Multiple balanced output switch - shielded cable noise

  Hi!
I'm working on a summing mixer with a 3-mode switchable output, here a simplified schematic :


  1st mode is a that 1646 balanced output
  2nd mode and 3rd mode are two different output transformers

At first I used 3 shielded cables and linked the 3 shields together to the chassis, it created some kind of HF loops (something that I couldn't measure with my scope, maybe some GHz..) in the output signal as well as on the ground and PS rails. When I removed the link the HF noise disappeared.
  I guess it's a kind of parasitic interaction caused by the capacitance of the cable. Of course the solution is to not link the shields, but could you help me to understand the phenomenon when the 3 cables' shields are tied together?

There should be no problem from tying all shields to chassis ground.

If that causes instability, the chassis ground current may be corrupting the signal somehow.

JR

How long are these cables? I'm wondering how much capacitance they might have signal-to-signal and signal-to-shield.

Try grounding these cable shields one at a time. Glancing at the 1646 datasheet, it should drive just about anything (or worst case not act bad while trying), but I wonder if some extra capacitance caused by grounding the shield gets reflected by the transformers to the outputs of the other two opamps, and causes oscillation. Presuming that's the cause, a fix would be a small resistor (maybe 100 ohms) in series with each primary.

I would think the 1646 is not the culprit.
Transformers have a primary-to-secondary capacitance that loads the output of the stage that drives the xfmr.
What benb proposes would probably avoid the unstability, but would also ruin the LF performance.
Try inserting a choke between the stage and the primary.
Check the schemo on this page
http://www.jensen-transformers.com/wp-content/uploads/2014/08/jt-11-ympc.pdf
they put ferrite beads, for the very same reason.

The cable are not long (15cm max), and one interesting thing to note is that the oscillation occurs when the shields are tied together, even not to chassis ground or anything else..

I would think that it is the stability of the Balanced OpAmp driver is the problem.  You may want to add some output resistors to each side as a first try, say 50-100 ohms. The stray C  of the cable may be causing the instability. The resistor build out will isolate the stray C.
Duke

Audio1Man said:

I would think that it is the stability of the Balanced OpAmp driver is the problem.  You may want to add some output resistors to each side as a first try, say 50-100 ohms. The stray C  of the cable may be causing the instability. The resistor build out will isolate the stray C.
Duke

Click to expand...

Thank you! However I don't really need to shield theese small pieces of wire, it was more to try to understand what was going on...

saint gillis said:

Thank you! However I don't really need to shield theese small pieces of wire, it was more to try to understand what was going on...

Click to expand...

Which pieces of wire are you shielding exactly? The signal wires running to the drivers, the signal wires running from the drivers to the output / transformers, or the shields of the output connections?

Audio1Man said:

I would think that it is the stability of the Balanced OpAmp driver is the problem.  You may want to add some output resistors to each side as a first try, say 50-100 ohms. The stray C  of the cable may be causing the instability. The resistor build out will isolate the stray C.
Duke

Click to expand...

I am of a totally opposite opinion. Output resistors are already built-in the 1646. But transformers do have capacitive coupling between primary and secondary (unless built with a Faraday shield - very seldom). The effects of pri-to-sec capacitance are well known, introducing capacitive loading of the driver stage, load-dependant HF CMRR, asymetrical HF response.
Anyway, the proof can easily be demonstrated, by leaving the shields connected and disconnecting in turn the drivers.

squarewave said:

Which pieces of wire are you shielding exactly? The signal wires running to the drivers, the signal wires running from the drivers to the output / transformers, or the shields of the output connections?

Click to expand...

the output connections

abbey road d enfer said:

Anyway, the proof can easily be demonstrated, by leaving the shields connected and disconnecting in turn the drivers.

Click to expand...

Indeed

saint gillis said:

squarewave said:

Which pieces of wire are you shielding exactly? The signal wires running to the drivers, the signal wires running from the drivers to the output / transformers, or the shields of the output connections?

Click to expand...

the output connections

Click to expand...

Are they tied together at the chassis or do you tie them together somewhere else and then run a single wire back to the chassis? It could make a difference because you will be changing the inductance of the shield conductor and resonance at high frequencies is almost certainly going to be affected by inductance.

Also, try plugging it into another piece of gear. If you terminate the remote + and - with 600R resistor and leave the shield unconnected at the remote end, do you still get HF noise?

Otherwise maybe your contact with the chassis isn't good enough? Is it anodized aluminum?

There is definitely something wrong because you should not get noise like that. Maybe a small amount if there's a lot of EMI like you live next to a cell phone tower or something like that but it should not be enough to really see on a scope unless you zoom in to 1mV / div.

Then again, my scope doesn't do anywhere near GHz so I don't suppose I would ever see such a thing if it existed.

As I said before I get noise even if the shields are tied together but not to the chassis!

There is are vital pieces of info missing.

What are you feeding?  Is it an unbalanced input?  How are you seeing the noise (which sounds like oscillation)?  Scope?  If so, where is the scope connected and where did you connect its ground?

Ok I 'll share a more detailed schematic

Are the shields connected to the destination?
This could be a simple ground loop.
If there is a fully balanced interface there is no need to connect the shield at both the source and the destination,  and many good reasons to only connect the shield at the source.

Hey guys, please read it all

1st of all I'm not asking for a solution, I wont tie these short cables' shields together, I shared this case only for the interest of understanding what's going on!

Secondly, the oscillation occurs when these 3 shields are connected together, even if not connected to any ground..