Hafler G300 clipping on negative portion of waveform

[Swampy]

Hi folks,

This amp came in with cooked R59 / R53, also CR11was open. R52 was charred a little measured ok...

I've replaced R52 | Q31 | R55 | Q33| R56 | Q32 | R53 (the whole bias string bar the trimmer).

Also replaced R59 and CR11 (used a 85C15 as I had one at hand) + CR12

I'm unable to adjust trimmer to bias into cutoff (which the left channel is capable of)

The waveform (image attached) clips from 30v pk-pk upwards (measuring output across a 4 ohm load).

The clipped waveform can be traced to pre-VAS, but is probably visible here due to nfb.

After dissapating 26 watts across 4 ohm load for 20 mins heatsinks get very hot to the touch.

V+ rails are @ + and - 67 v with no signal.



PRO2400 & Digitech G300 Schematic.pdf


My hunch is Q18 and/or19 is bad as clipping is just on negative portion of waveform. Any suggestions?

Ta
Matt

[Jazz P Bass]

Do you get a resistance reading across the bias pot?
Does it change when you turn it?
That numbering system is a PITA to follow.
I think you jumped channels on the Q18/19 thing. (Q38/39)
I would suspect the drivers. Q34 & 35.
Take some readings on the driver bases & emitters.(Vdc)

[Swampy]

Do you get a resistance reading across the bias pot?
Does it change when you turn it?
That numbering system is a PITA to follow.
I think you jumped channels on the Q18/19 thing. (Q38/39)
I would suspect the drivers. Q34 & 35.
Take some readings on the driver bases & emitters.(Vdc)

Yah bias pot reads 1K, voltage on base of Q13 changes when I turn it.

Yes bracketed numbers are a dog of the female variety.

Alas...

With bias trimmed to minimum (full counter-clockwise)

Replaced the drivers Q34+35 today, along with R60 , R54, R57, R30 (in zobel network...it was charred also?) + R32.

Difficult chassis/pcb to work on so thought I'd do a clean sweep over the resistors while it was accessible.


Q34 base = 1.37 v
Q35 base = -1.27 v

Q38 + Q39 + base = -0.62 v
Q36 + Q37 + base = 0.85 v


Running a 500hz signal, I measure 11 v pk-pk @ base of drivers Q34+35 before negative portion begins to square.

Under these conditions

Q38 + Q39 + base = -1.56v
Q36 + Q37 + base = 0.03 v

Q34 base = 0.55 v
Q35 base = -2.0 v



So why the asymmetry?

Matt

[Enzo]

If I had to guess, which is what I am doing, I'd suspect your negative side outputs are unable to drive the load, and the differential input is trying to correct for that. So it drives the whole mess offcenter. Are you running without a load? If you are loaded, remove the load from the output and see if anything changes.

Have you checked all the MOSFETs? And the gate resistors?

[Swampy]

If I had to guess, which is what I am doing, I'd suspect your negative side outputs are unable to drive the load, and the differential input is trying to correct for that. So it drives the whole mess offcenter. Are you running without a load? If you are loaded, remove the load from the output and see if anything changes.

Have you checked all the MOSFETs? And the gate resistors?

Enzo,

replaced gate resistors R63 R64, also Q38 / 39 but still having the same issue.
I don't understand why Q37 and Q36 being pulled into cutoff cause the negative portion of the waveform to clip square - as are they not responsible for the + portion of the waveform?

Is is safe to run a signal without load? Or are we interested in differences in idle dcs?

Matt

[Jazz P Bass]

It's quite alright to run the amp without a load.
Sometimes a clue can be found if the loaded output collapses but the unloaded one will not.
Solid state amps can be a real bear.
I have seen them puke up the output when the one of the input differential pair was faulty.
That would more than likely throw off your Vdc readings, which appear fine.

[Swampy]

It's quite alright to run the amp without a load.
Sometimes a clue can be found if the loaded output collapses but the unloaded one will not.
Solid state amps can be a real bear.
I have seen them puke up the output when the one of the input differential pair was faulty.
That would more than likely throw off your Vdc readings, which appear fine.

Alright, so runs fine with no load, nice trick to have up ones sleeve. Does this suggest that the malfunctioncan be isolated to output mosfets?

Reconnected load and pushed her a little hard..Q39 up in smoke. Makes sense..if bias is pulled neg. Q38 + 39 will conduct more.

JazzP - DC levels are good but as I mentioned in first post I'm unable to bias into cutoff, which the left channel can do.


I'll see what happens with a new set of output mosfets.

Matt

[Swampy]

Woops..I forgot scope was attached to + and - nodes, hence the incident this morning.

Anyways, with a new set of output mosfets installed, and under load the same scenario takes places. Voltage at the gates of Q38 + 39 pulls down, negative portion of waveform clips.


With no load I measure up to approx 100v pk-pk signal at output mosfet gates and output node. Bias voltages stay put, +0.6 and -0.35 volts on Q37 + Q38 respectively.


Is there some other experiments I can try to isolate the cause of this problem?

Matt

[Steve Conner]

The symptoms are weird. I've never seen anything like it. The only thing that would fit would be a faulty protection circuit shutting the negative side off prematurely, but according to the schematic there is no protection circuit! Or an extremely long shot: You used a diac instead of a zener diode.

Your scope shot is unclear because of double triggering. Can you try taking another one that shows the waveform clearly?

[Swampy]

The symptoms are weird. I've never seen anything like it. The only thing that would fit would be a faulty protection circuit shutting the negative side off prematurely, but according to the schematic there is no protection circuit! Or an extremely long shot: You used a diac instead of a zener diode.

Your scope shot is unclear because of double triggering. Can you try taking another one that shows the waveform clearly?

Just wanted to give this one a bump as I'm still yet to find cause of the problem.

So to reiterate, all is well unloaded, with load attached negative side collapses. I can output almost 100watts into 8 ohms, seconds later the bed is soiled and I back off the input signal to let her cool.

Interestingly, voltage at base of negative driver Q35 heads south (measured -12 v [briefly before quickly attenuating input] ) as does the voltage at base of positve driver Q34 (approx -10v),

Voltage at output goes up to approx 3v, correct me if I'm incorrect but it would seem bottom MOSFETS resistance increase and top MOSFETS decrease, thus voltage divider between + and - rails is no longer balanced and output rail is closer to + supply.


Power supply is fine as L channel outputs up to 180 W into 4 ohms.


If the diff amp was acting up would this not manifest when output is unloaded ?


ANd why is the negative side squaring when + MOSFETS are being pulled into cutoff?

Steve, image below is not in focus but best I could do in the moment.



Matt