Hi Richard,
Hey, that's pretty good! I'll have a look at it in more detail. I was considering something close to this in order to ensure equal base currents, but didn't get to the comparing current part due to measuring the emitter current and comparing that to the collector current. You are right in that high gain transistors should have a low error.
Thank you Richard, Chris
Hey, that's pretty good! I'll have a look at it in more detail. I was considering something close to this in order to ensure equal base currents, but didn't get to the comparing current part due to measuring the emitter current and comparing that to the collector current. You are right in that high gain transistors should have a low error.
Thank you Richard, Chris
That matches beta only. If you need to match Vbe only, or if you need to match both Vbe and beta, you need a different test fixture.
People who have studied an Algorithms textbook will recognize this as a problem with O(N^2) computational complexity. If you have 6 transistors A,B,C,D,E,F you have no choice but to test all (6^2 - 6)/2 possible pairings:
AB, AC, AD, AE, AF, BC, BD, BE, BF, CD, CE, CF, DE, DF, EF.
People who have studied an Algorithms textbook will recognize this as a problem with O(N^2) computational complexity. If you have 6 transistors A,B,C,D,E,F you have no choice but to test all (6^2 - 6)/2 possible pairings:
AB, AC, AD, AE, AF, BC, BD, BE, BF, CD, CE, CF, DE, DF, EF.
A similar complimentary matcher was posted in this thread (post #14):
Complementary transistor matcher?
Complementary transistor matcher?
Hi Mark,
Best of the season to you as well, and have a great New Year!
That response is simply due to the fact that I've examined matching transistors for decades as I attempt to improve my own work. Matching vbe has been strongly shown to be without any positive assistance at all. Matching beta on the other hand really improves the operation of diff pairs and outputs (drivers too if there are more than one). If you can match compliments, that has been very helpful as well. I've seen matching vbe pushed for years, but I can't figure out why if anyone has actually put some time into studying matched parts.
I find the idea that you should match transistors by their emitter - base drop very annoying because it is without merit and continues to be pushed. Maybe it's a case of folks not having the tools and / or knowledge, so all they can do is match vbe. There is the thought that if you do something, it has to be of value.
Anyway, even the manufacturers of transistors indicate that you match beta. Vbe is simply a byproduct of the fact that you have a PN junction. One thing that is important to know is that temperature variations completely invalidate any readings you may obtain. So that variable must be controlled in some way both in the measuring process and in the application.
-Chris
Best of the season to you as well, and have a great New Year!
That response is simply due to the fact that I've examined matching transistors for decades as I attempt to improve my own work. Matching vbe has been strongly shown to be without any positive assistance at all. Matching beta on the other hand really improves the operation of diff pairs and outputs (drivers too if there are more than one). If you can match compliments, that has been very helpful as well. I've seen matching vbe pushed for years, but I can't figure out why if anyone has actually put some time into studying matched parts.
I find the idea that you should match transistors by their emitter - base drop very annoying because it is without merit and continues to be pushed. Maybe it's a case of folks not having the tools and / or knowledge, so all they can do is match vbe. There is the thought that if you do something, it has to be of value.
Anyway, even the manufacturers of transistors indicate that you match beta. Vbe is simply a byproduct of the fact that you have a PN junction. One thing that is important to know is that temperature variations completely invalidate any readings you may obtain. So that variable must be controlled in some way both in the measuring process and in the application.
-Chris
@reddish75: I am not sure which schematic specifically you are referring to. There is a video w/ schematics for transistor matching here (small PNP or NPN pairs only). Some people like it, some people don't, but it gives a good idea if the transistors have similar hfe gain at a given Ib, Ic and Voltages. Some people don't like it because the transistors are not necessarily biased at the same conditions of the real circuit they need. In the end we get what we pay for. See at 53m00s.
YouTube
See also this one at 7m15s. Tony will have another video on transistor matching with a nice solution. Yet to happen.
YouTube
YouTube
See also this one at 7m15s. Tony will have another video on transistor matching with a nice solution. Yet to happen.
YouTube
Hi Chris,
I think it is fair to say that whether matching Vbe or beta is most important depends a lot on the application (e.g., the particular circuit and where in the circuit we are talking about). Also some applications are much more sensitive to beta and/or Vbe mismatch than others.
