Funny observation in matching BJT power transistors

[yungman]

TL;DR Summary

This is not homework.

Funny that if I pull the transistor out and plug it back in, the Vbe decrease by over 1mV

THIS IS NOT HOMEWORK

I design audiophile power amps as hobby. I need to match power transistors. It is A LOT HARDER than I thought. I built a test fixture to hold up to 50 transistors at a time. I try to match sets of 13 to about 1mV which is very tight as Vbe change is about -2mV/deg C. I have to move transistors around to get matching set all in a roll, not from far apart.

I set Vc=1.4V, emitter resistor is 910ohm and goes to -5.6V to get about 5mA of current through. I keep current and collector voltage Vc low so I won't cause big rise in temperature. Also I keep current low so the contact resistance of the proto plug in board will not cause problem. Below is the pictures to show:

One thing I notice that is very strange, if I pull the transistor out, then plug it right back in, the Vbe decrease by about 2mV. If I repeat on the same transistor again, it's will further decrease. It will take about 10mins sitting with power on to stabilize back to the original value.

I don't think my fingers will cause the temperature to rise so fast, I measure, pull out, plug in and measure, Whole process is less than 10sec. Another thing more important that convince me it's not the problem is it will take at least 10 minutes to go back to normal. It it is thermal issue, the heat up and cool down time is very different, 10sec to heat up, 10 minutes to cool down. I don't buy that at all.

Any explanation on this?I can assure anyone that is interested, matching is A LOT harder than it sounds.

 

[Baluncore]

Funny that if I pull the transistor out and plug it back in, the Vbe decrease by over 1mV

Any explanation on this?

You warm it in your fingers.
-2.5 mV per degree centigrade for silicon.
Fast conduction heating between fingers, slow convection cooling in air.

 

[yungman]

You warm it in your fingers.
-2.5 mV per degree centigrade.

I just edit to describe I don't believe it is. I know it's temperature sensitive, but it takes 10mins to go back to normal, I don't believe my finger heat it up in 10sec, and takes 10 minutes to cool down.

 

[yungman]

BTW, I know it's temperature sensitive, I hold the transistors by the side that is plastic. I avoided touching the back where it goes onto the heat sink. There is no way the plastic can conduct the heat in 10 sec. to raise 1deg.

 

[Baluncore]

The rate of heat transfer is proportional to temperature difference.
Your fingers are significantly hotter than the transistor environment, so heat travels into the package as a wave, reaching the semiconductor over a longer period.

If it heats by 1°C, then it will cool very slowly, since it is close to the temperature of the environment. The test would be to use a tool at the same temperature to remove and return the device from testing.

 

[yungman]

The rate of heat transfer is proportional to temperature difference.
Your fingers are significantly hotter than the transistor environment, so heat travels into the package as a wave, reaching the semiconductor over a longer period.

If it heats by 1°C, then it will cool very slowly, since it is close to the temperature of the environment. The test would be to use a tool at the same temperature to remove and return the device from testing.

I think we cross post. I know about the temperature issue, I grab the transistor by the side that is plastics, I avoid any metal part of the transistors. It's hard to believe it can go this fast. I am talking about I measure the voltage right away.

The weather is quite hot, in the 90s, I am not sure my finger is that much hotter than the surrounding.

 

[Baluncore]

The test would be to use a tool at the same temperature to remove and return the device from testing.

Use a pair of pliers, and do the experiment.

Heat accumulation is an integral function of temperature difference, so the heat from the plastic does not all arrive at the junction at the same time.

 

[yungman]

Ha!

Maybe it's this fast. I use the pliers to pull and plug back, it rise 0.2deg C.

Thanks, I can't believe it's that fast as I hold onto the plastic only for a few seconds.

Thanks

AlanEDIT: I am really surprised how can those non metal material conduct heat so fast? If you look at the picture, there is no metal on the side. I was so careful about it.

 

[Averagesupernova]

I hate to be such a downer here but what's the point? Yes, I get it that what you and some of the rest of us have learned is interesting but I do believe there is a good reason that we were taught in school how to design bjt amplifiers that are stable over temperature as well as from one to the next transistor of the same type.

 

[Baluncore]

I hate to be such a downer here but what's the point?

The point is that when you need to match a pair of transistors, you must be very careful about how you stabilise the temperature, and the power used while making the comparative measurement.

Matched transistors are required in some instrumentation, such as current mirrors and logarithmic converters, but they should not be needed in audio amplifier power output stages.

 

[Averagesupernova]

The point is that when you need to match a pair of transistors, you must be very careful about how you stabilise the temperature, and the power used while making the comparative measurement.

Matched transistors are required in some instrumentation, such as current mirrors and logarithmic converters, but they should not be needed in audio amplifier power output stages.

