Power supply noise measurement load ideas

[matthuszagh]

I'd like to build a load for power supply noise measurements. I've heard it's important to adequately load the power supply to get an accurate representation of the noise it produces. What sort of load do people use for this?

Thus far I've been just hooking up the supply to suitable 50W resistor. However there are a few problems with this:

1. the setup isn't shielded
2. the connecting cable isn't twisted and shielded over the full length, which adds some loop inductance and external noise pickup
3. the load resistor is highly inductive

I have noticed my lack of shielding is a problem; I still see a fair amount of noise when I turn off the supply but leave the cables in place. I suppose the solution for that is easy: cut a whole in a metal box and place the load in it with the interconnecting cable running through the hole. This should help with external noise pickup. But, that doesn't address part of 2 and all of 3.

I've seen some nice projects online for resistive load boxes (e.g., this one: https://www.angelfire.com/electronic/funwithtubes/LoadBox-1.html). But, those still suffer from inductive loads. I have an active DC load, but I don't want to measure it's noise.

If I could load the DUT less I could use smaller, less inductive resistors, but maybe this doesn't assess the DUT noise under realistic scenarios.

Or I guess I could stick with power resistors and accept that the noise at high frequencies might be somewhat exaggerated.

What do other people do?

[blackdog]

Hi matthuszagh

When I do noise measurements on Power Supply's it is always through coax connections, no scope probes!
Also make sure you have some low and high pass filters on hand.
With cutoff frequencies suitable for your applications.

Your measurement environment should be "clean", and by that I mean no equipment around you that generates strong interference fields.
A mains isolation transformer is also useful and also consider the interference level of many LED lights, it can be a pain in the ass. 

Modern "Active DC Loads" are themselves nice sources of interference.
This is because most use a PWM signals for driving.
Rigol, Siglent, Korad, most below 1000$ use this technique and are therefore not suitable for low level interference signal measurements on especially linear Power Supply's.

You can build yourself an Active DC Load with a very low noise level and then use a modern processor cooler on to cool the transistors or MOSFets.
You also need a fan, if you would make a controlerfor that fan, do it linearly, do not use PWM, bin there, done that. 

The project you show uses resistors that are not induction poor and switches that do not have high reliability, especially at the somewhat larger currents.

There are several manufacturers that supply resistors with very low inductance.
Vishay, Isabellenhute, Bourns to name a few.
These are often in TO220, TO247 etc housings.
Choose a few resistor values that is convenient for your application.
You can then mount these resistors back on a modern processor cooler.

I khope this helps a little...

Kind regards,
Bram

[Vovk_Z]

A wired low-Ohm resistor is quite a low-noise load itself. Its inductance doesn't matter much, because all loads may have some inductance too.
As for me, the large concern is a noise-meter itself. If it have all needed properties (frequency and voltage range, resolution, etc)  then you'll find all answers about a load with it, if you are interested enough.

[srb1954]

I've seen some nice projects online for resistive load boxes (e.g., this one: https://www.angelfire.com/electronic/funwithtubes/LoadBox-1.html). But, those still suffer from inductive loads. I have an active DC load, but I don't want to measure it's noise.

If I could load the DUT less I could use smaller, less inductive resistors, but maybe this doesn't assess the DUT noise under realistic scenarios.

You could use power metal foil resistors for minimum inductance. These are available in ratings up to 50W and generally come in something like a TO-220 package requiring mounting on a suitable heat sink to achieve their specified power rating.

Or you can get specialty WW resistors where the coil is wound so as to cancel out most of its inductance. These may be somewhat more difficult to source and quite expensive.

If you want to stick with conventional wire wound resistors for low cost and ease of supply you can compensate their parasitic inductance to the first order by paralleling the resistor with a series RC network. As the frequency rises the falling impedance of the series RC network compensates the rising impedance of the WW resistor. The crossover frequency of RC network, where it starts to look more capacitive than resistive, has to be matched to the crossover frequency of the WW resistor and its self inductance. For example, if you have 10 WW resistor with 10uH inductance you can use a series connected 100n capacitor and 10 resistor across it to cancel out the inductive effects.

Since the DC is blocked from the paralleled resistor it can be a low power, low inductance film resistor but you need to be careful not to apply high amplitude, high frequency signals to such a combination as all the high frequency currents will be diverted through the low power film resistor burning it up. With the component values mentioned above you should not apply high amplitude AC signals of more than 50kHz.

[matthuszagh]

Thanks for the recommendations! Based on this I found the PF2205 (https://riedon.com/media/pdf/PF2200.pdf), which seems well-suited to the purpose: 50W, <10nH and thin film. I couldn't find any metal foil up to 50W - just up to 10W or so. There are others that are thick film but I prefer thin film for low noise linear regulator measurements.

I'm considering adding a parallel RC compensation network for high frequencies, as mentioned. I would like to measure up to low GHz noise for switching converters. But, maybe trying to keep the impedance curve flat up to those frequencies is unnecessary and overkill; 10nH already seems pretty decent.

I tried my HP 6050A load to see how it would fare in terms of noise and it was extremely noisy in the constant resistance mode, so that's out.

[CamJam]

Hello,

Apologies to unearth this old thread, after watching a fundamental Friday by Dave and searching the forum, this thread seemed appropriate. I am an amateur at analog/physical measurements of circuits/equipment, so I appreciate all feedback and insight no matter how mundane or fundamental.

Trying to measure/confirm the rms noise of a B&K 1760A DC power supply which is quoted to have ≤ 1 μV rms (no specified bandwidth so I presume 20 MHz). I attached the datasheet .pdf below.

Used the 2 channel oscilloscope differential mentioned at the 21:45 timestamp of the video but with the circuit shown at timestamp 40:00 (used two 50 BNC and a 50 terminator resistor at the oscilloscope inputs). I attached the images of the BNC connections below.




Finally here is a snippet of my reading, used the trigger mode as AC, bandwidth limited both osc channels to 20 MHz and AC coupling.



Does anything seem inherently wrong with what I did to measure the power supply noise? (Can I even measure DC power supply noise using this differential method?)

Note: It was quite difficult to get a "stable" signal from either input channels... had to set the trigger to "AC" and  play around with the frequency scale knob of either input channels to be < 1 MHz for these results. If I set the frequency scale to be > 1 MHz the signals would become extremely "messy" and the resulting differential signal would "bounce" around the osc screen. Below is an image of the osc reading when the trigger was set to channel 3 and frequency scale > 1 MHz.

[JustMeHere]

Those long leads you have "flapping around in the breeze" will be picking up noise. 

[CamJam]

Haha, great analogy.

I will try with smaller (< 2.5 feet) BNC cables and report back.

Thanks,

[CamJam]

Replaced the 2.5 feet BNC leads with 6 inch BNC leads.

Acquired an image at high > 1 MHz



and low < 100 Hz horizontal scale.



Honestly still somewhat confused as to why the rms value of the difference between the osc 2 channels change as I turn the frequency knob...