I'm building two amplifiers that are similar in many ways, and would appreciate some advice.
One is a normal audio amp based partly on my own design, partly on Douglas Self's book "Audio Power Amplifier Design 6th ed.", with input from articles in this diyAudio forum and Bob Cordell's "Designing Audio Power Amplifiers 2nd ed." and from other sources.
The other is a specialized power supply whose core is a simplified version of the audio amp. Vout is +-50V, Iout is +-6A for any Vout. This can be called a bipolar power supply in terms of being able to source or sink current at any Vout.
Both use more or less the same general topologies in the core gain and output stages. I'm using LTspiceXVII to simulate.
This my first cut at building a discrete amplifier of this power, but I have 35 years of experience with major electronics companies designing analog ICs for various purposes. I'm considering the power supply to be kind of a dry run for the audio amp since it's quite a bit simpler.
At this point I'm trying to decide what transistors to use in the high voltage sections. For now I'll ignore the audio amp and just consider the power supply (which is essentially a specialized audio amp).
The VAS and output stages are based on Self's "Blameless" amplifier (cf. Self p. 311). The VAS is conventional. The output is what Self calls a CFP (Complementary-Feedback Pair).
My tentative choices are:
VAS devices: ksc3503ds ksa1381estu from ON
Drivers for power devices: mje340 mje350 from ON
Power devices: 2n3773 2n6609 (8 of each) Ebay/China special
For circuitry not in signal path: ksc2690AYSTU ksa1220AYS
Alternatives:
VAS: ON ksc2690AYSTU ksa1220AYS OR Toshiba ttc004b tta004b
CFP driver: Toshiba 2sc5171 2sa1930 OR Sanken 2sc4883A 2sa1859A
Other relevant specs:
Rails: +-54V to +-62V depending on load
Power transformer: Antek 600VA 45V (2 secondaries)
Bridge Rectifier: Taiwan Semiconductor GBPC5002 50A 200V
I will appreciate any comments.
Alan
One is a normal audio amp based partly on my own design, partly on Douglas Self's book "Audio Power Amplifier Design 6th ed.", with input from articles in this diyAudio forum and Bob Cordell's "Designing Audio Power Amplifiers 2nd ed." and from other sources.
The other is a specialized power supply whose core is a simplified version of the audio amp. Vout is +-50V, Iout is +-6A for any Vout. This can be called a bipolar power supply in terms of being able to source or sink current at any Vout.
Both use more or less the same general topologies in the core gain and output stages. I'm using LTspiceXVII to simulate.
This my first cut at building a discrete amplifier of this power, but I have 35 years of experience with major electronics companies designing analog ICs for various purposes. I'm considering the power supply to be kind of a dry run for the audio amp since it's quite a bit simpler.
At this point I'm trying to decide what transistors to use in the high voltage sections. For now I'll ignore the audio amp and just consider the power supply (which is essentially a specialized audio amp).
The VAS and output stages are based on Self's "Blameless" amplifier (cf. Self p. 311). The VAS is conventional. The output is what Self calls a CFP (Complementary-Feedback Pair).
My tentative choices are:
VAS devices: ksc3503ds ksa1381estu from ON
Drivers for power devices: mje340 mje350 from ON
Power devices: 2n3773 2n6609 (8 of each) Ebay/China special
For circuitry not in signal path: ksc2690AYSTU ksa1220AYS
Alternatives:
VAS: ON ksc2690AYSTU ksa1220AYS OR Toshiba ttc004b tta004b
CFP driver: Toshiba 2sc5171 2sa1930 OR Sanken 2sc4883A 2sa1859A
Other relevant specs:
Rails: +-54V to +-62V depending on load
Power transformer: Antek 600VA 45V (2 secondaries)
Bridge Rectifier: Taiwan Semiconductor GBPC5002 50A 200V
I will appreciate any comments.
Alan
Besides deciding transistor type, I consider such a power supply a high power Class A amp or worse, much worse.
Imagine it delivering 5 or 10V out @ 6A: relevant half will be dissipating up to 300W **continuously**
, without the relief of switching load 50% of the time to the other half (Audio duty) nor the relaxing effect of providing a sinewave to load.
It will all be brutal DC conditions.