I agree. That's why I said it was very interesting. I know full well there are places that matched transistors are used. No matter how golden an audiophile thinks their ears are, this is a silly thing to be doing from a practical point of view in audio.
-
That being said, incredibly stupid things can bring stupidly large amounts of money in the audiophile world. I'm wondering if the OP has purchased any matched pairs and compared characteristics against those he has graded himself.

 

[Joshy]

One of my lab assignments at school was to just touch the transistor and turn it on…. we saw the same thing. I’ll see if I can find that assignment.

edit:

Oh here was the one I was thinking about (on the bottom)

 

[Opera]

Back in the late 60's I repaired audio equipment. Transistor manufacturing was still in its infancy. They would sell a family of transistors with each number increase a higher breakdown voltage. I did a lot of car radios and the output transistors had a high failure rate. I bought the lowest voltage transistor and tested them on a voltage breakdown tester I built. I used the low voltage ones in car radios and the others in home stereos. 90% of them tested to the highest voltage.

What was interesting is that once in a while there was one which had a large abnormality in the curve at a particular voltage. I figured those should be tossed.
I had the pleasure of knowing Saul Marantz. He could tell some really funny stories. By training he was a graphic artist. That's why his stuff looked so good.

 

[yungman]

I hate to be such a downer here but what's the point? Yes, I get it that what you and some of the rest of us have learned is interesting but I do believe there is a good reason that we were taught in school how to design bjt amplifiers that are stable over temperature as well as from one to the next transistor of the same type.

I design very high end hifi power amps. I use 12 pairs of output transistors at bias current of over 100mA each. Without getting into detail, the best crossover characteristics is when the emitter resistor only drop about 28mV. This means if the transistors are NOT matched, even if it is off by like 5mV, there will be significant difference in current between transistors.

The thing is tempco of BJT is -2mV/deg C. The hotter it is, the lower the Vbe, so the one with the lowest Vbe will draw more current, in turn Vbe drops further and draws more current and can go into thermal run away.

It is NOT that simple when you get into extreme high end design. My amps are competing with ones that are over $4K power amps. It's nothing like those low bias class AB power amps that use only 2 to 3 pairs of transistors. Those you can design blind folded.

Believe me, when comes to sound quality, the more pairs of output transistors, the better it sounds. You cannot quantify by measuring THD, those means nothing. It's more important to lower the output impedance( NOT by closed loop feedback) particularly open loop output impedance.

 

[yungman]

That being said, incredibly stupid things can bring stupidly large amounts of money in the audiophile world. I'm wondering if the OP has purchased any matched pairs and compared characteristics against those he has graded himself.

There are a lot of snake oil, but there are a lot of intricate stuffs that people never even think of. I had been designing electronics for 30 years, from analog IC, RF, data acquisition and all sort of things. I though audio power amp is easy and good for hobby after retirement. Believe me, it's a lot more complicate than people think.

I ended up designing the amp with BW about 500KHz, stable with at least 10,000pF across the output terminals and slew rate over 30V/uS. I might be over kill a little, but it's not very far off. There are very good reason for these. You cannot tell by measure THD with a resistor load, that's the most deceiving spec in audio world. Remember speakers are very reactive particular having LC crossover. Good speaker cable can have high capacitance. I've seen famous amps like Threshold singing wild with only 1500pF across the terminals. That's why a lot of name brand amps cannot even use some of the good speaker cable.

 

[yungman]

I finally finished building my amp. Here are 3 pictures:

Even though this is home built, it's all pcb, terminal blocks, crimp wires. Everything done like production products. I compare my amp to commercial amps even in built quality. I have no plan in making it a business. There's no money in very high end audio stuffs. Cost me a little over $1,000 just for parts to build one amp(no snakeoil stuffs, components are just from Digikey, wires are the cheapest ones from ebay). It has 12 pairs of power transistors per channel. I have 4 separate power supplies, two for each channel, separate regulated supply for the frontend and big supply for the output section. But I really enjoy this.

 

[Averagesupernova]

...components are just from Digikey, wires are the cheapest ones from ebay).

You better not let that info out if you want to do as I said:

That being said, incredibly stupid things can bring stupidly large amounts of money in the audiophile world.

Not that your projects are stupid, there is a great value in doing these things.

-
This is what counts:

I really enjoy this.

 

[yungman]

You better not let that info out if you want to do as I said:

Not that your projects are stupid, there is a great value in doing these things.

-
This is what counts:

There is no money in high end stuffs, sales volume is too low. Only the mid-fi stuffs are selling by the millions. Like those class D amps, they are quite decent in the price range of $1000 or so.

Also, there's a lot of snake oil in high end stuffs, like special metal speaker cables, all the expensive capacitors and all that. You cannot explain enough to justify the design.

