DIY Injection Transformer for Power Supply Control Loop Response Measurements

[diyaudio]

I have been using a Jensen VB-1BB isolation transformer used for analog video. Not perfect, but not bad assuming that 10hz is not a necessity. It does ok down to rather low frequencies though. I paid $20 on eBay. BNC 75 Ohms in/out

http://www.jensen-transformers.com/product/vb-1bb/

I tried to use it with the Keysight power application on my MSOX6000 but was quickly stopped by it's inability to deal with the noise in SMPS. The newer software on the 3k's and 4k's seems to be a lot better at filtering and allowing a LUT for the injection voltage to get a better SNR on a range of frequencies. I was able to get measurements, but manually stepping through the frequencies, injection levels, averaging, and math filtering. A bit slow, but I think I was able to get believable measurements. After getting some pricing on dedicated FRA's, I am determined to figure out how to get what I need on the scope. I will design about 4-5 SMPS's per year so I need this capability for a grand total of maybe a week per year as I work out the supply design.



Having a real VNA would be nice, but certainly not a must.

-- Johnnie Hancock
Yes, we recently updated the firmware in the 3000T and 4000A X-Series. We implemented a completely different measurement algorithm that improves the dynamic range of frequency response measurements significantly. We also implemented a much better user-interface, as well as overlaying the gain and phase traces on the same plot. We plan to update the firmware in the 6000 X- Series next year. I don't know exactly when, but when it comes out, it will be a free upgrade.?

[rbola35618]

Hi EEVBLOGGER,

I posted this video awhile ago here on the blog. I wound my injection transformer using a P42/29 core. I basically replicated a Venable injection transformer that we have at work. I get good results with the transformer. I now use the Cleverscope to do my FRA measurement and do not need an injection transformer. The Cleverscope has an isolated function generator output and all I have to do is connect across the injection resistor. Since there is no injection transformer used, I get a bandwidth all the way to DC. The Cleverscope work well for me.

[diyaudio]

I found a unused complete Panasonic EMI Line filter core (from the junk box 10 years old), doesn't look  bad at all.
http://www.datasheetspdf.com/PDF/ELF18D290x/530267/3

[rx8pilot]

Hi EEVBLOGGER,

I posted this video awhile ago here on the blog. I wound my injection transformer using a P42/29 core. I basically replicated a Venable injection transformer that we have at work. I get good results with the transformer. I now use the Cleverscope to do my FRA measurement and do not need an injection transformer. The Cleverscope has an isolated function generator output and all I have to do is connect across the injection resistor. Since there is no injection transformer used, I get a bandwidth all the way to DC. The Cleverscope work well for me.

Robert -

Are you allowed to describe how the Cleverscope signal generator is isolated? Guessing it is not a transformer and some sort of isolation amplifier? The wide band electronic isolators used for this purpose are double or triple the cost of the the transformers, but seems unlikely they are terribly difficult to design and manufacture (could be sadly mistaken of course).

On the more traditional route, how critical is the transformer performance relative to the goal of FRA on an SMPS control loop? I do not have much knowledge of designing magnetics, but I can intuitively understand there is a challenge making a very wide bandwidth transformer like commercial models. If there is no need to test active PFC's, the low frequency is seemingly unimportant. I am not sure why you would need to go to 45Mhz on the high end either. In a practical sense, what bandwidth would you say is reasonable for the budget minded?

BTW - thank you for your video lessons on YT and the offline pointers you have shared.

[Bud]

. The transfer function of the transformer is not part of the measurement. You are measuring on either side of the transformer with respect to ground.

What is the point of having an "isolation" transformer if the measuring inputs not isolated?

Or theay are?

[rbola35618]

The designer of the Cleverscope is Bart Schroder. I approached Bart several years ago about implementing the FRA option. He redesigned the function generator so that the grounds on the function generator would be isolated from the scope grounds. He used an opto-isolator to do the isolation. You can float 300Vrms on the function generator. The isolation ideal was originally proposed to Bart by a colleague of mine Robert Raulk.

Regarding the loop measurement, if you are doing PFC then you need to go down to 1 to 5Hz. In my case, I do not do PFC since my design are only DC to DC space environment. So when I do a loop plot, I do them from 100Hz to 100KHz. My crossover frequency are around 1KHz to 10KHz. So I do not really need that high of a frequency. The Venable injection transformer state a bandwidth from 10Hz to 10MHz.  So 10MHz is well above the frequency that I am interested.

In addition to the FRA, I use the Cleverscope as a LCR. Bart implemented a LCR option. I built a jig with a current sense resistor and I am able to characterize my magnetics, and capacitors. You need very high frequencies, (The Cleverscope goes up to 65MHz) to characterize the magnetics and capacitance. When characterizing the magnetics, I am interested in the parasitic such as the leakage inductance, ESR, the parasitic capacitance (self resonance). The parasitic usually can only be seen at very high frequency where they become dominate.