A place where Vbe mismatch in a conventional Thompson amplifier topology is if the two transistors in the current mirror load are not matched in Vbe and the current mirror has no emitter degeneration. That will cause the collector currents in the differential pair to be different even if the differential pair itself is perfectly matched for both Vbe and beta. Different current in the diff pair transistors will cause a transconductance mismatch and cause some of the distortion we have been trying to minimize. It will also cause an operating Vbe mismatch in the diff pair.
However, dgenerating the current mirror greatly reduces the dependence on Vbe matching in the current mirror and tends to reduce its noise contribution. I shudder every time I see a current mirror load that is not degenerated. As a start, I often use the same amount of degeneration in the current mirror as in the diff pair.
Cheers,
Bob
Hi Bob,
This is one thread I was watching.
It would be great to put this information and (the one you wrote up in the other thread about the same questions) up on your web site, that way it does not get lost or buried in time. I will take these two notes you wrote up on device matching and save them on my computer for my own reference.
Cheers
Rick
This is one thread I was watching.
It would be great to put this information and (the one you wrote up in the other thread about the same questions) up on your web site, that way it does not get lost or buried in time. I will take these two notes you wrote up on device matching and save them on my computer for my own reference.
Cheers
Rick
Hi Bob,
I normally do match the current mirror load as well when it exists. I have never not matched the diff pair since all my experience shows that doing this does drop the distortion. This is my ultimate goal even when simply repairing an amplifier channel. I learned a long time ago that if the output of an amplifier sticks high, it will create a reverse bias E-B breakdown of one of the two transistors in a diff pair. I have seen this make that transistor noisy and even change the characteristics. Therefore, I normally replace the diff pair if I am repairing a defective channel in an amplifier.
When using a closely matched pair of transistors, and emitter degeneration resistors will also have to be matched or they will throw off the balance of the diff pair you worked so hard to get a tight match with.
-Chris
I normally do match the current mirror load as well when it exists. I have never not matched the diff pair since all my experience shows that doing this does drop the distortion. This is my ultimate goal even when simply repairing an amplifier channel. I learned a long time ago that if the output of an amplifier sticks high, it will create a reverse bias E-B breakdown of one of the two transistors in a diff pair. I have seen this make that transistor noisy and even change the characteristics. Therefore, I normally replace the diff pair if I am repairing a defective channel in an amplifier.
When using a closely matched pair of transistors, and emitter degeneration resistors will also have to be matched or they will throw off the balance of the diff pair you worked so hard to get a tight match with.
-Chris
I guess what Chris is saying, is to change them when you change the IPS bjts to a matched pair, since most old electronics hardly ever used precision resistors (except for some phono stages), let alone a current mirror for the IPS load.
I would think that 1% is all that is required, not sure if higher precision is necessary.
and you are saying to use precision resistors because they are cheap these days which makes a lot of sense.
Are we going to discuss what brand and types of resistors to use ? I have been using Vishay MBB02070 as they are pretty economical imo.
I would think that 1% is all that is required, not sure if higher precision is necessary.
and you are saying to use precision resistors because they are cheap these days which makes a lot of sense.
Are we going to discuss what brand and types of resistors to use ?
I have been using Vishay MBB02070 as they are pretty economical imo.Cheapest in stock that day from reputable supplier is my algorithm. No point caring about who makes them, life is short!
BTW for a differential pair into a current mirror load only the current mirror symmetry matters for linearity, the diff pair matching only affects the offset voltage, which may not be very important - then you can select the diff pair devices for low noise / high gain without the constraint of matching Vbe.
BTW for a differential pair into a current mirror load only the current mirror symmetry matters for linearity, the diff pair matching only affects the offset voltage, which may not be very important - then you can select the diff pair devices for low noise / high gain without the constraint of matching Vbe.
Hi Mark,
Right. I don't bother matching vB-E at all. The few times I have checked that in circuit, they tend to be very close anyway. A couple mV isn't going to upset DC offset in anyway that matters to audio. "Close enough for rock and roll".
When degeneration resistors are involved, they will at the very least become 1% metal film parts. If I have the diff pair matched to less than 1 mV (not hard to do), I'll simply match a pair of 1% resistors.
