alternative for STK3102-IV

The STK3102-IV is one of the biggest weak spots of some Marantz amps, like the PM80(SE). I had to replace the one in mine once but I was wondering if someone has ever come up with an alternative.

Does anyone know if such an alternative exists? E.g. a replacement made up of discrete components?
Surface mount parts are discrete, and everything eventually dies.

You can build any design you want with surface mount versions of 2N5551/5400 and MJE 340/350 and make it small enough to fit.

Crest uses a module they build in their v1500 that will drop right in, don't know what it costs.
Here is something interesting :

We all expect one to replace the 3102 , check soldering , check the known relays issue ,replace a few capacitors , set the bias and done...

I also expect "techs " that simply replace the 3102 and give away .

Point is that in the first case and when one wants to set the bias will look at the service manual and bias the amplifier accordingly .The problem is that the service manual is written WRONG !!

It states that at bias points bias should be 18mv over a resistor 0.18R and that equals to 50ma ......Obviously this wrong since this equals to 100ma
farther more in the high bias condition 198mv over a resistor of 0.18 that equals to 500 ma Obviously this is also wrong and equals to 1A

Pm 80 features both neg and pos test points and what the missed to print is that the sum of the two redings ( pos +neg ) should be 18mv and 198 mv respectively .

Cannot tell if more than double the bias will stress the IC but obviously in total the heat will stress the amp and produce fatigue of parts and soldering .

Now days fake IC exist from PMC /SAN which the quality i am not aware off so i replaced one , added a far better heat sink and bias the amplifier properly to see how far i will go ....

Enjoy a print of the schematic

Kindest regards


  • PM80 Settings.png
    PM80 Settings.png
    295.7 KB · Views: 383
Depends on how you interpret Marantz's instructions.
The way I interpret the instructions is that you connect the probes across BOTH resistors. That way you measure across 2 x 0R18 and the printed values do make sense after all (although 198 mV across 0R36 is, of course, more than the 500 mA mentioned :rolleyes:).

The reason I believe this is because many contemporary (and later) Marantz amps like the PM40 and PM82 used 2 x 0R18 resistors in a single three-pin housing. You were always instructed to measure across the outer pins, i.e. across both resistors.


From PM7200 service manual.


From PM82 service manual.
Last edited:
Try and use an appropriate chip amp instead, matching the voltage and output.
The STK modules were 2 chips in a single housing with a common power supply, so use two suited mono chips higher up or on a bigger heat sink.
More up to date, less hassle later, much more reliable, though TDA7294 also seems to blow.

Initially, wiring up and physical fitting will be tedious, but it should work, better than a Chinese substitute, or a pulled one.
These particular STKs were just font ends - designed to be used with discrete output stages. More reliable than the “all in one” units that handle high power, but still subject to the problems that all these hybrids did. The originals were chip and wire, with just a glued on plastic cover for environmental protection. Eventually they all fail from moisture ingress. These were relatively easy to replicate using discrete packaged parts mounted on a little daughter board. Even some of the complete power modules have been successfully replicated using discretes, but that is a lot more work to make the heat spreaders and pretty much have to be mostly SMD except for the outputs - because the output pair itself takes up half or more of the allotted room for the whole thing.

TDA7294’s operated at 100 watts are no more reliable. Failure mechanism is different, but blown chips are the rule rather than he exception. You’ve really got to run chip amps backed off of maximum power dissipation to get reliability. When you do so, they are quite reliable.