Hi Mace Horny,
Measure your emitter - base voltages to see if they are in reverse breakdown or not. This condition damages the transistors and will increase noise. It can also change beta and cause changes to the other characteristics too.
Sometimes it helps to draw pictures of the parts. Do one for each and pinout errors will be a lot easier to avoid. People tend to get stuck on simple mistakes, we all do it.
-Chris
Measure your emitter - base voltages to see if they are in reverse breakdown or not. This condition damages the transistors and will increase noise. It can also change beta and cause changes to the other characteristics too.
Sometimes it helps to draw pictures of the parts. Do one for each and pinout errors will be a lot easier to avoid. People tend to get stuck on simple mistakes, we all do it.
-Chris
Thanks for the reply.
Could you give me a short explanation how I measure them ? Because I already measured the Vbe of all transistors when the power supply is on with my DMM. One pole on the Emitter and one on the Base of each transistor.
And what does reverse Brakedown mean ? The faulty Transistors have a Vbe of 0,53 V and its decreasing as they ge hot ..
Last edited:
Hi Mace Horny,
If a transistor is reverse biased from emitter to base, you will typically read voltage drops on the order of 6.5 to 7.5 volts. A forward biased junction will measure anywhere from 0.55 to 0.8 volts. Normally at room temperature you might expect 0.55 to 0.6 volts. This voltage drops as the junction temperature increases. Signal transistors have a very small structure, so the thermal time constant is quite short. That means the part can be damaged before you know it.
The tester places a pair of transistors in a differential pair configuration. There are (in my case) a pair of 100 R resistors with 0.1% tolerance (but you can use 1% parts for less accuracy, it will still work) from collectors to the supply voltage (around 11 VDC). I used a pair of 10K resistors (again 0.1%) from the bases to the common ground. The two transistors "sit" on a constant current source with adjustable current. Anywhere from 1mA to 6 mA is normal. Make it the same as the circuit uses that you are grouping the transistors for. These days I run the CCS around 3 mA in general.
Place a foam "hat' over the transistors, then a box over that. Your meter will be connected across the two collectors. A low to 0.00mV reading indicates a good match. Higher readings mean these are not a good match, depending on what the tail current is set for. For example, a 50 mV difference isn't a good match. If you are careful, you can use the hFE (beta) function in a meter to pre-sort the transistors. Try to use tweezers or pliers to keep the parts near room temperature. If you are getting a large number of poor matches with the same hFE readings, it means your presort wasn't very good. One thing you will see is how bad matches made with the hFE function of a meter can be. Your matches with Vb-e will also show to be very poor. If you measure the Vbe of a mounted pair of transistors after the settle in, you will see they match as well.
When you install pairs matched like this, you may have to match the degeneration (emitter) resistors to preserve the match. You should also put some thermal compound between the two transistors and hold them together with some heat shrink tubing.
Try to keep in mind that with inexpensive meters, the mV function may be wildly inaccurate, as well as low DC volts scales. The last digit is certainly there for show, you can't trust it. The best meters to buy these days are Keysight / Agilent / HP, or Fluke. Those hold their calibration, have very good high frequency response and low voltage ranges you can trust. WHen making a measurement even with a good meter, the last digit is only accurate to a few counts either way. With the cheaper meters, the second last digit may be off by a couple counts. That's why I say that the last digit is there for show.
-Chris
If a transistor is reverse biased from emitter to base, you will typically read voltage drops on the order of 6.5 to 7.5 volts. A forward biased junction will measure anywhere from 0.55 to 0.8 volts. Normally at room temperature you might expect 0.55 to 0.6 volts. This voltage drops as the junction temperature increases. Signal transistors have a very small structure, so the thermal time constant is quite short. That means the part can be damaged before you know it.
The tester places a pair of transistors in a differential pair configuration. There are (in my case) a pair of 100 R resistors with 0.1% tolerance (but you can use 1% parts for less accuracy, it will still work) from collectors to the supply voltage (around 11 VDC). I used a pair of 10K resistors (again 0.1%) from the bases to the common ground. The two transistors "sit" on a constant current source with adjustable current. Anywhere from 1mA to 6 mA is normal. Make it the same as the circuit uses that you are grouping the transistors for. These days I run the CCS around 3 mA in general.
Place a foam "hat' over the transistors, then a box over that. Your meter will be connected across the two collectors. A low to 0.00mV reading indicates a good match. Higher readings mean these are not a good match, depending on what the tail current is set for. For example, a 50 mV difference isn't a good match. If you are careful, you can use the hFE (beta) function in a meter to pre-sort the transistors. Try to use tweezers or pliers to keep the parts near room temperature. If you are getting a large number of poor matches with the same hFE readings, it means your presort wasn't very good. One thing you will see is how bad matches made with the hFE function of a meter can be. Your matches with Vb-e will also show to be very poor. If you measure the Vbe of a mounted pair of transistors after the settle in, you will see they match as well.
