Need suggestions on very clean AC-DC(0.1mV pp) supply design

Hello Richie - That sounds like an interesting challenge. You are correct that SMPS is inherently noisier than a linear approach - the SMPS is also inherently lighter and smaller - but that is a different story. Don't forget that at the levels you are working at the SMPS isn't the only source of noise you need to worry about - any digital control logic or displays which you have will also generate noise.

One strategy I've used in the past is to take multiple measurements - if this is done so that any noise is asynchronous with the measurement - then the noise will be averaged out-

Measuring pA currents isn't easy either and I assume that you are interested in a low frequency measurement.

I would make some general comments

Use a soft switched topology - a Transition Mode flyback for example to generate +/- 12V directly. One issue here is that Common mode effects due to the switching action can inject significant noise currents. VERY careful design of the transformer construction - winding placement, dummy noise cancelling windings etc. - can help here. It goes without saying too that VERY careful PCB layout is needed too.

Assuming that there is a microcontroller somewhere in the system, then have the microcontroller turn the SMPS off while the ADC is taking the measurement, this will eliminate any SMPS generated noise from the measurement. While you are at it have the microcontroller go into a sleep state - preferably with the clock off until the ADC has done its work. You will need of course to provide an adequate energy storage during the time the SMPS is off - I'm guessing that this would be less than 1msec. If this could be made to work then I would use this approach. Of course I don't know if you are designing the entire system or just the SMPS section...

Let me know what you think - if you can fill in some details on the application I will think some more about this, but as I said, if you can turn the SMPS off while taking the measurement - that would be my first choice. Regards Colin

Hi Collins

Thanks for your reply, which is very informative to me. Your suggestion is quite brilliant! But we only have a small research team with limited research capacity. So currently we are planning to buy some commercial DC converters which can produce 12V directly from 220Vac. And then change the +12V supply into various DC voltages. 

Our current plan is to use combinations of power filters(R, L,C) and different LDOs, since in the application, voltages of: +-12V, +-5V, +3.3V are needed...

Could I have your comments? If the description is not clear, please tell me and I am very willing to explain it further.

Thanks again.

with best respect, 

Richie 

Hello Richie

OK, I think I understand - So here are a few ideas - just general comments for now.

Establish a baseline performance for your measurement system by running it off a battery - 12V lead acid cells are not too expensive - either for a Car or Motorcycle. Use linear regulators to give you the various 5V and 3v3 rails you need. this would let you check the performance of the measurement system in the absence of any external noise.

Check on line for low noise SMPS - I did a quick search and found some specialist companies offering these units - although I suspect they may be expensive.

How important is universal operation to you - how about using a linear PSU with a 115/230 input selection switch? If you had to you could build a circuit to automate the switchover so getting a pseudo-universal input supply. Linear supplies are inherently lower noise.

I think LDOs are have poorer line rejection ratios than regular linear regulators - remember that the low dropout feature isn't going to be of any benefit if Vin = 12V and Vout = 5V. So first choice here would be a standard linear regulator.

If you need a SMPS, see if you can get one based on a symmetrical front end, a double ended forward stage for example. Symmetrical input stages will generate less common mode noise than asymmetrical stages - eg a flyback. Remember too that it is very difficult to avoid converting a CM noise signal into a DM noise signal. A true CM signal won't add to your measurement noise, but a DM signal will.

Of course any power filter you use should be properly damped to prevent unwanted resonances from generating noise.

Do you need to worry about the interference effects of nearby E or H fields - this would affect the physical layout, screening etc of the product.

You can never spend too much time worrying about a PCB layout - especially for an application like this.

Hope this helps, I know that the above are only general comments - please feel free to contact me on this if you have further questions. Best of Luck, Colin

if you want to do this design i suggest that you use a linear dual tracking solutions with op amps as error amplifier and a TL431CLP as a voltage reference and BD136,BD135 with clip-on heat sink as ballast transistors...

Dear Collins

Thanks for your reply! As I have little experiences designing power circuits.. Probably I will start by some simple experiments.. By the way I just  measured the power supply from laptop adapters and they have ripples of 400mV...

Thank you very much!

Best regards, Richie

Yes, laptop adapters aren't low noise! - Have fun. Let me know if you have any questions or just want someone to review an idea.

Regards

Colin