MJL21193/94 replaced mica insulator with silicone and now a hum and DC

and proper (not CPU) heat sink grease (thermal compound)
Hi, what's wrong with cpu thermal paste ? It's it because it's viscosity is high?.
I have used Artic mx-4 that i mix with a tiny amount of another thermal compound the result it's a thermal paste not to thick not to " runny ". I use it both on my CPU, GPU, and transistors with great temp performance.
I'm genuinely asking why not CPU thermal paste ?.
- Bruno.
 
Right, but my point is that you can't isolate one characteristic from the rest. An electronic part or chemical each has a collection of characteristics.

You've got to balance the group of characteristics in each application to determine which is the most suitable. Like everything in engineering, it is a compromise to get the best overall outcome. In order to get the very high thermal conductivity of computer pastes you have to change the mix. Everything from dielectric constants, conductivity, voltage breakdown. Everything has to be considered. In addition the physical properties (like viscous fluid) have to be considered. Even something like mold growth can be important. The original thermal compound for power transistors and modules has been highly engineered over decades. You can't really do better without introducing some unknowns, or really high costs.

With computers you have very low potential differentials, and no signal voltage on those surfaces. The opposite is true in most industrial and audio applications. So the design of the thermal compounds have very different driving factors. Also your mounting pressures are very different. Anything caught between the two mounting surfaces has to equalize, and at high pressures you have the power to distort the metal or plastic body of the component. I had seen broken cases due to mounting pressure, and warping a metal (TO-3 or TO-220) case can either break the die inside, or cause it to detach from the heat spreader inside. That is invisible to the person mounting the part. Motorola came out with an excellent application note on mounting power semiconductors. I think every single technician should read it at least once. On Semi has it on their site.

I still use Wakefield 120. Industry standard, it lasts and most of our thermal expectations are predicated on the use of Wakefield 120 or an equivalent product.
 
  • Thank You
Reactions: xXBrunoXx
Yes, one can measure 7 meg ohms but still have a dielectric somewhere that is on the edge of flash over. Just because something isn’t conducting doesn’t mean it won’t with enough applied voltage. I have used conductive pastes carefully and had things stay isolated, but I’m not doing that on an amplifier running on +/-100 volts.
 
Hi delverboy,
Can you test the transistors out of circuit? I want you to make sure you didn't damage the actual transistors.

It's hard to describe proper mounting pressure, but I would make them snug plus 1/4 turn. That is not cranked down, the transistor should not move laterally with light to medium pressure. You'll see the grease squeeze out form the edges when you're close to the right pressure.

About grease. More is not better! You want a thin, even coat on the transistor, then on the insulator or heat sink mounting area. I use a #2 artists brush to apply it. Take care to keep everything clean, no bits of dust or anything else.
 
I haven't removed anything yet. Am about to take some readings while powered up and see if I can find the culprit. Failing that I will probably just replace and hope.
I did a diode test of all the transistors and didn't see anything that stood out.
The amps came with paste but I used that up and was planning to use the CPU paste I have. Thanks for the tip recommending not to use that.
 
Unmounted you may find those outputs heat up rapidly and destructively. Maybe temporarily use the CPU paste applied in a very thin, even coat and just snug them down for troubleshooting.

Once you figure things out use the proper grease and mounting techniques.
 
Checked vbe on Q9,Q10 and they were both equally .614v. Q6,Q8 were both equally .604v.
I removed the board from the heatsink and checked components for heating. Q12 quickly got hot, Q11 barely warm. Everything else barely warm.
Think I'll replace Q12 and see what I get.
 
Hold on.
The current through Q12 is going somewhere and you know it isn't Q11. Not unless the DC offset is very positive. So if you have a low DC offset, the voltage across Q11 and Q12 will be similar and they would run around the same temperature. If the drop across Q11 is low, it may be leaky or shorted.
 
Boy, I searched with Google and ended up with all kinds of incorrect stuff.

Here is the link to On Semiconductor : https://www.onsemi.com/download/application-notes/pdf/an1040-d.pdf

This should be required reading, I have it in a databook from Motorola. Am I really that old ???!

Edit:
Now there is another one, AN1083. I haven't read it yet - but not on On Semiconductor. I wonder if it is legit.
 
  • Like
Reactions: alain98
The differential pair wants to make it's inputs equal, same as an op amp. So Q1 and Q4 want to see their bases at the same voltage. How close depends on their match.

Measure the voltage drops across R22 and R23, they should be very nearly the same. This voltage divided by the resistor tells you the bias current. If you have a negative offset, Q4 will conduct less or be turned off. All the tail current would then go through Q1 and that would turn Q6 on harder which would tend to raise the base voltages on Q9 and Q10. Q5 and Q8 are constant current sinks. The voltage across R13 will tell you how much current it is conducting. It won't change much. If Q10 or Q12 is conducting too much, Q9 and Q11 will fight them, drawing far more current than they should through the entire output and driver stages.

So, follow the current flow. Check to make sure the transistors passing current have base-emitter voltages that would explain it (clearing those parts). If you have a transistor conducting a lot with a zero, reversed or very low base-emitter voltage, it is probably defective.