When you install pairs matched like this, you may have to match the degeneration (emitter) resistors to preserve the match. You should also put some thermal compound between the two transistors and hold them together with some heat shrink tubing.
Try to keep in mind that with inexpensive meters, the mV function may be wildly inaccurate, as well as low DC volts scales. The last digit is certainly there for show, you can't trust it. The best meters to buy these days are Keysight / Agilent / HP, or Fluke. Those hold their calibration, have very good high frequency response and low voltage ranges you can trust. WHen making a measurement even with a good meter, the last digit is only accurate to a few counts either way. With the cheaper meters, the second last digit may be off by a couple counts. That's why I say that the last digit is there for show.
-Chris
Last edited:
With the issues of Mace Horny I was also wondering a bit if all the different pinouts may have been the issue but as he checked them already.
If it works with the 2n5551 but not the BC550. Is there a chance they are not ok?
Hi,
would you mind sharing a schematic of this. I am currently using one to measure Hfe individually but since it starts changing immediately with temperature I am not sure about the results.
Also, may I would like to throw in my earlier question again: How important is absolut Hfe?
I have obtained Philips NOS 2n5401 and 2n5551. Couldn't find 2N5401 from Fairchild or Onsemi. The old Philips are quite a bit lower in Hfe compared to some newer equivalents from Diotec Semi or Taiwan Semi sold here (30-50%).
The Philips are very nice by appearance and despite quite a variance in Hfe I could match (<0.5% @2mA) good pairs but I am still wondering if high Hfe is important?
Thanks,
Dirk
If it works with the 2n5551 but not the BC550. Is there a chance they are not ok?
Hi,
would you mind sharing a schematic of this. I am currently using one to measure Hfe individually but since it starts changing immediately with temperature I am not sure about the results.
Also, may I would like to throw in my earlier question again: How important is absolut Hfe?
I have obtained Philips NOS 2n5401 and 2n5551. Couldn't find 2N5401 from Fairchild or Onsemi. The old Philips are quite a bit lower in Hfe compared to some newer equivalents from Diotec Semi or Taiwan Semi sold here (30-50%).
The Philips are very nice by appearance and despite quite a variance in Hfe I could match (<0.5% @2mA) good pairs but I am still wondering if high Hfe is important?
Thanks,
Dirk
Hi Dirk,
There is a thread now on that matcher. I'll look for a link. Here it is:
http://www.diyaudio.com/forums/equipment-tools/307138-matching-transistors-measuring-results.html
The absolute gain isn't typically important, but there range the transistor is in can be important. Transistors are often ranked by the gain range they are in and this mark should be on the body of the part. Normally a letter code, in North America these parts are often not graded, or the entire part number is different for gain ranges.
The higher the beta or hFE is, the more the transistor operates like an ideal transistor. However, different circuit locations may demand a part with lower gain. The best thing you can do is to install a new transistor that was similar to the gain the original part had. Differential pairs (long tailed pairs) generally work better with high gain. Be on guard for parts that show gains well below or above the average gain. If a part has much higher gain, it is probably defective in some way. Best to avoid these parts, and tossing them out is the best way to ensure they don't end up in a project or repair.
That's about it Dirk, Chris
There is a thread now on that matcher. I'll look for a link. Here it is:
http://www.diyaudio.com/forums/equipment-tools/307138-matching-transistors-measuring-results.html
The absolute gain isn't typically important, but there range the transistor is in can be important. Transistors are often ranked by the gain range they are in and this mark should be on the body of the part. Normally a letter code, in North America these parts are often not graded, or the entire part number is different for gain ranges.
The higher the beta or hFE is, the more the transistor operates like an ideal transistor. However, different circuit locations may demand a part with lower gain. The best thing you can do is to install a new transistor that was similar to the gain the original part had. Differential pairs (long tailed pairs) generally work better with high gain. Be on guard for parts that show gains well below or above the average gain. If a part has much higher gain, it is probably defective in some way. Best to avoid these parts, and tossing them out is the best way to ensure they don't end up in a project or repair.
That's about it Dirk, Chris
Hi Chris,
thanks a lot for the details provided and linking this thread.
So far I have tried a circuit provided to a German board by Michael Bittner matching Hfe and for fun I have also tried relative Vbe matchingas described by Ian Fritz. Seems to be quite hard to get both matched but probably way beyond what I need.
Still was fun and glad to get some more advice provided by you.
Thanks again,
Dirk
thanks a lot for the details provided and linking this thread.
