Can either of you remember from which seller you bought the comparators if not pulled from a spare L.A. PCB?
I pulled my replacement 1NB4-5036's from a dead LA board.
Looking at MarkL's photo with 1NB4-5036 removed I don't quite get the stated pinout for pin 29 which looks like a comparator input not GND. These IC's seem to have 9 external inputs and one common (to DAC) input for threshold reference. In the 54622D would this leave one spare comparator that could be bodged to replace a dead comparator input? If 1NB4-5036 is no longer obtainable would a MAX9693 piggyback bodge be reasonable? Of course either of these solutions are going to be awkward if timing is critical...
You are quite right on pin 29. It is an input. And I was also wrong on the test clock signal level. It's actually 50mV and not 5mV. I should have realized 5mV is too small for a clock input as I was typing it in. Embarrassingly, I already found these errors months ago and failed to post the corrections. Very sorry about that. I will post an update to that thread.
I like your idea of substituting an unused comparator section, but unfortunately it sounds like you need two. So depending on where you have dead channels and working comparator sections, you may end up with jumpers between the two chips. Don't let them get so long that they compromise high frequency performance.
I don't think the MAX9693 would work. From the scope traces in the referenced thread, the 1NB4-5036 outputs are >0V, and the MAX9693 has an ECL output (<0V). Maybe the 54622D biases the output differently. The MAX9693 may also have different delay characteristics which could cause skew problems with fast clock inputs. Plus the self-test, assuming the 54622D does it in a similar way to the LA cards, is going to fail.
On this MSO the dead channels will not toggle for any threshold setting when tested with probe comp. output and logic pod, they have a unloaded DC output of -2V for one and -200mV for the other with +/-10mV on good channels.
Would you agree that the most likely failure is comparator internal input protection diode defective due to having been hit with excessive negative voltage?
If such is the case is there any chance of fusing it away (lab PSU between input and negative supply rail) without killing the whole I.C?
It could be the input diodes. As I recall, the logic analyzer cards are fairly well protected before the signal gets to the comparator. I suppose someone could have put some high voltage signals directly into the connector (DIY probes, or whatever).
What is the swing on the good inputs and outputs?
If the seller is allowing you to dig in a little, you could try lifting the input and output legs on the bad channels and try your substitution idea above. I have little doubt that you can get your hands on a 1NB4-5036 either now or when more become available, but if there is a problem with the digital acquisition input, that is going to be a bigger problem to repair. So, if you really like the scope, the idea is to make sure the acquisition input is ok.
I would not bother to try to fix it by trying to blow the dead protection diodes. You could blow up more than you wanted and kill the whole chip.