Vintage amp (HarmanKardon HK6500) mystery hiss - after full recap

[ExtraThiccBoi]

I have an extra question, unrelated to this preamp noise issue.
When i got this amp, it was set to a bad secondary and the power supply voltage was 4.5V higher than supposed to. On the main amplifier board, there are 2 transistors that are already being run near limit (50V) when the voltage is correct. Could this incorrect voltage setting cause these transistors to go bad? I haven't done the maths but i suppose having main voltage 54V instead of 49.6V (or thereabouts) would cause the voltage at the base of this 2SA999L (50V/50V/6V) to go over the rated 50V. Schematic included

[David Hess]

Unfortunately, it does not list the author's name.

It is an old National Semiconductor application note and like most companies then and now, National did not allow the authors to put their name on work products.  Linear Technology and Analog Devices are unusual as exceptions to this and the difference indicates the company's attitude toward employees.  Think about the connotations of the term "human resources".

I suspect more than one author was involved but the application note does say "the author".  (1)

I 've also encountered small signal bipolar transistors that are noisy and have significant leakage.

I understand that the leakage and noise performance of a bipolar transistor can be permanently affected by reverse biasing the B-E junction to breakdown.  I wonder if this is what is happening here.  It's easy enough to overdrive the input of a power amp when it's off if driven from a preamp.

High gain (hfe) transistors are the most susceptible to damage from base-emitter breakdown and these are also the best transistors for low frequency low noise applications.  Lowing the gain (hfe) directly increases the input current noise.

Personally, I would change them and maybe add reverse base-emitter breakdown protection if only to see if it makes a difference.  Calculating the expected noise just from the schematic is not difficult but measuring it is.

Could this incorrect voltage setting cause these transistors to go bad? I haven't done the maths but i suppose having main voltage 54V instead of 49.6V (or thereabouts) would cause the voltage at the base of this 2SA999L (50V/50V/6V) to go over the rated 50V. Schematic included

It is the voltage across the transistor terminals which matters.  Q413 and Q415 are emitter followers which drive the following voltage gain stage and *that* stage is what sees the high voltage.  Q413 and Q415 only operate at relatively low voltages.

(1) Author found.

[schmitt trigger]

Unfortunately, it does not list the author's name.

It is an old National Semiconductor application note and like most companies then and now, National did not allow the authors to put their name on work products.  Linear Technology and Analog Devices are unusual as exceptions to this and the difference indicates the company's attitude toward employees. 

And nothing exemplifies this better than the late Jim Williams' app notes.
Not only did he sometimes employ colloquial language and subtle humor to explain a complex idea, but he would very often include deliciously absurd footnotes: "Local historians can't be certain, but this may be the only IC pin ever named after a person.", and preposterous-sounding app note names: App Note 25, "Switching regulators for poets: A gentle guide for the trepidatious" or even better: App Note 45, "Measurement and control circuit collection: Diapers and designs on the night shift"

And let's not forget his trademark tour de force, a hand-drawn cartoon on the last page, which could be related or not to the topic at hand.

[duak]

David, thank you for your comment on low noise transistors.  Do you have any further information on this particular effect?  I remember an app note for something or other from the 70s discussing this .  It used a 2N4401 with the base disconnected as a 5 or 6 V, low leakage current transient suppressor to protect the inputs of some precision device.

I should have been a bit clearer about damaging low noise transistors in audio systems.  In my case, the power amp was close to the speakers with long RCA phono ended cables from the preamp.  It's also got a safety ground that ties to signal ground through 100K in parallel with some capacitance.  The preamp was in a rack with various other equipment like a tuner, CD player and VCR; all of which have only a two pole line connection - none had a safety ground.  The big problem with RCA connectors is that the center signal pin makes contact before ground does.  When connecting the power amp, it's a damn good idea to not have it powered on or you could frap a speaker so this clearly indicates there is significant voltage there.  I suspect the voltage was high enough to damage the transistors, maybe not all at once but over a period of time.  Since I built the preamp too, I was often making changes and disconnecting and reconnecting the cables.    I mentioned I had a VCR in my system.  It was connected to the cable TV network.  I found that the VCR was injecting a 60 Hz hum into the system because it had a 5 VAC 60 Hz at 100 mA interfering signal coming from cable.  I put a couple of back to back 75 to 300 ohm transformers on the cable to break the ground loop and hum went away.

If the OP has found noisy transistors in the preamp, it would be more difficult but not impossible to have something similar happen.  The overvoltage could come from just plugging in an audio source that has an electrostatic charge on it.

Food for thought, anyway.

[David Hess]

David, thank you for your comment on low noise transistors.  Do you have any further information on this particular effect?  I remember an app note for something or other from the 70s discussing this .  It used a 2N4401 with the base disconnected as a 5 or 6 V, low leakage current transient suppressor to protect the inputs of some precision device.

I found some discussions though a search like "base emitter bipolar transistor breakdown hot carrier".  My understanding is that the breakdown produces hot carriers which have enough energy to cause dislocations in the crystal structure of the junction damaging it.  These dislocations allow minority carriers to recombine lowering minority carrier lifetime lowering current gain.  Gold doping also lowers minority carrier lifetime reducing hfe which is why RF and fast saturated switching transistors have low gain.

The damage can be repaired by annealing the junction at high temperature but of course this is hardly feasible in a typical application.  I wonder though if any space applications where radiation was an issue did this.

Since gold doped transistors and transistors operating at high current density already have low current gain, they are not affected significantly.  But high current gain transistors and transistors operating at low current density are especially susceptible.

I should have been a bit clearer about damaging low noise transistors in audio systems.  In my case, the power amp was close to the speakers with long RCA phono ended cables from the preamp.  It's also got a safety ground that ties to signal ground through 100K in parallel with some capacitance.  The preamp was in a rack with various other equipment like a tuner, CD player and VCR; all of which have only a two pole line connection - none had a safety ground.  The big problem with RCA connectors is that the center signal pin makes contact before ground does.  When connecting the power amp, it's a damn good idea to not have it powered on or you could frap a speaker so this clearly indicates there is significant voltage there.  I suspect the voltage was high enough to damage the transistors, maybe not all at once but over a period of time.  Since I built the preamp too, I was often making changes and disconnecting and reconnecting the cables.    I mentioned I had a VCR in my system.  It was connected to the cable TV network.  I found that the VCR was injecting a 60 Hz hum into the system because it had a 5 VAC 60 Hz at 100 mA interfering signal coming from cable.  I put a couple of back to back 75 to 300 ohm transformers on the cable to break the ground loop and hum went away.

Single ended consumer audio stuff always has problems with hum unless care is taken with the configuration.  Powering everything off of one outlet strip is a good start.

Protecting the input transistors can be as simple as placing a diode anti-parallel with the base-emitter junction of each transistor and some integrated differential pairs do exactly this:

https://www.analog.com/media/en/technical-documentation/data-sheets/mat12.pdf
http://www.leadelectronics.com/product_specs/LM394-DATASHEET.PDF
http://www.alfarzpp.lv/eng/sc/AS394CH.pdf

[duak]

Thank you David.  I had known about gross over current thru a zener, AKA "zener zapping" where the current is high enough to cause metal migration and shorting of the junction.  Your comments indicate that damage or degradation can occur at far lower energy levels.  I don't design anything for anyone except myself these days.  However, it's satisfying to learn something about a small mystery.

While doing some research (frobnicating on the 'net) I ran across this article on reverse avalanching in bipolar transistors: http://www.kerrywong.com/2014/03/19/bjt-in-reverse-avalanche-mode/  I had always wondered if there was anything special about the 2N4401 but it appears that there isn't; most any general purpose BJT appears to do this.  The 2N4401 seems to be similar to the 2N3904 but has a greater current rating.  I can see from the article that the effect demonstrates negative resistance which I think could be an advantage as a transient absorber, ie., it will reduce the duration of the pulse applied to the following circuitry.

[David Hess]

Thank you David.  I had known about gross over current thru a zener, AKA "zener zapping" where the current is high enough to cause metal migration and shorting of the junction.  Your comments indicate that damage or degradation can occur at far lower energy levels.  I don't design anything for anyone except myself these days.  However, it's satisfying to learn something about a small mystery.

Bob Pease wrote some articles discussing it also.

I had always wondered if there was anything special about the 2N4401 but it appears that there isn't; most any general purpose BJT appears to do this.  The 2N4401 seems to be similar to the 2N3904 but has a greater current rating.  I can see from the article that the effect demonstrates negative resistance which I think could be an advantage as a transient absorber, ie., it will reduce the duration of the pulse applied to the following circuitry.

I wish manufacturers gave more details about the fabrication of different devices.  A little can be learned from the National Semiconductor Discrete Databook.  The 2N3904/2N3906 are *not* just a smaller 2N4401/2N4403; they use a different process and a different layout.  Of course different manufacturers make different parts with the same part number and JEDEC specifications.

2N2222 1-17 20   NPN Medium Power (gold doped)
2N2906 2-10 63   PNP Medium Power

2N3904 1-20 23   NPN Small Signal (gold doped, interdigitated)
2N3906 2-13 66   PNP Small Signal (gold doped)

2N4401 1-21 13   NPN Medium Power
2N4403 2-14 63   PNP Medium Power

[elecdonia]

I have an extra question, unrelated to this preamp noise issue.
When i got this amp, it was set to a bad secondary and the power supply voltage was 4.5V higher than supposed to. On the main amplifier board, there are 2 transistors that are already being run near limit (50V) when the voltage is correct. Could this incorrect voltage setting cause these transistors to go bad? I haven't done the maths but i suppose having main voltage 54V instead of 49.6V (or thereabouts) would cause the voltage at the base of this 2SA999L (50V/50V/6V) to go over the rated 50V. Schematic included]

If the power amplifier sections are working on both channels then there is no problem with the circled transistors.  These operate at low collector current compared to other transistors in this amplifier circuit. So they shouldn't get warmer than room temperature. The fact that the power supply voltage might be slightly higher than the voltage rating of these transistors isn't likely to be an issue.

This is a rather complicated power amplifier circuit. HK was known for that during the 1980's especially.  HK designed their circuitry for very low distortion and very wide bandwidth.

This amplifier is capable of sounding good with most speakers.  But there might be a handful of speaker systems that it doesn't sound good with. Speaker systems with extremely wide ranges of impedance, from low to high at different frequencies might cause this amplifier circuit to oscillate intermittently. That could create a "harsh" or "tinny" sound quality.  This would depend on the relationship between the amplifier and the speaker.  In particular, some of the multi-driver speaker systems of UK design and origin might present issues when driven by this power amplifier.

RE: The power amplifier section:  It is very important to adjust the output offset voltage to a value as near 0.0 volts as possible. Also there should be a detailed procedure in the HK service manual for adjusting the "idle current" of the large output transistors. There are a total of 4 adjustment potentiometers:  Each channel (left, right) has 2 potentiometers:  One potentiometer is  used to adjust the DC output voltage at the speaker terminal to 0.0 volts. The other potentiometer is used to adjust the "idle current."

