You need a semiconductor analyser.
I designed my own transistor matcher.
It just uses a inverting transistor stage.
A PIC micro outputs a ramping base voltage and the output is read in by A2D then passed on to a PC for display.
The PC program can display up to 5 transistors curves at the same time for matching.
I designed my own transistor matcher.
It just uses a inverting transistor stage.
A PIC micro outputs a ramping base voltage and the output is read in by A2D then passed on to a PC for display.
The PC program can display up to 5 transistors curves at the same time for matching.
Thank you for your quick response, but I was referring to the board, a device developed by Codeniac. How to use it?
Hi Prohodimez,
Could you link to that device please? If it is what I am thinking of, I designed it and it is only designed to match NPNs or PNPs together. You seem to be asking how to match complimentary devices.
I did design in the ability to measure base current as well as collector current. If you installed the matched pair and got them running at the same tail current, they would be stable enough to measure the currents and you can then calculate the beta for each pair. Then simply put the pairs with the same beta together. Keep in mind that complimentary devices will never actually match, but getting the beta the same will help. It is the best you can do.
To "match" the output transistors, make a matched pair and follow the same procedure. They need to be matched so that the temperature term drops out and they reach an equilibrium where the device has a stable voltage drop across it, and the emitter current is also steady. Then you will also have matched channels for performance (a small difference that shouldn't matter in a practical sense).
Normally you would install transistors in physical contact and isolate them from the air with a foam "hat". Connect meter leads between the collectors and place a box or bowl over the entire thing to break air currents. Wait for things to settle down, we are talking about thermal equilibrium here so it takes 5 minutes or so. Your meter is a null device and you should be measuring in mV. The closest to zero you can get indicates a closer match between the parts.
This isn't something that is done quickly. It is boring but the results are well worth it. You can get sub 1% matches repeatedly. Note that the collector resistors (100R0) and base resistors (10K0) are 0.1% parts. You can use 0.01% if you want, but it gets expensive and I don't think that practically it would buy you anything. The sockets are high quality ones from Digikey or Mouser. Using the ones available from Ebay or Alibaba will be a frustrating experience and they greatly affect the performance of the jig. Do not economize on the sockets. They aren't terribly expensive anyway.
-Chris
Could you link to that device please? If it is what I am thinking of, I designed it and it is only designed to match NPNs or PNPs together. You seem to be asking how to match complimentary devices.
I did design in the ability to measure base current as well as collector current. If you installed the matched pair and got them running at the same tail current, they would be stable enough to measure the currents and you can then calculate the beta for each pair. Then simply put the pairs with the same beta together. Keep in mind that complimentary devices will never actually match, but getting the beta the same will help. It is the best you can do.
To "match" the output transistors, make a matched pair and follow the same procedure. They need to be matched so that the temperature term drops out and they reach an equilibrium where the device has a stable voltage drop across it, and the emitter current is also steady. Then you will also have matched channels for performance (a small difference that shouldn't matter in a practical sense).
Normally you would install transistors in physical contact and isolate them from the air with a foam "hat". Connect meter leads between the collectors and place a box or bowl over the entire thing to break air currents. Wait for things to settle down, we are talking about thermal equilibrium here so it takes 5 minutes or so. Your meter is a null device and you should be measuring in mV. The closest to zero you can get indicates a closer match between the parts.
This isn't something that is done quickly. It is boring but the results are well worth it. You can get sub 1% matches repeatedly. Note that the collector resistors (100R0) and base resistors (10K0) are 0.1% parts. You can use 0.01% if you want, but it gets expensive and I don't think that practically it would buy you anything. The sockets are high quality ones from Digikey or Mouser. Using the ones available from Ebay or Alibaba will be a frustrating experience and they greatly affect the performance of the jig. Do not economize on the sockets. They aren't terribly expensive anyway.
-Chris
Hi Prohodimez,
I surmise that your amplifiers are probably working aside from the objectionable 300mV offset at the outputs--- correct? If the offset could be trimmed to 0V, would you be satisfied?
I am deeply suspicious of the network R24, R25, T12, etc. I imagine the intent was to allow adjustment of bias input offset currents of the MAT03/MAT03 transistors by varying P2, but I'm skeptical that it can do that reliably. I'm going to predict that the output is minus 300mV with respect to ground, that trimming P2 affects offset and can make it even worse, but not trim it to 0V?
