Working on a 1 amp power transistor matcher.
It works around a transistor stage with base resistor and collector resistor with differential amplifiers measuring voltage across resistors.
From this it sends the plot to the pc via USB to display up to 5 different transistors at the same time.
A lot of buy in matchers are battery powered and often just use a couple of milliamps to measure the transistor.
This isnt a real world test as power transistors often work at amps rather than milli amps.
Most of the circuit is about generating a voltage ramp from a microcontroller then reading base and collector currents through differential amplifiers.
The microcontroller end is made as simple as possible and the heavy coding is done at the pc end.
You can zoom in on the 5 plots and run cursors along the plots to get Ic, Ib, Vc, Vb and Hfe at any point.
It works around a transistor stage with base resistor and collector resistor with differential amplifiers measuring voltage across resistors.
From this it sends the plot to the pc via USB to display up to 5 different transistors at the same time.
A lot of buy in matchers are battery powered and often just use a couple of milliamps to measure the transistor.
This isnt a real world test as power transistors often work at amps rather than milli amps.
Most of the circuit is about generating a voltage ramp from a microcontroller then reading base and collector currents through differential amplifiers.
The microcontroller end is made as simple as possible and the heavy coding is done at the pc end.
You can zoom in on the 5 plots and run cursors along the plots to get Ic, Ib, Vc, Vb and Hfe at any point.
Nige, with all due respect, a pic of the PCB is the least interesting info on this project! Specs, how do plots look, etc. is much more of interest!
Jan
Jan
I havent got pcb in yet !
Once i build one up I will add some plots.
Its just a simple curve anyway. It can display up to 5 at once.
If it isn't pulsed, it'll be useless. Any test current that is high enough to heat the transistor will just heat it, and you will be measuring stuff like thermal resistance and torque of the screw holding it to the heatsink instead of actual transistor properties. Likewise if the temperature of the transistor isn't known, it's useless. If you hold it in your hand five seconds longer while installing it, it'll be hotter, and its Vbe will be different.
Okay, okay, that was negative criticism, here's the positive one:
How I did pulsed transistor measurements. I'm not going to describe the exact setup I used, rather what it should have been to deliver the best results. But I used a board designed for something else.
Setup:
The idea is to keep the transistor idling at a low Vce (5V) and low Ie (100µA). It is held by a clamp back-to-back on top of another TO220 transistor which acts as temperature probe, with a sheet of keratherm red between the two. This ensures best thermal coupling and fastest thermal time constant between the DUT and the temperature probe transistor. It is possible to use an "official" temperature sensor chip, but these don't have a metal tab, so the thermal time constant is much slower, and coupling is worse.
This is placed in a thermally insulated enclosure (ie, a rolled-up kleenex held by tape). Of course one could devise a better looking apparatus.
Process:
First the DUT is self-heated while tracking its temperature with the sensor. This is simply done by applying high Vce and Ic. Once it is near temperature, dissipated power is reduced, and the software approaches the target temperature more slowly. Then it keeps it at the target temperature for a few tens of seconds, to ensure everything is isothermal, including the plastic package, etc.
Then, power is removed, and pulsed measurements are done as it cools down. This is because it's not possible to keep it isothermal while heating, but it is possible to do so while cooling. Without any power dissipated in the transistor, there is no thermal gradient between it and the temperature probe, so it tracks very accurately. Dissipation from pulsed measurements is negligible, and the insulation around the transistor ensures slow uniform cooling over a few minutes.
Implementation:
Base (or gate) is grounded via a resistor which also serves to measure base current.
Collector is driven by a cascode transistor which sets idle Vce, then a MOSFET to switch to measurement Vce.
Emitter is connected to a voltage source via a switched resistor. In my case, the voltage source is a USB-programmable bench power supply, but you could use a microcontroller DAC driving an opamp and a power transistor. It doesn't matter. Several resistors are mounted, switched by MOSFETs, to offer a wide range of currents without having to keep the voltage across it super accurate. Basically the voltage ranges from 5V to 30V, then it switches to another resistor.
A low current source (10~100µA) keeps the transistor on so its Vbe can be monitored at all times.
If a BJT is tested, and if we assume we know its Vbe versus temperature, then it is not necessary to use the temperature probe. MOSFETs have widely varying Vgs tempcos, so the temperature probe is mandatory.
