Replacement power transistors for old JC-120?

[J M Fahey]

Well, I started out trying to help some guy fix a JC120, and ended up learning 5 things I never knew about power BJTs myself

I'm also here to see and learn.
If, on the way, i happen to be able to help somebody, good.
As of where this thread is heading, I *think* that many are being carried by perfectionism , trying to replace anything with an even better, improved part. (I'm also guilty, Your Honor).
The problem is, this amp was already *very* over-engineered to begin with !!!!
Those beefy parts for what's an 60W into 8 ohms power amp !!!!! Too much !!
I can easily get those same 60W with 60's technology 2N3055's !!
If I am allowed to say something, I'd think as a serviceman, not an "enhancer" or whatever.
A couple MJ15023/24 driven by Tip 32/32C already are overkill, even if no "extended beta", mesa technology or whatever.
And what about beta?: I find "typical values" usable. Designing only with statistically improbable minimum values, and stacking them in a row as if all would happen at the same time, is no practical.
Well, maybe in a Venus orbiting Space Station when the nearest service technician lives 60 Million miles away that is the safest bet, but in my own shop I would just repair one with regular parts and test it. If is measures up (which it will do on 90/95% probability), fine. If not, I'll replace the suspect.
I might buy a couple extra parts and , measure and use the best, if necessary, and have the amp through my front doors in short time.
Using typical values:
>44V/6.5 ohms (minimum impedance)= 6.5 A (Yeah , I *know* the 44V are only instantaneous until the supply caps discharge somewhat and besides that, I'll loose at least 4 V between a couple VCEsat and VBE losses that stand in the middle)
>MJ15023/4 beta at 6.5A = 32 - Base current =6500mA/32=203mA.
>TIP31C/32C beta at 200mA: around 160 Base current=200mA/160=1.25mA.
>The "VAS" (we call it the "Class A driver transistor", which it is), must supply those 1.25mA or more.
VAS idle state current: the same as the current in its load resistors (10k+5k6 or +4k7, depending on amp version)= 44000mV/15600 ohms=2.82 mA , more than double the amount needed.
I'd repair the amp with robust and easy to find "Swiss Knives" MJ(E)15023/24 and Tip31/32C, would fire it up, load it with a 6,5 ohms 75W or 100W resistor and drive it to the onset of clipping.
If I get nice, symmetrically clipped 60W (or 55, it's the same),I close the amp and write a bill.
If the positive top of the waveform clips much earlier than the bottom one (say 2 or more volts earlier), I try to better the upper half beta, either replacing the output transistor, the driver, or both.

[R.G.]

I was going to use the MJ15024/5 for my current project, but I think I'll get some of the NJ parts with the tracking diodes instead. Farnell have those, albeit as US stock, or it might even be possible to get small quantities straight from ONSemi. Good job I didn't drill the heatsinks yet!

Let me know if obtaining them from Farnell turns out to be onerous. I have orders from Mouser coming in almost every week.

As of where this thread is heading, I *think* that many are being carried by perfectionism , trying to replace anything with an even better, improved part. (I'm also guilty, Your Honor).
The problem is, this amp was already *very* over-engineered to begin with !!!!
Those beefy parts for what's an 60W into 8 ohms power amp !!!!! Too much !!
I can easily get those same 60W with 60's technology 2N3055's !!
If I am allowed to say something, I'd think as a serviceman, not an "enhancer" or whatever.
A couple MJ15023/24 driven by Tip 32/32C already are overkill, even if no "extended beta", mesa technology or whatever.

You're absolutely allowed to say so! As several observant readers here have remarked, there is practically nothing for which I can't come up with some embellishment, some additional watzit to add to it.

Much of this is a knee-jerk reaction to an impoverished technical childhood. The transistors I had to design with when I was an engineerling were marginal for practically everything: gain, current rating, power rating, voltage rating, SOA, frequency response, everything. So I developed this fondness for rugged, burly, barefisted barroom brawlers of power transistors.

