Can you use an electronic load when doing smps node plots?

[georgeharker]

I’m testing stability and optimizing a buck step down power supply. To do so I’m doing some bode tests, but I have to put the system under load. I have a siglent sdl1024x-e but I’m finding that plot points below about 1kHZ are all over the place phase wise. My signal injection transformer is flat enough from about 28Hz to 5kHz. I’m using a 20A Texas Instruments buck converter which frankly seems stable outside the test setup and has about 30mV peak to peak ripple max. However when I run bode tests I get noisy results below 1k ish. I have tried using varisweep to up the signal in that range but to no avail. I’m now wondering if using an electronic load is an issue - I was in constant current mode at about 5A. I’ve also tried constant resistance.

Scratching my head a bit. I believe I can get the 0dB crossing and at that end it’s clean enough. But I’d love a clean graph to help me believe I’m not missing something major at lower frequencies.

[Smokey]

maybe put the load in constant resistance mode?

[Berni]

This is a common problem with electronic loads. They have regulation loops internally to constantly monitor and adjust the load to match what you set it for.

For doing any sort of stability analysis on power supplies it is best to use simple power resistors as loads. You can build yourself a chain of 1 Ohm power resistors that you can tap off at various points to get various kinds of load.

Then to do stability testing get a MOSFET and drive it using a signal generator. Use this power transistor to switch in a extra resistor as load, letting you get a step response of your power supply under test. If you want to see it as a freqency plot, that step response can be run trough differentiation then FFT.

[georgeharker]

I suspected as much.

I might grab some power resistors to act as a secondary load - I don’t have any and I need to pull 10 A so big standard ones won’t do.

It looks like a combo of updating the scope firmware, doing a self calibration, setting the load to constant resistance and using only hold gain after careful setting has much more sensible graphs coming out. I’d course the 0db crossing is pretty much the same but I’m now much more confident that there’s not something nasty lurking in the low frequency regions.

Thanks

George

[Smokey]

maybe put the load in constant resistance mode?

funny.  I spent part of the day trying to figure out why the high side switch board I'm designing was going into current limit mode even through I had my electronic load set to constant current way lower than the current limit threshold of the switch...

The solution.... constant resistance mode on the load...