[sdiy] Eurorack Grounding Question

[Michael E Caloroso]

My career as a system engineer gave me a lot of valuable experience
with grounding practices.  I've read plenty of references on grounding
practices in the control field and the audio field, and there's quite
a bit of common ground (pun not intended).

All grounds should be referred to as "references" (AKA "0V") with
respect to power source.

It is important to understand that the different references are not
identical.  Failing to appreciate this causes ground interferences
which manifests itself as noise in circuit signals, which can be a
b*tch to track down.  Diagnosing ground problems separates the men
from the boys.

Tony has the right approach for panels.  If I may clarify:

The power cable "earth" should be connected to the front panel and to
the sleeve of all sockets.  This reference should be referred to as
"chassis".  Neutral of the power cable should NOT be directly
connected to earth or anywhere else except the power transformer
primary (electrical standards regarding treatment of neutral varies
from country to country).

"Chassis" should be used as a shield for audio cabling between panel
and circuit boards.  Connect the the shield at the socket, but leave
the shield at the other end at the cable unconnected.

0V reference leaving the power supply should be referred to as "ground".

>From the power supply, all rails and 0V references should be
distributed to modules via a "star" pattern.  A power buss may be
convenient for assembly but this is a serial distribution which can be
potential trouble.  Power distribution was one of my solutions to cure
tuning problems in my old Minimoog.  As my studio system expanded, the
noise also increased.  Changing to a "star" AC power distribution
solved this problem.

In my experience I cannot trust ribbon cable interconnects for power
distribution.  The wire gauge in ribbon cables is not large enough to
handle large currents, and the crimp connection at the plug is not a
reliable enough connection.  Ribbon cable interconnects are fine for
logic signals where their inherent noise margins make them impervious
to flaky connections, but not for for critical voltages such as power
supply or CVs.

I have encountered malfunctioning systems using ribbon cable power
distributions.  The flaky crimp connector resulted in the digital
power rail dropping below optimal level.  That caused logic circuits
to malfunction.

For multiple power supplies in modulars with multiple cabinets, the
earth connections should be common between cabinets but leave the
power supply grounds isolated.

>From power supply entry at the module, "ground" separates into "analog
ground" and "digital ground" throughout the circuit board.

"Chassis" at the panel should NOT be directly connected to "ground"
ANYWHERE.  Brian's RC coupling is the universal technique for coupling
them.  I've seen systems with much lower resistance than 1M, and with
that high a value you may want to watch for excess current exceeding
the I^2 times R power product of the coupling resistor.

AC coupling at power pins of each logic IC is a long established
practice to shunt switching noise.  There is no established standard.
OEMs recommend a solid tantalum cap in parallel with a ceramic cap at
each IC (values vary depending on app notes - and the device), with a
large electrolytic cap at the power supply entry at each board.

Any high current high frequency switching such as LED pulsing is
notorious for dumping noise on the supply rail.  These circuits should
have their own supply rail decoupled from the source rail, usually
consisting of a series resistor (IE 100 ohm) with decoupling caps to
ground on both sides of the resistor.

Source signals with rapid transients - such as logic signals, clocks,
and LFO square waves - have a nasty habit of capacitively coupling
across circuit board traces and interconnect cables between panel and
circuit boards.  Put LFO square waves too close to an audio path with
the wrong impedance and you will hear audible clicks at the LFO
frequency.  No grounding solution will solve this (although current
loops are impervious to EMI).  Pay attention to cabling and circuit
board trace layout to avoid this problem.

EMI solutions involving balanced differential signals is a whole other
topic involving opamp CMRR and impedance matching.

The big problem is mixing modules of different makes that do not
follow the recommended practices.  Reference of sequencer clock and
control signals at the panel jack SHOULD be digital ground via a
socket with insulated sleeve, but all bets are off with other modules
that do not follow that convention.  Connect digital ground from one
module to another with socket sleeves that now connect to chassis, and
the potential for system noise arises.

A final note regarding XLR interconnects.  At the device, connect
analog ground to pin 1.  At the cable, connect pin 1 to to the shield
wire, but do NOT connect the shell (which is chassis) to the shield
wire or to pin 1.  This is the established practice long recognized by
the AES and the audio/radio industry.

MC

On 12/29/20, Brian Willoughby <brianw at audiobanshee.com> wrote:
> On Dec 29, 2020, at 01:30, Steven Cook wrote:
>> Hi, I'm designing a Eurorack module which uses a microcontroller. The PCB
>> has separate analogue and digital grounds joined at the power input
>> header, but I've realised since designing it that both grounds will also
>> be shorted together by the front panel as there are input jacks connected
>> to the digital ground and output jacks connected to the analogue ground. I
>> assume this would create an ground loop. Would I be better off not
>> bothering with separate grounds?
>
> Why do you have "digital" signals on the front panel? Aren't they all analog
> signals? Maybe you have a USB jack on the front.
>
> There are a few papers stating that separate grounds are pointless. Not
> everyone agrees on that. Perhaps the valid point is that it's still possible
> to do a bad job, even with separate grounds. I happen to think it's worth
> separating analog and digital, but you have to be really careful not to let
> signals cross ground planes because that's one way things can go wrong.
>
>
> On Dec 29, 2020, at 01:38, Tony Allgood wrote:
>> Create a third 0V connection (ground) from the power cable for the panel,
>> pot brackets and sleeve connections on all sockets. Don't use this for
>> anything else.
>
> Tony has the best advice. If you have a star connection for the various
> "ground" nodes, then there hopefully won't be any noise crosstalk between
> them. Keep them all separate until the point on the PCB where the power
> cable grounds them all.
>
> Consider this third 0V node to be "shield" rather than "ground" - maybe even
> use the "chassis" symbol to keep it separate from signal ground and power
> ground.
>
> When you connect "chassis" to "ground" you can use a high resistance in
> parallel with a capacitor. This will keep the shield near 0V without
> conducting noise into signal and power. I typically use 1 MΩ in parallel
> with 0.1 µF. I tried using an inductor, but had poor results with that
> approach.
>
> Brian Willoughby
>
>
> _______________________________________________
> Synth-diy mailing list
> Synth-diy at synth-diy.org
> http://synth-diy.org/mailman/listinfo/synth-diy
> Selling or trading? Use marketplace at synth-diy.org
>