So far I have tried a circuit provided to a German board by Michael Bittner matching Hfe and for fun I have also tried relative Vbe matchingas described by Ian Fritz. Seems to be quite hard to get both matched but probably way beyond what I need.
Still was fun and glad to get some more advice provided by you.Thanks again,
Dirk
Thanks for the information! I have used 9.2 kOhm instead of the 100 kOhm to increase current to a value closer to the final working current. Not sure if this changes things?
However, I have by now decided to stick to the Hfe values I have measured with attached circuit.
(Q7 and Q1 are the DUTs and one measures voltage (Vmeas) across the 10k resistors. R1 & R7 set the Iq. Hfe= 10k x Iq / Vmeas; originally provided by M.Bittner to the German AAA forum but not sure if this is the very original source)
However, I have by now decided to stick to the Hfe values I have measured with attached circuit.
(Q7 and Q1 are the DUTs and one measures voltage (Vmeas) across the 10k resistors. R1 & R7 set the Iq. Hfe= 10k x Iq / Vmeas; originally provided by M.Bittner to the German AAA forum but not sure if this is the very original source)
Attachments
Last edited:
Hello,
I also build at the moment this amp and will use the following capacitors
c14 10pf mica - will use 10pf Keramik
C2 und C7 100pf mica - will use 100pf Keramik
C3 und C4 330pf mica - will use 330pf foil capacitors
Can I use this capacitors, or is this a problem?
A question regarding the transistors
Toshiba 2SC5200/2SA1943 this transistors are shown in the pictures, I use this also.
But in the part list are other transistors described. Have both the same values?
Pls can you help me.
Thanks.
Regards
I also build at the moment this amp and will use the following capacitors
c14 10pf mica - will use 10pf Keramik
C2 und C7 100pf mica - will use 100pf Keramik
C3 und C4 330pf mica - will use 330pf foil capacitors
Can I use this capacitors, or is this a problem?
A question regarding the transistors
Toshiba 2SC5200/2SA1943 this transistors are shown in the pictures, I use this also.
But in the part list are other transistors described. Have both the same values?
Pls can you help me.
Thanks.
Regards
Hi dsch,
The compensation capacitor value changes with the type of output transistor used. This was a long time ago for me, but the best way to figure this out is to do a frequency sweep at a 1 watt level and watch for peaks in the response. If it's pretty flat, then you are okay. If you want to get the maximum bandwidth, vary the compensation capacitor until you get peaking, then bring the value back until the peak is gone. Then heat the amp up with music and try the sweep again to see if the peak came back and adjust as needed if it did. I used MJW0302 and MJW0281 and I think 10 pF (could have been lower) and everything was very stable.
So, you need an oscilloscope and a sweep generator. Put the ramp into the X input of your 'scope and the amplifier into the Y input, put the scope in X-Y mode. That will give you a sweep of frequency amplitude vs time. If you have a spectrum analyser, use that.
-Chris
The compensation capacitor value changes with the type of output transistor used. This was a long time ago for me, but the best way to figure this out is to do a frequency sweep at a 1 watt level and watch for peaks in the response. If it's pretty flat, then you are okay. If you want to get the maximum bandwidth, vary the compensation capacitor until you get peaking, then bring the value back until the peak is gone. Then heat the amp up with music and try the sweep again to see if the peak came back and adjust as needed if it did. I used MJW0302 and MJW0281 and I think 10 pF (could have been lower) and everything was very stable.
So, you need an oscilloscope and a sweep generator. Put the ramp into the X input of your 'scope and the amplifier into the Y input, put the scope in X-Y mode. That will give you a sweep of frequency amplitude vs time. If you have a spectrum analyser, use that.
-Chris
Hi Dirk83,
I haven't seen that circuit before. Interesting. I do now that in order to get matched beta, you need to put the two transistors in a diff pair configuration. What you have there wouldn't do that.
Why not just build the circuit others have used successfully? It is simple and very direct in the way it operates and all you need is some kind of null detector. The meter used will measure in mV and the only critical thing is an accurate zero. Any voltage measured is a degree of mismatch, and that is only marginally important. I did match my transistors using the jig, and it turned out great. My DC offsets are below 5 mV without any adjustments. That's pretty good. The distortion was also very low but I can't remember what the numbers were. I know I was very happy with the performance. The diff pair determines what distortion performance you have, so it would seem a valuable thing to set up.
The collector resistors don't have to be 10 ohm exactly. What is important is that the two resistors are very close to each other in resistance. You're making a radiometric measurement. You could probably use a centre tune meter from an old receiver (but short it when changing transistors!). For me, ordering the resistors as 0.1% made life simpler, so that is why I used them. I can also then calculate the hFE pretty accurately as the base resistors were also 0.1% tolerance parts. That represents the most expensive part of the matcher unless you buy the PCB from someone.
