# Measuring Output Impedance of Power Supply

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.

#### flg

Well, I don't know about that paper I just gave up on but, where I work we just take a deltaV/deltaI calculation. Say the supply in question outputs 1.2V @ no load. At 50A load it's output V drops to 1.15V. (1.2-1.15)/50=1mohm.

#### korneluk

That is an impedance calculation at a single (mains line) frequency. I am more interested in measuring the power supply impedance across the entire audio band.

-- josé k.

#### lcsaszar

The method described in the article is good. An alternative would be to use a current output oscillator or just a large resistor in series (decoupled from the power supply through a capacitor). A constant current through the internal impedance of the PSU would give directly measurable voltage that is proportional to the impedance. This method could be applied in-circuit, that is with the amplifier stages in place. The oscillator could be replaced with a noise generator and the DVM with a FFT analyzer...

#### jan.didden

Paid Member
oshifis said:
The method described in the article is good. An alternative would be to use a current output oscillator or just a large resistor in series (decoupled from the power supply through a capacitor). A constant current through the internal impedance of the PSU would give directly measurable voltage that is proportional to the impedance. This method could be applied in-circuit, that is with the amplifier stages in place. The oscillator could be replaced with a noise generator and the DVM with a FFT analyzer...

That's what I have used also very often. For instance, if you insert a 1A current into the supply output, the measured signal at that output is directly the Zout in ohms. Insert 1A, and if you measure say 53 mV, the Zout = 53mOhms. You can do that at several frequencies if you don't have a signal sweeper/analyzer. Record the Zout at each freq, put it into a table in Exel and presto, you have a nice graph of Zout vs. frequency....

Jan Didden

A decent dynamic supply impedance monitor could be made pretty simply. Attached is schematic. Monitor supply with scope to see how much it dips under transients created by signal generator.

#### Attachments

• supply_impedance.jpg
24.2 KB · Views: 1,100

#### Hi_Q

When I worked in a calibration lab we had an adjustable constant current active load which was set initially to fully on (Static Load) and adjusted for a particular current drain e.g. 1 Amp. We then switched the load on and off (Active Load) with a square wave signal and monitored the amount of ripple produced by the square wave. The higher the supply impedance then the larger the ripple produced. By knowing the static current first applied it is easy to calculate the supply impedance by measuring the square wave amplitude. In our example, a 150mV ripple would indicate an output impedance of 0.15 Ohms. The square wave can be set to any particular frequency of interest but in our particular application it was usually set to either 400 Hz or 1kHz.

#### lcsaszar

An alternative (and most logical) method would be to use the actual amplifier as the load. Drive it with white noise into a dummy load and tap the PSU with an FFT analyzer. It needs some sort of calibration, though.

#### jan.didden

Paid Member
Hi-Q said:
When I worked in a calibration lab we had an adjustable constant current active load which was set initially to fully on (Static Load) and adjusted for a particular current drain e.g. 1 Amp. We then switched the load on and off (Active Load) with a square wave signal and monitored the amount of ripple produced by the square wave. The higher the supply impedance then the larger the ripple produced. By knowing the static current first applied it is easy to calculate the supply impedance by measuring the square wave amplitude. In our example, a 150mV ripple would indicate an output impedance of 0.15 Ohms. The square wave can be set to any particular frequency of interest but in our particular application it was usually set to either 400 Hz or 1kHz.

That method gives you a feeling for the ringing and stability of the supply, but it doesn't give you a quantitative result against frequency. If you do it with a sine wave, at several frequencies, and graph it (or let Exel graph it), you have something that can be very accurately compared to other supplies or to changes you make. You also see clearly where it rises with frequency so you can concentrate on that issue if you wish.

There are many valid methods; often the crux is to find the method that helps you to design/improve it most effectively.

Jan Didden

#### jan.didden

Paid Member
oshifis said:
An alternative (and most logical) method would be to use the actual amplifier as the load. Drive it with white noise into a dummy load and tap the PSU with an FFT analyzer. It needs some sort of calibration, though.

What would that give you?

Jan Didden

#### lcsaszar

janneman said:

What would that give you?

Jan Didden
It gives the plot of dynamic output impedance vs. frequency of the PSU. At least it indicates if it is flat in the audio band.

#### jan.didden

Paid Member
oshifis said:
It gives the plot of dynamic output impedance vs. frequency of the PSU. At least it indicates if it is flat in the audio band.

Well, with noise as the test signal it doesn't tell you what the actual test signal in amplifude vs frequency is, so how can you interprete the fft? Am I missing something?

Jan Didden

#### SY

Maybe I am. With white noise, you know the spectrum (flat). If you feed it to the output of the PS through a known impedance, then measure the spectrum of the supply (normalizing out the supply noise), you should get the impedance versus frequency because of the voltage divider formed by the source impedance of the test signal and the output impedance of the power supply.

So, what am I missing?

#### Robert McLean

Rather than white noise, I would input pink noise. Then put a constant Q variable filter on the ouput to read the level. With white noise you would want a constant bandwidth filter, which I believe are more difficult to make.

#### jan.didden

Paid Member
SY said:
Maybe I am. With white noise, you know the spectrum (flat). If you feed it to the output of the PS through a known impedance, then measure the spectrum of the supply (normalizing out the supply noise), you should get the impedance versus frequency because of the voltage divider formed by the source impedance of the test signal and the output impedance of the power supply.

So, what am I missing?

Yes SY, I understand the method. But the noise sprectrum is only flat when integrated over time. So, for it to work, you would need to do a spectrum plot integrated over time. Would just averaging many periods be enough? How difficult is it to measure the RMS value (or average maybe) of the noise signal (because that would be your reference)? How does that relate to the amplitude you see in the spectrum plot?

Jan Didden

#### SY

In a sense, a finite length measurement IS an integration over time (think of an RTA). Now, that falls apart for frequencies below 1/t (with t being the sampling time, i.e. 1/sample rate times number of samples per measurement), but that's a matter of appropriate choice of sampling time and sampling rate.

It also appears to me that you could use an MLS or impulse to do the same thing.

#### EC8010

All this noise sounds like a really hard way to make the measurement; use a sine wave and vary its frequency.

#### SY

Or three seconds of MLS, eh? I know you like turning the big dial (and it DOES feel good), but the multiplex advantage can't be denied. And if you limit the range to 50-100Hz and higher, the measurement noise can be reduced for the same measurement time by signal averaging.

#### EC8010

I want big dials, levers, valves, lots of noise, and a whistle. Ah, I seem to have asked for a steam engine!

#### jackinnj

Refreshing this -- there's a company in Pennsylvania which makes measurement apparatus for batteries and fuel cells -- some good information their website (such as use of twisted pair instead of coax, use of Faraday shield, etc.) -- New Page 1 -- Gamry has some nice charts with measurements into the microhms from DC to a MHz.

I've found that milliOhms, hundreds of uOhms aren't too difficult but require care. 10's of uOhms aren't that easy. My hat is constantly off to Janneman and WJ who seem to have gotten this nailed down (impedance measuremetn probably accounts for my tonsure looking like that of a Franciscan monk) -- WJ to one uOhm:

An externally hosted image should be here but it was not working when we last tested it.

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.