3

I'm laying out the PCB for a small and simple power supply to feed some audio amplifiers. It will feature a DC -DC converter module and I'd like to filter the output for 1) extra smoothness 2) learn about power supply filtering. So I found this courtesy of TI:-

schematic

I'm using a TRACO converter rather than TI, but otherwise the circuit is the same. The switching frequencies are similar. They're both 2W - 3W power. So a 100uH ferrite inductor looks like:-

inductor

If I were to have a ground plane on the backside of the PCB, do I back it off from the inductors (and converter)? Consider that there will be inductors inside the converter. That's three inductors. Do I put three holes through my ground plane, or keep it solid?

I looked at one question that suggests keeping the ground plane solid, whilst the answer for this one suggests a hole. I'm confused again.

Further confusion arises from some cheepo converters I have from on-line. You'll have probably seen one of these fairly ubiquitous thingies:-

ebooster

These have a solid ground plane on the reverse side. Is that just 'cos it's easy /simpler or are cutouts unwarranted? Clarification would be most appreciated.

Paul Uszak
  • 7,327
  • 5
  • 37
  • 69
  • 1
    Solid ground is probably better. Of course there must be holes for the through-hole pins. When people say solid ground, they don't mean that there are no vias or holes for component pins. They just mean that you don't route any other nets on that layer. You are probably focusing on the wrong thing here, though. The key with switch-mode supply layout is to understand where the high current paths are, and make them as short as possible to minimize the loop area. – user57037 Jun 11 '17 at 04:38

2 Answers2

3

That round Siemens inductor will have lots of external field lines, because there is no closed path. The air path is 1cm or 2cm (10mm or 20mm)

The cheapo switcher inductor looks like a black tub with central core, and the flux only has 1mm or 2mm air path, thus much less external flux to be a bother.

That round Siemens inductor will induce voltages (Faraday Law induction) all around on the PCB.

Find a closed-ferrite-path inductor to use.

analogsystemsrf
  • 33,703
  • 2
  • 18
  • 46
  • To be clear, you're advocating one of those toroidal shaped inductors, rather than a straight shaft based one? For example : http://uk.rs-online.com/web/p/leaded-inductors/1048436/ ? – Paul Uszak Jun 11 '17 at 13:05
  • Yes. I've read lots of discussion on 'diyAudio' about their choices for 60Hz power transformers, so as to keep the 60Hz (and diode-rectifier spikes) out of the 1nanovolt or 10nanovolt trash-floor-goal RIAA frontends. Their choice, after lotta experiments, is expensive race-track power transformers; these keep primary and secondary far apart tho on the same core AND the flux path is totally enclosed (no air gaps, no I+E or C+I construction of mechanically-mismatched soft iron laminates.) Again----provide a full path for the flux. Avoid air gaps of any sort, whether inductor or transformer. – analogsystemsrf Jun 11 '17 at 22:42
  • Do you have an example of the kind of transformers you're talking about? Or a link to those discussions? I think I understand, but a search for "race-track power transformers" got me pretty much what I expected, and I don't think it's what you meant. – Samuel Edwin Ward Jun 12 '17 at 01:59
1

Here is photo of French-made racetrack power transformer

enter image description here

The overt separation of primary and secondary provides a strong reduction in capacitive coupling between those 2 windings; electrostatic shields are in place, from what I recall, and you need to find a "return path" for those shields.

Another purpose for the core was near-total-containment of the 50Hertz flux (and any diode-rectifier high-frequency surges, which easily couple into the RIAA frontend, because the fast edges ----- some few microseconds, if not mitigated ----- of the diode surges provide high dI/dT and thus lots of magnetic coupling.)

In your situation, you just need to provide non-air-path flux paths, to minimize the flux interacting with GND or VDD planes and inducing switch-reg frequency trash into the planes.

The propagation delay of standard 1 ounce/foot^2 copper foil ----- pushing electrons THRU the foil ------- is 160 nanoseconds, about 1,000,000X slower than the speed of light. This delay comes from SkinDepth math, where severe attenuation occurs as electric and magnetic waves attempt to impose voltage gradients inside copper metal.

Slow E&M waves will penetrate the foil (your planes) and appear ON THE BACK SIDE to cause interference. Waves (or edges) faster than 4MHz get progressively better attenuation as edge rates speed up. Thus fast SwitchReg clock rates and fast switching edges ( > 250 nanoseconds) are more easily shielded.

analogsystemsrf
  • 33,703
  • 2
  • 18
  • 46