Even if you do demonstration, it's hard. When comes to top quality sound, everything matters, the room, the speakers, the placement of speakers(very important), the power amp, the speaker cables, everything matter, the sound is only as good as the weakest link. It's very hard to demo. Imagine you have to ship all the equipment to a show, hoping you can find a good placement. It's a lot of money to market the amplifier.

I have friends offered to buy my amps, I just don't even want to go there. I ended up giving one to a good friend of mine since I have quite a few, more than I ever need.

There are so many snake oil in high end audio, also some big brand actually hire well known people to give wrong information. Believe it or not, one guy advocate testing speakers and amp by listening to one single speaker(single mono channel)! For cry out loud, the whole thing about high end system is being able to match the phase of the two speakers so you don't hear any phase shift sitting in the middle. Believe it or not, One of my major test is using the system to listen to news channel! If I get the human voice right, I already go very very far. You listen to how solid the voice, clarity that you can understand the words. That the person like standing in front of you speaking to you. If you can get that, you already go very far! You have to make the stereo to sound like that. Anything wrong, you can hear it. You will notice the voice always sounds better if you only use one side. The key is to make the voice sound just as solid, as clear in stereo with the person like standing in the middle of the room. We call this sound stage and imaging.

I do everything by engineering, I can explain most of the things in electronic theory. Only one thing I cannot explain is different front ends sound different even thought they all are ultra low THD. My lowest THD front end is 0.0015% worst case at almost full power at all frequency. The one that is 0.006% sounds better! Why? I have no idea. I did double blind test and people agree! Human ears are really complicate and of cause the frequencies of sound and music is very complicate. That's part of the reason I design my amps to have -3dB at about 500KHz so there is minimum phase shift in audio band. Also 500KHz BW will give me loop gain of about 25(theoretically) at 20KHz so there is enough negative feedback to keep the output impedance low.

You guys are talking about buying matching transistors, can you imagine how much it cost to buy 12 matching transistor set? Sadly, I just tested the amp today, it's NOT much better if any than my older one with only 9 pairs but higher bias current. That's kind of disappointing. I guess there is a limit to increase number of pairs of output transistors. I can tell you there is a big difference between 7 pairs and 9 pairs, maybe, 9 is the optimal. If the 12 pair is better, it's only very slightly better, it can well be psychosomatic, that I want to hear the difference!.

Anyway, I talk too much already, at least people here can understand what I am saying.

 

[Baluncore]

I finally finished building my amp. Here are 3 pictures:

Well done. It looks very professional.

A chain is only as strong as its weakest link.
Now you must work on your choice of music, the company you keep, the listening space, and the loudspeakers.

 

[yungman]

Well done. It looks very professional.

A chain is only as strong as its weakest link.
Now you must work on your choice of music, the company you keep, the listening space, and the loudspeakers.

Oh yeh, the speakers, speakers placements, speaker cables are just as important. This is my living room:

I bought the speakers used, still $6000USD, new price in 2003 was over $15K. Notice I have black tapes at the foot of the speakers, I had to move inch by inch, turning the speaker angles a little at a time to optimize the sound stage. I used to build my own speaker cables. It's not the wires that make a difference, it's how many smaller insulated wires that make the difference. I used 6 pairs of 16gauge wires to make the cable twisted together. Against the common believe that the resistance of the wire is important, it's the INDUCTANCE of the wire that is important. If you look at the inductance per unit length of wires of different gauges, you'll see a 12 gauge wire is only slightly less inductance than a 16gauge. The way to reduce inductance is use many small wires in parallel, tightly couple together to minimize inductance. Just draw the equivalent circuit of the output impedance of the power amp through the cable to the LRC equivalent of the speaker, you will see the the importance of lower inductance of the cable. You also will see the resistance of the wire is NOT even in the picture. Now I use a Chinese copy of Kimber Kable speaker cable, same idea, just better looking for the living room. It is not sounding any better than my home built cheap ebay copper clad aluminum(cheapest kind) wires. I concluded the kind of wire is not even important as resistance is not even in the picture. Calculate the reactance of the wire and you'll see.

I just use any DVD and CD players, I don't think the low signal level section is important at all. same as the interconnect coax cables. I just buy the thicker ones from Amazon. It's only from the power amp to the speakers that is very critical. I built my own preamp, just an audio jFET opamp to buffer the signal from the DVD players with a volume control, nothing mystery in that. Funny I find jFET opamp sounds better, but still ala Digikey! About like $2 ea. I am born CHEAP, I don't spend an extra penny if I don't have to. The only thing that is expensive in my amp is the CHASSIS, believe it or not, it's $500USD including shipping from China. The shipping today is more than the chassis itself!

In the picture, you see 3 amps I built.

Too bad I don't know anything about speaker design, or else I would not pay that kind of money. Also, I don't think my big boss like to have a home built speakers in the living room. She's picky and my carpentry work is less than desirable. I know my limits.