Replacement power transistors for old JC-120?

[uvacom]

I'm looking for some transistors to repair my very old ('79-ish) JC-120. It's actually a JC-120A, it was for the asia-pacific market and I got it from this guy who was in Little River Band (australian band). The power transistors are Toshiba 2SC1625 and 2SA814 (a complementary NPN/PNP pair - datasheets attched), both of which are long out of production and from I can tell virtually impossible to find on the obsolete market. However, being as they are power transistors I think they should be fairly replaceable by some modern part, and I've found the 2SC4793/2SA1837 which seem to be compatible in every spec, but I don't have much experience in component selection so I wanted to check here first. I've linked to the datasheets for anybody who wants to help me out.

2SC4793
2SA1837

My biggest concern is the Vcesat(max) spec - it's much higher (1.5V) than the original transistors which were rated at .5V. Don't I want the saturation voltage of Vce to be as low as possible to allow greatest headroom?

[Phostenix]

It looks like those are the driver transistors, not the power xstrs. I'll let the more experienced techs give you a definitive answer, but I'm a little concerned about the lower SOA on the replacements. They look like a more robust xstr based on the voltage ratings, but the SOA takes a dive 10 volts earlier. I'm not sure if drivers run that close, but that's what jumped out to me.

[oc disorder]

I think Phostenix is right they are the driver transistors.
See the attached (Roland-Layout-'79)they (green dots)are mounted on the circuit board near the emitter resistors (red squares) for the real output transistors.
The output transistors should be mounted on the chassis and wires connected to them should terminate on the terminals I've marked with blue dots.
Not sure exactly which terminals but you have it in front of you and I dont.
In the past I have used TIP 41C and TIP 42C general purpose replacements
for the drivers and had no complaints.
The serial number should help you track down the correct schematic.
I notice on the '79 one the driver transistors (hard to read)are listed as
2SC1624 and 2SA814 and the output transistors are listed as 2SD425.
On the '82 model the drivers Q18 & Q20 are listed as 2SB649 and 2SD669
The outputs listed as 2SD845 or 2SC3182,2SC2773 ,2SC2564(pick any one).
Likewise on the '84 model the outputs are listed as 2SD845 or 2SC2773
(they dropped the other two off the equivalents list).
The '79 one's outputs 2sd425 should be in a TO3 package and an equivalent
that comes to mind is the 2N3773.See the data sheet for a picture of the TO3 package.
The 2SC2773 outputs are in a different package rectangular and Sanken make replacements.
Hope the above is useful and not just a rant!
I recall seeing LRB when they had the Rolands they had ditched Glen as a singer and had John Farnam as they were trying to take over the LA freeway music market. Well I suppose you could play their songs on a dirt road in the outback ! All their backline were wired for 120 volts so they carted around giant stepdown transformers when they toured down under.Much preffered Glen as a singer he had a more country approach and had on board influences from "The Band" Music from Big Pink etc .

[uvacom]

Good observation, maybe I'll need to shop around for transistors with a lower Vce(max) but better SOA curves. There just is not very much in terms of selection for transistors with this kind of power handling with a high enough hfe. Although, I am wondering how crucial the hfe specification is in this circuit.

What really defines a "driver" transistor vs. a "power" transistor anyway?

[oc disorder]

What really defines a "driver" transistor vs. a "power" transistor anyway?

Price !

Their position in the circuit ... usually the power trans are higher current and voltage.
Same goes for motor servo drivers etc

You could design a circuit using the drivers above as outputs.
But it wouldnt have the wattage of the above just with a pair.

[uvacom]

I think Phostenix is right they are the driver transistors.
See the attached (Roland-Layout-'79)they (green dots)are mounted on the circuit board near the emitter resistors (red squares) for the real output transistors.
The output transistors should be mounted on the chassis and wires connected to them should terminate on the terminals I've marked with blue dots.
Not sure exactly which terminals but you have it in front of you and I dont.
In the past I have used TIP 41C and TIP 42C general purpose replacements
for the drivers and had no complaints.
The serial number should help you track down the correct schematic.
I notice on the '79 one the driver transistors (hard to read)are listed as
2SC1624 and 2SA814 and the output transistors are listed as 2SD425.
On the '82 model the drivers Q18 & Q20 are listed as 2SB649 and 2SD669
The outputs listed as 2SD845 or 2SC3182,2SC2773 ,2SC2564(pick any one).
Likewise on the '84 model the outputs are listed as 2SD845 or 2SC2773
(they dropped the other two off the equivalents list).
The '79 one's outputs 2sd425 should be in a TO3 package and an equivalent
that comes to mind is the 2N3773.See the data sheet for a picture of the TO3 package.
The 2SC2773 outputs are in a different package rectangular and Sanken make replacements.
Hope the above is useful and not just a rant!