Have you considered proper cooling?
And SOA?
Not sure simulators consider dissipation or heatsinking.
EDIT:
Imagine it delivering 5 or 10V out @ 6A: relevant half will be dissipating up to 300W **continuously**
, without the relief of switching load 50% of the time to the other half (Audio duty) nor the relaxing effect of providing a sinewave to load.It will all be brutal DC conditions.
Have you considered proper cooling?
And SOA?
Not sure simulators consider dissipation or heatsinking.
EDIT:
No no no no no.
Last edited:
Besides deciding transistor type, I consider such a power supply a high power Class A amp or worse, much worse.
Imagine it delivering 5 or 10V out @ 6A: relevant half will be dissipating up to 300W **continuously**
It will all be brutal DC conditions.
Yes indeed.
Yes. I have honking great heat sinks, and have used the heatsink simulator recommended by Heatsinks USA to make sure that cooling is adequate at maximum power dissipation in the output devices. There are 8 npns and 8 pnps, half of which will be on at any one time. That's 0.75A per device with worst case about 70V across them, for upwards of 50W per device. SOA is marginal but ok. However, in practice I can't think of many situations where the machine will be stressed this hard.
You have to be smart about it. Like I said, Heatsinks USA suggests a simulator.
EDIT:
Why?
Like I said, this is my first attempt at doing this. I know that a lot of Ebay transistors are fake, but are these particular devices really that bad?
I had bought 20+20 devices for $70 some time ago. Proper(?) devices from Digi-Key / NTE Electronics would cost close to $200.
I suppose I'll learn, one way or another. But that's why I'm reading this forum.
The GBPC5006 Bridge rectifier is a 600v device for about the same price.
The thing is, nobody knows how bad they will be. You might find a gem, but more likely you'll end up disappointed with performance or reliability issues. Ebay and the like are clearing houses for manufactured parts that do not meet the specifications.
If you buy from a reputable dealer, you know that they should be good parts.
The 2n3773 and 2n6609 can be had for around $4 each, which is pretty reasonable.
The thing is, nobody knows how bad they will be. You might find a gem, but more likely you'll end up disappointed with performance or reliability issues. Ebay and the like are clearing houses for manufactured parts that do not meet the specifications.
If you buy from a reputable dealer, you know that they should be good parts.
The 2n3773 and 2n6609 can be had for around $4 each, which is pretty reasonable.
Actually that many in parallel is more than enough SOA. In practice, four would be enough. I ran +/-60V rails on several home brew amps back in my college days, and they used banks of 4 NPN and 4 PNP. Either D424/B554 or 3773/6609 outputs. Got a really great deal on 80VCT 8 amp transformers for $15 each and built a bunch of amps out of them. Similar power to crown DC300A, except I used bigger heat sinks. DJ’ed with them, even bridging into 4 ohms and not a lick of trouble with them. Blew many speakers with them - some intentionally.
I would be worried about E-bay transistors. You can’t get an original 2N6609 anymore. Most are outright fakes. You can get copies (ie, Mospec) and they seem to work ok, but aren’t really any cheaper than just buying MJ15004’s. You still have to get them from a reputable supplier or destructively test samples to be sure they’re good.
MJE340/350 are not quite strong enough to use as drivers. Not here, anyway. You will over dissipate them. How do I know? Those amps I was telling you about earlier used output triples with MJE340/350 predrivers. I had a mistake on one of the PCBs where one driver transistor was shorted base to emitter and was effectively out of the circuit. The amp still worked for 2 years, but had a lot of distortion with loads of less than 4 ohms on that one channel. The “predriver” ran blazing hot and eventually clued me in on why that channel acted weird. Removed the short on the board, replaced the predriver for good measure, rebiased it, and all was well.
If they are fake it’s a good bet that even 8 of them in parallel won’t be enough. I had a batch of fake MJ15024’s that wouldn’t current share. They didn’t even have the same dies inside them - even though markings were identical. Yes, I cut them open to see.
The chances of a fake 2N3773 are better than 50%. The chances of a fake 2N6609 are much closer to 100%, unless what you do find are old Motorola pulls. I do find them occasionally, and when I do I buy them. All of them, because it’s usually just a handful. Every time I’m in Skycraft I check a short list, and sometimes get lucky. E-bay? Never.