I can see where Keysight can also implement the LCR option into their 3000T series. This would great if they could do that. I have the regular 3000 series and it can not do the loop analysis. I asked Keysight if they could add the loop analysis to the regular 3000 but they said the FPGA was not powerful enough. But that is OK since I have the Cleverscope.
   

[rbola35618]

Hi Bud,

The input do not need to be isolated. The ground of both channel 1 and 2 are connected to the power supply ground while the positive probes are connected to an injection resistor. Keep in mind that the resistor is 10 ohms to 100ohms and if you are monitoring a 5V output, then both end of the injection resistor will be close to 5V. You then need to connect a signal source across the resistor which as I mentioned is riding on 5V. So if you connect a non-isolated function generator across the resistor, you will end up shorting your 5V output to ground thru the function generator's ground. Therefore you have to use an isolation transformer. I hope this make sense. Robert Bolanos

[Bud]

Thanks Robert. Basically you are saying transformer is a DC block. I interpreted In the context of this thread that the purpose of transformer is galvanic isolation from the power supply for safety reason.

[Kleinstein]

The transformer is there to add the signal to the 5 V output with all its higher frequency components. So it is not just to block a clean 5 V, it is there to inject a differential signal. The links to the electronic replacements could help to understand the purpose:

They have an amplifier that adds the driving signal and outputs it to the other side.

For a low frequency circuit this could work quite well. But with a faster circuit the amplifier needs to be pretty fast, not add a significant phase shift that might upset the DUT. This might even apply to frequencies higher than you are interested in. So for slow circuits this electronic way might be an alternative to the injection transformer. The amplifier way also has another limitation: it does only work at low voltages, so the transformer can provide this level including DC.

[rbola35618]

The only problem that I see with the Keysight is , if I am not mistaken, the resolution of the ADC is only 8 bits. The dynamic range and the ability to get the signal out from the noise is not as good as a regular FRA or the Cleverscope.

My Cleverscope has a 10bit ADC and I can upgrade it to a 14 bit ADC. So theoretically, the Cleverscope should be more sensitive. I have done some test where I can a good loop plot using only 1mV on input stimulus signal. I then tried doing that on out 30K Venable and it was not able to get a good plot. 

I think some of the higher end Keysight may have a higher resolution ADC. I would assume that those with higher ADC would work much better.

RB

[rx8pilot]

The only problem that I see with the Keysight is , if I am not mistaken, the resolution of the ADC is only 8 bits. The dynamic range and the ability to get the signal out from the noise is not as good as a regular FRA or the Cleverscope.

My Cleverscope has a 10bit ADC and I can upgrade it to a 14 bit ADC. So theoretically, the Cleverscope should be more sensitive. I have done some test where I can a good loop plot using only 1mV on input stimulus signal. I then tried doing that on out 30K Venable and it was not able to get a good plot. 

I think some of the higher end Keysight may have a higher resolution ADC. I would assume that those with higher ADC would work much better.

RB

To be fair, the resolution of the ADC is only a useful spec if the analog circuits in front of it are better. With full manual control on the 8bit Keysight using averaging and filtering I was able to pull out some really small signals.

[rbola35618]

Hi rx8pilot,

What would be an interesting test would be to see how low of a stimulus signal will get you a good plot.

If you get a chance and are interested in finding out, do some plots and decrease the signal and see where it starts to plot noise. I would be interested to find out how sensitive it is.

RB

[Kleinstein]

The 8 Bit converters in the scopes are really fast. So they can do a large amount of oversampling to reduce noise. An 8 Bit resolution looks low, but at GHz sampling the noise density is not that bad. It is well possible that the noise from the (switching) supply itself is higher than the ADC / Scope noise. Often 1:10 probes already give more noise than the scope itself. There is not much to gain with a low noise scope, if there is more noise from the DUT.

The important part could be how good averaging and harmonic analysis is done and thus what is the effective bandwidth. A slower measurement with longer averaging on each point can reduce the noise - especially if it is slower for the points with poor S/N only. So it could be just a question of how much time is spend for the curve. It's only improving like the square root of the time, but there is still a last resort in using extra time - up to the point of over night.

[rx8pilot]

Hi rx8pilot,

What would be an interesting test would be to see how low of a stimulus signal will get you a good plot.

If you get a chance and are interested in finding out, do some plots and decrease the signal and see where it starts to plot noise. I would be interested to find out how sensitive it is.

RB

Using the built in application, there is essentially no solution for SMPS. By the time I got a usable SNR, the loop will have long since been clipping. The 6000 scopes do not yet have the new power application that the 3000's and 4000's have. I could try to setup some manual tests to see what happens on one of my boost converters. I will be in the CNC shop for the next few days, so possibly on Sunday.

It would be great to throw the results and method into this discussion. Not a whole lot of conversation to be found on loop response. TALK_nerd(1000) = AUDIENCE_small.

