Power Transistor Current Gain Change

[bnchwrmr]

Hello to you all,

I am repairing a mid 70's Sansui stereo power amp, and found the
two paralleled NPN power devices are still good, but have a beta
of 15 and 22, while their two PNP compliments are 67 and 75.
I don't know if current gain in transistoers can change with age
and heat exposure. If so, does it affect one polarity more than
the other?? Those of you who have worked in electronics as your
career may have experiences to shed light on my ignorance about
this. Even though these two transistors "work", I don't want to put
them back in, because of their low hfe numbers---I'm thinking they're
headed for failure sooner than the PNP devices. What do you think
about this?...does beta in any transistor change with age and heat??

Thanks for your help in advance.
bnchwrmr

[PRR]

> beta of 15 and 22,

At what current??

If you test big 5,000mA devices with a multimeter Hfe check at perhaps 1mA, the reading is not very useful.

And I can't think of a way the Hfe will "go weak".(*) Transistors usually fail catastrophically. Usually as dead-short, though if the power supply has ooomph they may be found blown open at autopsy.

I think the true Hfe plot of the NPNs may be 20 at 1mA, 40 at 10mA, 80 at 100mA through 2A, and then falling. It is not impossible that similar vintage PNP will have better Hfe at 1mA, though worse above 2A. And it is very likely that any Hfe over 20 (at operating current which is surely nearer 50mA idle) will amplify fine.

Heat will cause a rise in Hfe, but it falls back when the device cools.

(*) In super-precision input stages, where a match-pair have Hfe equal within 1%, a momentary emitter breakdown which does not melt the junction can sometimes cause Hfe match to drift to 2% or 3% difference. In very high impedance low-level measurement, this is not good enough for government work. But all loudspeaker output stages are built with ample reserve of base drive, Hfe is not a critical parameter, and overall NFB makes the drive current do the right thing so the speaker voltage is correct.

If you are terribly worried, stick them in, load with 8 ohms, check that it will swing full power. If it is really short of Hfe on the NPN side, the positive swing won't come close to the rail voltage under load. Be aware that pre-FTC audio amps may not be safe at FULL power for more than a second or three.

But personally, I would not abuse a 1970s Sansui this way. If it will play Bob Seeger really loud, without gross sonic insult, it's fine.

[PRR]

Slow memory...

Late 1960s, if you could find ONE big tough device, you were in clover. Quasi-complementary was king, and sometimes with two Germanium PNPs even after the rest of the amp was Silicon.

In 1976 we could get "complements" which differed only 2:1 at most useful currents.

In 1970, there were no true complements in Power transistors. We hunted out two very different devices with similar maximum ratings, tolerable cost, and accepted (worked around) 4:1 differences in gain at working currents. Because the two devices might not be even approximately similar layout and doping, even this rough match would be way-off at very low currents such as a DVM's Hfe tester.

You can build an Hfe tester for "realistic" currents, but it is tricky, and can be risky.

If the amplifier uses a bipolar (+ and -) power supply, the connection below will test at about 200mA 40V. Keep the transistor ON the heatsink. Leave the collector connected to its supply voltage (I have assumed emitter-follower outputs). Lift the base and emitter. Base to ground via 200 ohms, emitter to negaive rail via 200 ohms, this must be a 10W part (and may stink if the rail is over 45V).

Using an un-grounded (battery powered) DC meter, read the voltages across each resistor. The resistors are equal, so the ratio of voltages is the ratio of currents, which is Hfe. (If it is convenient to use a 220 resistor at base, the ratio is 10% off; you could allow for that but it is really unimportant.)

I would want 2V or less across the base resistor, and assuming the rail is roughly 40V, that indicates plenty of Hfe for 1970s designs. I would probably expect more like 1V or 0.5V across the base resistor at ~~200mA on a 1970s power device, Hfe nearer 40 or 80.

For PNP, flip everything over. Don't obsess about match; that luxury was not available in the early 1970s. And if a specific pair of devices was found to be reliable, it might have been used even after "better match" complements became common. Some old crude devices were mysteriously less failure-prone than many "better" devices.

