measuring the noise of a power supply

Status
Not open for further replies.
Hi guys, I have two questions actually. First, how do you measure the noise on a power supply. Often people talk about noise components, and so my assumption is, since we are assuming a DC supply is perfectly dc, any AC would be noise, right? Would a very quick and simple way to measure the noise be to use a multimeter on AC volts across the dc output of the power supply? Similar to how you measure the DC component of an amplifier by measuring DC, instead of AC.

I have a sense of how it is done with an oscilloscope, but if someone wants to explain that, I'm sure I could learn some. I know this wouldn't tell you where the noise is, frequency wise, I imagine its just the largest AC component being represented? I also know it would change with load, but I can simulate loads too with dummy loads, right?

As for load regulation, when people mention the percent, is that basically what percent of change takes place from minimum load vs maximum load? For instance, if under idle current I have 61 volts and under peak load I have 57 volts, I have something like a 7 percent change. Would that be 7% load regulation?
 
Hi,
Your assumptions on noise are correct.
An oscilloscope is nice, but only if you work on a switch mode PSU.
A good true RMS multimeter should be enough.
Hope this helps.

This is exactly what I am doing at the moment. I have 200mV pk-pk of 100Hz (50Hz mains here) without the regulation loop. It all turns awful when the regulation kicks in. I can see the 200mV 100Hz being squashed and replaced with high frequency noise. Plus, the ferrite transformer starts to buzz. I need to work on the loop.
 
Thanks guys, I'm glad I was on the right track at least. I was measuring very low levels of noise then, but it sounds like, based on that TNT article, that its probably lower than I even could measure because of the resolution of my method. Makes me happy to know the supply isn't making a lot of noise.
 
We have really really nice multimeters in college, sat on the output of my PSU on no-load, it autoranged to <10mV.

The correct method of oscilloscope is to put it on DC mode, and move the line down to the 0-line. Then just adjust the voltage/division and time knobs until you can see it. (making sure it's triggered properly)

Then just take typical methods

L
 
There are 2 components- ripple and noise.
Ripple in a smps is simply based on the LC on the output stage. The majority of ripple is delta I in the output choke * ESR of the caps. Double the caps and you reduce the ripple by 2. Double the switching freq and you 1/2 the ripple.

If there is a low freq component then the loop needs tweaking- as the other poster mentioned.

The noise component is generated by the switching transients and is usually in the 1Mhz and up range. This is why snubbers are important.
 
Noise is all over the spectrum in a converter output...
Typically the spectrum analzer is used to look at noise... This is what i use at work to meet noise requirements... Also we inject signal into the input BUS and must meet specific rejection requirements at the output...
You need one that has a good noise floor at least -120dB ...
We normally use an LNA in the front end of the spectrum analyzer...This will drop the effective noise floor to like -140dB...
You need to account for the Gain of the LNA minus the noise floor of the LNA added on to the noise floor of the spectrum analyzer, this will be the new effective noise floor you will be measuring from..... If you just use a AC volt meter, then you will only be measuring the amplitude of the fundamental ripple and not seeing the noise floor as well as switching spikes...There are resonances at roughly half the switching frequency in some current mode converters as well other resonances that change depending on Duty-Cycle...

Chris
 
Status
Not open for further replies.