Consulting on an old Evans Custom Amp

[R.G.]

I got a call from an amp-tech friend of mine. He was fixing on an old Evans Custom amplifier owned by a pedal steel player. The amp had failed at a gig. The first symptom was, as he put it, "a small campfire started where the speaker was." The speaker was apparently a vintage unit that's highly prized by pedal steel players, and has no existing re-cone/repair kits.
The power amp had failed spectacularly, as early; 1970s amps were likely to do. All four output devices were shorted, the complementary drivers were shorted, and the voltage amplifier transistor was shorted. The input differential pair was damaged, and he could not get the amp to get a DC offset less than about 2V. This is a typical chain failure from the time of the first generation of differential-input power amps.

I got into it when my friend asked about the stubborn offset. He sent me the schematics and I started calculating. The amp has +/- 50V supplies and is rated for a 4 ohm speaker, which calculates out to about 300W, neglecting the minor losses. 250W probably. The amp uses MJ802s for all four outputs. It uses TIP31s for voltage amp and positive side driver and the TIP32 complementary device for the lower side/PNP driver. All of these devices are rated for (wait for it) 90 and 100V BVceo. The thing was running on the edge of its Vceo ratings the day it was manufactured.This is not even taking into account the safe operating area limitations of the TIP and MJ802 devices; SOA was not well understood in the late 1960s and early 1970s when this was designed. There was no particular current limiting or V-I limiting built into the circuits; the only attempt at this was a variant of "inherent current limiting" that relies on reverse biased diodes paralleled with the bias string and the voltage across emitter ballast resistors to give a quickie limit that might let the primary fuse blow. I think the idea was that a speaker short circuit could be held up by the 30+30A current ratings of the MJ802 output stages.

This version of the Evans amps apparently has a reputation for spectacular failures. I understand why.

I pointed the amp tech to the MJ15003 for outputs, even though they are "only" 20A rated; they are also rated for 150BVceo and a much bigger SOA. I suggested the MJE15030 and its complement for the drivers and voltage amplifier; these are also well over the 100V power supply.

I also rediscovered a part of why the offset voltage was so bad. A big thing in getting the offset down is to match not Vbe on the input transistors, but the hfes. I used to know that, but haven't played output amp design for a while. The input stage is a simple PNP differential pair with an emitter "current source" of a 22K resistor. The voltage amplifier drive stage is a simple 680 ohm resistor on the collector of the first input transistor; the second transistor has its collector connected to -50V. The diffamp in this kind of circuit is inherently unbalanced, so offset can be expected.

There is always the issue of fixing a power amp too much and getting complaints about it sounding too hifi or not as good as it did. The unbalanced input diffamp, the single voltage amplifier stage, and the quasi-complementary output stage certain contribute to a certain level of distortion. The bias spreader diodes are touchy, and it has a tendency to run hot, which can also "contribute" to the sound. That's all left as is. The updated amp will be much more reliable with modern and non-edge-of-its-teeth voltage rating devices.

I could not control my urge to fix it some more. The entire power amp circuit less the TO-3 outputs and one bias diode are on a 3 x 5" PCB which plugs into a 15 position edge connector. All the emitter resistors are on the card, so the edge connector contacts are carrying 6 and 7 amp peaks from the output stage. Longevity can't be that great. But I found myself doing a reinvention PCB that has the same PCB edge connector and much the same circuit, but with added current and/or V-I limiting circuits added.

It was a fun run through 50 year old technology.

[Jazz P Bass]

Nice post.
Thanks.

[Jon Snell]

MJ802 is the prefered component, high Hfe 200W dissipation and very robust.
I doubt there will be more than 75volts across it at any time so always use the original spec devices.
The model ... you don't disclose, so I cannot comment as to why it failed but replacing the output stage without finding out what caused it has its drawbacks. I don't like doing a job more than once!

I expect DC appeared across the speaker and fried the coil.

[drewl]

Sometimes output devices just die?

Got a Peavey Session 500 where just one of the T03's shorted.
Everything else tests fine. I replaced all eight in case others were stressed.

Actually have TWO Session 500's this week, the 2nd one was just a loose connector.
Heavy suckers!

[glebert]

Sometimes output devices just die?

Heavy suckers!

Pedal steel: where the amps are heavy, but the instruments are heavier! (sometimes)

[R.G.]

Sometimes output devices just die?

They do. They're subject to the same kind of bathtub-curve failure distribution that everything else is.

You have early-life "infant mortality", which drops to a low and nearly constant failure rate for quite a long time, then a steadily rising failure rate where more and more of them die with time. The bathtub curve model of failures is a heuristic, a recognized pattern that seems to fit the way the real world works. There has been a great deal of work on semiconductor reliability. In general, the hotter something gets, the more heat-cool cycles it has, the more mechanical stresses it has/contains, and the closer internal power peaks come to its limits, the faster it wears out. This is exponential - the "limits" are where even one peak causes time to fail to drop to microseconds.


The link below is a sleep- and headache-inducing intro to how some of this works.

https://www.renesas.com/us/en/docume...-failure-rates

Yeah, they do just wear out and sometimes randomly die. The skill of designers in keeping temperatures down and keeping operation away from the edges of their limits is what generally determines longevity.

[R.G.]