The jig uses a few 0.1% resistors in key locations. And you're right, they aren't too expensive when you are only using a couple here and there. I used 8 when I built my jig. 4 - 100R0 for collector loads and 4 - 10K00 resistors for base to common connections. I was using a few HP 34401A meters to do measurements with, and now I have a Keysight 34465A, so those resistors in 0.1% actually made sense to use. A perfect match would see 0.0 mV between collectors as the meter is mostly used as a null indicator. If you feel like measuring the voltage drops you can figure out what the percentage match is. By knowing the emitter currents, collector currents and base currents you can figure out what the β is at that temperature and current level.
It would be fun to try and calibrate a center 0 meter for this. One from a scrap FM tuner might work well. You could mark it in mA or % (or even both).
-Chris
Right. I don't bother matching vB-E at all. The few times I have checked that in circuit, they tend to be very close anyway. A couple mV isn't going to upset DC offset in anyway that matters to audio. "Close enough for rock and roll".
When degeneration resistors are involved, they will at the very least become 1% metal film parts. If I have the diff pair matched to less than 1 mV (not hard to do), I'll simply match a pair of 1% resistors.
The jig uses a few 0.1% resistors in key locations. And you're right, they aren't too expensive when you are only using a couple here and there. I used 8 when I built my jig. 4 - 100R0 for collector loads and 4 - 10K00 resistors for base to common connections. I was using a few HP 34401A meters to do measurements with, and now I have a Keysight 34465A, so those resistors in 0.1% actually made sense to use. A perfect match would see 0.0 mV between collectors as the meter is mostly used as a null indicator. If you feel like measuring the voltage drops you can figure out what the percentage match is. By knowing the emitter currents, collector currents and base currents you can figure out what the β is at that temperature and current level.
It would be fun to try and calibrate a center 0 meter for this. One from a scrap FM tuner might work well. You could mark it in mA or % (or even both).
-Chris
Surely the opposite is true, specifying a particular series is least likely to last - if you give the specification only and let people figure out their own BOMs they can redo that in the future as component series and even manufacturers come and go...
"33k 1% metal film 1/4W 0207" is all you need, or "33k 1% thin-film 200mW 1206 SMT" or whatever.
For something much fussier like inductors the precise component part number may be relevant, but for resistors?
Hi Mark,
I strongly doubt a 0207 resistor can dissipate the same 1/4W the 1206 package is rated for. That, and we now worry about voltage ratings for resistors, even through-hole varieties.
In fact, for low noise or high power applications, you pretty much have to specify the right part.
Given that most folks will choose the capacitors and resistors dujour, I wouldn't want to leave them to their own devices <sic>.
-Chris
I strongly doubt a 0207 resistor can dissipate the same 1/4W the 1206 package is rated for. That, and we now worry about voltage ratings for resistors, even through-hole varieties.
In fact, for low noise or high power applications, you pretty much have to specify the right part.
Given that most folks will choose the capacitors and resistors dujour, I wouldn't want to leave them to their own devices <sic>.
-Chris
I just made those example up on the spot.
For low noise "metal film" or "thin-film" does the job.
For capacitors "polypropylene" or "polypropylene foil" will usually be fine.
If voltage matters you specify it, if power would be borderline for the size, specify it - all this is common sense, specify what is wanted.
Just saying the series is losing information (why that series?) and makes substitution when unavailable more tricky.
Perhaps you also have to say "not unbranded junk"
There's no reason not to give a BOM too, but that's a more transient thing.
For low noise "metal film" or "thin-film" does the job.
For capacitors "polypropylene" or "polypropylene foil" will usually be fine.
If voltage matters you specify it, if power would be borderline for the size, specify it - all this is common sense, specify what is wanted.
Just saying the series is losing information (why that series?) and makes substitution when unavailable more tricky.
Perhaps you also have to say "not unbranded junk"
There's no reason not to give a BOM too, but that's a more transient thing.
Hi Mark,
I don't know. I always look at the specification before just sticking any old part in. But then, I'm use to repairing good test equipment. The specification leads you to a proper substitute part. I would rather have the general spec on the schematic and the parts list give you a specific part you can chase.
-Chris
I don't know. I always look at the specification before just sticking any old part in. But then, I'm use to repairing good test equipment. The specification leads you to a proper substitute part. I would rather have the general spec on the schematic and the parts list give you a specific part you can chase.
-Chris