Ampeg BA115HP Bizarre happenings!

[Sowhat]

I have an Ampeg BA115HP in for repair… complaint is that it's dead. The customer told me that he brought it into another well respected shop a few weeks ago and they fixed it with the same complaint and it worked for about a week before it blew again. He did not want to bring it back to them so it came to me. I went through it and several things were bizarre. Before I get started, here's the schematic:

457SCH_0.pdf

First it had a 3 amp fuse, 120VAC 60Hz calls for a 4 amp fuse. Power FET's Q8 and Q2 were shorted… the ones removed in the following picture. Most bizarre was the primary wiring for the power transformer… it was just WRONG!!! It was some mismatch of 240 volt wiring but not exactly, it actually looked randomly wired as if someone was in there mucking around and forgot which wires went where.



First I put in a 4 amp SB fuse and replaced Q8 and Q2, actually I replace all 4 output FET's… they're cheap. I tested the power resistors, diodes and driver transistors, all looked good so I wired the transformer up according to the schematic and before I hooked up the secondaries I tested the scheme which worked correctly now it gave me a little under 100 volts AC at the secondary, that's what it should be. After attaching the secondary I ramped up the variac slowly and everything looked good, as I ramped up it exhibited some small millisecond voltage drops but that settled down the higher the voltage got. I attributed that to flakey contacts in my well used variac. Once up to 120 volts I shot a sine wave into the input and got a bigger sine wave out of the output, pretty much what you would expect. I had not set the quiescent current yet but on the scope I noticed something a little strange with the waveform.



The crossover distortion artifact was way up from the mid point and there was a little notch at the peak of the positive going portion of the wave. I hooked up to the .33 ohm 10 watt source resistors as instructed and I should adjust the trimmer for 10 millivolts. Problem here was I could not read any millivolts at all. I decided to tweak the quiescent current trimmer anyway and I could see the crossover distortion increasing and decreasing, all of a sudden SNAP!!! My variac fuse blew. It's a 5 amp fuse and the amp has a 4 amp fuse which didn't blow… whatever. Post mortem is that the same power FET's blew… Q8 and Q2. Now I know you might say: "why didn't you use your current limiter" and at first I did but it limits at 2 amps max and hampers the proper operation of this 4 amp amp, so I bypassed it… whoops!, it looked like it was OK to go straight on the mains, guess I was wrong about that.

There's something taking these two FET's out and I believe it has something to do with that strange output waveform, problem is as to what could be wrong… any ideas?

[The Dude]

It could be many things, but I'll say this: It's my experience that the bias trimmers in those amps are horribly intermittent. Sometimes cleaning them works, other times, I've had to replace them.

[g1]

Not sure what 'instructions' you were referring to about setting the bias and using source resistors?
It sounds like you were adjusting the trimmer under load with signal?
The schematic calls for setting the bias with amp cold and no signal applied.
Like The Dude mentioned, these bias trimmers are often problematic and any glitch with signal and load could easily lead to fuse blowing or worse.
All that said, I don't think the funny waveform glitch has to do with bias adjustment.

[Sowhat]

You are right and I had signal flowing but I was referring to a little snooping I did after I initially went to set the quiescent current for the output MOSFET's per the instructions on the schematic, no signal, no load. When this did not go to plan I decided to take a glimpse and shot some signal into it where I noticed the radically offset crossover region. Before I could shutdown, the inevitable happened, that's all... actually, that's not all, let me explain:

This amp has it's power supply spread across both the main board and the preamp board making isolating rails a bit of a creative chore without blowing something up. First I disconnected all the +-50 volts junk by basically removing the 4 power MOSFET's. The driver transistors still had +-50 on them but their bases were biased by the +-16 volt rail mainly used for the preamp. I checked the +-16 volt rails and I got +2 volts and -16 volts, this explains the offset crossover region I saw on the scope. I replaced the two 16 volt 5 watt zeners that create this rail... no change. I then removed the two driver transistors and like magic my +16 volt rail came back. So, it's the driver stage that is at fault, but wait a minute, I already checked those transistors and they check out OK, at least with my basic component tester which only has 9 volts to work with. As it would have it I don't have any replacements in the shop so... the weekend is upon me and I have a Markbass LM 2, I think, service manual and schematic to ponder. Tuesday morning I will pickup a couple NTE's on my way to the shop and this plot will thicken.

