Damping transformer ringing (Zobel network question)

[ccaudle]

Signal + is referenced to signal -.

The reference point for audio common is at the power supply output 0V.

This is one of the statements which caused confusion. Abbey and I were talking at cross purposes a little bit there. From a system standpoint, the downstream differential receiver considers as the signal the difference between signal + (aka hot) and signal - (aka cold). In that sense the hot output references the cold output.

When I referred to reference node and what the hot signal referenced, I was speaking in terms of circuit design. Since this is a tube design, the output signal is controlled by the grid to cathode voltage, so the output signal reference node is determined by the cathode connection. That is probably going to be called "gnd" in the schematics, although gets a little bit trickier to think about when you have a circuit powered with negative voltage rail, or an op-amp follower which only has connections to the power rails.
But in the case of most tube circuits it is pretty straight forward, the output signal is controlled by the grid voltage as measured referred to the cathode circuit. Not some idealized concept called "gnd," not some random point in the chassis which has a conductive path back the cathode, literally the cathode pin of the tube is the reference point for the grid which controls current through the tube. So the point where you should connect the resistor-capacitor combination which will then connect to pin 3 should be right beside where the cathode resistor connects to gnd.

Either way the receiving end pin 3 sees 27R to chassis.

So the point of the pin 3 connection is not the impedance to chassis, but the impedance back to the point where any voltage change will cause a change in the pin 2 output signal.

Say you connect the device to another device with XLR cable, and there is a slight difference in protective earth voltage which causes a slight difference in the chassis voltage between the two pieces of equipment. That voltage difference will drive current through the shield of the connecting XLR cable, then through pin 1 onto the chassis, and around the chassis back to the protective earth connection where the power line comes in (assuming that the device is not a double insulated design with no power entry protective earth).
The resistance and inductance in that path will cause a voltage to develop at pin 1. If you connect pin 3 components to that point, now that voltage is directly on the pin 3 signal, but not on the pin 2 signal, so it is differential, to the downstream input just part of the signal.
If instead you connect the pin 3 components directly beside the cathode resistor connection, now any voltage generated by that interference current causes the entire circuit gnd node to ride on that noise voltage, so both the pin 2 signal and the pin 3 signal have that same noise voltage. Since it is common to both signal lines the downstream receiver will be able to reject that noise.

 

[nsalis]

There is definitely some confusion. You want to make an impedance balanced output from an unbalanced signal. Audio common should not go to pin 3. It should go to pin 3 via a 27R resistor. Pin 1 should go directly to chassis at the XLR. At one point in your system Audio common will be connected to chassis (usually at or near the power supply).

Edit: Here is a simple picture to illustrate the difference.

Cheers

Ian

We learned from Bill Whitlock that a complete duplication of source impedances on the two output legs is necessary.

In order for the impedances to track at lower (as in 50/60 Hz mains) frequencies, the same value coupling capacitor must be in series with the resistor in the grounded side. Otherwise, the interface CMRR will be degraded at mains frequencies - where the highest possible rejection is usually needed.

See attached dwg.

Nick Salis

 

[abbey road d enfer]

In order for the impedances to track at lower (as in 50/60 Hz mains) frequencies, the same value coupling capacitor must be in series with the resistor in the grounded side. Otherwise, the interface CMRR will be degraded at mains frequencies - where the highest possible rejection is usually needed.

This is debatable. In a live situation, mains frequency is seldom an issue, and in many cases, a HPF can be applied to minimize the issue.
I've found that keeping tight CMRR is more important at "buzz" frequencies, such as those generated by the dreaded phase-controlled dimmers.
However, considering the low cost of adding a capacitor, I don't see any reason not to do it (except maybe space).

 

[Gold]

Okay, my conceptual error was thinking chassis was the reference 0V. It was a disconnect because I know chassis is not audio common. Once I was able to picture audio common as the reference voltage it made a lot more sense.

 

[ruffrecords]

This is debatable. In a live situation, mains frequency is seldom an issue, and in many cases, a HPF can be applied to minimize the issue.
I've found that keeping tight CMRR is more important at "buzz" frequencies, such as those generated by the dreaded phase-controlled dimmers.
However, considering the low cost of adding a capacitor, I don't see any reason not to do it (except maybe space).

The only other disadvantage is it raises the -3dB point of the output stage and its load by an octave and potentially increases the distortion in the electrolytics. Generally not an isue with 10K bridging loads but if you want this output to drive a 1K load you might have a problem. The simple solution is to double their values.

Cheers

Ian