[Moved]: Very low ripple psup design.

Hi

In a new project I will participate in we are in need of a very low riple powersuply design. We talk about less than 50uV ripple at full load.
I know that this in most cases eliminates switchmodes and such if the design should not be to complex, thus leaving linear regulation as best option.

I do not have the full spec yet as this is a project that I will get involved in next year somewhere, but I assumes, based on wild qualified guesses, that full load means something like 120A pulses in up to 1ms, or less (or more... not important right now). However, my question is not for you to solve the design, more a request for good sources to read up on power supply designs relevant to some of my requirements. (Maybe I have misunderstood something as-well, but that's my problem entirely.)

I know that PSU's with less ripple has been done before, but I guarantee you that it has never been done by me, ever. And that is the core-problem here i guess :-(.
But I can learn, but I need something to learn from. So, tutorials, open designs that I may study, or ref. to other relevant information would be highly appreciated.
Radar designs comes to me as good ref. designs here, (or maybe some of NASA's neat probes to measure very weak signals in space).

Batteries as main sources will be considered, but afaik it's preferred to have regulated PSU's.

Here's an example source I found, it might trigger someone...: **broken link removed**


Thanks in advance

Do you have any idea what the output voltage would be. and is that 50uV P-P. or P.

- - - Updated - - -

Here is a formula from G Chrysis power supply book, page 178.

https://www.amazon.com/High-Frequency-Switching-Power-Supplies/dp/0070109516

The output ripple is determined by the output capacitors ESR.

ESR = ripple voltage (i think in P-P) / .25 x load current.

.000,050 V / .25 x 120 A = .000,001,67 ohms = 1.67 micro ohms

To me this does not sound realistic.

- - - Updated - - -

It looks like my notes are wrong. I am not sure were i got that first formula from but i looked in the Chrysis book again on page 179 and it says.

ESR max = Delta Vout/Delta Iout

.000,050 V / 120 A = .000,000,417 ohms = .47 micro ohms.

That's even worse.

Give up now and avoid the holiday rush!

Maybe many interleaved power supply's is the answer.

The first thing to do might be to get some responsible person to come up with a full list of requirements and specifications, and get it in writing.
That may require a planning meeting to discuss requirements, and try to get justification for each point finally decided.

There is a very big difference between peak, average, and rms values for measuring ripple and noise, and the relative importance of the measurement method depends on the intended final use and application.

Specifications for voltage regulation and transient response may be of importance as well. The more constraints there are, the more difficult the problem.

Only when a final requirement has been positively nailed down can you start to consider all the relevant factors, and plan some sort of a solution.

Well, No argue against that, As mentioned I have little information, and less knowledge, so well see when I get more.

P to P it is, that I know for sure. I just got an early warning that I was about to be involved in the topic and wanted to spend the time to learn.
I do not think that is the worst thing to do. Ill, check out the book...

Offcourse there will be meetings, but it would be vice for me to touch up on the topic before the meetings to be able to ask all the right critical questions. Thats one of the reason I wanna do this.

Fine, peak to peak means no narrow voltage spikes, so a switching supply is probably best avoided.
The choice then is probably between either a linear regulator, a battery, or a very potent multi stage LC filter, or perhaps a combination.

I was thinking switch mode to get you in the zone, then perhaps filter its output heavily so you get your ripple spec...better than a total linear design.
Of course, your ripple depends on the load transients as well as the smps...because if your load suddenly pulls a load of current from being nothing, then how on earth you are going to keep the output to within 50uV I will never know.
I suspect that 50uV is an error at this stage...unless perhaps the load is totally totally constant..no transients whatsoever......another point is the pcb traces when there's a transient, they could easily drop more than 50uV.

I cant imagine what the application is, because with most equipments, psu ripple is not a problem, you can just get a dsp chip to blank it out.

I would think ambient noise could put more than 50uV of noise in your equipment, even of the psy had zero ripple

Yes, its probably something very special and quite obscure.

Its getting right down into the noise floor for many things.

Sounds like a class H amplifier is what you want. Basically it uses a SMPS followed by a linear regulator biased with a low dropout voltage to preserve decent efficiency.

as per mtwieg's comments, we have done these up to 600A 12V 4 quadrant. 50ppm current control

and the 50ppm current control, we presume that enables the ripple spec of 50uV?

We had large inductive loads so the load ensured the volt ripple was very low and the control electronics and linear stage ensured the current was within a 100ppm band - so yes ripple less than 30uV rms, 0 - 12VDC out, 0- 600A out. Water cooled linear stages with high precision DCCT to measure current (shielded from Earths mag field)