Yes, it must be insulated from the chassis. 
The shielded jack you posted must have been wired wrong or defective.  It should have acted the same as the other did when it was grounded to chassis.
Also, the noise you experienced must be due to the dress of the wires.  Where you routed it through the back of the chassis must be a sensitive area.
An insulated jack mounted where the original speaker wire went through the chassis should not pick up any more noise than the original wires did in the same spot.

What do you mean by 'shielded jack'?  You should use a plastic type insulated jack so the bushing does not have any continuity with the chassis.
There shouldn't be any more problem with EMI than the existing speaker wires would have.  It may have to mount where the speaker wires route out of the chassis.
Not sure if the amp is capable of driving an additional speaker without disconnecting the internal speaker.

The 'original' zener was just a defective regular diode.  Someone very smart realized it could be controlled and utilized.  Like a lot of discoveries, I guess. 
Always analyze your mistakes. 

Those resistors are shown as 1 ohm on the schematic.  If so, you have idle currents of around 88mA per output device.  That seems kind of hot, 10 to 20mA would be a better target.  Adjust bias to get around .01 to .02V across each of those resistors.
Then again check for DC at the output.  If it stays around the 10mV you have now, that is good.

With no load connected, is there any DC at the output?  What is the DC voltage drop across each of R737, R738 ?

How much each output device conducts is related to the bias circuit.
As you have made modifications, the stock schematic does not show the whole picture.
Post a schematic that includes all your circuit changes.

Now the edits have rendered my question redundant. 
In any case, your answers in posts #2 and #4 are all correct.

Yes that is it, touching 2 component leads with the same probe.
But also, the meter is reconfigured as a short when you are in the current (Amps or mA) ranges.  That is why we must 'break' the circuit and insert the meter in series to measure current.
If you do a current measurement, then forget and try to measure voltage with the meter still set for current, you are essentially putting a piece of wire across the points your 2 probes connect to.  Hopefully the meter fuse will save the piece of gear you are working on.

Good work.  Mainly I wanted to impress the idea of always asking 'from where?'
Whether you are talking about driving distance, or voltages, the point of reference should always be the first thing established.
With voltages, it is usually assumed to be 'with respect to ground', but it doesn't hurt to ask.  And it's very helpful to be able to juggle the reference point around in your head when you are visualizing a circuit.

Quote from: saturated on March 24, 2025, 02:10:57 PMI'm thinking it has something to do with my meter's (internal resistance?) 

Yes.  You could look it up, or you could solve for X: 
8.76/12.80 = x/(x+6)

edit:  my mistake, should have read 13.8, not 12.8, see below.

A nice visual reminder that voltages are always in respect (or reference) to something else.  In this case it was ground (common) but it does not necessarily have to be.

For bonus points:
1) what are voltages at A and C with respect to B.
2) what are voltages at B and C with respect to A.
 

Quote from: saturated on March 19, 2025, 08:41:56 PMI got to thinking I guess these guys flash sixty times a second. 

The LED elements run on DC.  Under the plastic cover of the bulb is a circuit board with the electronics.

Here's one version of schematic.  4 transistors, driver transformer, output transformer. 

Quote from: DPM309 on March 15, 2025, 11:57:50 AMThe schematic shows them as 2N5988 (PNP) and 2N5991 (NPN). The markings on the transistors are TIP42B (PNP) and TIP41B (NPN). The ones I purchased were TIP42B and TIP41B which are good replacements.

I responded to your post over at MEF forum, but will include it here for others who may be working on these amps.
The TIP41 and 42 have BCE pinout.  The original 2N5988 and 5991 have ECB pinout, so the TIP's will not work as replacements.
There are fake datasheets out there showing ON Semi 2N5988 with TO3P case and incorrect pinout, but those transistors do not exist.
The real 2N5988 and 5991 were TO-225 case with ECB pinout.

Edit:  received response that this amp had been working and came with TIP41 and TIP42 outputs.
So can only conclude there are more than one version and be sure to check where the traces of the output devices are going, as there seems to be only 1 version of schematic available, and it does not match the later version.

Quote from: J M Fahey on February 20, 2025, 05:42:01 PMThat´s why each of the channels is already BTL, you can´t undo that.

Now if you wire both BTL amps in parallel, you can get twice the current but peak voltage does not change.

I'm still wondering if Ch.1 and Ch.2 can be put together in regular bridged configuration, rather than the parallel PBTL.
If so, it would give full chip output into 8 ohms (rather than the 2 ohms of PBTL).
The reason I raise the question is that if it were possible, that may be what the Pandora is doing and how they get that high power figure into 8 ohms.  If so, I guess the amp would be marked 'minimum load 8 ohms'.  (edit:  minimum load 8 ohms confirmed here:  https://lichtlaermaudio.com/shop/pandora )