2N3055 inside - commercial famous amplifier models, quasi complementary power output

The original rating for the 2N3055 was 60 VDC C-E. I used to design with them in the 1970s, and I worked at a place that sold parts to industry. I studied parts and their characteristics and still do. The original data sheets I have were marked "Lamda". They might still be around here somewhere, but I am not going to attempt to search for them. I referred to them many times. I also have some original leather bound Motorola transistor data books. Yeah, those are old, and they are not reprints.

Price is never, ever a deciding factor in servicing equipment or building new. Why? A couple bucks saved and take out several times that in components or even the PCB itself. What's that worth? Then you still have to buy the same thing again, but hopefully you are now smart enough to buy either the real (known chain of custody) device. Even better, a device better suited to what you are doing. I should also mention that your time in troubleshooting what went wrong, and the labour involved in correcting your goof should be enough for anyone to avoid making that choice.

"SE" was a Motorola designation for a special device. It may be selected for higher breakdown voltage, beta or leakage, it may even be a modified device in some fashion. You will not find a data sheet on those, and the company would never release information on it. They were supplied under special contract to the OEM, so no technical information would ever be released to anyone but the customer and internal engineers. If others used that designation I wouldn't know, but in North America it was specifically a Motorola thing.

The 2N3442 was used in better commercial amplifiers (earlier production) until better parts came out (2N3773 for example).. The 2N5631 (and 2N6031 was the compliment). Another thing to consider is the cost of getting matched sets of output transistors. I can't tell you how many I had to buy to get a matched set of 4 MJ15022 /23 / 24 / 25 transistors. The spread of beta was pretty wide, Japanese parts were better and I used tons of 2SD424 and 2SD555 (and their compliments). When you had to match 8 of each number, things got real expensive. Did matching matter? It sure did and still does. The MJ/MJL series parts from OnSemi match out of the tube a lot better than I could get doing it by hand, and getting better matches does cost nearly as much. So what am I saying here? The new "expensive" parts are actually less expensive to use than the classic, cheaper parts! That and performance is much better, so how much more do you need to know?

Yes, the metal case parts are more expensive. The production quantities are falling, and materials much more expensive. They are heavier, so shipping thousands at a time is more expensive too. The cost of rejects is therefore higher and it all adds up to much higher costs in getting the good quality part in your hand. Of course they are much more expensive!

-Chris
 
Actually the SOA is quite good - full power to about 80 volts. The fT spec'ed in the RCA manual is 80 kHz. No typo, not 800. Motorola always spec’ed a MHz, but they HAD a high SOA epi process. Don’t think it was quite as real-world rugged as RCAs homo. These could be anything, but I’d suspect closer to the lower end if they came out of a power supply - manufacturers would have used the best tool for the job. If you had used high fT (ie, switching types like 2N6547) as drivers, and a good set of predrivers (say 2N3440 and2N5415) in a QC3 you might have been able to make it listenable.
 
@anatech (and others) - This is a bit of a tangent to the reply above, but when you measure hFE of a typical power transistor for matching purposes, what are the test conditions? Manufacturers seem to use 4 VCE quite often (and 1 or 2 amps for a 15 A part) but is this applicable to audio usage? When I match parts I use a curve tracer to fully characterize the parts but I know that is not necessary - but what is the recommended minimum of matching?

Thanks,

Hal
 
wg_ski, so you're saying that I'd better replace the lone driver pairs by much faster devices in order to approach some real builds? Would MJ15023's do?

Best regards!
Yeah, a 4 MHz device would be better. It’s significantly faster than the outputs. If you found some “junk“ switching types they would be better still. Ive got a lot of old pulls that would work well. I’d never pay full price for them, of course. And you could use any modern TO-3P audio type as well since the form factor of the drivers and outputs doesn’t need to match. Good use for the ones that fall out of matching on more serious projects.
 
Hal- that depends!
If you are matching all NPN or all PNP it is common to find a closely matched pair over the current range from samples of devices from a lot of say 6 or so. Better than 5% in most cases.
If you are matching an NPN with a PNP it is generally more difficult.
My approach is to match at a nominal current somewhere in the middle of the expected range e.g. 1 or 2A for a 50W 8 ohm amp (or indeed 100W amp with parallelled pairs). I typically ain to get to within 20% but you often find a lot of NPN and a lot of PNP have close matches to their own polarity - but sometimes all miss more than 50% so you don't get a tight match at all for any pair.
Commercially matched pairs were offered at about 25% for a single polarity or about 50% for complementary pairs, but modern devices can achieve better.
As for the Vce then the region to avoid is quasi-saturation. I typically use 5V for up to about 3A for most modern devices, but choose a Vce which avoids it on the worst of the two (usually the PNP).
I agree with anatech that it is still worth matching as closely as possible even with modern transistors. Distortion is always lower, even if it is already low.
 