I find it fascinating to realize that Mouser Electronics sells a 2N3055 for $1.14 in theory, but they're out of stock. So is another company's $1.47 2N3055. The 2N3055's you can actually buy are $2.47 each. The 150V/150W NJW0281G and complementary NJW0302G sustained-beta devices are in stock at $1.40 each. The big brothers, 250V/200W NJW3281G and NJW1302G are in stock at $1.70 each. The 2N3055 has ft=2.5MHz, the sustained-beta devices have ft of 30MHz. I'm a sucker for power devices that are rated for more power, more voltage, more current, more SOA, and are as cheap or cheaper to buy. You *save money* by putting in the big ones? How often does that happen?

And what about beta?: I find "typical values" usable. Designing only with statistically improbable minimum values, and stacking them in a row as if all would happen at the same time, is no practical.
Well, maybe in a Venus orbiting Space Station when the nearest service technician lives 60 Million miles away that is the safest bet, but in my own shop I would just repair one with regular parts and test it.

You are absolutely correct. In general, the average part is just fine. Absolute worst case design (it has a formal name!) is something that space and military organizations use, and for exactly the reason that sometimes you can't haul that box into the repair store tomorrow. Your space probe or howitzer shell fuze may have to work today.

Did I mention that I as a child I always wanted to design space electronics? I did so want to design things with explosive bolts in them...

[uvacom]

In regards to the "worst case" design philosophy, I can attest that in my engineering program they advocate against it, instead using a monte carlo analysis on the design to determine whether it will operate correctly given component values randomized over a defined statistical distribution. Orcad PSPICE offers this capability (although I have yet to master it).

However, I think one distinct advantage to DIYing is that one can instead adopt a "best-case" philosophy by hand-selecting/matching components and using resistors of a tighter tolerance than would be profitable in mass production.

[uvacom]

I'm also here to see and learn.
If, on the way, i happen to be able to help somebody, good.
As of where this thread is heading, I *think* that many are being carried by perfectionism , trying to replace anything with an even better, improved part. (I'm also guilty, Your Honor).
The problem is, this amp was already *very* over-engineered to begin with !!!!
Those beefy parts for what's an 60W into 8 ohms power amp !!!!! Too much !!
I can easily get those same 60W with 60's technology 2N3055's !!
If I am allowed to say something, I'd think as a serviceman, not an "enhancer" or whatever.
A couple MJ15023/24 driven by Tip 32/32C already are overkill, even if no "extended beta", mesa technology or whatever.
And what about beta?: I find "typical values" usable. Designing only with statistically improbable minimum values, and stacking them in a row as if all would happen at the same time, is no practical.
Well, maybe in a Venus orbiting Space Station when the nearest service technician lives 60 Million miles away that is the safest bet, but in my own shop I would just repair one with regular parts and test it. If is measures up (which it will do on 90/95% probability), fine. If not, I'll replace the suspect.
I might buy a couple extra parts and , measure and use the best, if necessary, and have the amp through my front doors in short time.
Using typical values:
>44V/6.5 ohms (minimum impedance)= 6.5 A (Yeah , I *know* the 44V are only instantaneous until the supply caps discharge somewhat and besides that, I'll loose at least 4 V between a couple VCEsat and VBE losses that stand in the middle)
>MJ15023/4 beta at 6.5A = 32 - Base current =6500mA/32=203mA.
>TIP31C/32C beta at 200mA: around 160 Base current=200mA/160=1.25mA.
>The "VAS" (we call it the "Class A driver transistor", which it is), must supply those 1.25mA or more.
VAS idle state current: the same as the current in its load resistors (10k+5k6 or +4k7, depending on amp version)= 44000mV/15600 ohms=2.82 mA , more than double the amount needed.
I'd repair the amp with robust and easy to find "Swiss Knives" MJ(E)15023/24 and Tip31/32C, would fire it up, load it with a 6,5 ohms 75W or 100W resistor and drive it to the onset of clipping.
If I get nice, symmetrically clipped 60W (or 55, it's the same),I close the amp and write a bill.
If the positive top of the waveform clips much earlier than the bottom one (say 2 or more volts earlier), I try to better the upper half beta, either replacing the output transistor, the driver, or both.

This is a refreshing post, I've been obsessing over replacement parts so it's good to have my fears allayed through your practical experience.

