Usually Opamp circuits have high PSRR, thus one may ignore the power supply. But here the power supply Vin has a 1:1 (sytematically) influence on the output voltage.
The same is with the sawtooth. This is no problem as long as you know this.
So your simulation may ignore this and keep Vin and the sawtooth "perfect" or
You do realstic simulation including Vin and sawtooth (and comparator timing)

Indeed the most "phase shifting part" in a real circuit should be the output LC filter.

Klaus

 

[Magnethicc]

Well I do have a Newtons4th PSM1700 Frequency Response Analyzer so I guess this is pretty good start.
Akanimo - That settles it. As he shows it is possible I feel comfortable doing it myself as well. Thanks!

 

[Warpspeed]

The problem with a filter is that it introduces new ... and maybe unknown ... phaseshift.

Its very easy to check, just connect both channels to the same source and do a sweep. Phase and gain should sit at 0 degrees and 0 db. I am using 100Khz quartz crystals in my two filters, and its not difficult to tweak a matched pair for both frequency and Q to close enough limits to be still be very useful.

That is what the commercial gear does anyway. Sweep both channels together, store the result in memory. Then sweep the circuit under test, the difference between the stored and measured becomes the displayed result.
A well screened analog filter is the only way you are going to get sufficient dynamic range out of something like this.

Once all the crap has been filtered out, the rest is straightforward.