[DummyLoad]

But personally, I would not abuse a 1970s Sansui this way. If it will play Bob Seeger really loud, without gross sonic insult, it's fine.

in some circles, that could be considered cruel and unusual punishment, maybe even torture. thanks! now i have "night moves" stick in repeat in head...   :P

 ;)

[bnchwrmr]

Thank you,PRR, for the information and helpful insight.
and for the light-hearted humor...good laugh with that
and ISOtone's reply...thanks fellas!

Yep, your apparent guess that I might have tested the
devices with a meter's "Hfe" function was dead on. It's
all I've got, and mostly used for small transistor
work.

The TO-3 big boys in the unit are 2SA745A's (PNP, 120V,
8A, 70W), and 2SC1403A's (no datasheet, assume same as
PNP's), with two devices paralled to form an equivalent
"super transistor". All devices tested good with junction
resistance tests using an analog FET-VTM, and the Hfe
checks on a DVM. All of the transistors checked "good",
however, the PNP's had betas of 67 to 80, while all 4
NPN's read from 15 to 23!! I replaced them with MJ15015's
(NPN, 120V, 15A, 180W) all with betas in the 60's and
70's. One of the MJ15015's in the batch has a beta of 18.
And 20 is supposed to be the low end value, interesting!

I made the replacements to keep the loads on the drivers
somewhat balanced. Just based on an assumption.

I'm glad to learn the the power devices beta increases with
heat, that means the drivers run a little cooler than would
appear. Let the TO-3's handle the heat, they have the massive
heat sinks after all. The driver's heatsink is mounted to the
driver board, and isn't very big.

I'm keeping the low beta NPN's and will use your circuit
to test them...with my +/-34V (actual) home-built
supplies. The Sansui supplies are +/-54V.

Never did like those quasi-sysmetrical designs anyway!

Don't have any Bob Seeger, but I've got some Three Dog Night
to try things out with!! Ready to test with dummy loads.

Thanks again PRR!!
All The Best,     bnchwrmr

[PRR]

> I replaced them with MJ15015's

That's so 1980s!

And MJ15015 is -very- much slower than 2SA745A. Like 10 times slower. Which is liable to ruin Sansui's careful frequency compensation. It may oscillate.

> 20 is supposed to be the low end value

At 3 Amperes! (2SA745A spec min 30 at 3A)

The gain at 1mA (or whatever the multimeter uses) may not be the same. It will probably be less. TIP41, a significantly newer device, shows gain at 1mA about half of the peak at "useful" currents. Older devices had more low-current droop. Not that I can find one, but I have Andy Grove's analysis of a clean junction BJT. (Things were much worse before Grove's work.) The low-current fall-off is inevitable, sumthin about recombination rates.

You have not proven the old devices are "bad", and I have real doubt about their speed in an amplifier optimized for 2SA745A.

[bnchwrmr]

First, I want to thank you, PRR, for making me dig alot deeper into things
and putting the ol' thinking cap back on. It's gathered some dust over the
past 20 to 25 years! You're way ahead of me, and it can appear that your
knowledgeable and experienced posts are meant to intimidate, but I don't
believe that's where you are coming from. I believe you are honestly trying
to show us all that we can "come up higher" from where we are, and are out
to inspire growth and learning, and not to shame others for not knowing what
you know. I had no way of knowing that the MJ15015 was "so 1980s!", (my
most recent Sam's Transistor Spec Manual is dated 1978!), and I've never
heard of Andy Grove, let alone read any of his material. Since the early 80's
I have been just a "paper electronics hobbyist" with little hands-on or new
information to grow and experiment with. I have only been aware of
any of the Motorola MJ15xxx devices this year, and have not seen them in any
schematics, I only downloaded their datasheets recently. I would love to be
able to buy actual 2SA745A replacements...but where??...and how much??

Secondly, you are right that the MJ15015 has a much lower frequency limit as
compared to the -A745A type which is listed in my Spec Manual as rated to 15MHz!
The Motorola MJ15015 data says the gain-bandwidth product can vary from 0.8MHz
to 6MHz, if I'm reading it correctly. And you are right again that I have not
proven that the 745A's are really bad/inadequate under real working conditions.
I just wanted to get this thing repaired and outta my hair! I was only looking at
the electrical specs of the devices, luckily I even had them on hand. I need to
enlarge my stable of useful TO-3 transistors for replacement and messing around.
I will fire this up anyway just to see what happens...but I'll put the 2SA745A's
back in if I have to after testing them per your circuit, and post the findings.