-EB
 

[ExtraThiccBoi]

I have an extra question, unrelated to this preamp noise issue.
When i got this amp, it was set to a bad secondary and the power supply voltage was 4.5V higher than supposed to. On the main amplifier board, there are 2 transistors that are already being run near limit (50V) when the voltage is correct. Could this incorrect voltage setting cause these transistors to go bad? I haven't done the maths but i suppose having main voltage 54V instead of 49.6V (or thereabouts) would cause the voltage at the base of this 2SA999L (50V/50V/6V) to go over the rated 50V. Schematic included]

If the power amplifier sections are working on both channels then there is no problem with the circled transistors.  These operate at low collector current compared to other transistors in this amplifier circuit. So they shouldn't get warmer than room temperature. The fact that the power supply voltage might be slightly higher than the voltage rating of these transistors isn't likely to be an issue.

This is a rather complicated power amplifier circuit. HK was known for that during the 1980's especially.  HK designed their circuitry for very low distortion and very wide bandwidth.

This amplifier is capable of sounding good with most speakers.  But there might be a handful of speaker systems that it doesn't sound good with. Speaker systems with extremely wide ranges of impedance, from low to high at different frequencies might cause this amplifier circuit to oscillate intermittently. That could create a "harsh" or "tinny" sound quality.  This would depend on the relationship between the amplifier and the speaker.  In particular, some of the multi-driver speaker systems of UK design and origin might present issues when driven by this power amplifier.

RE: The power amplifier section:  It is very important to adjust the output offset voltage to a value as near 0.0 volts as possible. Also there should be a detailed procedure in the HK service manual for adjusting the "idle current" of the large output transistors. There are a total of 4 adjustment potentiometers:  Each channel (left, right) has 2 potentiometers:  One potentiometer is  used to adjust the DC output voltage at the speaker terminal to 0.0 volts. The other potentiometer is used to adjust the "idle current."

-EB
 

Yes i adjusted it as the spec says, idling current is 41.70mV in both channels (40mV spec). I reduced DC offset as much as i could but due to that capacitor in feedback loop, it tends to drift couple tens of mV.
The power amplifier works flawlessly, sounds good and is very dynamic.
Thanks a lot for your help!

[magic]

So how is the preamp doing? Did input pair replacement fix its noise?

[ExtraThiccBoi]

So how is the preamp doing? Did input pair replacement fix its noise?

Nope it didnt fix it.
I guess my only option would be to redo that preamp stage with a op-amp, i have an old mitsubishi which is made to run at 23V for exactly these sorts of applications, but i don't have a fricking clue how to build that. I guess i would have to attach on the feedback lines and then deal with the input and output resistors and capacitors.

[ExtraThiccBoi]

I have an extra question, unrelated to this preamp noise issue.
When i got this amp, it was set to a bad secondary and the power supply voltage was 4.5V higher than supposed to. On the main amplifier board, there are 2 transistors that are already being run near limit (50V) when the voltage is correct. Could this incorrect voltage setting cause these transistors to go bad? I haven't done the maths but i suppose having main voltage 54V instead of 49.6V (or thereabouts) would cause the voltage at the base of this 2SA999L (50V/50V/6V) to go over the rated 50V. Schematic included]

If the power amplifier sections are working on both channels then there is no problem with the circled transistors.  These operate at low collector current compared to other transistors in this amplifier circuit. So they shouldn't get warmer than room temperature. The fact that the power supply voltage might be slightly higher than the voltage rating of these transistors isn't likely to be an issue.

This is a rather complicated power amplifier circuit. HK was known for that during the 1980's especially.  HK designed their circuitry for very low distortion and very wide bandwidth.

This amplifier is capable of sounding good with most speakers.  But there might be a handful of speaker systems that it doesn't sound good with. Speaker systems with extremely wide ranges of impedance, from low to high at different frequencies might cause this amplifier circuit to oscillate intermittently. That could create a "harsh" or "tinny" sound quality.  This would depend on the relationship between the amplifier and the speaker.  In particular, some of the multi-driver speaker systems of UK design and origin might present issues when driven by this power amplifier.

RE: The power amplifier section:  It is very important to adjust the output offset voltage to a value as near 0.0 volts as possible. Also there should be a detailed procedure in the HK service manual for adjusting the "idle current" of the large output transistors. There are a total of 4 adjustment potentiometers:  Each channel (left, right) has 2 potentiometers:  One potentiometer is  used to adjust the DC output voltage at the speaker terminal to 0.0 volts. The other potentiometer is used to adjust the "idle current."

-EB
 

I see you know a lot about audio amps and the circuit designs, would you be able to tell which one of those is better? One is HK6500, the amp that i have. Second is PA2200, a power amp i can get a deal on. It's rated higher but the european version of the 2200 is cut down, so it will most likely provide similiar RMS power to the HK6500.

[magic]

I guess my only option would be to redo that preamp stage with a op-amp

Or insert an opamp between the volume pot and the preamp. Its output impedance will be zero and the preamp will be happy.
Doesn't need to be a high voltage one, just add a 78L15/79L15 pair and some capacitors.

You will probably need a FET opamp. I would try OPA2134 for starters or TL072 if you feel cheap. Bipolar opamps like NE5532 may still have too much current noise for this volume pot. No idea what your Mitsubishi opamp is.

Good idea to use a socket, I guess. Just mount it on a piece of perfboard, connect and and see what happens. Beware that it may change nothing if the noise comes from the volume pot itself rather than the input transistors, but it probably comes from the trannies.

[David Hess]

Adding a JFET or CMOS voltage follower to buffer the potentiometer is probably the easiest way.

Or the bipolar differential pair could be replaced with a JFET differential pair.

[magic]

CMOS certainly isn't audiophile-approved
Not even sure if there are any suitable CMOS amps out there - low noise, high voltage, good CMRR, low distortion...

Discrete JFETs are perfectly halal but the reduction in open loop gain would have unknown influence on distortion of that stage. The degeneration resistors could be removed, but OP doesn't even have a scope to test for stability.

[ExtraThiccBoi]

I guess my only option would be to redo that preamp stage with a op-amp

Or insert an opamp between the volume pot and the preamp. Its output impedance will be zero and the preamp will be happy.
Doesn't need to be a high voltage one, just add a 78L15/79L15 pair and some capacitors.

You will probably need a FET opamp. I would try OPA2134 for starters or TL072 if you feel cheap. Bipolar opamps like NE5532 may still have too much current noise for this volume pot. No idea what your Mitsubishi opamp is.

Good idea to use a socket, I guess. Just mount it on a piece of perfboard, connect and and see what happens. Beware that it may change nothing if the noise comes from the volume pot itself rather than the input transistors, but it probably comes from the trannies.

I have a mitsubishi m5220l SIP. Pulled from an old amp. It's a bipolar opamp, would that atleast help? The thing is most modern opamps arent bueno for such high voltage rails from the datasheets i've seen.
I will have to run a jumper from output to inverting input right? Is that a voltage buffer operation?
Should be easy enough to freewire it.
Sorry, i am not really good with designing circuits and even worse with building them from scratch.

[David Hess]

CMOS certainly isn't audiophile-approved
Not even sure if there are any suitable CMOS amps out there - low noise, high voltage, good CMRR, low distortion...

The performance of the existing design is already compromised enough that using a good CMOS operational amplifier will not be a disadvantage.

Discrete JFETs are perfectly halal but the reduction in open loop gain would have unknown influence on distortion of that stage. The degeneration resistors could be removed, but OP doesn't even have a scope to test for stability.

That is why this option is more complicated compared to just adding a voltage follower; performance testing and some changes would be necessary although it is possible to just compute the new source degeneration resistor values.  With the same adjusted input stage transconductance, everything should work the same except the distortion characteristics will be different.

[magic]

OPA2604 is the only JFET opamp coming to my mind that can run on ±22V. A genuine one, because a "rebranded" TL072 could blow up
It may be cheaper to add 15V regulators and get a wider choice of opamps.

