[sdiy] Signals leaking into the PSU?

[Neil Johnson]

Hi Mattias,

> Would it be worthwhile to completely avoid using the ground for anything else than a signal ground reference, and power all current-drawing circuits from the +/- supplies? If ground would be needed here, it would be a buffered ground created by an op-amp buffer that draws its currents from the +/- supplies, not from the ground it buffers. (Does anyone do it like this?) :-)

No, you still need power ground for power circuits.  A rule of thumb I
generally start with is N+1 ground lines, for N power lines.  So in
Eurorack for example, assuming we ignore the +5V rail, we have 4 power
lines (2 @ +12V, 2 @ -12V), so we would need at least 5 ground lines,
ensuring the impedance of the ground starts at a low level.  Then
measure and determine if more ground lines (lower ground impedance) is
needed.

Of course there are other issues at play as well.  As well as a common
for the power circuits, we also need a signal common (because
unbalanced) and also what to do about the panel metalwork?

There are generally two steps to sorting noise out:
1) Don't generate it in the first place (see constant current LED chains)
2) Minimise the spread of what's left (so put RC filters on the power
inlet to keep the AC noise current in a local loop and not spreading
everywhere else, and for really noisy parts of the circuit add more
local noise control)

>> As Roman suggested earlier IIRC, a much better way to drive LEDs is a
>> constant current chain between the two supply rails, with switches
>> (electronic or mechanical) across the LEDs to short them out when you
>> want them off.  This has two distinct advantages:
>> 1) it keeps the current draw constant, so no sharp edges of current on
>> the power lines;
>> 2) it keeps noise out of the common ground line.
>
> Is there any recommended go-to circuit design for implementing a good simple constant-current LED driver (preferably one with CV control of brightness as well, for non-binary indication)? Feels like important circuits like this would be great to have "standardised" in the community. Oh, I'm then assuming a 25yr+ EE design that is guaranteed to be "better", not the "bad" ones that other people rave about.  O;-)

LOL!  Yeah the simple on/off is either a mechanical switch or
transistor.  There are plenty of example circuits in mixer manuals.
Again: read widely and weigh up the pros and cons of the various
solutions, ONCE you understand what the original circuit was doing.
Variable brightness is definitely more involved: one solution I have
used in the past is to drive the LED from two opamps: one for the
level control and one for a local virtual ground.  This keeps LED
noise out of the ground, and if you're driving the LED with fast edges
you can isolate the opamp in its own power domain.

> The constant-current designs for LED-bar level meters found in mixing console schematics are nice, perhaps the simpler circuits above are hiding there as well?

Go and explore.....

>> There is a veritable treasure trove of design knowledge in the form of
>> analogue mixing desk schematics and service manuals.  Hundreds of
>> them, from Mackie, AHB, Soundcraft, Neve, SSL, Studer, DDA, and so on.
>> Many of them can be found with a little bit of googling.  And you'll
>> see how to switch LEDs quietly, how to power opamps for low noise, and
>> how to distribute power in a large bussed audio system.
>
>
> Darn, need to switch (silently!) to googling instead... :-)

For a laugh try this:
https://hottconsultants.com/techtips/maxemission/

:)

Neil