If this is correct, I suggest removing T12. If you're lucky, offset may decrease and be acceptably small. If there's improvement but the offset is still excessive, I can suggest simple board mods to allow trim-to-zero behavior.
Good luck!
I surmise that your amplifiers are probably working aside from the objectionable 300mV offset at the outputs--- correct? If the offset could be trimmed to 0V, would you be satisfied?
I am deeply suspicious of the network R24, R25, T12, etc. I imagine the intent was to allow adjustment of bias input offset currents of the MAT03/MAT03 transistors by varying P2, but I'm skeptical that it can do that reliably. I'm going to predict that the output is minus 300mV with respect to ground, that trimming P2 affects offset and can make it even worse, but not trim it to 0V?
If this is correct, I suggest removing T12. If you're lucky, offset may decrease and be acceptably small. If there's improvement but the offset is still excessive, I can suggest simple board mods to allow trim-to-zero behavior.
Good luck!
Last edited:
Hi Prohodimez,
You know the MAT02 and MAT03 are closely matched - right? That has more of an effect on distortion than playing with caps and other silly things. Also DC offset, and now I know why yours is so high.
All bets are off on that build Prohodimez. I don't know why you are even trying to improve something as deeply flawed. Until you get the proper pairs installed, there is zero point in listening tests or efforts to fix the offset, which is a situation you created using those transistors.
I'm not trying to be hard on you, just letting you know what the effects of substituting those input pairs.
-Chris
You know the MAT02 and MAT03 are closely matched - right? That has more of an effect on distortion than playing with caps and other silly things. Also DC offset, and now I know why yours is so high.
All bets are off on that build Prohodimez. I don't know why you are even trying to improve something as deeply flawed. Until you get the proper pairs installed, there is zero point in listening tests or efforts to fix the offset, which is a situation you created using those transistors.
I'm not trying to be hard on you, just letting you know what the effects of substituting those input pairs.
-Chris
If you're seeing well under 1mV, maybe you declare victory! Your other channel is likely to behave similarly.
The mod I was going to suggest (probably not necessary): referring to your original schematic, remove T12 and R26. Cut a trace if necessary to make P2's wiper available as a third terminal; connect top of P2 to +15V so that P2 has +/- 15V presented across its resistive span; lift the end of R25 currently tied to the nearby "A" test point and instead connect to P2 wiper; finally, connect a non-polar cap from T12's "collector pad" to signal ground so that the junction of R25 and R25 is well filtered. Now P2 will introduce an adjustable offset at the input. You'll probably find the pot way too twitchy, so make R24 and R25 larger to get easier tweaking.
If you want to servo for the fun, your schematic is probably a bit more complex than needed--- and IMO replicating the R31, R32, T12 network is asking for trouble.
Have fun!
The mod I was going to suggest (probably not necessary): referring to your original schematic, remove T12 and R26. Cut a trace if necessary to make P2's wiper available as a third terminal; connect top of P2 to +15V so that P2 has +/- 15V presented across its resistive span; lift the end of R25 currently tied to the nearby "A" test point and instead connect to P2 wiper; finally, connect a non-polar cap from T12's "collector pad" to signal ground so that the junction of R25 and R25 is well filtered. Now P2 will introduce an adjustable offset at the input. You'll probably find the pot way too twitchy, so make R24 and R25 larger to get easier tweaking.
If you want to servo for the fun, your schematic is probably a bit more complex than needed--- and IMO replicating the R31, R32, T12 network is asking for trouble.
Have fun!
Last edited:
HI BSST,
His DC offset is 300 mV ...
This is a pretty standard gain stage. Sensitive to unmatched diff pairs.
-Chris
His DC offset is 300 mV ...
He didn't use the MAT matched pairs and stuck transistors in there, not matched. Then he made a circuit change ...Please tell me how to match the transistors in this scheme. I tried using a Chinese multimeter, but nothing worked, the offset is 300mv on both channels.
He replaced the series signal resistor with a capacitor, there is already a capacitor in series, so who knows what he really did.Ingeniously! I removed the board with the T12 transistor, put Nichicon BP in the R2 gap, the offset on the channel without a capacitor is 0.4 mv, with a capacitor of 0.2 mv after 15 minutes of warming up.If you can suggest simple board mods, it would be great.
