I've replaced output devices, schottky diodes, driver devices and input devices; I guess most all types of semiconductors have failed on one channel or the other, except for small signal diodes. In my younger years I built and played around with numerous amps and was not always as careful as I should have been. It is possible some of the failed devices were due to my fumble fingering something or other. At this point I don't recall - I can just see what was replaced. Given my intended use, which would be driving "normal" speakers at sane levels, I'm guessing a properly rebuilt Ampzilla would outlast me. I'm still on the fence, but leaning towards some level of fixing/rebuilding, so really appreciate everyone's comments on device selection.
Per the ON spec sheets, hfe looks better on the MJ21195/96, although spec'd at slightly different Vce:
MJ21195/96: 5V/16A - hfe>8
MJ15022/23: 4V/16A - hfe>5
Too heavy! It would break my trash bin and my refuse collector may fine me for excess weight!
Interesting. I guess I have not spent enough time looking at SOA curves. Assuming an 8-ohm load, 70V rails (mine are actually 69 and would probably droop a little but I'll ignore that), the voltage is equally spread across the stacked output devices (and I can still do simple math), then ... at an output power of 130W I hit the 1sec SOA curve (Vce=58, I=4). I guess 200W is a little optimistic, like everyone has been saying.
Output transistors have more gain now than 1974, so you don't need the high rail voltages to achieve max wattage limited by soa. Then OT's could get output about a third of the rail voltage, now devices can go about half rail voltage. (35v^2)/8=153W, divided by two output pairs, is 76W per pair. sqrt(153/8)=4.3 amps, kind of pushing dc soa at 70 v.
BTW, a 2n5631/6031 datasheet from motorola shows 0.95 A soa at 70v.
A PV-4C is 120 W/ch @ 8 ohms, and has 4 pairs of MJ15024/25. And a 5" fan.
So assuming your temp sense IC's are good (and I don't know how to test them) you could improve reliability of ampzilla by cutting rail voltage. Either regulate, which you don't have heat sinks for, or increase output pairs which you don't have heat sinks for, or a different transformer, which is $75 with freight. 55 v rails gets 6.9 A on two pairs or 3.5 each, down in the DC soa limit. I use the rail voltage when I'm reading the soa table incidently. Inductive speakers can increase the current over resistive loads. A +-55 v rail comes from a 41 vac nominal transformer with two windings, since they are sold by full power average voltage.
If changing to a toroid transformer, put a surge limiter NTC resistor in series with primary windings like a GE CL50, because they surge pretty badly at turn on. Toroids also pass lightning strike surges worse than E-frame too, IMHO, so I put a 275 vac MOS supressor primary in parallel after the fuse. Cuts the pop volume of a nearby strike, anyway. I solder these accessories on a cinch 4 terminal strip near the AC fuse.
And stuff some speaker DC protection in there too, see the $5 protection board thread. My speakers cost $300 each used, the amp was $100 + repair work.
BTW, a 2n5631/6031 datasheet from motorola shows 0.95 A soa at 70v.
A PV-4C is 120 W/ch @ 8 ohms, and has 4 pairs of MJ15024/25. And a 5" fan.
So assuming your temp sense IC's are good (and I don't know how to test them) you could improve reliability of ampzilla by cutting rail voltage. Either regulate, which you don't have heat sinks for, or increase output pairs which you don't have heat sinks for, or a different transformer, which is $75 with freight. 55 v rails gets 6.9 A on two pairs or 3.5 each, down in the DC soa limit. I use the rail voltage when I'm reading the soa table incidently. Inductive speakers can increase the current over resistive loads. A +-55 v rail comes from a 41 vac nominal transformer with two windings, since they are sold by full power average voltage.
If changing to a toroid transformer, put a surge limiter NTC resistor in series with primary windings like a GE CL50, because they surge pretty badly at turn on. Toroids also pass lightning strike surges worse than E-frame too, IMHO, so I put a 275 vac MOS supressor primary in parallel after the fuse. Cuts the pop volume of a nearby strike, anyway. I solder these accessories on a cinch 4 terminal strip near the AC fuse.
And stuff some speaker DC protection in there too, see the $5 protection board thread. My speakers cost $300 each used, the amp was $100 + repair work.
Last edited:
The significant SOA violations, especially with the original output devices (2N5631/6031), certainly lends some understanding as to why Ampzilla failures are common, especially in commercial applications. With my very low power usage the last decade or so I'm sure I never violated SOA, but after the earlier rebuilds I did a lot of high power testing so may have weakened the devices right from the start. Thanks for your help and insight. I worked for 40 years as a EE (RF comm equipment), retired a few years ago, and miss the daily technical challenges.
The original thinking was that each 5630 would see only 35 volts or less at any significant current. At 35V they are good for about 5A. But at 5 or 10 amps the gain falls enough that the base current upsets the bias thru those 2 watt resistors, and it doesnt divide evenly anymore. Put a heavy and reactive enough load on it and poof. The trouble with second breakdown is that you don't even have to run it hot to run into danger. With a standard parallel stage, it will divide the current evenly if you have a reasonable vbe match and a high enough emitter resistor value - even if the stage is running out of beta.
Leach solved tha problem in the Suoeramp by using a triple. But as many of them oscillated as didn't - it is real sensitive to construction technique.