One thing to note is that that circuit is also for the JC-160, so perhaps that's partially the reason for the excessive available gain.

[J M Fahey]

Hi RG and uvacom.
Interesting posts.
Although it might seem we are drifting into pilosophical discussions, we are really fine tuning *very* practical approaches to problem-solving, whether purely electronic or not.

[uvacom]

And hey! Since we're talking about it, I have some questions about the engineering of the JC-120 (contrasting with e.g. Self's Blameless topology). Note - all components discussed reference the "top" channel on the 79 JC-120 schematic.

1. In the JC-120 schematic the emitters of the drivers and the biasing stage are connected to the feedback loop. Additionally, the biasing stage is connected to the output stage through 2.2k resistors. Why?

2. The side of the load that would normally be connected to ground goes through a couple .3ohm resistors, and is also connected to the NFB loop (I think). Is this some kind of output short protection?

3. It looks like (with proper circuit analysis) it would be a fairly simple modification to improve the input pair's performance by eliminating R32 and adjusting the values of R31 and R30 to achieve balanced collector currents. Possibly emitter degeneration resistors could also be incorporated, if the board layout allows. Am I correct in assuming this?

4. What is the purpose of R33 and R42? Self's topology does not incorporate anything similar.

5. Why is there an asymmetry to the driver stage, where Q11 has a resistor R45 at the emitter and resistors R48/49 are of unequal values?

6. Would there be any negative side effect to lowering the values of R52/53, as prescribed by Self to minimize class AB distortion?

[J M Fahey]

Hi uvacom.
*Very* short answer: don't worry, be happy, leave it alone.
Slightly longer answer: even if Douglas Self himself (mind you, one of my "idols") were personally modding this amp, the results would be absolutelly inaudible to any panel made of professional, very well seasoned guitarists.
Going from .00404% to 0.000017% at 15 kHz

(.00404%) when scaled down for a realistic C/L gain of 23, is reduced by a factor of (23/1.47)2 = 245, giving a negligible 0.000017% at 15 kHz.

is absolutely inaudible when the sound source is a guitar, and the output transducers are guitar speakers.
Is Self's work useless? Not at all, it would be great to design an SS (guitar) power amplifier that had at least 60Hz to 8kHz flat response, less than 5% distortion, OPEN LOOP, feat easily achieved by an AC30 type output stage, with a good quality output transformer.
PS: the "extra" components you see appear because the Roland (and any other guitar amp) is a "real world" product, not a theoretical example.
Thanks for the link, I'll download and print it to be read whenever possible.

[R.G.]

Hi RG and uvacom.
Interesting posts.
Although it might seem we are drifting into pilosophical discussions, we are really fine tuning *very* practical approaches to problem-solving, whether purely electronic or not.

Yeah. The unavailability of explosive bolts always yanks me back to my senses.

When I wake up from the space reveries, I tend to look at replacement parts in light of bang for the buck (sorry, explosive metaphor crept back in... ). The pure-replacement lines like NTE are quite expensive if you use them, and they are only an approximate match, as they do not exactly replace all those devices they cross reference to.

In the case of power amplifier output transistors, you can nearly always use higher-voltage, higher-current, higher-power, higher-gain, higher frequency, bigger-SOA devices of the correct polarity and adaptable (if not identical) mounting in any position. This is very freeing, because you don't have to find an exact match. You just need to get enough, and on top of that, "enough" is getting cheaper as we go!

Bottom line, replacing output devices has gotten easier, cheaper, and more reliable, which is A Very Good Thing for the repair shop. I would be tempted, were I a repair shop owner, to stock the high-power sustained-beta bipolars in bulk for the best price and use them in anything, even if it was overkill. They'll certainly work, they're as cheap or cheaper than the identical device, and they would always be in stock, at hand, no waiting for parts. So you could charge *higher* service fees for fast turnaround repairs.

And hey! Since we're talking about it, I have some questions about the engineering of the JC-120 (contrasting with e.g. Self's Blameless topology). Note - all components discussed reference the "top" channel on the 79 JC-120 schematic.

It's kind of like telling your doctor "it hurts when I do ... this..." and having him tell you "Well then don't do that."

Self's "Blameless" setup is a full bore assault on lowering distortion. Good if you're critically listening to music through very high fidelity sources and speakers. But it's fundamentally the wrong approach for guitar amps. There's almost no point in trying to get super-low distortion if you will then use a distortion pedal in front of it, or drive speakers with 3% distortion themselves.

I haven't dug through the JC-120 design, but many solid state amplifiers intended for musical instruments include features to deliberately avoid the super-low-distortion nature of hifi designs in favor of pleasant-sounding but inaccurate distortions like musicians like.

Which is a long-winded way of saying the same thing that J M Fahey says:

*Very* short answer: don't worry, be happy, leave it alone.

If it's just intellectual curiousity, I can do some digging, but the real answer is that Self's designs are pursuing a different and somewhat contradictory goal.

[teemuk]

1. In the JC-120 schematic the emitters of the drivers and the biasing stage are connected to the feedback loop. Additionally, the biasing stage is connected to the output stage through 2.2k resistors. Why?

The first question: Because the driver and the power transistors form standard Darlington (upper) or Sziklai (bottom) circuits that way.
Second question: I think you are mistaking the current "VI limiter" to biasing circuit. The biasing circuit is the “VBE multiplier” circuit built around transistors Q5 and Q6. It will produce a voltage potential between the bases of the driver transistors.
The limiting circuit, on the other hand, will decrease the driver base current whenever the voltage potentials at the bases of the protection transistors (Q8 & Q11) increase. The reference is sampled over the emitter resistors of the power transistors so the reference tracks the output current, limiting drive whenever output current rises too high.

2. The side of the load that would normally be connected to ground goes through a couple .3ohm resistors, and is also connected to the NFB loop (I think). Is this some kind of output short protection?

No. It’s “current feedback”. The load current is sampled and used as a secondary feedback signal. In essence it will lower the amplifier’s output impedance and consequently the amplifier’s voltage gain becomes dependent on load impedance instead of being a constant value, as defined by the generic voltage feedback. The performance of the circuit is closer to that of generic tube amps that way and the amp gives a nice gain boost at the resonant frequency of the speaker.

3. It looks like (with proper circuit analysis) it would be a fairly simple modification to improve the input pair's performance by eliminating R32 and adjusting the values of R31 and R30 to achieve balanced collector currents. Possibly emitter degeneration resistors could also be incorporated, if the board layout allows. Am I correct in assuming this?

Yes. The audible difference might not be astounding though.

4. What is the purpose of R33 and R42? Self's topology does not incorporate anything similar.

Normal RC filter for the dual power supply. With the expense of a drop in supply rail voltage the input and the VAS stage will have a “cleaner” power supply with less effects from hum, sag, supply noise etc.

5. Why is there an asymmetry to the driver stage, where Q11 has a resistor R45 at the emitter and resistors R48/49 are of unequal values?

Well…. If you take a look, in fact, the whole output stage is asymmetric instead of consisting of two complementary Darlington transistor pairs. You see, the power transistors have the same polarity so the other one must be converted (by additional circuitry) to behave like it would have an opposite polarity. The upper pair of driver and power transistors is a Darlington circuit, the lower is a Sziklai circuit. The Sziklai circuit inverts the polarity of the complete pair. Thus, two different circuits need two different setups to behave about similarly.

6. Would there be any negative side effect to lowering the values of R52/53, as prescribed by Self to minimize class AB distortion?

Yes. The emitter resistors help in limiting current and achieving a thermal balance. Too low values might make the amplifier thermally unstable. It would also mess up the operation of the designed current limiter.

[Steve Conner]

I'm also here to see and learn.
...I'd repair the amp with robust and easy to find "Swiss Knives" MJ(E)15023/24 and Tip31/32C, would fire it up, load it with a 6,5 ohms 75W or 100W resistor and drive it to the onset of clipping.
If I get nice, symmetrically clipped 60W (or 55, it's the same),I close the amp and write a bill.
If the positive top of the waveform clips much earlier than the bottom one (say 2 or more volts earlier), I try to better the upper half beta, either replacing the output transistor, the driver, or both.