This is a pretty standard gain stage. Sensitive to unmatched diff pairs.
-Chris
Hi Chris,
I wasn't sure what to make of the "R2 gap" and assumed a typo or something benign. R2 replaced with a big cap would make unity gain at DC and large gain above--- assuming the amp didn't oscillate.
The DC gain is only about 6.6, or about 45 mV offset referred to the input--- a lot. There's quite a bit of emitter degeneration at the MAT sites, so even with discrete, unmatched input transistors, my theory is that the R24, R25, T12 network is responsible for the big offset; whatever voltage presents at T12's collector has much higher gain along the R24 path to the input than does the R25 path. R24 dominates significantly and it can only pull the non-inverting input low, can't push to positive voltage in the event it's what's needed. Only my opinion of course.
Hi Prohodimez,
Would you confirm discussion about R2 site? You could install a large cap in series with the 1K resistor to reduce gain to 1 at DC, but must retain the 1k to have flat gain at audio frequencies. But as an experiment, would you retain R2 = 1K and report output offset? Thanks.
Steve
I wasn't sure what to make of the "R2 gap" and assumed a typo or something benign. R2 replaced with a big cap would make unity gain at DC and large gain above--- assuming the amp didn't oscillate.
The DC gain is only about 6.6, or about 45 mV offset referred to the input--- a lot. There's quite a bit of emitter degeneration at the MAT sites, so even with discrete, unmatched input transistors, my theory is that the R24, R25, T12 network is responsible for the big offset; whatever voltage presents at T12's collector has much higher gain along the R24 path to the input than does the R25 path. R24 dominates significantly and it can only pull the non-inverting input low, can't push to positive voltage in the event it's what's needed. Only my opinion of course.
Hi Prohodimez,
Would you confirm discussion about R2 site? You could install a large cap in series with the 1K resistor to reduce gain to 1 at DC, but must retain the 1k to have flat gain at audio frequencies. But as an experiment, would you retain R2 = 1K and report output offset? Thanks.
Steve
Last edited:
Hi Steve,
You could be right. I looked at the basic circuit and ignored the rest. Prohodimez stated he didn't know how to use the matcher and has not confirmed the one he tried to use. I can only assume the diff pairs are not matched at all.
I do know from experience that unmatched diff pairs in this circuit can lead to high DC offsets (300 mV is extremely high) and high distortion (compared to what the circuit is capable of).
Now, looking at T12 and the servo, it is a poorly thought out circuit. Pulling T12 could only help matters. I guess I am curious as to how this circuit came to be?
-Chris
You could be right. I looked at the basic circuit and ignored the rest. Prohodimez stated he didn't know how to use the matcher and has not confirmed the one he tried to use. I can only assume the diff pairs are not matched at all.
I do know from experience that unmatched diff pairs in this circuit can lead to high DC offsets (300 mV is extremely high) and high distortion (compared to what the circuit is capable of).
Now, looking at T12 and the servo, it is a poorly thought out circuit. Pulling T12 could only help matters. I guess I am curious as to how this circuit came to be?
-Chris
Hi Prohodimez,
Thank you very much.
I guess that shows just how poorly designed some of these magazine projects can be. In my view, disabling and depopulating (to save the parts for something useful) the DC servo section would be a great first step. It shouldn't need a DC servo to begin with. Just use a decent coupling capacitor. All active audio circuits should use a coupling capacitor. I have seen enough DC servos that have failed, causing the power amplifier to eat speakers.
The match on those input pairs matters a lot to how much distortion and the sound quality you get. I would simply buy the correct dual transistors (not from Ebay!) and use them for this project. Alternatively, you pretty much need to use the jig (if it's my design) in order to get them matched tightly enough to replace the dual pairs used in the project. This is a tedious process that I only do to make larger numbers of matched transistor pairs. I buy 50 to 100 pieces of each, then match transistors for a couple days.
-Chris
Thank you very much.
