In case I sound like I know what I'm doing: I don't. Can you elaborate a bit for the uneducated?
And FWIW, the Atlas DCA55 (that I use for my initial sorting into "beta" groups) tests single transistors only.
Hi Mike,
You have to use tweezers to handle transistors on your presort. Otherwise you will have less well matched pools to work with. This comes from experience. You also can't breathe on them until after you have the beta reading.
On close beta readings I still get wide differences between collectors, but the close matches tend to measure at the same beta if you can keep them at the same temperature. That is nigh impossible to do with single testing.
Even when I have the same beta readings, they may come in with 10 mV difference or more. The meter beta readings are very approximate. I'll test all the parts within a range of beta, and many different beta readings are actually excellent matches.
Depending on your test current, a 1 mV difference is a wildly good match, and I tend to have most of mine come in there. I do tend to allow them a longer time to equalize thermally, like about 10 minutes or longer.
I line the transistors beside their listed beta, or stick them in foam and use numbered references that tie to beta. I'll test one against the pile of close readings, then on to the next against the rest. It is slow going at the beginning, but it goes a lot quicker as you progress. If I have a balance below 1/2 mV I'll just pair the two and on to the next.
Understand that with matches this close, you must match the degeneration resistors in the circuit (if used). If you don't, the resistances will throw your match off. These matches are silly good.
-Chris
You have to use tweezers to handle transistors on your presort. Otherwise you will have less well matched pools to work with. This comes from experience. You also can't breathe on them until after you have the beta reading.
On close beta readings I still get wide differences between collectors, but the close matches tend to measure at the same beta if you can keep them at the same temperature. That is nigh impossible to do with single testing.
Even when I have the same beta readings, they may come in with 10 mV difference or more. The meter beta readings are very approximate. I'll test all the parts within a range of beta, and many different beta readings are actually excellent matches.
Depending on your test current, a 1 mV difference is a wildly good match, and I tend to have most of mine come in there. I do tend to allow them a longer time to equalize thermally, like about 10 minutes or longer.
I line the transistors beside their listed beta, or stick them in foam and use numbered references that tie to beta. I'll test one against the pile of close readings, then on to the next against the rest. It is slow going at the beginning, but it goes a lot quicker as you progress. If I have a balance below 1/2 mV I'll just pair the two and on to the next.
Understand that with matches this close, you must match the degeneration resistors in the circuit (if used). If you don't, the resistances will throw your match off. These matches are silly good.
-Chris
I sort them out on a sheet of paper that's got a grid drawn on it, with zero in the middle, and squares numbered 1-99 to the left and right. I pick a reference transistor with a beta somewhere in the middle of the pack. (Using a cheap transistor tester) The transistors are tested in the jig, and then placed on the grid according to how far away from the reference they are. After a while, you start to get little piles of transistors that are close to each other, and these candidates can be tested more thoroughly for matching.
The good news is that when you buy qty=10 pieces of 1% resistors, you are guaranteed to find two of them that match each other within (1/10)th of 1%. Assuming you have an ohmmeter and test fixture that can repeatably resolve an 0.1% difference.
And, as it turns out, ten 1% resistors cost less than two 0.1% resistors. Plus, 1% resistors are available in many, many more resistance values.
And, as it turns out, ten 1% resistors cost less than two 0.1% resistors. Plus, 1% resistors are available in many, many more resistance values.
OK, I have a more practical idea—than an insulated glovebox—for regulating the temperature of the DUTs, and bringing them to stability quickly, at a consistent, repeatable temperature. Expounding on Diego Mike's heatsink-die approach... Start with a cylinder of aluminum or copper, and machine a recess, precisely the shape of two TO-92 transistors pressed face-to-face. An oval recess deep enough to fit down over the transistors. (The circuit board would need minor modification, flipping one of the transistors 180 degrees.) If the machining is precise, and the physical tolerance is close, then the two DUTs should stay very close to the temperature of the surrounding metal, without any need for thermal paste. A way is needed to heat the die to a temperature that would be typical in the equipment its intended for. (My experiments suggest that the temperature is not critical as long as it's above 30C or so.) There are thermostat/relay boards available for cheap on ebay. A small power resistor can be affixed to the die as a heater, and the thermostat's thermocouple embedded in the die near the transistors.
Hi Chris,
Mount the driver transistor on the other side and make use of its heat output too. It should give two heat sources, one on each side to heat things more evenly.