The MJ15015 datasheet shows the test conditions for DC current gain as follows:
     (Ic = 4.0 Adc, Vce = 2.0 Vdc)  hfe = 10 min, 70 max
     (Ic = 4.0 Adc, Vce = 4.0 Vdc)  hfe = 20 min, 70 max
     (Ic = 10 Adc, Vce = 4.0 Vdc)  hfe = 5.0 min, -- max
Why did they use such a low Vce for these DC current gain tests?? Seems
to be useless to me. Your proposet test conditions seem to be way more
useful. Your circuit has an Ie of about 200mA, so I will test with that
value, while also testing at lowering Vce values, i.e., 20v, 10v, etc..

What do you mean about things being much worse before Grove's work. I
don't know of the impact of his work, so this statement is of no real
benefit to me.

"Older devices had more low-current droop." Wow!! Where do you find
this kind of information?? By "low-current droop", are you talking
about current gain decreasing with diminishing Ic?? Sometimes these
things you say need explanation and expansion for us not-so-in-the-
know strugglers here. Please help us by stepping into our shoes and
see what clarifications would be helpful...Thanks in advance!!
And how long ago determines "older transistors"??

Now I know to be on the alert about "device speed" when it comes to
replacement and analysis of proven designs. I'll be alot more careful
in the future....I appreciate the "heads up"!

Thank you for putting up with my long-winded posts!
Keep ya posted!   bnchwrmr

P.S.--I mean no offence in any way....you have built and honed your
knowledge and experience continuously over the years, and I take my
hat off to you. What you have can't be found in basic engineering
texts...I have a couple of them and they didn't mention low-current
droop in transistors, or electron-hole recombination effects on gain,
frequency, or anything else that I remember.
Thanks again for challenging me to think and grow.
Thanks for helping us all.

[PRR]

> worse before Grove's work

Early transistors were VERY variable. Production yield on Point Contact was few percent. Alloy was more art than science. Theory did not explain why real transistors were crappy. Theory was about happy perfect junctions floating in an infinite sea of crystal. Grove dug into surface contamination, diffusion gradients, corner geometry, and showed how these could be engineered and controlled. This made good transistors, and multiple transistors (chips), possible.

> I've never heard of Andy Grove,

He dinked around labs a while, and then got a job with a 2-person start-up company, escaped from a large company. Worked his way up, founded new business, got to be pretty big. Today he is better known as ex-president/CEO/chairman of INTEL.

> Ic = 10 Adc, Vce = 4.0 Vdc
> Why did they use such a low Vce for these DC current gain tests??

10A, 4V is 40 Watts, enough heat to mess-up the parameter being measured.

The Hfe at higher voltage will be similar or nominally higher. The Ic/Ib curves are mostly flat, slight slant (Early Effect). So the low-volt tests are good enough for a guide.

Do you understand why speed is an issue on a high-NFB amplifier?

[bnchwrmr]

Thank you very much for the time and help here, PRR.

Sounds like Andy Grove isn't just known to those in engineering circles
only, so I will try to search for him and see what's out there...sounds
like he approached semiconductors with open eyes and an open mind. With
the birth of the triode tube and transistor, there was so much rivalry
and fighting going on...as well as patent suits, and so on.

Wow, yes, of course. If I had put 2 and 2 together (Amps and Volts), I
would have gotten 4 (Watts)!! Thanks again! They DID have to limit the
Power dissapation, after all. They sure couldn't ask a 70 Watt device
to dissapate 400 Watts (40 volts @ 10 Amps), now could they??....Duh!!

Man, once upon a time I could explain the Early effect, but I will have
to go back and look it up.

> "Do you understand why speed is an issue on a high-NFB amplifier?"

