Yamaha MX-1000 Class-H stuck rail, cant figure out why

im working on a Yamaha MX-1000 right now and got a weird problem i couldnt fix despite trying for hours.

This amp uses a class-h topology with +-40V and +-90V rails.
On the right channel, the negative supply is stuck at -HB (-90V allways on the transistors) for some reason. it should be -LB.
+LB and the other channel work just fine at +-40V

ive taken out all transistors multiple times, measured them, bought new ones and put gain matched ones in there (+-10%)
i looked at every solder joint, redone the whole right channel in the end..
tested every diode.
also checked the resistors and connections inbetween as best as i could.

the power transistors are new, all tested and gain matched. 3 PNP, 3 NPN for output, 2 PNP 2 NPN for high rail switching.

This is the schematic of the Right amp board:

this is my problem area:
1, 2, 3 ,4 are identical to the other channel.
the Rail that feeds Q320 Q324 and Q328 on the left is stuck at -90V, it should be at -40V.

i checked (and later replaced for good measure) Q350 and Q346.
also measured every resistor on this board, all are fine.
All diodes are fine in their normal direction, all zeners are fine in their reverse direction.
no caps shorted, non are noticably leaky.

Q342 is good
Q338 is good aswell, so is D324 (12V zener) and Q332.
aswell as D328 (24V zener) and D332

(flipped image..)
on the thermal image, you can see the right lower gets warm, thats the Left channel -HB APS
its dead cold on the left side (right channel) but the output transistors get quite warm at 90V and 22mA biasing..

i wrote down all differences to the other channel

Right channel | Left Channel
across D319/D320 87V vs. D316/D315 23V
across R392 71V vs R391 7.7V
across Q332 Vbe 0V? vs Q331 Vbe 0.61V
(but weirdly when i measure from ground to Q332 base its 87V, but to Emitter its 80V. maybe the internal multimeter resistance is playing tricks?)
across Q332 Vce 80V vs Q331 Vce 24V
across R400 and 396 0V vs across R399 and R395 50V

seems like Q332 isnt turning on.
its a 2SC2705
when i take it out, it measures just fine using every transistor tester i have. tried with a new one, no luck.
maybe because Vbe is 0.. but where should it come from?
Both diodes in parallel are 1s1885.
and measurements past R392 are identical to the working channel.
i checked the positive HB side aswell maybe it affects the negative, but didnt found any obvious problems.

Board pics: (too big to upload here)



im stuck, can anyone help?
OK, to backtrack and get a little history, what was the reason you replaced all the transistors in the first place? Was this a preventative rebuild attempt, or was it a repair from the get go?

I ask because, in the case of ‘80’s and early ‘90’s Yamaha and Accuphase gear it’s strongly recommended to reflow all solder joints, as they are the prevalent cause of virtually any failure modes. It is however not recommended to “recap” them unless a failing component is identified, not only due to the heuristics involved.

Ylli beat me to it - I’ll try to find my schematic and have a look in the coming days.

P.S. I had one of these and it was a beast of an amplifier.
I got this one really cheap because it was sold as broken.

One Amplifier board failed and blown some components including one output transistor. luckily all of the expensive matched dual-transistor chips survived.

after replacing many transistors with matched pairs and triplets aswell as some resistors, i found that the right channel PNP / Highside gets quite hot under not that much of a load.
not only did they have a bad heatsink contact due to a Burr on the tapped-hole,
but i found that one of the rails is stuck aswell.
this stuck rail failure very likely occured at the previous owner and caused the transistor to run too hot and blow eventually.

Will test shorting Q332 tomorrow to see if it enables -LB. enough solder fumes for today.

The schematic can be found here, https://elektrotanya.com/yamaha_mx-1000_u_full.pdf/download.html

i reflowed all joints on this board, still need to do this on the other boards aswell.
and replaced most caps that are not "MUSE", these ive tested fine. ESR is good, no excessive leakage to speak off, capacity on spec.

Its certainly a beast. my back hurts everytime i gotta lift it on my bench.
but i hoped for dual-mono design which it is not. its just two transformers for + and - rails and caps shared for both amps.
im gonna replace the output terminals aswell, the original ones are not that great.

allready found a damn good deal on a matching cx-1000 preamp to go along with it. should arrive soon.
  • Like
Reactions: 1 user
just tried your suggestion.
Shorting C-E of Q332 did indeed switch it to Low-rail. D324 and everything after gotta be alright.

something isnt driving the base, may have to look further into HB+ section of the APS.
the only part connecting them together is R388 (22k) which measures fine.
measured a few things again on both channels to compare.
drop across R388 is 32.6v, R387 is 32.9v
drop of R388 and R387 to GND is both 16V.
drop of D328 and D327 is both 23.9V
drop of R392 is 73V, R391 is 7.6V. both measure 100k as they should.
drop of R390 and R389 is both 7.9V

after a few times disassembling/assembling i blew Q332 by having a tiny solder ball stuck in the thermal pad shorting out one APS transistor.. luckily no part got damaged except q332...
i didnt have any 2sc2705 on hand and put in a 2sc2383 to keep on testing.
gotta order some parts. again.

cant imagine they need to be matched even closer. yamaha states "Pair" on Q340 and Q342 and Q314 Q316. but doesnt state what tolerance.

@mbz yea, im gonna take a look into it once it arrived. heard of the solder joint problem too because people tend to plug their 5kg voodoo rca cables on it flexing the whole chassis. old rubycon caps gotta go aswell.
with Q338 removed
i measure on Q332
Vbe 0.6V
Vce 41V
(g = ground)
Vgb -40V
Vge -41V

i measured the transistor (2SA1145) i just pulled, it was fine. replaced it temporarily with a 2SA1013 but didnt change anything.
measured R404, was also fine. 1k.
also checked if the silkscreen is wrong, but its all correct.

i noticed the 100k resistors R390 R392 and so on are not in the parts-list anywhere. but they measure 100k, all of them.
i moved all transistors from the good board to the bad and its still stuck at -HB
Q341 -> Q342
Q339 -> Q340
Q337 ->Q338
Q335 -> Q336
Q331 -> Q332
Q329 -> Q330

gotta try diodes next.

i hope it doesnt depend on the compilmentary transistors gain or something.
the good channel has (averaged)

the bad channel has (averaged)

still got a few 2SC3281 here with lower gain, might swap those out to have less gain than PNP just like on the good channel..

what a weird issue
Don't think transistor match is going to make any significant difference. I put the circuit into LTSpice (SIM) and this is what it looks like with Q332 removed:
Screenshot 2023-09-16 111533.jpg
A shorted D324 Zener could get Q332 Emitter near -41V. What voltage do you see at the junction of R392 and D328?
you removed Q338, not Q332.
i measure (With all transistors in the circuit) -26V on the R392 D328 junction. which is werid, i once measured a 24v drop across D328 before but that doesnt add up to the stuff i measured here... gonna replace it, its just a zener.

edit: replaced all, all measured fine out-circuit but meh just give it new ones. no difference.
Last edited: