200W Marshall power amp oscilations

[Markus2006]

I've build this power amp (Marshall 3520): http://www.drtube.com/schematics/marshall/3520pwr.gif . I replaced output MOSFETs with 2SK1058 and 2SJ062. It works fine but at low input signal levels (~200mV) the amp oscilates on the positive half-cycles. The only component that could influence this problem is C3 capacitor but when I increase it, the problem still remains. What can be the reason for the problem?
Another question is about quiescent current of the amp. It is not specified on the schematic. Currently it is 30mA per transistor. Is it a correct value?

Mark

 

[lineup]

try with a zobel filter across output
this can be needed for some inductive speakers

10 Ohm small resistor + 100nF filmcapacitor, polyester
in series, from output end of the output inductor to GND

if this does not help,
not even when using 47pF, 100pF or even 220pF for C3,
then I do not know
---

Important that R16, 17, 22, 23 resistors (470 ohm) are not too far away from the transistors GATE pins.

30 mA could work
but generally MOSFET likes a bit more .. maybe 100, 150 mA per each MOSFET

How much is your power supply voltage?

 

[Markus2006]

The supply voltage is +/-52V. I tried Zobel filter but this didn't help. I will change C3 to much higher value and report back. The value was clearly modified on the schematic but I cannot see what was the original value.

Mark

 

[lineup]

you can try to add High Freq lowpass RC filter at input.
1 kohm in series with C1 and one 220pF or 470pF cap across R1 (47kohm)
I do not think this will help, though.

One 47pF across R5 (4k7) fb resistor. One other try.

Is there one resistor in parallel with the output inductor?
If not, try with one 10 Ohm or 4R7

 

[Markus2006]

Thanks. There is no resistor in parallel to the inductor. I will try all your suggestions in the evening.

Mark

 

[Markus2006]

I did some tests. Unfortunately I found out that none of the changes helps. But... I found out that the problem does not appear when the amp is cold. It starts after a short warm-up period. Then I checked the transistors used in the amp. And it appears that MPSA92 (used instead of MPSA93) is hot. The simulation in LTSpice shows that the power dissipation of this transistor is 300mW at full power. The datasheet says that maximum power dissipation of this transistor is 625mW. Could it be the reason of the problem? I don't have any PNP transistor to make the test today.

Mark

 

[east electronics]

it could be

a pcb issue as well .... do you have your own pcb ???? beyond that when amp is getting hot

on the other hand have you performed the circuit as is ???? cause only by looking at it i would say that decoupling and by pass caps are not there ....

are they omited for simplicity ???? or they are omited any way ...

another thing i would do is to for a minute remove the VI limiter to see if make any diference

also simple design like that at this rail voltage might be quiet ready to oscilate with out a reason .....to explain that i mean that if its a marshal design they would have in mind that this amp will work with 50cm speaker cable as used on quitar amplifier ....

longer leads is another issue

there is also a very good posibilty that this oscilation might might be generated from your measuring setup as well
( hapend to me and then realized that my signal generator and scope created a ground loop through my metal made bench )

but then again i was testing a mosfet pcb .....5 years of operation and various tetsings on bipolar boards or comercial amps never seen this ....but with the mosfet it was there

 

[quicknick]

2SK1058 and 2SJ162 have different gate-source capacitance (lower for the 2SK). Amps using this output pair usually have some means to equalise this; you may try fitting 220 pF caps between gate and source on the 2SK1058 or increasing their gate resistor value. Hope this helps.

 

[nigelwright7557]

I've build this power amp (Marshall 3520): http://www.drtube.com/schematics/marshall/3520pwr.gif . I replaced output MOSFETs with 2SK1058 and 2SJ062. It works fine but at low input signal levels (~200mV) the amp oscilates on the positive half-cycles. The only component that could influence this problem is C3 capacitor but when I increase it, the problem still remains. What can be the reason for the problem?
Another question is about quiescent current of the amp. It is not specified on the schematic. Currently it is 30mA per transistor. Is it a correct value?

Mark

I usually start with no bias volts and then turn it up until the crossover distortion goes. I test it on a scope with about half working AC on output. This gives sufficient bias to get rid of crossover distortion but not so much you are wasting power into the heatsink.

 

[nigelwright7557]

I've build this power amp (Marshall 3520): http://www.drtube.com/schematics/marshall/3520pwr.gif . I replaced output MOSFETs with 2SK1058 and 2SJ062. It works fine but at low input signal levels (~200mV) the amp oscilates on the positive half-cycles. The only component that could influence this problem is C3 capacitor but when I increase it, the problem still remains. What can be the reason for the problem?
Another question is about quiescent current of the amp. It is not specified on the schematic. Currently it is 30mA per transistor. Is it a correct value?

Mark

I noticed on my own similar design to this that I got quite a bit of hum with that power supply setup. I put a 330R in series with the positive supply to the VAS and LTP stage and a 470uF cap to GND.

 

[quasi]

Try this

Hi Markus2006

Try inserting a series RC network between the collectors of the input and feedback transistors. This will reduce the high frequency gain of the input stage and improve stability. Use a 33 ohm resistor and a 330pF capacitor.

Also make sure you have good power supply de-coupling as close to the FETs as possible. This should be around 220uF from the rail to the central ground on the PCB bypassed with 0.1uF poly / MKT etc.

Cheers
Quasi

 

[Markus2006]

Thank you very much for your input. I will try all the suggestions.
Sakis, the board is mine but it's very simmilar to the original one (slightly smaller). Which caps are missing? The oscillations are both with a genarator and an osciloscope, and with guitar (so this is not a problem of the measurement equipment).
What is funny is that I have a working prototype with different MOSFETs used - 2SK412 and 2SJ113. They are not lateral but they don't have the problem with oscilations . The quiescent current rises after a warm-up period by 50% but after that it stays stable. Another difference between this amp and the final version is that 2SK412/2SJ113 require higher voltage to be biased correctly and the R8 (100 Ohms) resistor on the schemaic is changed to 1.2k. Also, as far as I remeber, the TR3 transistor does not get hot. The question is whether this is caused by (lack of) oscillations or by change of the R8 resistor.
Quasi, the power supply is not perfect (I use 40 cm wires) but the prototype does not have problems with this. I also tried adding some capacitors directly on the board but this didn't help.
Quicknick, I also have a feeling that the problem may be related to output transistors.

Mark

 

[Markus2006]

Re: Re: 200W Marshall power amp oscilations

I usually start with no bias volts and then turn it up until the crossover distortion goes. I test it on a scope with about half working AC on output.

Nigel,

I tried this method but I decreased the quiescent current from 30mA to 5mA per MOSFET and I still cannot see any crossover distortion on the scope. What frequency are you using? Should I go down below 5mA?

I didn't have much time yesterday to perform other tests. I only added a small heatsink to MPSA92. It still gets hot. I checked that the power dissipation of this transistor is 300mW, which compared to 600mW in the datasheet is maybe to much. I will replace this transistor with MJE350. The same transistor in the prototype is also quite warm. It gets only slightly lower voltage (47V instead of 51V) due to different MOFSETs used in the prototype.
And the oscillations happened at least once so I don't think that the problem is solved. But it's more difficult to reproduce it.

PS: I found out that this amp is almost the same as described in this document by Nelson Pass: http://www.passdiy.com/pdf/diyopamp.pdf . Interesting .

Mark

 

[lineup]

As a last try
you can test with using different value on those GATE resistors, 470 Ohm
I would first try with 220 ohm .. then 100 ohm
(If this doesnt help, I would give 680/820 ohm a test)


Another thinking.
not often we see a BASE stopper in VAS transistor, 100 ohm for TR3.
What happens if you remove this one (put a jumper instead)?

 

[Markus2006]

OK, I will try it. One thing that worries me is the fact that the currents in the differential amp are not equal. This is exactly due to the T3 transistor. Nelson , in his document, describes these currents as equal but it is hard to believe since T3 transistor is connected to one of the transistors (and it draws additional current). Do you think that this is correct?
There are few other components in this amp that are different than I would expect. For example the feedback resistors are 4k7/120 instead of 47k/1k2 as I feel would be more correct.
Also the R11 resistor (4k7) seems to be strange; it's value causes that the input transistors have very small current (500uA). I think that this is not enough and I plan to increase it (to get at least 1mA through input transistors). What do you think about this change?

Mark

 

[Markus2006]

you can test with using different value on those GATE resistors, 470 Ohm
...
not often we see a BASE stopper in VAS transistor, 100 ohm for TR3.
What happens if you remove this one (put a jumper instead)?

I made some tests. Change of the gate resistors did not help (I tried both 220 Ohms and 1k).
Jumper on base stopper resistor also did not help. The same with a capacitor and a resistor between collectors of the transistors in the differential amp.
Finally I replaced output transistors (2SK1058, 2SJ162) with transistors from the prototype (2SK412, 2SJ113) and amp works correctly . I had to increase one resistor (R8) to provide correct MOSFET biasing. But the quiescent current is to high - it's 100mA. I have to check why it is higher than in the prototype (it was 30 mA). Anyway, the oscillations are gone. It looks like transistors which are not lateral and shouldn't be used in the amp work better than lateral MOSFETs (which should be a direct replacement for 2SK135, 2SJ49). Strange.

Mark

 

[The Chump]

Hi There,

you could try putting a 10pF cap from gate to drain of each 2SK1058 - right on the legs.

 

[Markus2006]

The Chump, you are the champpion . Yes, this solved the problem. I put 15pF just on one of 2SK1058 from gate to drain and the oscillations are gone. Thank you very much. But I don't know how this can be explained. Is it possible that someone intentionally put incorrect value of C3 (47pF) on the schematic? Or, does 2SK1058 differs so much from 2SK134?
Maybe one last question: what should be the quiescent current in this amp? Currently iit is set to 30mA but the output transistors get quite warm. I can significantly decrease the current without visible distortion. Even with 5mA the distortion is still not visible. Any recommendations?

Thanks again,
Mark

 

[nigelwright7557]

The Chump, you are the champpion . Yes, this solved the problem. I put 15pF just on one of 2SK1058 from gate to drain and the oscillations are gone. Thank you very much. But I don't know how this can be explained. Is it possible that someone intentionally put incorrect value of C3 (47pF) on the schematic? Or, does 2SK1058 differs so much from 2SK134?
Maybe one last question: what should be the quiescent current in this amp? Currently iit is set to 30mA but the output transistors get quite warm. I can significantly decrease the current without visible distortion. Even with 5mA the distortion is still not visible. Any recommendations?

Thanks again,
Mark

I always set the bias on mine by when the crossover distortion goes. After that I test across the whole voltage range of input that there is no crossover distortion.

 

[lineup]

it is obvious that
using 2SK1058 instead of 2SK134 in this circuit gives problems

and that the solution is difficult to find without some additional simulation/oscilloscope testings

some capacitances and some bandwidth data of these mosfets are causing this issue

Shows once again, that one amplifier built with one kind of Output Devices in mind
is not necessarily going to be better with other Output Devices
and in some cases this will give real problems to fix.

So be careful when 'upgrading' to output transistors with improved data.
May take a lot of trial & errors to adjust the amplifier circuit to the new parameters.
In a feedback amplifier, the output stage is a part of the amplifier circuitry as a whole.
Not, for example, like one separate power supply capacitor outside of the loop.
Which can be exchanged easily.

Lineup