-Chris
I haven't seen that circuit before. Interesting. I do now that in order to get matched beta, you need to put the two transistors in a diff pair configuration. What you have there wouldn't do that.
Why not just build the circuit others have used successfully? It is simple and very direct in the way it operates and all you need is some kind of null detector. The meter used will measure in mV and the only critical thing is an accurate zero. Any voltage measured is a degree of mismatch, and that is only marginally important. I did match my transistors using the jig, and it turned out great. My DC offsets are below 5 mV without any adjustments. That's pretty good. The distortion was also very low but I can't remember what the numbers were. I know I was very happy with the performance. The diff pair determines what distortion performance you have, so it would seem a valuable thing to set up.
The collector resistors don't have to be 10 ohm exactly. What is important is that the two resistors are very close to each other in resistance. You're making a radiometric measurement. You could probably use a centre tune meter from an old receiver (but short it when changing transistors!). For me, ordering the resistors as 0.1% made life simpler, so that is why I used them. I can also then calculate the hFE pretty accurately as the base resistors were also 0.1% tolerance parts. That represents the most expensive part of the matcher unless you buy the PCB from someone.
-Chris
Hi Chris,
thanks for your opinion. Indeed, I was looking for a simple circuit for a while but the web is full of different circuits and just confusing.
From your posting I read that you think direct matching is more favourable!? In particular after seeing the temperature related shifts in beta I would agree (if I read your post correctly) and would have preferred this. That's why I was looking for an alternative like the one from Ian Fritz (measuring Vbe)that I have build. Saying "Vbe" I also got confused since some people seem to prefer it over the Hfe matching but I am way way too unexperienced to judge on my own.
I will for sure consider to build the circuit you have linked. Thanks again for your helpful comments.The only thing holding me back right now is the time I have spend already with measuring (~500 transistors).
Best,
Dirk
Edit: Are the PCBs available somewhere? As mentioned, I am not that experienced and I wouldn't be sure that I can correctly build your designed/posted circuit correctly. Thanks
thanks for your opinion. Indeed, I was looking for a simple circuit for a while but the web is full of different circuits and just confusing.
The one shown in my post has been used by quite a few people in the German "Symasym Community" and has even been references in the German SymAsym Wiki, which is not online anymore. This are the reasons why I decided for this circuit.
The circuit is no direct matching of two transistors but one measures beta individually. The two spots allow measuring NPN and PNP.
From your posting I read that you think direct matching is more favourable!? In particular after seeing the temperature related shifts in beta I would agree (if I read your post correctly) and would have preferred this. That's why I was looking for an alternative like the one from Ian Fritz (measuring Vbe)that I have build. Saying "Vbe" I also got confused since some people seem to prefer it over the Hfe matching but I am way way too unexperienced to judge on my own.
I will for sure consider to build the circuit you have linked. Thanks again for your helpful comments.The only thing holding me back right now is the time I have spend already with measuring (~500 transistors).
Best,
Dirk
Edit: Are the PCBs available somewhere? As mentioned, I am not that experienced and I wouldn't be sure that I can correctly build your designed/posted circuit correctly. Thanks
Last edited:
I measure hFE using DMM to batch near devices ready for next stage.
I measure hFE in a jig (thermally clamped together) at the current I intend to use in the operating circuit.
I then take a group of these almost identical hFE devices and using one as REF and all the others as DUT try to find a pair that have low diferential collector current as the Vbe is swung through a range that may be used during operation.
I find that this final test completely refutes the claim by other Members that selection by hFE automatically finds matched pairs because Vbe depends on hFE.
I claim you need to test/match for BOTH.
I measure hFE in a jig (thermally clamped together) at the current I intend to use in the operating circuit.
I then take a group of these almost identical hFE devices and using one as REF and all the others as DUT try to find a pair that have low diferential collector current as the Vbe is swung through a range that may be used during operation.
I find that this final test completely refutes the claim by other Members that selection by hFE automatically finds matched pairs because Vbe depends on hFE.
I claim you need to test/match for BOTH.
Like this:
To add a photo, files or non standard files.
First click "go advanced" in the box below the "quick reply" message box. Doesn't matter if you decide half way through a message to do that, it carries it forward.
Then click "Manage attachments". Maximise the new Window so that you can see all the text.
Click browse in the first box at the top and find your picture. Repeat for any more pictures.
Click upload... a message appears "uploading"
When complete the files will show as being attached. Now click the small text that says "close this window"
The pictures should now be attached and when you submit your post they will appear.
Make sure your pics aren't too big, a couple of 100k is plenty, and many members object when they are massive and it alters the margins
It tells you in the attachments window what max sizes are allowed.
If you want to attach a file that has a non standard format for example excel, circuit simulation etc then try putting the files in a zipped folder and attaching that.