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

[Spiros Makris]

Hello everyone,
The answers given are mostly true by themselves, but all those methods
address different problems associated with a power supply section.
So, trying to make the distinction here, the main tasks you need to
accomplice so that the module stays safe:
1. Guard against inverse polarity. Just a shrouded connector is not a
robust solution, you need electronic protection against it. MOSFET
transistors or shottky diodes placed in series are the best options in
most cases
2. Protect the module from an overcurrent condition. That could be a user
caused sort circuit or a failed part of the system, which is not caused by
reverse polarity. To be fair, in most sanely designed modules that is
unlikely to happen, so it could very well be omitted. AFAIK, most
commercial modules don't even bother. Fortunately, you can estimate the
maximum safe current draw with a fair amount of certainty, so you can pick
an appropriate fuse (most probably rated under 200mA) if you so wish.
3. Protect from an overvoltage condition. Again, this is a very unlikely
fault in most modular systems that follow one of the defined standards.
Even then, most of our electronics go up to 36V, therefore you need a major
screw up for things to smoke.


The functions that relate to how well your module will operate (regarding
noise, oscillator drift etc) are:
1. Provide regulated and reasonably "clean" supply voltages for the bulk of
the system's circuits. That includes opamps or other ICs that benefit from
maximum headroom.
2. Provide regulated, "clean" and  *accurate*  voltages that will be used
as a reference (tuning controls etc) or supplied to circuits with high
supply noise sensitivity. Usually those are 2.5V-5V.

A supply voltage is a DC voltage + various AC stuff you don't want there,
like 50/100Hz hum, noise from the regulators, high frequency ripple ("hum")
from switching mode regulators, slow fluctuations because of temperature
drift or other factors. Your goal is to put all of the problems within
spec, ie suppress them enough that they won't be noticed.
You should regard any output that doesn't come from a circuit dubbed as
"precision" a voltage that is only reasonably stable and with at least some
AC noise. So your case power supply might say 12V and come from a 7812
regulator, but you should never assume it to be *exactly *12V (or any other
measured value).

Opamps generally don't require much in terms of supply filtering because
they have power supply rejection (called PSRR- power supply rejection
ratio). Slowly drifting DC is no problem to them either because the gain
and bandwidth of various stages is always set in place with passive
components and is not depended on the power supply. Of course, the supply
voltage has to be larger than whatever you want to amplify. Other than
correct bypassing of each IC, it's rather pointless to do more for them in
most scenarios. If the board includes +/-10V (or 9V) regulators for other
reasons, it will indeed be better to use those. Otherwise, the benefits
hardly justify the cost in board surface, heat generation and efficiency
loss.

Single-ended transistor circuits will *always* be prone to supply noise.
It's always a matter of how much it matters in that specific context and
how much that noise will be amplified. For example, a single transistor
overdrive circuit will have a gain of 1 for supply noise. Essentially,
whatever ripple there is in the supply will be passed to the output intact!
Now in such cases, a long term stable voltage is of little importance- what
really matters is that we keep the frequency of those fluctuations low
enough that they will stay unnoticed (you really can't hear an 12V go to
11V within one hour or more). Therefore, we have two options: we either use
a regulator IC or a passive filter and our goal is to filter frequencies
over 50Hz, which is the lowest frequency of ripple/noise that could appear
in a system. Generally speaking, on low current draws (<10mA), a simple
resistor and capacitor can be satisfactory, but comes at a considerable
cost in terms of size, because a low resistor means you have to use a
pretty big capacitor for this to have any point. However, it can
theoretically have very low voltage drops. I would argue that in a system
with 24V total supply voltage headroom is hardly an issue, so using the
regulator IC is the best choice in technical terms for most cases I can
think of.

Control potentiometers and circuits related to CV are the most important
and usually the most sensitive. Unlike amplitude, you can detect even
slight detuning very easily, and the result can be detrimental. You have to
both provide a *precise* voltage (a "reference"), and a clean one (low
noise, in any frequency). Common examples are the tuning knob of a VCO, the
Vee supply of an OTA (especially when biased with a resistor and not a
current source), the Positive supply of an exponential converter made with
NPN transistors (supply fluctuations are translated to reference current
and pass directly to your VCO output frequency) etc. Those are all cases
where long term drift causes very audible results and makes you retune you
oscillators or appears as FM/AM modulation.
To set things straight, a resistor and a capacitor can only suppress some
high-frequency components but don't regulate at all. Common voltage
regulators don't do the trick either because they are not designed to be
stable over temperature, or over time. Your 5V reg can read 5.1V tomorrow
and still be within spec. Instead, you have to use a reference voltage
generator, that is advertised as "precision" and look at its spec sheet.
There you can find numbers that justify if it will be good enough, or you
need to aim for something even better. Depending on the IC used, those
voltages could also supply whole circuits and not just control or other
references. Always check with the datasheet to verify that the current draw
you need is compatible with the stability you are asking for.

There are many dimensions to this question and one has to get a grasp of
how a circuit is affected by a supply voltage. You can simulate for that by
putting an AC source in series with your DC voltage supply, grounding the
inputs (if there are any) and scoping the output.

Spiros



On Wed, Jan 8, 2020 at 7:46 AM Pete Hartman <pete.hartman at gmail.com> wrote:

>
>
> On Tue, Jan 7, 2020 at 10:16 PM Quincas Moreira <quincas at gmail.com> wrote:
>
>> Yes, Yusynth and MFOS modules for example use such protection, and I can
>> vouch that it does work, as I've burned a few myself when starting out due
>> to inverted opamps. The 10R burned before anything else could be damaged on
>> the boards. Erica Synths uses resetable fuses for the same purpose and it's
>> a better solution since you don't then have to replace the fuse, it just
>> interrupts the flow when excessive current is drawn, but doesn't burn out.
>>
>
> "Usually".  I've seen it done :-D   By others in this particular case.
> The board was in such bad shape by the time I saw it, I have no idea what
> they did to melt the PTC fuse.
>
> The problem I have with the 10R solution is that if it's a VCO and it
> draws any significant current *and* uses the rails for reference (which a
> lot of those using this method seem to do), it can lead to an unstable
> VCO.  It doesn't take much to be audible, and I have encountered this with
> "real" designs in the DIY space.
>
> Pete
>
>
> _______________________________________________
> Synth-diy mailing list
> Synth-diy at synth-diy.org
> http://synth-diy.org/mailman/listinfo/synth-diy
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://synth-diy.org/pipermail/synth-diy/attachments/20200115/f1a56dc1/attachment.htm>