I frankly have no idea how discrete transistors compare to those in bipolar opamp input stages in terms of current noise. If they are comparable, a bipolar opamp ought to have some advantage over discrete due to about 10x lower input bias current. Whether it will be sufficient for you I don't know.

Yes, you would connect IN+ of each channel to the signal and IN- straight to OUT. Some 47-100Ω resistor between OUT and the cable to the preamp board may be a good idea to avoid risk of oscillation. Add 100nF ceramic plus a 10-100uF electrolytic between V+/V- and that should suffice for a quick noise test. I'm not even sure if it makes sense to upgrade to separate capacitors from each rail to ground later, the output current is very low.

And I came up with another option: if you set volume to max and balance to center, the preamp board will probably be more or less shorted to the input jacks. So it will see whatever impedance the external source has. So you could set the volume to max and use an external box with a low impedance (like 5~20k) volume pot. Or even an active preamp or whatnot.
It would take exerimentation to see what input impedance can be tolerated. So take some resistors and insert them into the RCA jack (from hole to ground) and see how much hiss remains when you turn volume up. Don't touch the input jacks with volume at max

[ExtraThiccBoi]

OPA2604 is the only JFET opamp coming to my mind that can run on ±22V. A genuine one, because a "rebranded" TL072 could blow up
It may be cheaper to add 15V regulators and get a wider choice of opamps.

I frankly have no idea how discrete transistors compare to those in bipolar opamp input stages in terms of current noise. If they are comparable, a bipolar opamp ought to have some advantage over discrete due to about 10x lower input bias current. Whether it will be sufficient for you I don't know.

Yes, you would connect IN+ of each channel to the signal and IN- straight to OUT. Some 47-100Ω resistor between OUT and the cable to the preamp board may be a good idea to avoid risk of oscillation. Add 100nF ceramic plus a 10-100uF electrolytic between V+/V- and that should suffice for a quick noise test. I'm not even sure if it makes sense to upgrade to separate capacitors from each rail to ground later, the output current is very low.

And I came up with another option: if you set volume to max and balance to center, the preamp board will probably be more or less shorted to the input jacks. So it will see whatever impedance the external source has. So you could set the volume to max and use an external box with a low impedance (like 5~20k) volume pot. Or even an active preamp or whatnot.
It would take exerimentation to see what input impedance can be tolerated. So take some resistors and insert them into the RCA jack (from hole to ground) and see how much hiss remains when you turn volume up. Don't touch the input jacks with volume at max

Following your recommendation, i will probably do a dirty fix. I will run the wire from the preamp board to a DPDT switch. The other set of contacts will carry signal from extra RCAs at the back with 1K resistors (harman kardon reverse engineering). If i want to run it as power amp for desk use, i will run directly from those rear RCAs into the power amp, with my Magni 2 Uber as preamp (it has buffered output). If i want to run it as integrated amp, i will just switch the DPDT switch and put signal into the power amp from the preamp and whole integrated BS-ittery. Might be the easiest solution.

[floobydust]

The JFET muting circuit is common to the inputs to both channels, you might want to check its health. I have seen a failed JFET generate noise for two channels.
You can also use freeze spray to track down a noisy BJT or component. Before you blame the design. HK was pretty good.

[ExtraThiccBoi]

The JFET muting circuit is common to the inputs to both channels, you might want to check its health. I have seen a failed JFET generate noise for two channels.
You can also use freeze spray to track down a noisy BJT or component. Before you blame the design. HK was pretty good.

The power amp is good. I just finished the mod and after checking with the volume knob, the power amp generates very little noise. If i crank the Magni 2 Uber to the max (ear-shattering scenario with -18dB RMS digital output from my Modi 2 Uber DAC), there is a little tiny bit of hiss but MUCH quiter by what is generated by the preamp. I won't blame the power amplifier design, after reading some quick articles, i can tell even with noob's eye it's good. In fact, i have purchased another power amplifier incorporating basically the same design but with twice the current headroom.
The JFETs from my understanding are used to short the inputs to ground when the protection circuit detects fault, afaik shorting inputs means turning off the power transistors (??).