Ok, let's see if I get this one. With a high ratio of negative feedback,
that is, with the output signal amplitude being much,much greater than
the amplitudes both at the feedback point as well as the amp input,
the changes in output amplitute would have to change fast enough so that
the changes at the feedback point would be able to change as fast as the
changes at the input, in order that the differential pair would maintain
control over the signal integrity, i. e., stable closed-loop gain, phase
angle, accurate signal reproduction , etc. This requires the devices in
the voltage section as well as the output drivers and final output devices
to have faster slew rates, especially as these all must handle the higher
swings in amplitude at all frequencies in the passband, than the differential
pair are required to. Slew rate is the rate of voltage change during a
chosen time interval, usually 1 uS--S.R. = Volts/uS. Seems like slew rate
is limited by how fast base current can charge up junction capacitances,
or something like that. In some feedback networks, you will find a small
capacitor across the resistor connected between the output point and the
feedback point. I have seen this referred to as a "hurry up" capacitor,
and helps extend the closed loop upper frequency response...if I remember correctly.........is this right??

I remember seeing a datasheet on a Damn Fast Amplifier that had a very
amazing 6000 V/uS slew rate, how pitiful it makes the 741 op amp look.

Finally, I hooked the Sansui up under no-load conditions with scope,
and got a reproduced triangle up to about 50 volts Pk to Pk for about
10 to 15 seconds (~400 Hz), before the smell of smoke ended things.
No sign of any oscillation.

Thanks again for your help and information, PRR.

All The Best,
bnchwrmr
(Where there's smoke, there's Bench Warming!)  :D

[PRR]

> They sure couldn't ask a 70 Watt device to dissapate 400 Watts (40 volts @ 10 Amps), now could they??....Duh!!

They "could", with pulse-testing.

But there is another problem, and it bedeviled audio amplifiers too long. A device might carry 10A at 10V (on the infinite heatsink), but at 40V it would fail at not much over an Amp. "Second Breakdown". Hot silicon conducts better. Current never spreads equally across the whole junction. Any part which gets more current gets hot, takes more current. You quickly have all current flowing in a small part of the device.

Look at TIP41 datasheet Fig 3. Although a "65W" device, it will not take 65V at 1 Amp on a bass-wave. While pure-resistor loads won't normally pass through here, reactive loads make oval load-"lines" and will often kick outside the SOA curve.

http://en.wikipedia.org/wiki/Safe_operating_area

Your essay on NFB and speed has substance. There's many ways to look at it.

You know that three C-R networks around one 12AX7 triode can oscillate, Tremolo Oscillator.

It looks like plate-grid NFB. But each C-R has phase shift. If each is only 60 degrees at some frequency, all the same, you have 180 deg phase shift at that frequency, and your NFB becomes PFB. If gain+loss exceeds unity, it oscillates.

EVERY stage is a low-pass (nothing amplifies to infinite frequency). A one-stage amp, with no complications, can only have 90 deg of phase shift, and is "unconditionally stable". Two stages can give 180 deg but only at infinite frequency where gain is zero; it is won't oscillate, but tends to wobble badly unless the two roll-offs are at very different frequency. And few real "2-stage" amps are truly 2-pole. There's always some stray roll-off. If it adds a few degrees, at a frequency where the two main poles sum to say 175 deg with a wee bit of gain left, that's an oscillator.

Good read:
The Monolithic Operational Amplifier: A Tutorial Study
www.national.com/an/AN/AN-A.pdf

The difference between a LM741 and a loudspeaker amp is mostly size and voltage. The frequency compensation "can" be the same, although audio designers are prone to be more daring than factories selling all-purpose chips.


> 50 volts Pk to Pk for about 10 to 15 seconds (~400 Hz), before the smell of smoke

That should not happen. What's wrong?

[bnchwrmr]

> They "could", with pulse-testing.
Yes, but, of course, that is not Class AB operation.

My 1969 RCA Transistor Manual is the only place where I could find any
info on Second Breakdown. They describe what you stated, saying the
current is focused in an area about the width of a human hair...it causes
a localized heating which may melt a minute hole from the collector to
the emitter and thus cause a short circuit. They also go on to say that
the liniting effects of second breakdown are more severe in power trans-
istors in which narrow base structures are used to achieve good high-
frequency response. This brought to mind the 2SC1403A NPN's with their
15MHz corner roll-off point.

Once again...where do you get this stuff?? I'm amazerd at what you have
learned. I have seen more tube Vp/Ip charts than actual transistor ones,
and I have never seen an oval loadline plotted for any reactive load,
whether capacitive or inductive. 'Bout all they "explain" loadlines with
are resistive loads, 'cause it's simple. How would you plot one, what
would it look like, and then once you get the oval loadline, would you
determine the angle of the slope of the curve at your operating point
to find the equivalent resistive slope at that operating point?...
or have I misunderstood something here??

> ...it will not take 65V at 1 Amp...
Yes, and I alwys got aggravated that you could not get the -A, -B, or -C
versions of the TIP-29 through -34 series through Rat Shack and other
sources...keep it low voltage, keep it safe seems to have been the rule.

> can only have 90 deg of phase shift
When does this occur?

What makes up your 'real "2-stage" ' amplifier?

I have not had the chance to tear into the Sansui yet, but I did
hook up your test setup using my 2 34V supplies, 2 100-ohm/10W res
and 2 100-ohm 1/2W res for the base, I matched the res pairs to
within one ohm. I tested 6 NPN devices, the 4 2SC1403A's, the low
beta MJ15015, and a 2N3055. Each device was tested on the DVM hfe
functuion. I subtracted the calculated base current from the
calculated emitter current to get the collector current, then solved
Ic/Ib to obtain the DC current gain. I allowed enough time for the 2
10W resistors to get hot to the touch (the transistors stayed cool).
My results were:

  Device #1 2SC1403A; beta(meter) = 13, beta(calc) = 92.68
  (Ie = .15397130A, Ib = .00164356A, Ic = .15232774A) 
 
  Device #2 2SC1403A; beta(meter) = 15, beta(calc) = 93.58
  (Ie = .155445545A, Ib = .001643564A, Ic = .153801981A)
 
  Device #3 2SC1403A; beta(meter) = 31, beta(calc) = 97.74
  (Ie = .155445545A, Ib = .001574257A, Ic = .153871287A)

  Device #4 2SC1403A; beta(meter) = 24, beta(calc) = 99.64
  (Ie = .155693069, Ib = .001547030A, Ic = .154146039A)

  Device #5 MJ15015G; beta(meter) = 18, beta(calc) = 72.26
  (Ie = .153960396A, Ib = .002101496A, Ic = .151858911A)

  Device #6 2n3055; beta(meter) = 174, beta(calc) = 653.17
  (Ie = .155445545A, Ib = .000237624A, Ic = .155207921A)

I will only use the DVM for small signal device tests. And I will
put the Japanese transistors back in when I get the stereo back
apart. I could sure ask alot more, but I'll end it here.
Thank you one more time, PRR, this is enlightening and puzzling
at the same time.
All the Best     bnchwrmr (the un-engineer)

[PRR]

>  Device #1 2SC1403A; beta(meter) = 13, beta(calc) = 92.68

What I suspected. These are BIG devices. Like using a 12" water valve: you don't get much control at dog-wash flow, but it works fine when wetting-down a whale. And conversely, a "do all" DVM Hfe test must be careful, maybe only hamster-wash flow, and be misleading about a whale-wash valve's true worth.  

Fine matching across the several devices. Mostly if it above the minimum for the design, it's cool; "matching" is not a virtue. Matching happens by follower action and emitter resistors, as long as the Hfe is "enough". And in power transistors, usually 20 or 50 Hfe is ample.

> Device #6 2n3055; beta(meter) = 174, beta(calc) = 653.17

This looks like a mis-measure or mis-marking. Certainly on an old '3055, I'd be very suspicious. It might not be gain, it might be all leakage. Small (50mA) devices sure can be processed and selected for peak gain above 600, but the usual thing on POWER devices is you let gain slip in favor of robustness. IIRC, 2N3055 only promised Hfe>20 (at some hefty current). Low intrinsic gain makes it less likely to tear itself apart under stress. I know the old "2N3055" has been through some changes, but 10-30X rated gain seems like a long way to come without some re-specification and perhaps a new type-number.

Oval loadlines are shown in Radiotron Designers Third, "Relation Between Power Stage and Loudspeaker".