I know of lots of equipment that does run around 25°C. What is it about 30°C and above that is so special? In better designs the diff pair may only have 10 ~ 15 V across them. Self heating is next to non-existent in that situation. If you're looking at J-Fets, then expect 7V or less.
-Chris
Mount the driver transistor on the other side and make use of its heat output too. It should give two heat sources, one on each side to heat things more evenly.
I know of lots of equipment that does run around 25°C. What is it about 30°C and above that is so special? In better designs the diff pair may only have 10 ~ 15 V across them. Self heating is next to non-existent in that situation. If you're looking at J-Fets, then expect 7V or less.
-Chris
The cylinder would be vertical, with the transiostor holes in the bottom, and I thought to mount the heater resistor on top of the cylinder, centered perfectly symmetrically. Maybe in a machined groove filled with thermal epoxy.
Nothing special about 30C, but in my observations, the DUTs track really well between about 25C and 80C, and not so well below 25C. I reckon 30C is a good spot along that continuum.
I think this jibes with your observation that they match quicker at higher currents like 3ma. More self-heating and so their temperatures get up into this more linear region above 25C.
Nothing special about 30C, but in my observations, the DUTs track really well between about 25C and 80C, and not so well below 25C. I reckon 30C is a good spot along that continuum.
I think this jibes with your observation that they match quicker at higher currents like 3ma. More self-heating and so their temperatures get up into this more linear region above 25C.
Hi Chris,
Certainly possible. I've been more focused on using something reliable than exploring all possible effects.
I still have my first one thrown together on perf board as I needed a better way ASAP. Later I added the PNP section. Of course, not before I needed it. One thing for sure, there is a huge difference between a tightly matched pair and a pair that isn't very close. So much so that it doesn't make sense to install a pair that isn't matched.
-Chris
Certainly possible. I've been more focused on using something reliable than exploring all possible effects.
I still have my first one thrown together on perf board as I needed a better way ASAP. Later I added the PNP section. Of course, not before I needed it. One thing for sure, there is a huge difference between a tightly matched pair and a pair that isn't very close. So much so that it doesn't make sense to install a pair that isn't matched.
-Chris
Hi Algar,
Those are the current sources. The LEDs are use as the voltage reference and should be surrounded with closed cell foam up the base of the TO-126 devices. The 3mm LEDs then poke up through the mounting hole in the TO-126 device so that it somewhat tracks with temperature. As a current source, they work pretty well as a temperature compensated source / sink.
-Chris
Those are the current sources. The LEDs are use as the voltage reference and should be surrounded with closed cell foam up the base of the TO-126 devices. The 3mm LEDs then poke up through the mounting hole in the TO-126 device so that it somewhat tracks with temperature. As a current source, they work pretty well as a temperature compensated source / sink.
-Chris
Hi Algar,
I used some Japanese devices I had, but you can use any Pro-Electron or JEDEC devices you want as long as the current gain is reasonably high. So any European or North American parts should work.
I normally measure the emitter to collector voltage and make that close to 10 VDC, only because some datasheets use that as the test voltage. A bipolar 10 ~ 12 V supply should be more than okay. In fact, you can use a high voltage as long as the current source transistor and resistors can be run that high, also the DUT's voltage limit.
-Chris
I used some Japanese devices I had, but you can use any Pro-Electron or JEDEC devices you want as long as the current gain is reasonably high. So any European or North American parts should work.
I normally measure the emitter to collector voltage and make that close to 10 VDC, only because some datasheets use that as the test voltage. A bipolar 10 ~ 12 V supply should be more than okay. In fact, you can use a high voltage as long as the current source transistor and resistors can be run that high, also the DUT's voltage limit.
-Chris
The transistors Anatech used in his original Beta Matcher design are no longer available. Per his recommendation, the TO-126 transistors I used for the constant current source are Fairchild KSC3503 (for the NPN) and Fairchild KSA1381 for the PNP. These are both available through Digikey.
I plan to post a complete parts list soon.
I plan to post a complete parts list soon.
I’d like to summarize, up to this point, a few things I think I know about using Anatech’s Beta Matcher rig.
First, maintaining the transistors under test at the same temperature is critical for successful matching. Anatech’s tried and true approach of using some thermal paste between the devices, forcing them in contact with heat shrink, and covering the test rig to minimize the effects of stray air currents, has served him well. It is a very slow process and recently, I and others have been doing thought experiments and trying actual physical techniques to study and improve the transistor matching method. So far, nothing has proven to be better, but the effort continues.