That helps somewhat, actually. My amp is the same rev as the 79 schematic that's out there, but the 2SC1625 is used instead of the 1624. I think the main difference is in the max Vce allowed. I should note these transistors are the "y" variant with an hfe range of 120-240. I'd be more than happy to use TIPs if the much lower hfe isn't a big deal.

[uvacom]

That makes sense. I was calling them power transistors because that's what we use them for in school, but for much smaller power applications.

Question: In looking at datasheets, why does the graph for hfe vs. Ic show much higher hfe values than the minimum specification? Is it simply because those are "typical" values vs. the minimum values?

Anyway, after thinking about it a little bit more there shouldn't be a problem with the Vce saturation voltage difference. The rails are at +-43v, so the input voltage swing would have to be absolutely huge to saturate those output transistors. The design probably doesn't allow for that but even if it does, I won't ever play it that loud. This JC-120 has JBL E120s in it, it's ****ing loud! I'm starting to think the hfe is probably not that important either, it might just draw a little more base current. Am I wrong?

[Steve Conner]

Yes you are correct. The Vce(sat) doesn't make much difference.

The graphs do show typical values, but any good designer would ignore them and use the worst-case ones.

On the subject of hfe, in a good design there just needs to be "enough", in other words, more than the worst-case figure for the original part. Otherwise the amp might not be able to make its rated power. Real life is rarely so simple, though. In practice, some designs will go unstable if you use transistors with too much hfe or Ft. Where "too much" means more than any transistor the designer tested it with...

Audio power transistors have improved by about four generations since the 70s. I've heard the MJ15024/5 recommended as bombproof TO3 super transistors that work in just about anything. There are also quality TO220 driver transistors in the MJ range.

Edit: The driver transistors I was thinking of are the MJE15030, 1, 2, and 3 (NPN, PNP, different voltage ratings)

I looked at the datasheets for the original parts, and they have remarkably high hfe for that kind of transistor. I don't know of anything like that with a guaranteed minimum gain of 120. The MJE15032/3 are only guaranteed to have a gain of 50, and likewise for the TIP parts. The original poster did a good job hunting down replacements with high gain, but good luck actually finding them for sale! :-(

I also notice the originals are the older mesa type transistors. As Bob Pease mentions in his Analog Troubleshooting book, that process is now obsolete and everything is planar epitaxial, which gives a higher Ft.

[uvacom]

Hmm, it kind of sounds like I'm stuck between a choice of using TIPs or those MJEs and hoping the design can handle the higher ft (and possibly reducing the power output of the design because of lower hfe), or getting my hands on those newer toshibas somehow and hoping they don't blow up when I crank it because of the less robust SOA specs.

I don't get it. If audio transistors have "improved" so much over the last 30 years, why can't any modern ones match the high gain, low vce(sat), low cutoff current, and wide SOA of these ancient beasts?

[Steve Conner]

Blame the trade unions. :P

If you look at the original datasheets, you'll see that the gain could be anywhere from 70 to 240. At the factory, they selected these transistors into two gain ranges.

But nobody selects transistors for gain any more, because the time it takes to measure them now isn't cost-effective. By the time you work out the interest on loans used to buy the test equipment, wages of the guys needed to keep it running, rent on the floor space it takes up, etc, every second that a part spends on an automated tester adds about 30 cents to its price.

50 is the minimum, but the transistors you buy could be much higher. Nothing to stop you buying a dozen and selecting them yourself, except that if they came from the same batch, they'll probably all measure the same.

What's important is the product of the two hfe's of the driver and output transistors, not either in isolation. In a standard power amp, look at your schematic to see what current the VAS is biased at. That's how much is available to drive the driver transistor. Now divide the rail voltage by the rated load impedance: that's how much the output transistor needs to supply.

So, if we have a 44V rail and an 8 ohm load, that's about 5.5 amps. If the output transistor has a gain of 50 at 5 amps, then the driver needs to source 110mA. If the driver also has a gain of 50 at 110mA, then it needs 2.2mA into its base. If the VAS can supply that, you're good to go.

Oh, and according to Pease, mesa transistors were known for their good SOA. They were slower, but tougher. I believe TV line output transistors are still mesa type, even if nothing else is. They went obsolete because they were more expensive to manufacture: some needed cooking in a diffusion furnace for 20 hours.