Leach solved tha problem in the Suoeramp by using a triple. But as many of them oscillated as didn't - it is real sensitive to construction technique.
Last edited:
I couldn't believe the output transistor stacking was real. I thought perhaps though some off schematic wiring the output transistors were really parallel. 2n5631/6031 was only Vceo of 140 v, but with MJ15024/25 ON semi solved that problem, with 250 v Vceo. In 1974 there were various theories why transistors blew up - that was put to rest with Motorola's famous App note of about 1980 where 2nd breakdown & soa limits were introduced (no I don't have it).
Danschy, I'm a physicist, who haven't even read Doug Self or Bob Cordell's amp design books. I'm sure you could pass me up if you did. But people tend to obsess over 3rd decimal place harmonic distortion and various ideals, whereas I find some commercial products in the PA market quite acceptable sounding. Considering that speakers generally distort so badly. I'm not golden eared, but I am silver with 14000 hz response still age 67. I can quite easily hear the 1% hd of my tube amp, and have moved on to transistors and op amps in the 0.3% HD class & below. And if I want to hear some real sound, my piano is sitting right there, totally authentic 0.000% HD.
Danschy, I'm a physicist, who haven't even read Doug Self or Bob Cordell's amp design books. I'm sure you could pass me up if you did. But people tend to obsess over 3rd decimal place harmonic distortion and various ideals, whereas I find some commercial products in the PA market quite acceptable sounding. Considering that speakers generally distort so badly. I'm not golden eared, but I am silver with 14000 hz response still age 67. I can quite easily hear the 1% hd of my tube amp, and have moved on to transistors and op amps in the 0.3% HD class & below. And if I want to hear some real sound, my piano is sitting right there, totally authentic 0.000% HD.
Last edited:
In 1974 140 vceo was high for a PNP. So was a 30 volt 2nd breakdown point. They had NPNs that were more rugged, but they were only NPNs and had fT around 200 kHz. There were other ways of blowing them up - playing it loud and running multiple speakers wouldn't do it, but running a 20 kHz square wave with no load sure would. The Mj15024 came along and solved that problem. but the pnp still took a while - it wasn't available immediately.
I once tried building a stacked output stage out of all NPNs. Don't try it. You can't make it stable. There's a reason why no one does or did it.
I once tried building a stacked output stage out of all NPNs. Don't try it. You can't make it stable. There's a reason why no one does or did it.
That was a long time ago where transistors were scarce. Today, we have the luxury of transistors that can operate at +-70V rails.
I personally would change the output to a more conventional design with one pair of drivers and 4 pairs of outputs. I would change the bias as well.
This would of course mean that the amp needs to be re-stablized.
All versions of vintage Ampzilla heatsinks only have room for four output transistors per channel. All of the related amplifiers, Dynaco, SAE, and Ampzilla have the same output arrangement, they are what they are. There's not much that can be done without drastic mechanical changes that will ruin any value the amplifiers may have. Like I said previously there are plenty of all those amplifiers still out there working fine and people enjoying them. Obviously nobody would design an amplifier that way today but there was not much choice 40-50 years ago. Let's help the OP repair/rebuild his the way JB made it without redesigning this classic.
Craig
Craig
Thanks Craig. I have greatly enjoyed the discussion, and appreciate everyone's comments. The soa violations were new to me, so feel I now have a much better appreciation for the amp's limitations, and also suggestions for better devices. I agree with your opinion, and will likely do a minimal rebuild, recognizing that anything much above 100w-8ohm will potentially shorten it's life, but I don't see myself ever going there outside the test bench.
Have some other stuff going on so haven't had a chance to troubleshoot my Ampzilla, but I have a good idea for replacment power devices based on above comments. Any suggested for replacement for schottky diode 1N5823? Just use 2N5551/5401 for the input devices (UPSU06/56)?. A few years ago I replaced the fuse bulbs for the meter panel with some LED bulbs, but they have also died. Any recommendations? Thanks!
If you look at the 2N5630 and consider that the output stage is stacked it will handle
about 4A at 40V (roughly half of the supply) or 160W for just a minor SOA derating from
the 200W spec.
MJ15024 handles about 6A at 40 volts, 240W and has not hit the SOA limited area of the curve at that voltage.
Similarly the MJ21195 shows about 5.5A at 40V, 220W on the SOA curve and is also not limited.
Both of the newer parts should do their rated 250W and they would be like having 250/160 = 1.6 times as many of the older output devices.
Also consider that straight thermal derating is required, fan cooling is a big advantage over
other power amps without fans.
Bongiorno probably recommended the MJ21195/96 over the MJ15022/23 because of their higher speed and therefore a reduction in cross conduction.
about 4A at 40V (roughly half of the supply) or 160W for just a minor SOA derating from
the 200W spec.
MJ15024 handles about 6A at 40 volts, 240W and has not hit the SOA limited area of the curve at that voltage.
Similarly the MJ21195 shows about 5.5A at 40V, 220W on the SOA curve and is also not limited.
Both of the newer parts should do their rated 250W and they would be like having 250/160 = 1.6 times as many of the older output devices.
Also consider that straight thermal derating is required, fan cooling is a big advantage over
other power amps without fans.
Bongiorno probably recommended the MJ21195/96 over the MJ15022/23 because of their higher speed and therefore a reduction in cross conduction.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.