I am working on a board that contains a DC stage, an SBC module, mechanical relays (smd) and communication modules. The power supply is a cheap 24v AC-DC converter. It is a COTS product. The DC stage needs to provide a 12v, 5v and 3.3v.

I need to supply different kinds of systems simultaneously: communication modules (wifi, ethernet, cell, usb), an external tablet, an SBC, multiple ATtinys, RS422 drivers, sensors and other peripherals.

As I see it, the DC design is key both from power requirements of the board and also from EMI and noise suppression.

The previous version of the board was designed poorly in terms of schematic and layout.

My plan:

  1. Design a converter for the 24V to 12V drop.
  2. Use a linear regulator (non LDO) to step the 12V to 5V.
  3. Use an LDO to regulate the 5v to 3.3V.

My design guidelines are, of course, open for input.

My questions are:

  1. what kind of converter should I choose? will a flyback be sufficient?

    Can I trust a switched converter to supply communication modules and controllers?

  2. If I use a non-isolated converter (24V to 12V), will separating the grounds (analog, digital) help me suppress interferences on my PCB? Are there other methods I should implement?

  3. Should I place the DC stage on one small area on the board or put the regulators as close as I can to their relevant consumers?

  4. If I don't have size constraints, is it better to lay out in two layers or four?

I'd welcome other inputs and reading references (links, pdfs...) :)

Many thanks for your help!

  • 2
    \$\begingroup\$ You cannot be assured of any reasonable design to meet your expectations until you define them. i.e. write a spec. V, I, ripple, step load and step risetime, load regulation, source regulator, EMI near-field (crosstalk) and far-field, susceptibility to climatic and electrical stress (EMI, ESD, PLT). Once you list all of these in this question , highlight ** the ones you have concerns with. Never assume one method is better than the other until this is defined. But do look at commercial solutions for ideas. \$\endgroup\$ Mar 29, 2021 at 13:52
  • \$\begingroup\$ But do look at commercial solutions for ideas. (lots of them and reverse engineer with an EMI probe or shorted 10:1probe loop for near field noise >=20MHz and transient response.) \$\endgroup\$ Mar 29, 2021 at 13:55
  • \$\begingroup\$ First estimate the power cosumption on the 5V and 3.3V rails. If > 1W, reconsider that linear 5V regulator. \$\endgroup\$
    – user16324
    Mar 29, 2021 at 14:09
  • \$\begingroup\$ If you can’t define my specs, then you need to identify long cables which radiate noise without balanced cables to source impedance and balanced for common mode noise,you are probably better off buying a good COTS PSU which are cheap for 12,5V,3.3 and measure your radiated noise on the cables while designing like PCs, a good common earth ground for best cable EMI reduction. This is why All VGA cables have a ferrite BALun even with 75 matched loads. I don’t think you have enough experience to get everything right in a DIY design. \$\endgroup\$ Mar 29, 2021 at 15:30

1 Answer 1


This is a very comprehensive question and answering it within one post may be difficult. Since I cannot comment yet I guess I'll make a full post of the advice I have.

  1. Keep in mind what the loads consuming by each of your supplies will be. If the loads are high enough consider using buck converters instead. Unless you're not too concerned about power efficiency which could be the case.

  2. Definitely separate your analog and digital grounds if your analog is very sensitive to noise (i.e. high speed I/Os). However, if the frequencies of operation are not that high you may not be impacted by noise all that much from a power supply perspective.

  3. It's generally not a good idea to nest power conversion stages (i.e. a 12V to 5V stage followed by a 5V to 3.3V stage). This could be neglected if using straight LVRs/LDOs since efficiency is only Vin/Vout. But again check your load demands since DC/DC converters can be more efficient.

  4. For your question "Should I place the DC stage on one small area on the board or put the regulators as close as I can to their relevant consumers?" This falls in the realm of power delivery networks (PDNs) and it's very application specific. If you will have high dI/dt events (i.e. high load steps) then if you're too far away from the consumer, you'll be hitting parasitic inductances that will compromise the effectiveness of the regulator. Also, the longer the copper trace the more resistance you'll have.

I won't make this post longer unless you want to hear more. But for now these are the recommendations I'd give you.

  • \$\begingroup\$ thank you for answering. I see that the question I asked is too complex for one discussion. I will separate it to threads with your insights. \$\endgroup\$
    – metsik
    Mar 30, 2021 at 9:01

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.