[Sowhat]

As far as the bias... is it bias or is it quiescent? My freinds down under and across the pond say quiescent and my friends in the states say bias. If it's a tube job I say bias and if it's a SS job I say quiescent, they are really both the same thing, right? setting the idle current... anyway I was checking the quiescent current trimmer with my little bit of snooping and it looked to be working A-OK!

[Enzo]

How about this:

bias is the thing that controls idle current. In a tube the bias voltage on the grid controls the cathode current through the tube. So while we all understand what you mean if you say I have 38ma of bias on my 6L6, that is incorrect.

In a solid state amp, the current through the output emitter resistor is the idle current, and that is controlled by the bias transistor further back adjusting the driver base to base voltage.

In either case bias is what controls the current you are measuring. The current itself is not the bias.

[Sowhat]

It's been a week of interruptions but I have been able to work on this amp a bit. I have swapped out all the silicon in the power output section: Q1,2,3,4,5,7 and 8 as well as Diodes D6,7,8,and 9. I have checked all the resistors and capacitors and everything looks good. I can power it up and I can measure the idle current on source resistors R75 and R72 and set 10 millivolts across each of them and boy! is that bias pot touchy! Source resistors for Q2 and Q8 however do not measure so nicely... R73 (for Q2) measures barely anything while R74 (for Q8) measures over 20 millivolts and get's a little warm. I don't leave it on for any length of time and still power up on a limiter for the brief periods when I take measurements. I have even pulled out the microscope and examined the circuit board for cracks, none found. This amp will actually run for a while before it shorts out and this is probably what happened with the previous repair attempt by the other shop. There's something fundamentally amiss here and I'm just not seeing it... help?

[nevetslab]

A couple things....first, are you re-using the same insulator pads under the MosFET's? Once thermoset has taken place on those, replace them. I'd use fresh insulators just as a normal rule on power devices. Greased mica works fine, has better thermal conductance, though a pain due to being messy.

Second, I didn't hear anything mentioned about matched pairs. I'm normally servicing the SVT4-Pros and their cousins that use multiple MosFET's. It's always recommended using matched sets, which is, of course very expensive. I buy in qty, then grade them batch them for their Drain current at specfic Vgs and Vds settings, then measure their Vgs at a specific Drain current and end up with grouped & matched N-Ch and P-Ch MosFET's. The results vary wildly in the groupings, but I still get enough yield for similar N-Ch & P-Ch matched parts. The current sharing is far better than just taking them out of the static bag and plugging them in with a pair that are already installed.

Have you tried just using one pair of N-Ch and P-Ch parts and see what you get.

I'm also puzzled by the notch at the positive peak. Any chance that's oscillation? Granted, no RF envelope is present, but it has the trend one sees in instability.

[Sowhat]

A couple things....

1. Yes, I'm reusing the insulator pads: I cleaned them, inspected them, and re-greased them and stuck them on random FET's... not in the same order, and not by design, although come to think of it, maybe that's it, I don't know at this time, but something to consider as I proceed.

2. Ahhh, matched FET's, yeah, well... I have matched FET's for turn on voltages under load before but unfortunately, this time I didn't, you can call me a cheapskate, but there it is. If FET's are reasonably matched, and that's a loaded proposition, they still will bias up to a reasonable level, not radically caddy-whumpus like this. You're exactly right about matching, I just don't think that's the problem here.

