[sdiy] How to design out "usb noise"?

[mskala at northcoastsynthesis.com]

n Sun, 4 Feb 2024, brianw wrote:

> Granted, if the synth is a USB Device, then self-powered is a good
> option, but there is also a good argument from the users' perspective to
> design for Bus Powered. At the very least, you have one less wall wart
> to keep track of when packing for a gig.

That is not the scenario contemplated by the original question nor my
answer to it, both of which made clear that we were talking about a
synthesizer with its own power which connects to USB only to exchange
data.

> I do not think that it is possible to optically isolate in two
> directions. USB uses differential signaling, so both the D-/D+ are
> driven by the Device at some times, and by the Host at other times. In
> fact, that's one of the reasons why a Self-Powered Device still has to
> wait for VBUS, because it can't step on the Host driving those Data
> lines.

It's easily possible, by the technique I already described:  a transciever
on one side, powered by the USB bus power, and optoisolators (one per
line, and each line going in only one direction) between that and the rest
of the circuit, which is powered by its own power, which does not come
from the USB power.  The "transciever" is basically a glorified tri-state
buffer that can either drive the bus, or listen to it, with the proper
differential signalling - as well as detecting and reporting things like
the state of VBUS.  It connects to the rest of the circuit at standard
logic levels, several lines in each direction but each line going in only
one direction.  So insert an isolator in the appropriate direction in each
line and you're done.

The PIC24FJ64GB chip I used in my USB module design has a built-in
interface for connecting such a transciever.  In that particular case the
external transciever interface doesn't work, as a silicon erratum, and I'm
not sure exactly why one would want to use an external transciever when
there's one built into the microcontroller anyway.  But the feature
certainly exists and was advertised as something the chip could do, before
they errata'd it.  The Texas Instruments TUSB1106 is an example of a chip
that serves in the "external transciever" role; I don't know if it in
particular would work with my PIC24F chip in particular, if the interface
even worked at all, but the point is that this is a fairly ordinary
function and not something wacky I dreamed up.

Whether it would be a *good* idea to use an external USB transciever with
an optoisolator on every line between it and the rest of the circuit, is
another question, but it's absolutely possible and not particularly
unusual or surprising.  One reason someone might want to do it would be
what we're talking about here:  an effort to have as little circuitry as
possible be powered by the USB power.

I would rather do what I said in my first response to this thread:  have
an entire USB-MIDI to DIN-MIDI translater be on the "USB" side of the
isolation boundary.  Then there is only one bit of serial data that
needs to cross the boundary, in one direction, and the stuff on the
"synth" side of the boundary is just the same as in any ordinary DIN-MIDI
controlled synth.

With NO power or other galvanic connection between the two sides.

> If there is a chip that optically isolates the bidirectional
> differential USB Data lines, I'd be fascinated to look at a data sheet
> to see how it works.

I'm not suggesting that the optoisolator would connect directly to the USB
data lines with nothing else whatsoever on the USB-powered side.  I'm
suggesting that something on the USB-powered side translates the signal,
including by splitting the two directions of signal flow, into an
interface that can connect to an appropriate number of ordinary
unidirectional optoisolators.  I daresay that's how the "USB optoisolator"
as a packaged product, that someone else mentioned in this thread, works.

-- 
Matthew Skala
North Coast Synthesis Ltd.