# Measuring the Output Impedance of a (Large Capacity) Voltage Source

Discussion in 'Electronic Design' started by Anand P. Paralkar, Dec 29, 2013.

1. ### Anand P. ParalkarGuest

Hi everyone,

I am trying to test a MOSFET H-bridge at around 400V DC input. I need
to observe the MOSFET Drain-Source waveforms for ringing during the on
to off state (and vice versa) switching.

I observe that the drain to source waveform is affected by the DC source
that I use to supply in the input DC voltage to the MOSFET H-bridge. We
have three different sources (including one that I built using a simple
variac, rectifier and huge capacitor bank) and I feel that there is a
particular DC source (bought from a vendor) that does not work well
(gives horrible drain to source waveforms).

In order to be sure, I would like to see that the one that gives a bad
waveform has something in its output impedance that makes it misbehave
while the other two have something "good" about their output impedance
that make them give good switching waveforms.

Is there a (simple) method to measure the output impedance of a 600V,
15A DC voltage source?

I will use the suggested method to evaluate the three sources that I
have and hopefully, I will find the evidence to prove that the source
impedance.

Regards,
Anand

2. ### WimpieGuest

El 29-12-13 10:55, Anand P. Paralkar escribió:
I had such things in the past also (but with less power). You may
observe the transient (step) response of the supply by switching a
large low inductance resistor with a mosfet or fast IGBT. It was
strange to see that some power supplies were close to instability.

You could also use small signal measurement, but the large signal
response may be different from the small signal response.

Somewhat OT: I think it is good to have sufficient decoupling close
to the H-bridge so that the DC power supply doesn't "see" the
switching transients.

3. ### Fred AbseGuest

If you mean what I mean by "large capacity", why aren't
you using IGBTs?

If there's any inductance in the load, you need freewheeling diodes across
the FETs, don't trust the parasitic substrate diodes to do the job.
Firstly, look at the waveform at the power supply. It should be DC, with
no more than 20% P-P ripple.
Measure its open circuit voltage, then resistively load it for, say 1 amp,
and measure the voltage again. Calculate the source resistance from the
reduction in voltage, and the value of your load resistor. It's just a
resistive voltage divider.
Describe, or preferably show us, the waveforms.

4. ### Maynard A. Philbrook Jr.Guest

The one you're seeing unacceptable behavior is most likely a switching
supply or one with a very tight regulator in it.
The ringing or even sudden changes of current loads to no loads for
example is making it a little jumpy.

inductive ringing can generate more voltage than your supply and it's
also possible you are hitting a safety shut down because the caps being
used in the supply are small, since they are a switcher at most likely
much higher frequency than the ring you have, it makes it easy for your
H-bridge to upset it..

Try hanging some large low ESR caps on the output side of the supply or

Jamie

Den søndag den 29. december 2013 16.08.54 UTC+1 skrev Fred Abse:
unless you disable the parasitic diodes with a series diode I don't see
external diodes helping much

It seems most power fets have similar rating for the diode as for the fet
but they can be slow

maybe that is the problem, shoot-through triggering a current limit in the supply

-Lasse

6. ### Fred AbseGuest

That's the point.

7. ### Klaus KragelundGuest

Add a series low ESR capacitance to the output (high voltage type). Use a network analyzer to sweep the impedance coupled into the capacitor (if you do not have an impedance analyzer, apply a function generator with series resistance, and plot values a regular frequencies.

If the load is really low impedance, use a HiFi amplifier as the source. Ifthat proves to feed to little current into the output to measure the impedance, add a transformer to boost the current.

Sub 10mohm impedance measurements can be done this way

Cheers

Klaus

8. ### misoGuest

(if you do not have an impedance analyzer, apply a function generator
with series resistance, and plot values a regular frequencies.
If that proves to feed to little current into the output to measure the
impedance, add a transformer to boost the current.
Seems to me you should to meditate a bit about charging this series cap.
The instantaneous connection would put the full 400v into the network
analyzer. Perhaps a precharge is in order.

9. ### misoGuest

The first thing to consider that the power supply isn't linear. If it is
a switcher, then for sure it isn't linear. If it is a "linear" supply,
well linear circuits have slew limitations, and if you are slewing, you
are not linear. Further, there is usually protection circuitry in the
loop. Linear supplies have multiple control loops.

Thus you probably have to whack the supply with a load pulser and
observe the voltage via a scope with probe AC connected. But then the
next issue is are you testing the power supply or the load pulser. Most
electronic loads are pretty crappy, so you are watching the electronic
load settle as well as the power supply.

For datasheets, where you want to be 100% sure all the ugliness (and
hopefully lack thereof) is due to the device under test, you use as
passive of a load pulser as possible. Typically you go from no load or
in some cases minimum load current (i.e. high value resistor) to a low
value resistor switching it in by a power mosfet. I've built these
pulsers using PCB strips to solder a number of parallel carbon film
resistors in order to make a low inductance resistor. Generally one
mosfet is enough. You just make sure the on resistance is low compared
to the resistor array.

Use this somewhat passive load pulser to perturb your supply and see
what happens.

10. ### Klaus KragelundGuest

Precharge would be a good idea for this voltage. I have used it for lower voltage (30V), and the inrush surge current is manageable at that voltage

Cheers

Klaus

11. ### Anand P. ParalkarGuest

We have a MOSFET that can handle 20A and 1200V. That was large, atleast
for me.
No, we currently have only lamp loads (50 * 100W incandescent bulbs).
Surprisingly, the voltage, atleast as seen at the input of the H-bridge
PCB shows only a DC voltage.
I am not too sure this will hold good when we are using the DC to
produce a switching output (as in a PWM sine or rectangular wave).
Will try.

Thanks for replying and wish you a happy new year.

Regards,
Anand

12. ### Fred AbseGuest

That's small to medium capacity, to me. I'd still use IGBTs, though.

40A, 1200V IGBT FGA20S120M \$2.24 at Newark.

IGBTs don't have parasitic diodes, BTW.
Lamps have very low resistance cold. You'd be better using 5kW worth of
heating elements. 50 hundred watt lamps, in parallel, will be a fraction
of an ohm cold.

Never start an investigation thinking you know the answer. That's what
police do. Not engineers.
Measure at different load currents, up to maximum. Plot the results.

You need a pulsed load to go any further, so you can see rise times. Maybe

13. ### Maynard A. Philbrook Jr.Guest

Only at 25C otherwise, it's a 20 amp unit.
and also, they do have diodes.

https://www.fairchildsemi.com/ds/FG/FGA20S120M.pdf

Jamie

14. ### Guest

With an infinite heatsink.
With an infinite heatsink. Good luck with any practical cooling to
25C while dissipating 30W.

15. ### Guest

Good grief. ...and cool that with LN2? <sheesh!>

16. ### Fred AbseGuest

Dissipation is dependent on duty cycle.

Typical VCEsat 1.75V at IC=20A, TC=125C That's 35W at 100% duty.
PDmax is 348W at TC=100C

Note temperature quoted as TC, not TJ.

18. ### Fred AbseGuest

Sorry, that should be 174W at 100C

19. ### Guest

Sorry, try doing that in the real world.