Filtering power from DC-DC converters

[aion]

I'm looking at DC-DC converters such as the Recom RS3-1215D and Traco TEC 2-0923 to get +/-15V from 9V input. I am wondering how I should be filtering them for audio applications.

My understanding has always been that more filtering is better in audio. But all of these DC converters have a maximum capacitance load rating (some models as low as 47uF, some as high as 1000uF, average around 220uF) which is not something I'm accustomed to seeing, so I'm trying to better understand what's going on with these.

I've looked at about ten datasheets from different manufacturers and can't find a definitive answer to the filtering question. Most of them don't mention filtering at all. Some of them mention the module has an internal capacitor. Some of them recommend putting small capacitors on the input and occasionally the output, but nothing higher than 10uF.

For example, here's the Traco application notes recommending using inductors and capacitors on the input to meet EMI specs from EN 55032:
for instance: https://www.tracopower.com/media/793/download

So the question - what should I be doing here? This isn't medical or test equipment, just 40-year-old analog audio circuitry, which doesn't appear to be a primary market for these companies so I don't get the sense that their datasheets and recommendations are written with audio in mind.

- Should I really go light on the filtering when I use these, or am I misunderstanding the "max capacitance load" spec?

- Is the LC circuit from the app notes of any use to audio?

- The modules will generally go on a sub-board, so should I be doing anything to decouple the rails once it gets to the main signal board?

- And how should I handle grounding? The converter has its own isolated ground. I've seen some applications where the DC power (+V and gnd) is connected directly to the DC converter and then signal & chassis ground are connected to the DC converter's isolated (common) ground. Is this the right way to do it?

[imJonWain]

I was researching the same thing awhile back and couldn't find a whole lot of info either so I'm all ears.  I ended up making boards with space for both a 1W DC-DC brick and the ac multiplier setup to experiment with.  I kept the DC-DC and supporting circuit in a corner by itself connected to the main circuit ground at a single point.  I also used a jumper to play with isolating the DC-DC input and out grounds.

The capacitive loading spec is real, the inrush current can send the module into protection or kill it (I did this while experimenting lol).


I need to revisit my notes but I did the PCB as per below.  I remember without the inductors and just those caps there is about ~100mV of ripple at a frequency above audio.  I'll pull the board out tonight and play with it.  One manufacturer I remember had a much better data sheet than the others with more info about selecting parts.  CUI, I think? I'll see what I have saved.




AMZ has a short article on a similar module.
http://www.muzique.com/lab/24v_iso.htm

Murata app note.
https://www.murata.com/-/media/webrenewal/products/power/appnote/dcan-58.ashx?la=ja-jp

[antonis]

Never dealt in depth with SMPS but at a first glance you'll have to deal with high frequency interference, maybe via a HPF (of 2nd or higher order..)




Of course, 100/120 Hz ripple is also signicant so it will require the usual PSRR precautions for circuit amplifiers..

[Rob Strand]

. But all of these DC converters have a maximum capacitance load rating (some models as low as 47uF, some as high as 1000uF, average around 220uF) which is not something I'm accustomed to seeing, so I'm trying to better understand what's going on with these.

The common reasons are:
- Start-up issues.  The converter won't start as it starts up under a short.
  Time scales are 1ms to 50ms so adding inductors doesn't fix the problem.
- Feedback stability issues, which can often be fixed with an output inductor.
- Reliability issues due to start-up stress.

I've looked at about ten datasheets from different manufacturers and can't find a definitive answer to the filtering question.

LC filtering is a pretty common way to knock out some of the output ripple.

You can also add active filters.  Simple implementations with one transistor introduce a voltage drop.  Active filters don't need large caps on the output.

Beyond that is to use linear regulators.

For example, here's the Traco application notes recommending using inductors and capacitors on the input to meet EMI specs from EN 55032

All to do with EMI and doesn't help output voltage ripple at all.  However ...   

The input filters can prevent current pulses on the input causing noise problems.  It's not uncommon for ripple currents to cause more problems than the ripple voltage on the output.   The ripple current causes the 0V/ground wires to become noisy.

The cap from input to output can stop some types of EMI getting into the local audio circuits.

- The modules will generally go on a sub-board, so should I be doing anything to decouple the rails once it gets to the main signal board?

If voltage ripple is a problem it's best to get rid of it at the output of the DC-DC converter.   Any additional decoupling would depend on the circuit itself, to some degree independent of the fact you are using the converter.  Larger systems might use local LC filters for extra insurance.

- And how should I handle grounding? The converter has its own isolated ground. I've seen some applications where the DC power (+V and gnd) is connected directly to the DC converter and then signal & chassis ground are connected to the DC converter's isolated (common) ground. Is this the right way to do it?

There's no right way to do it.  There's definitely ways that reduce the risks.  Connecting the input and output grounds together means it's harder to remove problems caused by input currents.   If you wire the DC DC input supply directly to the input supply and with it's own tracks then the input current pulses  don't get onto your grounds.   If after that you choose to wire the 0V output to the system 0V then there's little risk of the input current pulses causing problems.   If on the other hand if you bundle the input 0V and output 0V together in the one place then you are promoting problems in that the current pulses can start making you ground noisy.   Some systems you don't care or you can get away with it, and that simplifies the layout, but for audio it's a risk.

Another source of noise is electromagnetic interference from the module getting into the audio circuits.  The worst case probably being from fields from the inductors and transformers in the converter.   Unless you know what a particular converter is like it's safer to keep the modules well away from sensitive circuits.   You can only work out how far to push the issue by experiment and if you use a different converter it could be completely different.   I have worked on non-audio products where unshielded inductors on motherboard and on converter modules created noise problems.

[imJonWain]

Murata had the better datasheet and also some helpful application notes on their site.

On my board using Traco PN: TBA 2-1223 with a .47uF input cap, a .47uF output cap per rail close to the module and a 22uF cap per rail a bit away.

I don't hear any noticeable noise worse then a normal supply would give.

Ripple measured with probe via paperclip method. ~35mV P-P at ~90kHz

[box]

Murata had the better datasheet and also some helpful application notes on their site.

On my board using Traco

I don't hear any noticeable noise worse then a normal supply would give.

Ripple measured with probe via paperclip method. ~35mV P-P at ~90kHz

Hi,
this circuit was used in maestro ps-1a and works flawlessly 

box