The P-Ch and N-Ch pair would be a good test but which one? Q2, an N-Ch is in the lead pair and Q8, a P-Ch is in the trailing pair and these two culprits are the ones that ain't behaving right. That's what's got me currently confused... Q2 has literally no bias voltage and Q8 has at least twice the bias of Q5 and Q7 which are pretty much dead on to each other, I initially set bias for 10mv which is the low end of the acceptable range.

The notch... with biasing like this anything's possible when you pump up some watts. I think that if I can get 10,10,10 and 10mv or anywhere near that I will have fixed the notch and crossover region problems.

I'm going to try just the first pair like you suggest anyway, that might reveal where the fault lies.

[Sowhat]

Well... finally got back to this one and decided to blow the dust off my matching rig. It's the one described by Rod Elliott of Elliott sound products in Australia, I made it a long time ago but frankly have not used it in a while... I even canibalized a switch from it for some other project in the past.



I only have 4 N channel FET's and 4 P channel FET's right now... more are on order. My matching tests are shooting for 100ma quiescent and here's my results on this small group of FET's:

IRFP104N (N-Ch)

1 109.2
2 1544.1
3 85.3
4 93.4

IRFP9140 (P-Ch)

1 184.2
2 600.5
3 189.0
4 194.3

FET's 1&2 of each type were the ones I had installed in the amp and it becomes obvious what's causing the imbalance... nevetslab, you hit the nail on the head. Actually I have not seen this big of a mismatch ever... I would have been better served by installing FET's 3&4 from each type. I'm going to wait until my big order of FET's arrive before proceeding on this amp, maybe I can find even better matches.

[Sowhat]

Now that I have got my big order of FET's and matched them, everything is now normal, the amp work properly although I have not yet burned it in yet, but I'm confident. Got nearly a perfect 10,10,10,10. I goosed that up to 15ma or there-a-bouts and that was that.

I don't want to leave the impression that you absolutely positively must match your ganged FET's. At one time it was death to not do so but manufacturing tolerances have steadily got better over the years until you could pretty much bet on having a close-enough match right out of the box especially from the same run. Not perfect but passable considering the average lifetime of cookie cutter musical instrument gear. This was the case for me so much that as you see I had to "blow the dust" off of my matching rig, it was not because I didn't work on ganged FET outputs, I do them regularly, this is actually surprising to me based upon experience but then again, reality is in no way obliged to conform to my misguided notions.

[ziggy007]

Wow, Extremely similar issue on same amp. I've just replaced the two P Channel Fets and whilst adjusting bias on cold amp for 15mV I noticed similar large differences across the resistors. Then the fuse blew!! Its still blowing, with the inrush Fuse starting to smoke before the main T2.5A fuse goes Pop. (I'm in oz). Thanks for the food for thought SoWhat. I'm now going to check that bias pot. Maybe its the culprit!

[J M Fahey]

Not talking about the biasing problem but the notch on top of the waveform: this is a well known artifact of using "switching/vertical" MosFets into Analo0g Audio (ClassAB) duty, they are not really suited for that and this shows it.

Basically such Mosfets need huge Vgs voltage to turn ON , we are talking 4.5 to 6V (or more) for 10A class current, if you can´t supply them Mosfet simply hits a brickwall and does not move on ... while those 4.5 to 6V are *ample* for the driver transistor to keep working and supply significant current.

So in the linear region MosFet provides all the muscle needed and current is acopy of input signal ... but when MosFet runs out of steam it simply saturates, becomes a very good imitation of a piece of wire ... and Driver current passes through it into the load.

But such current is out of phase so instead of the missing "mountain top" you get a notch.

There are a couple solutions for that, from excellent but expensive "extra higher voltage rail", say 5V more, feeding just the drivers, so they can go beyond main rails and properly saturate output MosFets, you see this on some amps (extra cost and complication) or cheaper (what I do) clamping driver some 5V below rail, so both driver and MosFet stop at about the same time= no notch which sounds buzzy , by ear it cand be confused with crossover distortion which again is ugly buzzy, but the scope tells the difference.