MJ802 is the prefered component, high Hfe 200W dissipation and very robust.

The MJ802 is a GREAT device, within its limits. I use it wherever I can. My issue is that the design flirts with the sudden-death limits of the MJ802. Referring to the OnSemi datasheet at
https://www.onsemi.com/pdf/datasheet/mj802-d.pdf
and squinting at the second-breakdown limits, you get into second breakdown starting at about 25V and 8 A, declining to roughly 350ma at 75V and zero current at 100V. The device can get into forward-biased second breakdown very easily with a 100V total power supply.

I doubt there will be more than 75volts across it at any time so always use the original spec devices.

In normal operation, at realistic levels (the thing is a ~300W amp after all) probably so. But any kind of funny inductive speaker load expands the ideal resistive load line out into an ellipse, so the current and voltage can get into second breakdown territory easily.
Actually, the lack of any specific protection circuitry is even worse. One of these at idle getting even an instantaneous short will have the output devices with 50V across them and only the wire resistances and emitter ballast resistors to limit current. The SOA limit on the MJ802 is a hair over 1A, so a short that causes as much as 2A to flow could tip both outputs into SOA death. In this combination, raw current limiting might not save even the 30A rated MJ802, as the current needed to get into SOA problems is so low. Single-slope V-I limiting might do it.

Using the original spec devices is generally the way to go, except when the original designer didn't have enough information to work with, as I believe happened in this case.

The model ... you don't disclose, so I cannot comment as to why it failed but replacing the output stage without finding out what caused it has its drawbacks. I don't like doing a job more than once!

I expect DC appeared across the speaker and fried the coil.

Near as I can tell, it was almost identical to the LB-500, possibly a later manufacture or modified version. The info you posted on a thread some time back was where I got started. I worked from top and bottom side photos and reverse-traced the circuit, then compared it to the LB-500 schematic. There were minor differences in component values and the biasing string, otherwise the same design.

I'm with you - just replacing the failed part without understanding the reasons for the failure can be a nightmare. In this case, every transistor after the input differential pair was dead, as was the campfire of a speaker. Understanding what happened to cause it was in fact why my amp-tech friend called me. He wanted to know how to figure out what started this. He really, really doesn't like doing the same job more than once, either.

I like the MJ802 and its complement the MJ4502 so much that I know their limits by heart, and the +/-50V supply led me to look at the component specs in detail. On most amps, this is a no-brainer; the outputs are operated safely inside their operating limits. For this amp, the outputs are right at the very edge of their BVceo limits, even if the power line doesn't happen to get a little higher than normal. But, as we all know, it does. This particular amp is so old that general understanding of SOA was not common, and the operating conditions of the original design flirt with SOA disaster.

You're absolutely right - DC on the output fried the speaker coil. "Campfire" was the term they actually used.

[Mick Bailey]

The pedal steel players I've done repairs for over the years have been remarkable in that no matter how burnt out or damaged an amp is, they want it fixed. To some degree, harp players I've known share the same view. Often I've explained that the amp will cost far more to repair than to replace with a working used example, but the customer has gone ahead with the repair anyhow.

[loudthud]

In normal operation, at realistic levels (the thing is a ~300W amp after all) probably so. But any kind of funny inductive speaker load expands the ideal resistive load line out into an ellipse, so the current and voltage can get into second breakdown territory easily.

Second breakdown territory easily ? With an overdriven guitar you can get nearly rail to rail Voltage at over 1 Amp.

Wanna see what an overdriven guitar looks like to a transistor's SOA ? Check the video files in this thread: Note; output devices are common Source MOSFETs with series diodes in the Drain leads to allow inductive spikes from the speaker.

https://music-electronics-forum.com/...-files?t=35493

You may need Windows Media Player to view.

For SOA, On Semi NJW21194G / NJW21193G are the best devices I've found.

[R.G.]

You're right - there are much better output devices than the MJ15003. The 21193/4 pair are great. The MJ3281 and its complement are another, as are the MJ15030 and its complement. Toshiba and Sanken have also produced some fine examples.

I have to admit, I was soft-pedaling the suggested transistor substitution to try to keep the amp tone as similar as possible to what it was with the MJ802s.

The higher SOA devices tend to be "sustained-beta" types, where their hfe does not fall off nearly as much at high currents as any older-style power transistors. From reading about them, I gather that the newer devices are heavily subdivided on the chip, so that the single transistor chip is effectively an integrated circuit with many paralleled smaller transistors. From reading, I gather that second breakdown is a big-transistor-chip problem and that subdivision into zillions of small devices in parallel seems to sidestep the hot-spot issue. This is my generalized impression. I'm not at all current with actual power transistor semiconductor techniques any more.

For what little it's worth, I gathered up my notes on the design and PCB layout. The link below should take you to it. I did order some protos of the card. I cheated a bit on the card. The original card was 3 x 5 inches. I got that into 3 x 4 inches. That is too short to pick up the mounting standoff in the chassis, but it's probably OK for a quick and dirty test on my friend's repair work in progress. It's not usable in the amps in general without more mechanical support. The inch shorter was to hit one of the edges in price at my favorite PCB house. Even 100mils longer would have made the PCB order another $100 more expensive.

https://www.dropbox.com/scl/fi/w1mj7...=8zo7x251&dl=0