[Kleinstein]

I looks like isolated generator is an $400 option. So if you have that type of scope, this is an attractive option.

If you want to build the transformer yourself, I would consider a more limited frequency range and than the option to chose from a few. It should be easier to make two of them for separate frequency ranges than one with very high bandwidth. Depending on the DUT you might have to choose the suitable one.

[Kleinstein]

Usually there is not need for isolated probes. It is just that with that special scope (cleverscope) you have the option to use the isolated generator option instead of the injection transformer and a normal (grounded) generator.

For a normal scope setup this is less of an option. It might be possible in the manual mode, as a very low cost option, to use an isolated (maybe even battery operated) generator - but this would be without special scope functions and one might still need the trigger signal to do at least multi period averaging. For the trigger a fast opto-coupler could be an option - so all the way down to DC.

[r0d3z1]

I have been using a Jensen VB-1BB isolation transformer used for analog video. Not perfect, but not bad assuming that 10hz is not a necessity. It does ok down to rather low frequencies though. I paid $20 on eBay. BNC 75 Ohms in/out

http://www.jensen-transformers.com/product/vb-1bb/

This is a similar product: Model 0017CC (10 Hz to 5 MHz, 50 ohm to 50 ohm, BNC to BNC)  by North Hills Inc.

[rx8pilot]

I have been using a Jensen VB-1BB isolation transformer used for analog video. Not perfect, but not bad assuming that 10hz is not a necessity. It does ok down to rather low frequencies though. I paid $20 on eBay. BNC 75 Ohms in/out

http://www.jensen-transformers.com/product/vb-1bb/

This is a similar product: Model 0017CC (10 Hz to 5 MHz, 50 ohm to 50 ohm, BNC to BNC)  by North Hills Inc.

Interesting option. Strange that these are targeted to video applications but 50 Ohms.

This model seems like a better fit than the Jensen

Sent from my SM-G892A using Tapatalk

[David Hess]

Picotest should redraw their diagram to be even more confusing.

A 60 Hz AC current transformer can work as inexpensive injection transformer.

[rx8pilot]

Picotest wants to get a lot of money for a transformer....

Sent from my SM-G892A using Tapatalk

[David Hess]

Picotest wants to get a lot of money for a transformer....

It is not a lot of money for someone who gets paid for the sort of work where the transformer would be useful.  Greater than $5000 DSOs are not expensive in that respect either.  Think of the cost in comparison to the salary of the engineer using it.

For hobbyists, there are other less expensive but more time consuming methods to use like driving the injection point with a transconductance output or using a 60 Hz AC current transformer in reverse.  I usually get by with just measuring the small signal and large signal load transient response.

[rx8pilot]

Picotest wants to get a lot of money for a transformer....

It is not a lot of money for someone who gets paid for the sort of work where the transformer would be useful.  Greater than $5000 DSOs are not expensive in that respect either.  Think of the cost in comparison to the salary of the engineer using it.

For hobbyists, there are other less expensive but more time consuming methods to use like driving the injection point with a transconductance output or using a 60 Hz AC current transformer in reverse.  I usually get by with just measuring the small signal and large signal load transient response.

True enough.
I am in the purgatory of power electronics design. I just don't do it enough to justify the bench full of specialized test gear, so I muddle through the best I can. I looked into renting the Bode100 VNA with all the accessories and it is about $750-$850 per month rental. About 8 rentals would pay for it. I would use it for 3-4 designs per year - so buying it breaks even at around 2 years for the dedicated FRA.

Measuring transient response will definitely reveal most control problems - I do the same over a range of operating conditions. I have been able to get some use out of the FRA built into my Keysight scope although it takes some doing when testing noisy switching supplies. The automated FRA application cannot deal with the noise, but manually I can step through the frequencies and use math filters and averaging to get a measurement of gain/phase in the control loop. Slow - but eventually gets to a reasonable answer of phase margin.

[Jay_Diddy_B]

Hi,

I described my injection transformer in another thread.

https://www.eevblog.com/forum/projects/dynamic-load-bode-plot-using-hp-35665a-dsa/

Here are some of the details:

Schematic:



Board Design:



Construction






I normally use an HP 3577A VNA for measuring control loop response. The HP3577A does a great job, but it is big and heavy.

Jay_Diddy_B

[rx8pilot]

Other than size - the 3577A is an interesting option since it goes to 5hz.

Any maintenance issues with it?

Your injector looks pretty cool.....off to read that thread now. Thanks for posting that one.

[Jay_Diddy_B]

Other than size - the 3577A is an interesting option since it goes to 5hz.

Any maintenance issues with it?

snip ...

The only issue that I have had with the 3577A is the CRT. The CRT assembly is the HP1345A display module. I changed mine with one I got from another instrument. If it happened again I would fit one the LCD retrofit kits.

It is a great instrument.

Regards,

Jay_Diddy_B