If you just want your amp to work again with a minimum of hassle, do what JMF says! I'd do exactly the same as a repairman.

But as a designer, I use the worst-case values. The cost of one recall in the field is enough to pay for all the extra worst-case beta, or whatever.

If you took the blue pill and want to see how deep the rabbit hole of amp-geeknology goes, that's a different story altogether. The answer is very deep indeed, and Douglas Self is lurking somewhere near the bottom clutching "Precious", his golden toroidal transformer, and hissing "Distortion... We hates it..."

Eventually you clamber out into the sunlight and it dawns on you that for musical instrument use, anything below 1% THD is fine. The exact equation for the acceptable amount is (1+N^2)% where N is the number of beers consumed before playing.

Re the open-loop thing: I once experimented with a hybrid amp, a pair of MJ15024/5 whose bases were driven by an EL84, through a little Champ OT. There was no feedback at all, just a floating power supply to bias the transistors into AB.

I intended it as a kind of retro-modern twist on the work RG did analyzing the Thomas Vox SS amps. It made over 60 watts, and actually still sounded remarkably good. Due to the lack of NFB, the output was almost a pure current source.

[teemuk]

As for mentions of Douglas Self and his Blameless low-distortion amps...

I can't agree that such designs would be overly beneficial in guitar amplifiers (sometimes it's even the opposite) but I do like the fact that he took the time to write abou the various distortion mechanisms and how to improve the circuits to compensate them.

IMO, it's better to know such practices exist and then make a free decision to ignore the ones that have less importance than be completely unaware of them.

That discussion also teaches a lot about generic operation of these types of "Lin" power amp circuits.
Too bad he chose to ignore all the rest.

[uvacom]

It's kind of like telling your doctor "it hurts when I do ... this..." and having him tell you "Well then don't do that."

Self's "Blameless" setup is a full bore assault on lowering distortion. Good if you're critically listening to music through very high fidelity sources and speakers. But it's fundamentally the wrong approach for guitar amps. There's almost no point in trying to get super-low distortion if you will then use a distortion pedal in front of it, or drive speakers with 3% distortion themselves.

I haven't dug through the JC-120 design, but many solid state amplifiers intended for musical instruments include features to deliberately avoid the super-low-distortion nature of hifi designs in favor of pleasant-sounding but inaccurate distortions like musicians like.

Which is a long-winded way of saying the same thing that J M Fahey says:

If it's just intellectual curiousity, I can do some digging, but the real answer is that Self's designs are pursuing a different and somewhat contradictory goal.

Well, it's intellectual curiosity, but not "just". I'm really hitting it hard to learn practical audio electronics design but I've still only got a basic understanding of how these circuits operate. I know what a sziklai pair is, for example, but I don't completely understand when to use one. So when a design deviates from some reference circuit, at my level it's nigh on impossible to tell the difference between a design choice made to improve performance, a design choice made to introduce some pleasing nonlinearity/tonal effect, and a design oversight which would have no desirable effect.

Now, whether it's worth it to e.g. swap some resistors in the input differential stage to improve its performance in a way that may not be noticeable is a worthy discussion. For me, getting some practice in on doing the calculations necessary to perform such a modification alone makes the exercise worthwhile, and if I'm working on the thing anyway swapping out 30 year old electrolytics and testing/replacing transistors, swapping a couple resistors is trivial.

[uvacom]

teemuk - thank you for taking the time to patiently answer my questions. If I may query in response:

The first question: Because the driver and the power transistors form standard Darlington (upper) or Sziklai (bottom) circuits that way.

How they are configured as such is not apparent to me. In a sziklai pair, the base of the pnp is connected to the collector of the npn, the collector of the pnp is connected to the emitter of the npn (which form the "emitter" output), the base of the npn is the "base", and the emitter of the pnp is the "collector".

Contrast with the pair in the JC-120, where the base of the pnp is the input (and presumably the "base" of the subcircuit), the collector of the pnp goes through a 100ohm resistor before it is connected to the emitter of the npn, the base of the npn is connected through the 10 ohm resistor (fuse?)to the collector of the pnp (and is not the signal input), and the emitter of the pnp goes through a 33ohm resistor before connecting to the NFB loop (on the other side of the current limiting .3ohm resistors.