[David Hess]

OPA2604 is the only JFET opamp coming to my mind that can run on ±22V. A genuine one, because a "rebranded" TL072 could blow up

44 volt JFET operational amplifiers are not very common.  My list only includes the old LF3xx and LF4xx series of parts.  Analog Devices only has the LTC6090.  The OPA604 from TI looks ideal but the LME49860 could also work.  There are some other potential parts from TI.

It may be cheaper to add 15V regulators and get a wider choice of opamps.

Or even better, shunt regulators in the form of a zener diode (or transistor base-emitter junction) and a resistor for each side.  The current demands are not great and I would not trust a 44 volt part working directly off of that supply anyway.

I frankly have no idea how discrete transistors compare to those in bipolar opamp input stages in terms of current noise. If they are comparable, a bipolar opamp ought to have some advantage over discrete due to about 10x lower input bias current. Whether it will be sufficient for you I don't know.

The only advantage integrated bipolar parts have is the availability of super-beta transistors.

[floobydust]

The JFET muting circuit is common to the inputs to both channels, you might want to check its health. I have seen a failed JFET generate noise for two channels.
You can also use freeze spray to track down a noisy BJT or component. Before you blame the design. HK was pretty good.

...
The JFETs from my understanding are used to short the inputs to ground when the protection circuit detects fault, afaik shorting inputs means turning off the power transistors (??).

The "muting protection" circuit has only output over-current detect, and power up delay timer, I don't see DC detect. I think it's mostly anti-thump.
You have to check coupling capacitors, when you recap old audio gear sometimes polarity is backwards. With age, DC offsets can change, or the orginal schematic can be wrong.
At the tone controls, C513/C514 is suspicious compared to C517/C518. Just to make sure cap leakage current isn't adding noise.

[ExtraThiccBoi]

The JFET muting circuit is common to the inputs to both channels, you might want to check its health. I have seen a failed JFET generate noise for two channels.
You can also use freeze spray to track down a noisy BJT or component. Before you blame the design. HK was pretty good.

...
The JFETs from my understanding are used to short the inputs to ground when the protection circuit detects fault, afaik shorting inputs means turning off the power transistors (??).

The "muting protection" circuit has only output over-current detect, and power up delay timer, I don't see DC detect. I think it's mostly anti-thump.
You have to check coupling capacitors, when you recap old audio gear sometimes polarity is backwards. With age, DC offsets can change, or the orginal schematic can be wrong.
At the tone controls, C513/C514 is suspicious compared to C517/C518. Just to make sure cap leakage current isn't adding noise.

Yep, it brings the amp up slowly and indeed, no pops or thumps.
All caps, resistors, transistors are correctly placed, i've checked that many times. I guess the tone circuit is just high noise design or the 2sc2320 are bad transistors.
I replaced the caps with values that were in the amp to begin with, the schematic had some differences.
Btw, that 22uF/10V cap was 2.2/50V in my amp, replaced it with 2.2 as well.
I managed to get absolutely minimal DC offset with tongue at the right angle and plastic pokey thingy at 5mV when idle around 15mV when playing (that cap in feedback loop causes this drift afaik).

[magic]

Following your recommendation, i will probably do a dirty fix. I will run the wire from the preamp board to a DPDT switch. The other set of contacts will carry signal from extra RCAs at the back with 1K resistors (harman kardon reverse engineering). If i want to run it as power amp for desk use, i will run directly from those rear RCAs into the power amp, with my Magni 2 Uber as preamp (it has buffered output). If i want to run it as integrated amp, i will just switch the DPDT switch and put signal into the power amp from the preamp and whole integrated BS-ittery. Might be the easiest solution.

Not exactly, my recommendation (and the easiest solution if it works) was to turn the volume to maximum and use an external passive/active preamp for attenuation, connected to the normal input jacks.