But another reason why I recommend those MJ/MJE parts is that they have a guaranteed SOA. Peavey used them in PA amps up to 600 watts. See the OnSemi datasheets for details.

[olddawg]

Just for thought....In all of my years on the bench I can't remember seeing a pair of complementary driver transistors fail without some kind of excessive current draw in the output section. I would check your outputs and bias. How did you verify that these transistors were bad? At this age I would be suspecting a lot of electrolytic caps in the area.

[Steve Conner]

Yeah, what Olddawg said too... The driver transistors aren't directly in the line of fire, as it were, so I'd expect them to fail as a consequence of some other part going.

One thing I saw happen here was that the wire feeding supply to the collector of an output transistor came loose. With the output dead in the water, the driver then ended up trying to drive the speaker on its own, which probably pushed it outside of its SOA and killed it.

[uvacom]

Well, when I bought the amp it worked fine for about a half hour and then the next time I turned it on, one of the channels was shot. I looked inside, found a burned out resistor, traced it back to one or both of the drivers, and pulled them. One of them fails a diode test. I don't even have the amp in my possession right now, I left it at my parents a few years ago. I just came across the transistors and decided that this is my holiday project. Ill pick the amp up next week.

I want to refurbish the whole amp because it sounds incredible, at least the channel that works does. I'm sure the electrolytics need to be replaced, they definitely could have caused problems. But part of me also thinks it might be just as fun (albeit much more work) to instead take the circuit and build a somewhat modernized version just using the original chassis, cabinet, speakers, and possibly trannies. Either way, I need to have a good understanding of the existing circuit to move forward, and this thread has helped quite a bit.

[uvacom]

What's important is the product of the two hfe's of the driver and output transistors, not either in isolation. In a standard power amp, look at your schematic to see what current the VAS is biased at. That's how much is available to drive the driver transistor. Now divide the rail voltage by the rated load impedance: that's how much the output transistor needs to supply.

So, if we have a 44V rail and an 8 ohm load, that's about 5.5 amps. If the output transistor has a gain of 50 at 5 amps, then the driver needs to source 110mA. If the driver also has a gain of 50 at 110mA, then it needs 2.2mA into its base. If the VAS can supply that, you're good to go.

Okay, those are just example values, right? I think that would yield a power output of about 240W per channel. The schematic (which I've attached) shows 22V across the 8 ohm speakers, so that's 2.75A, and those values give something pretty close to 60W which is what we want. Unless I'm missing something, then none of the below will hold true -

The output transistors are also mesa parts with high hfe (the datasheet specifies 2SD425-O which have an hfe of between 70-140 @ 2A). So that's good, right? That means the drivers should only have to source 2.75/70 = 39.3mA. The VAS will in turn have to source .786mA, which doesn't sound like much.

Well, whether that's right or wrong it is apparent to me how to calculate the VAS bias current requirements for this amplifier. Next question is, how can I figure out how the VAS is biased in this circuit? I think this might require actually having the amp in front of me, as the scan is not of a high quality (I think I might order a schem from roland). But anyway, Q7 is the VAS, correct? Are Q5 and Q6 there to bias Q7? I'm not familiar with that arrangement. They look like a darlington pair but Q5 is a PNP. I admit to being a little overwhelmed at the prospect of analyzing this circuit directly, since it seems that the bias current of Q7 is dependent on the base voltage of Q7, which is dependent on that differential amplifier that comes before it.

[Steve Conner]

Yes, those were just example values. I'm working on a solid-state hifi design just now, so they might be more related to that.

5.5 amps peak into 8 ohms is a little over 100 watts RMS. The 200W is the instantaneous value at the peak of the waveform, which is what you need for calculating the maximum transistor currents. So, you need to multiply your 22V and 2.75A by the square root of 2 to get your peak values.

I see from your schematic that the VAS (Q7) is biased with the bad old bootstrap method. So roughly speaking, the drive current available for positive swings is the positive rail voltage divided by (R39+R40). That's 40V divided by 15.6k: 2 and a bit milliamps.

The power amp circuit is perfectly standard, shouldn't be hard to fix up. Q5 and Q6 are the thermal compensation for the output stage bias: one of them should be mounted on the heatsink with the output devices. They don't actually handle any signal (note that they have a honking big electrolytic bypass cap across them)

The little 15pF cap from collector to base of Q7 is the compensation capacitor. It stabilizes the whole circuit. If you put different transistors in and the amp starts to oscillate and burn up, the first thing to try is making this cap a bit bigger.

As it's just an ordinary power amp, you could probably replace it with any other power amp circuit, without altering the overall tone of the amp.