Hi Hal,
Matching at lower current is more important since emitter resistors force current sharing at higher levels more effectively. I'll match around 100 mA to 250 mA. Matches at those low currents will behave nicely at higher current levels due to the emitter (source) resistors. As noted, modern device matching is much better these days. For complimentary matching all you can do is pickl units that are more close in beta, but do not expect them to match or track closely due to how differently doped silicon behaves. Closer is better than not.

My jig for power transistor matching will hold 4 devices on the same heat sink. I run it up to the same temperature every time and let it stabilise, then measure beta (or transconductance) for all four devices. Then I take the ones I figure match and mount them all for the truth. It's a pain but sometimes absolutely required (Counterpoint amplifiers for example).

To match signal devices you have to hold the temperatures exactly the same. I made a jig that puts them in a long tailed pair configuration, I run at the expected tail current, but 3 mA allows them to settle faster. The jig isolates them from the environment thermally and you merely measure the collector voltages. Your meter is then a null meter so it's easy. Differential voltages will be in the mV to sub mV levels. Essentially the devices are placed in the same circuit configuration that they are normally used in.
 
Hi Jan,
That figure was always 60 VDC C-E and 100 VDC C-B. It's possible that European manufacturers allowed some variance, but the JEDEC spec was pretty much the standard in North America.

I've been using Linear Systems devices for small signal where I can. Shipping to Canada costs a mint though! FO high voltage I have to match individual parts, and when the lead pinout doesn't agree.
 
Yes, it's a mess!

MJ2955.png
 
Hi Hal,
Matching at lower current is more important since emitter resistors force current sharing at higher levels more effectively. I'll match around 100 mA to 250 mA. Matches at those low currents will behave nicely at higher current levels due to the emitter (source) resistors. As noted, modern device matching is much better these days. For complimentary matching all you can do is pickl units that are more close in beta, but do not expect them to match or track closely due to how differently doped silicon behaves. Closer is better than not.

My jig for power transistor matching will hold 4 devices on the same heat sink. I run it up to the same temperature every time and let it stabilise, then measure beta (or transconductance) for all four devices. Then I take the ones I figure match and mount them all for the truth. It's a pain but sometimes absolutely required (Counterpoint amplifiers for example).

To match signal devices you have to hold the temperatures exactly the same. I made a jig that puts them in a long tailed pair configuration, I run at the expected tail current, but 3 mA allows them to settle faster. The jig isolates them from the environment thermally and you merely measure the collector voltages. Your meter is then a null meter so it's easy. Differential voltages will be in the mV to sub mV levels. Essentially the devices are placed in the same circuit configuration that they are normally used in.

Thank you very much - that hits the main points I was wondering about. As mentioned, just about everything I do involves low level signals and I really have no experience with higher levels as applied to audio and what is important to it specifically. My focus is in metrology and signals that are not only small but are either DC or sub-1KHz. I as well use a custom built measuring system (the 576s are just too big for the space I'm currently using) although mine is for one device only (and only 1A IC max) - everything is sent to a computer and I do the comparisons there. For environmental testing I just put the fixture into a chamber I made to characterize components in general - just use a large circular connector as a common interface. Need to make an adapter for higher power devices, though... I have several hundred TO3 NPNs of different flavors pulled from some large active loads to test. They should be all matched parts but how well matched is the question. Anyway, thank you for taking the time to reply; I'm getting a lot better feel for what is involved... and why.

Hal
 
Hi Hal,
I'm trained in metrology as well. Journeyman calibration tech. Temperature is mV/uV and DC, or RTD. I used an HP 3458A and several Fluke multifunction calibrators as well. Pressure, voltage and current. I did instrument repairs on top of calibration. I worked up to 1 GHz, we didn't have assets higher than that, and preferred to send things over a couple hundred MHz to our other lab. TUR was the problem.

You can group the TO-3 devices by leaving them arranged on the bench for hours. No heating or AC vents. Clip carefully onto the leads and measure beta. A Heathkit IT-18 works very well. Once you have them grouped, use a jig to really sort them out. You want low current to avoid self-heating during initial sort. I run 10 VDC C-E so you still need to wait for thermal equalization before you compare them. Time intensive ('cause temperature is involved).

You need to equalize case temperature, and air takes too long to settle - plus you cannot force the parts to have the same temperature unless thermally connected. Hence my heat sink for power devices. I have one for Mosfets, and one for BJTs. The signal transistor jig places the two parts in thermal contact with a thermal insulator over them, then a hard cover to block air currents. Yes, required. The die must be as close to equal temperature as reasonably possible.

On my bench I used HP 34401a's, and now 34461A and 34465A. I still use my 3457A's as well.
 
  • Like
Reactions: Mike Gergen