[sdiy] Module power - regulated or filtered (passively)?

[René Schmitz]

Hi Brian,

On 06.01.2020 02:10, rsdio at audiobanshee.com wrote:
> The idea of a couple of dozen modules that each have their own ±10 V
> regulators somehow seems wrong. I realize that it’s fully necessary
> for some modules, but is it really necessary for every kind of
> module?


VCOs are the likely candidates that benefit from having their own
regulators. (Especially if the rails are used as reference voltages...)

I have used seperate regulators for VCOs in a modular (+-15V System, but
the VCOs run at +-12.) My motivation was to eliminate residual
soft-syncing via the supply, and that worked fine.
I don't have much blinky-blinky-stuff in that box, so that was not a problem
for me.

> A good op-amp should have an excellent power supply rejection ratio,
> meaning that a nominal amount of power supply noise should not appear
> in the audio (or CV). In addition, you can’t really get 20 V
> peak-to-peak, input to output, from an op-amp that only has a ±10 V
> supply, so there’s a real advantage to giving the op-amp a few Volts
> more than it’s peak signal. I realize that most modular gear rarely
> deals with 20 Vpp (±10 V) signals - it’s usually more common to see
> 10 Vpp (±5 V).
> 
> So, I’m thinking, what’s wrong with simply connecting the ±12 V
> supply from the Euro power to the op-amps without precise regulation
> (linear or switched)? I’d want to at least have LDO diodes for
> protection against polarity reversal, and passive RC filtering to
> reduce the noise cheaply.

Opamps have terrible power supply rejection at higher frequencies. They
might reduce the fundamental of your 50Hz / 100Hz hum, but not the
harmonics. Ironically the more filter caps you add, the more you push
the ripple to higher harmonics. You get a nice buzz, instead of a low
hum....

> On the subject of passive filtering of the supply, even a linear
> regulator will dissipate 4 mW per 1 mA of current draw. So, if a
> module draws 20 mA to 30 mA, then on-board linear regulation will
> dissipate 80 mW to 120 mW. In contrast, using 10 Ω resistors for
> passive filtering would only dissipate 25 mW for a 50 mA draw. 10 Ω
> resistors would not even catch up to linear regulator dissipation
> until the module is drawing 100 mA, which is quite a bit for most
> modules. Using a separate passive filter for each op-amp might mean
> 200 mW dissipation for a 50 mA module, but that’s still on par with a
> linear regulator at the same current draw.


A passive filter won't be able to filter out slow fluctuations caused by
say an LED blinking at 1Hz. So there the regulator is clearly desirable.
Not to say passive filters local to a module are useless, because a
10ohm + cap can remove much of any higher frequency hash from the rails.

Another strategy might be to isolate the polluting modules, and have 
local regulation for them.

> I’m not going to think about switching regulation options, because if
> a module has the board space and needs a switching regulator then
> that’s probably not the kind of module I’m asking about in this
> thread.
> 
> Another question is whether the passive filter should be referenced
> to ground if the op-amp doesn’t even connect to ground. It seems
> possible to just put 10 Ω of resistance (or less) on each supply pin
> of the op-amp with a single capacitor between the + and - power pins.
> Some op-amp configurations do reference ground, or at least some sort
> of virtual ground, so it might make sense to reference a pair of caps
> to that ground, although it shouldn’t be necessary.

Your input and output signal is always referenced to that ground,
ideally near that opamp. I would put the caps to GND near the opamp.
That way the loop which any changing current passes is shorter and will
only involve one rail. (If you place the cap between the rails there
will be some current drawn from the opposite rail as well. And if you
place these 10ohms in there that will shake up the opposite rail every
time your load draws some current.)



Best,
  René

--
synth at schmitzbits.de
http://schmitzbits.de