Matched output transistors: does gain change with age?

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Hi:

I am restoring a 30+ year old Son of Ampzilla. The amplifier was still working/passed DBT check/reasonable DC offset prior to my decision to do a complete recap and replace meter lamps with LEDs. I know that the matched input differential pairs often drift apart with passage of time, but was surprised to find a significant variation in gain in the output pairs:

Channel A

-NPN- -NPN- -PNP- -PNP-
30 73 95 72

Channel B

-NPN- -NPN- -PNP- -PNP-
51 128 167 79

Yes, the transistors have original GAS 118/119 numbers; there was no sign of previous repair/replacement.

So question: is it normal to see such large variation in gain (the factory obviously had hand matched them 30+ years ago)?

Rgds
Mayank
 
Last edited:
Administrator
Joined 2004
Paid Member
Hi Mayank,
No, those were not hand matched, or not matched properly if they were matched. You will tend to see changes if that output stage failed and only the shorted or open transistors were changed.

New transistors have better performance, including being able to hold their beta up at higher currents. When I restore an old amplifier of mine, I do install a set of fresh output transistors that are hand matched. That's even if I ever get my dream amplifier - a Marantz 500 and all it's output transistors!

New transistors generally have less spread on their beta values, so it is easier to get matched sets while buying fewer over all to achieve that goal. All good things for me and you. I would install a matched set of outputs and try to match the complimentary driver transistors as well. The input diff pair is critical to match, and a jig is needed in order to get the really tight matches that really improve performance. Another pain in the rear end, but well worth the effort.

-Chris
 
Thank you for your comments Chris!

No, those were not hand matched, or not matched properly if they were matched.

The transistors did have the tell-tale red number markings which had led me to presume they might have been, but perhaps they weren't matched properly or have just drifted apart.

You will tend to see changes if that output stage failed and only the shorted or open transistors were changed.

Agree. I have found this to be in several "serviced & working" amplifiers with one/two outputs replaced.

I have already replaced the VAS and drivers with newer OnSemi devices. I will be replacing the outputs as well once I receive my package from Mouser.

Good luck on your hunt for the Marantz 500 !!

Rgds
Mayank
 
I am curious to know what is the actual ( measured ) benefit of pairing output transistors.
Are these amplifiers low feedback loop gain ?

I will defer technical clarification on this to Chris and other "Yoda's" here. I've always replaced all the output transistors with similar gain whenever I find 1 or 2 bad in the bank.

Attached is the schematic of the outputs of this GAS Ampzilla Son amplifier.

Rgds
Mayank
 

Attachments

  • Output Schematic.jpg
    Output Schematic.jpg
    417.3 KB · Views: 714
was surprised to find a significant variation in gain in the output pairs:

Channel A

-NPN- -NPN- -PNP- -PNP-
30 73 95 72 :eek:

Channel B

-NPN- -NPN- -PNP- -PNP-
51 128 167 79 :eek:

Yes, the transistors have original GAS 118/119 numbers; there was no sign of previous repair/replacement.

Your very high Hfe numbers make me think you measured at very low current, a few mA , probably a handheld meter or even the one included in a multimeter.
Those current levels are not realistic at all.

To be relevant you should measure power transistors at a current level compatible with what they will deliver to speakers, say 4 to 10A or thereabouts.
So question: is it normal to see such large variation in gain (the factory obviously had hand matched them 30+ years ago)?
Variation in gain?
Quite possible, but to be certain make a jig to measure them at high current.

Hand matched? ... maybe, but leave "obvious" out, you are just assuming it :)

I bet there was a reasonably wide window of tolerance, but in any case the great equalizer is the individual emitter resistor.

In any case no significant variation or degradation with time, what you have there is basically what was fit on Day 1.
 
if Vbe and hfe are linked then matching becomes more critical. The higher beta transistor will drive somewhat more power due to the emitter resistors which mitigate current hogging in one device.
I wonder
_what is dominant for load sharing Vbe or Beta ?
_is there a dependance. I mean do we have usualy hight beta with low Vbe or usualy high beta with high Vbe or nothing usual ?
_is 0.2 ohm emitter resistor, high enough to force load sharing safely ?
 
Administrator
Joined 2004
Paid Member
Hi Nigel,
Its just a crystal lattice which doesn't change.
I wish! Nope, it changes all right, but not like the cathode of a tube becoming depleted.
If you are buying in transistors then they usually come from same batch so will probably be close enough anyway.
Only if you have wide tolerances, but it is better than a mixed bag of transistors.

-Chris
 
Administrator
Joined 2004
Paid Member
Hi mchambin,
I am curious to know what is the actual ( measured ) benefit of pairing output transistors.
Up to a 20 dB reduction in distortion, as reported and measured by the fine folks at Motorola, and later at On-Semi. Measured by myself many times over the years as well.
Are these amplifiers low feedback loop gain ?
Yes and no. Every amplifier will benefit. Probably the low gain amps will benefit more.

-Chris
 
Administrator
Joined 2004
Paid Member
Hi mchambin,
I wonder
_what is dominant for load sharing Vbe or Beta ?
Beta. VBE is swamped by the voltage across the emitter resistors, and it becomes less effective as the temperature increases. At high currents, voltage across an emitter resistor will easily swamp the effects of vbe.
_is there a dependance. I mean do we have usualy hight beta with low Vbe or usualy high beta with high Vbe or nothing usual ?
The two don't have any relationship to each other that I have noticed
_is 0.2 ohm emitter resistor, high enough to force load sharing safely ?
Generally speaking, if your outputs are matched properly - yes. That is also assuming a heat sink without hot spots.

-Chris
 
Hi mchambin,

Beta. VBE is swamped by the voltage across the emitter resistors, and it becomes less effective as the temperature increases. At high currents, voltage across an emitter resistor will easily swamp the effects of vbe.
You seem to say that Hfe variatio0n is the main problem, but I think it´s exactly the contrary.
Hfe variation (within reason, such as a 2:1 or 1:2 factor) does NOT determine how much current passes through each transistor, because they are not fed "individually each the same base current" by any means, all bases are tied together so by definition are at same voltage, driver transistor supplies drive current of course, but in principle "each transistor pulls as much base current as it needs".
Once you introduce emitter resistors, each power transistor becomes "voltage controlled", output current becomes V across emitter resistor/its resistance and that voltage is Voltage present at the base minus Vbe.

You can have minor Vbe variations which as you noticed are swamped by voltage across resistors but you do not have a mechanism to inject a defined (and the same for each one) current into each base.

Suppose each power transistor is passing 5A into the load, each resistor will drop 1V .
You´ll need some extra 0.7V at each base, so drive *voltage* will be 1.7V

If you have, say, 50mV or even 100mV Vbe variation within transistors (which is gross), it´s clear that net voltage across resistors will vary very little, so current through them will also be quite uniform.

And if we have now equalized currents, what happens when Hfe varies unit to unit?
Even as much as 3:1 or 4:1?

Well, since collector current is quite uniform, Hfe variation means that *base* currents will be different ... but I don´t worry mkuch about that, main dissipation factor is Vce*Ic , which as we see is reasonably uniform; Vbe*Ib is also a legitimate contributor , but pales compared to the main one.

So I agree that different Hfe does cause *some* change in each transistor total dissipation ... but its effect is quite small.
Hope I have made me clear :)
 
I wonder
_what is dominant for load sharing Vbe or Beta ?
_is there a dependance. I mean do we have usualy hight beta with low Vbe or usualy high beta with high Vbe or nothing usual ?
_is 0.2 ohm emitter resistor, high enough to force load sharing safely ?
My answer above suggests that 0.2 ohm emitter resistors are great current equalizers, and neither Vbe or Hfe are *that* important as it appears at first sight.
 
New transistors have better performance, including being able to hold their beta up at higher currents. When I restore an old amplifier of mine, I do install a set of fresh output transistors that are hand matched. That's even if I ever get my dream amplifier - a Marantz 500 and all it's output transistors!
-Chris

Thanks for your post Chris, it inspired me to learn more about Marantz and the model 500. Good luck with your search for a 500:)


Here's a link for anyone that may be interested.

http://audiohistory.com/files/documents/marantzhistory.pdf
 
Administrator
Joined 2004
Paid Member
Hi JMFahey,
Clearly we disagree. My own personal observations support (and gave rise) to my beliefs. Beta variation makes more difference near the low current end maybe, that's where the contribution of emitter resistors is minimal. However, Vbe differences tend to be quite small, especially with beta matched transistors.

What should have grabbed your attention was the app note from Motorola and On-Semi where they confirm that matching beta can lead up to a 20 dB reduction in distortion in the output stage. You can see the same information on some data sheets for power transistors. The app note was written long before I made the connection, but I hadn't seen it until after I had come to this realization (but without numbers). That kind of validation feels pretty good.

I know that several engineers believe it is the voltage from emitter base that determines current flow, but I think that the impedances are low enough so that E-B voltage is actually translated into current flow. They just like something that is more linear, but then real life tends to be a little messier, doesn't it?

At any rate, the scientists who manufacture transistors seem to support the view that it is in fact beta that determines current sharing. Everything I have seen and measured supports this view.

If you have a good distortion analyser, you can examine this on your own time. Also, I have found that beta measured at about 1 mA tends to vary as the beta at higher currents. However, you have to hold your variables constant is as reasonably possible. Running transistors at high current introduces temperature as another very touchy variable. As long as you don't have hot spots, beta should still assist better current matching, even though at higher currents I suspect the match on the emitter resistors might play a bigger role.

-Chris
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.