But perhaps this is an alternate complementary arrangement?

Second question: I think you are mistaking the current "VI limiter" to biasing circuit. The biasing circuit is the “VBE multiplier” circuit built around transistors Q5 and Q6. It will produce a voltage potential between the bases of the driver transistors.
The limiting circuit, on the other hand, will decrease the driver base current whenever the voltage potentials at the bases of the protection transistors (Q8 & Q11) increase. The reference is sampled over the emitter resistors of the power transistors so the reference tracks the output current, limiting drive whenever output current rises too high.

Okay, that makes sense. I did think that Q5 and Q6 looked like a VBE multiplier, but I'm not familiar with that two-transistor configuration (which incidentally looks a lot like the driver-output PNP-NPN pair). I think somebody mentioned earlier that they served some kind of temperature compensation function, but I could have misunderstood.

No. It’s “current feedback”. The load current is sampled and used as a secondary feedback signal. In essence it will lower the amplifier’s output impedance and consequently the amplifier’s voltage gain becomes dependent on load impedance instead of being a constant value, as defined by the generic voltage feedback. The performance of the circuit is closer to that of generic tube amps that way and the amp gives a nice gain boost at the resonant frequency of the speaker.

Aha! Would it be fair to say that this would become the dominant nonlinearity of the power amplifier stage (since the rest appears to be nominally a linear amplifier save for some small error signal)?

Yes. The audible difference might not be astounding though.

Hmm. I'll probably try it anyway just for giggles, although apparently the distortion this induces is mostly HF.

Normal RC filter for the dual power supply. With the expense of a drop in supply rail voltage the input and the VAS stage will have a “cleaner” power supply with less effects from hum, sag, supply noise etc.

Again, this totally makes sense.

Well…. If you take a look, in fact, the whole output stage is asymmetric instead of consisting of two complementary Darlington transistor pairs. You see, the power transistors have the same polarity so the other one must be converted (by additional circuitry) to behave like it would have an opposite polarity. The upper pair of driver and power transistors is a Darlington circuit, the lower is a Sziklai circuit. The Sziklai circuit inverts the polarity of the complete pair. Thus, two different circuits need two different setups to behave about similarly.

See above, I am confused by this as they do not appear as any sziklai pair I have yet seen.

Yes. The emitter resistors help in limiting current and achieving a thermal balance. Too low values might make the amplifier thermally unstable. It would also mess up the operation of the designed current limiter.

Fair enough.

[teemuk]

See above, I am confused by this as they do not appear as any sziklai pair I have yet seen.

Maybe there's a bit of confusion going on on both sides. Anyway, this picture should explain the configurations way better than I can do in English.



Douglas Self discusses this stuff in quite a detail in his book but I'll rephrase some of it anyway. The basic push-pull emitter follower output stage (that uses a driver and output transistor pair) can be configured in three ways:

- Fully complementary symmetry: Both NPN and PNP sides are Darlington pairs
- Quasi-complementary: One pair is Darlington, the other is Sziklai. This scheme allows using the same polarity of power transistors, which used to be an important design detail in the past (and sometimes still is when MOSFET output devices are used).
- Compound follower: Both NPN and PNP sides are Sziklai pairs

Each of these output stage configurations requires a bit different bias as well as a different thermal tracking scheme to maintain the bias in safe levels.

[jrfrond]

MJ15024 is the Swiss Army TO3, and MJE1503x is the TO220 guy.

Amen brother! I stock hundreds of transistor types, but we consistently use MJ15024/25 for TO-3, TIP41/42 for TO-220 and 2SA1962/2SC5242 for TO-247, but also MJE1503x when the mood hits us, as well as Toshiba 2SB554/2SD424 for TO-3. Then there are the TO-3P xistors.

The bottom line is that you have a lot of latitude with bipolar transistor subs, which is how the whole SK/GE/ECG/NTE concept of limited-selection replacements came to be. Same with diodes. Just stock 1N4007, and don't bother with the -1 to -6 versions. Just use the highest-rated transistors you can and don't worry about it. Just make sure that the voltage and current ratings at least match.