Beyond the original method offered by Anatech, Phloodpants proposed using a fan to convectively cool the devices and bring them to the same temperature. This is quicker and simpler than the thermal grease/heat shrink method. However, it seems to have some issues. The test temperature seems to make a big difference in results. Phloodpants has provided data that shows that the C-C voltages are very sensitive and move around a lot at ambient temperatures. This seems to be much less of a problem at elevated temperatures. This suggests that the use of a fan is problematic because although it may do a good job bringing the two transistors under test to the same temperature, that temperature is going to be at or near ambient. So, while the fan approach may be useful for pre-sorting, it is probably not a good method for final pairing of transistors. I think Phloodpants has come around to agreeing with this (Chris, correct me if I’m wrong).
Another approach involved me building a small thermal cap from aluminum that fits over the two transistors under test and helps keep them at the same temperature. This has worked reasonably well but has not shown itself to be any better or easier than Anatech’s method.
None of these three approaches solve the issue of testing at elevated temperature. It makes sense to pursue somehow elevating the test temperature to some value above ambient. One approach discussed so far is to place the test rig into some sort of temperature controlled oven – probably too difficult for any but the most dedicated. Another method is to deliver heat to the transistors while mounted on the board. I’ve worked toward this a bit with my aluminum “thermal cap”. (More accurately, I’ve just retained the small amount of heat generated through the device self heating.)
There is a third approach which I don’t think has been discussed yet: moving the DUT’s off the board entirely, and making the temperature control to be easier. More on that later.
Back to temperature: based on some laboratory analytical work I’ve been involved with, I’ve learned that for stable instrument temperature control, you want to be significantly above ambient temperature to allow the thermostatic controller to do its job. I suggest that the final testing (matching) be done at between 90 and 100 F, which has the added benefit of simulating the amplifier internal temperature reasonably well.
So, summarizing all this, the two critical parameters in these tests are: first, to thermally couple the devices; and second, to perform the tests at elevated temperatures. Anatech has made us understand that the thermal coupling is the most important aspect of this testing. His approach of using some thermal paste between the devices, forcing them in contact with heat shrink, and covering the test rig to minimize the effects of stray air currents has served him well. By the way, I have taken his advice and adopted 300 mA as the test current in all my testing and experiments. I think his method might work even better if the test temperature were above ambient. Maybe it would only be quicker, but that’s worth pursuing. Everyone who has done this matching agrees it is a very tedious and time consuming process.
Let’s consider how closely we have to maintain our final test temperature. For the final matching, it probably doesn’t matter whether it’s 94 or 98F, as long as it remains steady (in order for the C-C voltage to stabilize). All along we have assumed that two well-matched transistors will track each other’s Hfe throughout the moderate range of temperatures encountered inside the amp, so whether they are matched at one or the other of those temperatures shouldn’t really matter. However, when pre-sorting the transistors by Hfe, the ambient seems to matter a great deal. This is why I think that Phloodpant’s fan approach could work very well, but only if the pre-testing is done indoor where the ambient temperature is steady.
One last thing – self heating of the devices under test: some very simple calculations I’ve done on transistor heating show about 3 Fahrenheit degrees heating of the transistor under our test conditions (300 mA tail current). Experiments with a new heated thermal cap (more on that later) indicate about 2.6 degrees heating, which agrees fairly well with the calculated value and supports the notion that the self heating is relatively small. Please note that these temperature data are for the 2N5111’s I’m working with. Your mileage may differ.
First, maintaining the transistors under test at the same temperature is critical for successful matching. Anatech’s tried and true approach of using some thermal paste between the devices, forcing them in contact with heat shrink, and covering the test rig to minimize the effects of stray air currents, has served him well. It is a very slow process and recently, I and others have been doing thought experiments and trying actual physical techniques to study and improve the transistor matching method. So far, nothing has proven to be better, but the effort continues.
Beyond the original method offered by Anatech, Phloodpants proposed using a fan to convectively cool the devices and bring them to the same temperature. This is quicker and simpler than the thermal grease/heat shrink method. However, it seems to have some issues. The test temperature seems to make a big difference in results. Phloodpants has provided data that shows that the C-C voltages are very sensitive and move around a lot at ambient temperatures. This seems to be much less of a problem at elevated temperatures. This suggests that the use of a fan is problematic because although it may do a good job bringing the two transistors under test to the same temperature, that temperature is going to be at or near ambient. So, while the fan approach may be useful for pre-sorting, it is probably not a good method for final pairing of transistors. I think Phloodpants has come around to agreeing with this (Chris, correct me if I’m wrong).