The chances of a fake 2N3773 are better than 50%. The chances of a fake 2N6609 are much closer to 100%, unless what you do find are old Motorola pulls. I do find them occasionally, and when I do I buy them. All of them, because it’s usually just a handful. Every time I’m in Skycraft I check a short list, and sometimes get lucky. E-bay? Never.
Ok, I ordered some. Since my rails are +-62V nominal worst case, and allowing for about 10% slop, shouldn't the 200V of the GBPC5002 be sufficient? Or should I want to leave more headroom?
Given the comments below, I can see why. I suppose "a penny wise and pounds foolish" applies here.
Ah, I see! I thought that most of the cheap stuff was just manufacturers getting rid of overstock on aging items. Or the outright fakes.
I found some at Digi-Key from NTE Electronics for $4.40 and $4.91 -- close to half the price of the devices from other manufacturers. Is there anything for me to be worried about if I buy them?
I also found some devices from other distributors like Newark from outfits like "ELECTRONIC PRO" and "SOLID STATE" for less than 2/3 the Digi-Key price. I know nothing of the reputation of Newark or these other outfits. I suspect that, on general principles, I should avoid them, no?
Not when I combine the SOA specs with the Heatsinks USA thermal simulator. I tried four and, even with my honking great heat sinks, temperature soared to about 250C. After going around and around a few times, I hit on eight devices, which even with forced air cooling resulted in about 70C temperature at the device locations. I might be able to find a better physical placement, though.
For me, that's before PCs.
Cool!
What about the ones from NTE Electronics that Digi-Key sells?
Which brings up the issue of What, exactly, is a fake? Way back in the early 1980s, Signetics came out with their NE5534 op-amp family, which was extremely expensive. Today, Signetics is gone, but everybody and his brother sells these devices. I know for certain that Signetics did not sell its IC process to a zillion manufacturers, so in effect, nearly all NE5534's today are "fakes". Comments?
Specwise, the MJ15003/MJ15004 and MJ15023/MJ15024 are roughly comparable to the 2N3773/2N6609 pair. Since my present project is a DC power supply, what do you recommend?
I'm trying to be careful to use LTspiceXVII to measure PD, and all critical specs, under all conditions. The work is not yet completed since I'm still learning the program.
If I understand you right, you used three MJE340s in parallel, but the mistake was that one device was out of the picture. So the two remaining devices ran way hot. Right? So how does this show over-dissipation in my circuit?
Well, there are fakes, and then there are fakes. I'd be happy to own a fake Shelby Cobra GT500.
Wow! Since I'm now stuck with 20 of these things, is it worth my time to try to test them?
A fake is a part they labelled as something it is not. A fake is not a part someone made to original specs but just happens not to be original.
The 5534 I buy from MOuser is not a fake, someone somewhere has the authority to be a "second source", and they also make real 5534s. A fake 5534 is where someone got a big box of 8-leg ICs, prints 5534 on the top and sells them online to naive buyers. Signetics was bought by Phillips. Now the 5534 is a generic and made by many manufacturers.
People used to make body kits for VWs that looked like fancy race cars, or my fave, the ROlls ROyce hood for the front. That isn't fake, it is a parody. On the other hand, if I took a Ford Falcon, and had someone make a body for it that looked like a Shelby, and ran ads online offering a Shelby, That would be a fake. You buy it for ONLY $300,000 and when it arrives, you find a fancy looking Ford Falcon instead of what you thought you bought.
My design work is not in high power SS amps, but in my experience, when I see TO3 outputs in a row, I often see a TO3 driver to serve them. ANy reason to use parallel multple smaller transistors like the MJE3xx types instead of a larger one?
The 5534 I buy from MOuser is not a fake, someone somewhere has the authority to be a "second source", and they also make real 5534s. A fake 5534 is where someone got a big box of 8-leg ICs, prints 5534 on the top and sells them online to naive buyers. Signetics was bought by Phillips. Now the 5534 is a generic and made by many manufacturers.