I guess that shows just how poorly designed some of these magazine projects can be. In my view, disabling and depopulating (to save the parts for something useful) the DC servo section would be a great first step. It shouldn't need a DC servo to begin with. Just use a decent coupling capacitor. All active audio circuits should use a coupling capacitor. I have seen enough DC servos that have failed, causing the power amplifier to eat speakers.
The match on those input pairs matters a lot to how much distortion and the sound quality you get. I would simply buy the correct dual transistors (not from Ebay!) and use them for this project. Alternatively, you pretty much need to use the jig (if it's my design) in order to get them matched tightly enough to replace the dual pairs used in the project. This is a tedious process that I only do to make larger numbers of matched transistor pairs. I buy 50 to 100 pieces of each, then match transistors for a couple days.
-Chris
...nothing worked, the offset is 300mv on both channels.....
300mV here is an error, not a mis-match.
Did you measure all the indicated "A, B, C..." points? Please list those numbers. I suspect one or more will be very far off from the example numbers.
Its possible to get a matched Hfe pair but they turn on at different voltages, completely defeating the object of matching.
You really need a curve tracer with Hfe measurements too to match properly.
You really need a curve tracer with Hfe measurements too to match properly.
Hi PRR,
Yes, you are 100 % correct. It is the servo that created the 300 mV DC offset. With the servo removed, the offsets returned to reasonable levels.
However, distortion performance will depend on good matches, as will DC offset. You can mitigate the DC offset issues with a coupling cap. For performance, that is a matching issue.
-Chris
Yes, you are 100 % correct. It is the servo that created the 300 mV DC offset. With the servo removed, the offsets returned to reasonable levels.
However, distortion performance will depend on good matches, as will DC offset. You can mitigate the DC offset issues with a coupling cap. For performance, that is a matching issue.
-Chris
Thanks for links, Prohodimez.
I still believe removing T12 with R2=1k and measuring output offset would be an insightful experiment.
If you wish to implement servo nulling of DC, my recommendation would be to abandon all T12 circuitry and simply connect a 56K resistor from R2, R3 junction to opamp output. Servo bandwidth will be about 1/10 of the R26/C13 corner frequency.
Steve
P.S.
Just saw earlier posts and concur.
Best to all.
I still believe removing T12 with R2=1k and measuring output offset would be an insightful experiment.
If you wish to implement servo nulling of DC, my recommendation would be to abandon all T12 circuitry and simply connect a 56K resistor from R2, R3 junction to opamp output. Servo bandwidth will be about 1/10 of the R26/C13 corner frequency.
Steve
P.S.
Just saw earlier posts and concur.
Best to all.
Last edited:
Hi Nigel,
Well, no actually. I guess you haven't seen the matching jig I designed. You can only match parts when they are maintained at the same temperature. That is how my jig works. It puts the devices in a long tailed pair configuration, and that automatically looks after E-B voltages and beta. Placing a meter across the collectors as a null indicator allows you to obtain extremely tight matches. Given that you are operating the DUT's under the same conditions as they will operate under in use, you are getting the best match possible.
-Chris
Well, no actually. I guess you haven't seen the matching jig I designed. You can only match parts when they are maintained at the same temperature. That is how my jig works. It puts the devices in a long tailed pair configuration, and that automatically looks after E-B voltages and beta. Placing a meter across the collectors as a null indicator allows you to obtain extremely tight matches. Given that you are operating the DUT's under the same conditions as they will operate under in use, you are getting the best match possible.
-Chris
> distortion performance will depend on good matches
Match will never be low-low-low THD null because current balance is not enforced. Instead we take an LED and a bunch of Vbes and 1%(?) resistors. If the "B" voltages are not all the same (and they never will be), then diff-pair currents are not exact-equal and Doug Self et al have shown the rapid rise of THD off-balance.
It's a very pretty drawing but very much of-its-time. And the Jensen 990 is just as old but started on a better path.
Match will never be low-low-low THD null because current balance is not enforced. Instead we take an LED and a bunch of Vbes and 1%(?) resistors. If the "B" voltages are not all the same (and they never will be), then diff-pair currents are not exact-equal and Doug Self et al have shown the rapid rise of THD off-balance.
It's a very pretty drawing but very much of-its-time. And the Jensen 990 is just as old but started on a better path.
- Home
- Design & Build
- Equipment & Tools
- Matching transistors & measuring the results