Another approach involved me building a small thermal cap from aluminum that fits over the two transistors under test and helps keep them at the same temperature. This has worked reasonably well but has not shown itself to be any better or easier than Anatech’s method.
None of these three approaches solve the issue of testing at elevated temperature. It makes sense to pursue somehow elevating the test temperature to some value above ambient. One approach discussed so far is to place the test rig into some sort of temperature controlled oven – probably too difficult for any but the most dedicated. Another method is to deliver heat to the transistors while mounted on the board. I’ve worked toward this a bit with my aluminum “thermal cap”. (More accurately, I’ve just retained the small amount of heat generated through the device self heating.)
There is a third approach which I don’t think has been discussed yet: moving the DUT’s off the board entirely, and making the temperature control to be easier. More on that later.
Back to temperature: based on some laboratory analytical work I’ve been involved with, I’ve learned that for stable instrument temperature control, you want to be significantly above ambient temperature to allow the thermostatic controller to do its job. I suggest that the final testing (matching) be done at between 90 and 100 F, which has the added benefit of simulating the amplifier internal temperature reasonably well.
So, summarizing all this, the two critical parameters in these tests are: first, to thermally couple the devices; and second, to perform the tests at elevated temperatures. Anatech has made us understand that the thermal coupling is the most important aspect of this testing. His approach of using some thermal paste between the devices, forcing them in contact with heat shrink, and covering the test rig to minimize the effects of stray air currents has served him well. By the way, I have taken his advice and adopted 300 mA as the test current in all my testing and experiments. I think his method might work even better if the test temperature were above ambient. Maybe it would only be quicker, but that’s worth pursuing. Everyone who has done this matching agrees it is a very tedious and time consuming process.
Let’s consider how closely we have to maintain our final test temperature. For the final matching, it probably doesn’t matter whether it’s 94 or 98F, as long as it remains steady (in order for the C-C voltage to stabilize). All along we have assumed that two well-matched transistors will track each other’s Hfe throughout the moderate range of temperatures encountered inside the amp, so whether they are matched at one or the other of those temperatures shouldn’t really matter. However, when pre-sorting the transistors by Hfe, the ambient seems to matter a great deal. This is why I think that Phloodpant’s fan approach could work very well, but only if the pre-testing is done indoor where the ambient temperature is steady.
One last thing – self heating of the devices under test: some very simple calculations I’ve done on transistor heating show about 3 Fahrenheit degrees heating of the transistor under our test conditions (300 mA tail current). Experiments with a new heated thermal cap (more on that later) indicate about 2.6 degrees heating, which agrees fairly well with the calculated value and supports the notion that the self heating is relatively small. Please note that these temperature data are for the 2N5111’s I’m working with. Your mileage may differ.
I decided to try and separate the transistors under test from the test rig. I attached six small gauge wires about 8 inches long to the male and female sides of two IC sockets. This serves as a sort of ‘extension cord’ for the transistors. I can move the transistor pair in their new socket away from the board and experiment with different ways to keep them at the same temperature, and also control the testing temperature.
I also added heating to my thermal cap by mounting two small power resistors (rectangular ceramic) to the top. They are tack glued with thermal paste between them and the aluminum block. So now, I have a heated thermal cap that maintains the transistor pair at the same temperature, and allows me to control that temperature.
With the foam insulating cover in place, I can achieve a block temperature of 96 F with a current of 125mA. (The resistors are 2-watt, 10R connected in series, with an external resistor dropping some of the 5 vdc power source.) This works out to about 630 mW. I’m measuring the thermal cap temperature with a Fluke handheld IR gun. I want to install a small thermocouple in the thermal cap to improve the temperature measurement accuracy. I am only beginning to experiment with this new version, but if it works, it’ll be a whole lot simpler than building an oven.
I also added heating to my thermal cap by mounting two small power resistors (rectangular ceramic) to the top. They are tack glued with thermal paste between them and the aluminum block. So now, I have a heated thermal cap that maintains the transistor pair at the same temperature, and allows me to control that temperature.
With the foam insulating cover in place, I can achieve a block temperature of 96 F with a current of 125mA. (The resistors are 2-watt, 10R connected in series, with an external resistor dropping some of the 5 vdc power source.) This works out to about 630 mW. I’m measuring the thermal cap temperature with a Fluke handheld IR gun. I want to install a small thermocouple in the thermal cap to improve the temperature measurement accuracy. I am only beginning to experiment with this new version, but if it works, it’ll be a whole lot simpler than building an oven.