People used to make body kits for VWs that looked like fancy race cars, or my fave, the ROlls ROyce hood for the front. That isn't fake, it is a parody. On the other hand, if I took a Ford Falcon, and had someone make a body for it that looked like a Shelby, and ran ads online offering a Shelby, That would be a fake. You buy it for ONLY $300,000 and when it arrives, you find a fancy looking Ford Falcon instead of what you thought you bought.
My design work is not in high power SS amps, but in my experience, when I see TO3 outputs in a row, I often see a TO3 driver to serve them. ANy reason to use parallel multple smaller transistors like the MJE3xx types instead of a larger one?
If you assume it is going to dissipate the maximum possible average power, as in Vcc^2/(10*RL) in each transistor bank, sure. You almost can’t buy enough heat sink to dissipate that much continuously. Real world average, even wen blasted, is far less. If it wasn’t, you can’t even supply enough current off a 15 amp breaker to power the amp. What is of interest is the maximum instantaneous, which is 480 watts with a 2 ohm resistive load (which is the same as 4 ohm at 60 degrees). That needs to be inside the DC SOA, but only happens for a few milliseconds at a time. The transistors will handle that as long as the average is low enough that the heat sink can handle it. Your heat sink does not have to handle 480 watts (or 480 x2), or even the 192 watts (x2 ) maximum average. Real world average will be more like 100 watts and you can certainly buy a heat sink that will do that. If you need more because you expect to be clipping the **** out of it, get a fan.
People who make 5534’s make them under license. You can also get fakes which might be a 741 marked as a 5534. What you get with power transistors is some cheap under-related part in a very poor package. They will fail quickly, if not instantaneously. NTE is a crap shoot. They may be real, and just re-marked, something made by a second source under license, or Chinese fakes that won’t last 20 seconds. Solid State IS NTE. Business address is the same. Multicomp Pro is Newark’s house brand, and can come from second sources such as Mospec, CDIL, Central, or Solid State (NTE). They are not outright junk, but may not always be up to snuff either. I’ve used Mospec (even 2N6609’s) and they hold up as well as you would expect any generic 150 watt power transistor. But won’t quite take the torture you can give a real 1990’s Motorola. Inside the die is full size, but obviously different construction from the original.
No, I had a single MJE340 driving a C2565 (really old 150 watt Toshiba high ft type) driver, driving four 3773’s (triple darlington). When the driver was shorted base-emitter, the MJE tried to drive the four 3773’s. It could al lighter loads. Didn’t have enough current gain at 2 ohms. Not only did it distort, but the poor 340 got hot enough to blister fingers. Never blew it, but I still replaced it anyway when the amp was corrected.
Last edited:
No,it´s not 3 transistors in parallel but an current gain block divided in 3 stages.
for the simplest to explain example, consider the "F" one:
you know how Darlington transistors work: they are connected so Hfe, current gain , multiplies.
Suppose you have a Darlington made out of two transistors (the classic example) and each one has Hfe: 20 at its operating current.:
When power transistor (say, 2N3773) delivers 20A into the load, its driver (say, TIP31) must supply its base with 1A .
WAY too much for a feeble MJE340.
Even if you parallel 3 x MJE340´s , you are still murdering them.
Enter the Triplet, a "super Darlington" if you wish, and which you will find in MANY "muscle" power amps:
Now "power transistor" is made out of many (4 to 20) parallel 2N3773 which will easily provide those 20A, still needing 1A (total) at parallel Base, supplied by a single 2N3773 used as a driver, which is doing an easy job, and which will need 1A/20=50mA at its base ... which is an easy job for a single MJE340.
Compare both cases and now you´ll see why MJE340 is never used as a "direct" driver on its own in high power amps.
Yes, I know there are a couple such designs floating around.
All I can tell you, by experience, is that they may simulate good but do not survive live sound conditions.
PS: just noticed the imgur graphic was actually a copy from Rod Elliot´s one.
More power to him and his EXCELLENT site
All rights acknowledged.
Yes. What I would do in your situation is bias it as a 3 amp current source (for example, set base voltage to give 6 volts across a 2 ohm emitter resistor), and hit it with the 62 volt supply at the collector for a couple seconds. Mount the transistor to a heat sink, with the fins in an ice bath. If it dies, pop the lid and see what’s inside.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.