At regular intervals, the micro does a pulsed measurement. First, it applies Vce. Then, it applies voltage on the emitter resistor. It waits 1µs, then acquires Vbe, Ib and temperature several times using the DMA burst sampling of STM32 to make sure the samples are as close as possible in time. It is not necessary to acquire Ie because the emitter resistor value is known, and the voltage across it is known. However, acquiring Ic may be useful to make sure the transistor capacitances have finished charging and it has settled before the acquisition.
Data is accumulated as the transistor cools down, and then it is plotted. So we get all the usual curves, at all temperatures.
The difficulty here is settling time of Vce and especially Ie. If Vce settles slowly, it's not too bad, because it is applied before the Ie pulse, when Ie is only 10-100µA. So if it's very slow and takes 5µs to settle, so be it. But the Ie pulse must have fast rise time, and settle to a flat top in under 1µs, because then the transistor is heating very quickly.
I used resistors and switched MOSFETs because I didn't want to bother with designing a fast settling time current source for the emitter. If you want to test both small signal transistors and big ones, then you need a current source with fast settling across a very wide range of currents. At low current this is not compatible with a big MOSFET pass transistor with large capacitance. So several current sources would be needed, with a way to switch between them to get rid of the capacitance. It's just a mess. It was much simpler to use a voltage source to charge a capacitor to the requires voltage, then switch resistors with MOSFETs. Since the MOSFETs doesn't have to work in linear mode, fast low gate charge low capacitance trench FETs can be used.
To keep voltages steady during the pulse, I just used capacitors. This slows down the change in voltage between measurements, but it's not too much of a problem.
https://www.diyaudio.com/community/threads/assorted-pulsed-transistor-measurements.373764/
Okay, okay, that was negative criticism, here's the positive one:
How I did pulsed transistor measurements. I'm not going to describe the exact setup I used, rather what it should have been to deliver the best results. But I used a board designed for something else.
Setup:
The idea is to keep the transistor idling at a low Vce (5V) and low Ie (100µA). It is held by a clamp back-to-back on top of another TO220 transistor which acts as temperature probe, with a sheet of keratherm red between the two. This ensures best thermal coupling and fastest thermal time constant between the DUT and the temperature probe transistor. It is possible to use an "official" temperature sensor chip, but these don't have a metal tab, so the thermal time constant is much slower, and coupling is worse.
This is placed in a thermally insulated enclosure (ie, a rolled-up kleenex held by tape). Of course one could devise a better looking apparatus.
Process:
First the DUT is self-heated while tracking its temperature with the sensor. This is simply done by applying high Vce and Ic. Once it is near temperature, dissipated power is reduced, and the software approaches the target temperature more slowly. Then it keeps it at the target temperature for a few tens of seconds, to ensure everything is isothermal, including the plastic package, etc.
Then, power is removed, and pulsed measurements are done as it cools down. This is because it's not possible to keep it isothermal while heating, but it is possible to do so while cooling. Without any power dissipated in the transistor, there is no thermal gradient between it and the temperature probe, so it tracks very accurately. Dissipation from pulsed measurements is negligible, and the insulation around the transistor ensures slow uniform cooling over a few minutes.
Implementation:
Base (or gate) is grounded via a resistor which also serves to measure base current.
Collector is driven by a cascode transistor which sets idle Vce, then a MOSFET to switch to measurement Vce.
Emitter is connected to a voltage source via a switched resistor. In my case, the voltage source is a USB-programmable bench power supply, but you could use a microcontroller DAC driving an opamp and a power transistor. It doesn't matter. Several resistors are mounted, switched by MOSFETs, to offer a wide range of currents without having to keep the voltage across it super accurate. Basically the voltage ranges from 5V to 30V, then it switches to another resistor.
A low current source (10~100µA) keeps the transistor on so its Vbe can be monitored at all times.
If a BJT is tested, and if we assume we know its Vbe versus temperature, then it is not necessary to use the temperature probe. MOSFETs have widely varying Vgs tempcos, so the temperature probe is mandatory.
At regular intervals, the micro does a pulsed measurement. First, it applies Vce. Then, it applies voltage on the emitter resistor. It waits 1µs, then acquires Vbe, Ib and temperature several times using the DMA burst sampling of STM32 to make sure the samples are as close as possible in time. It is not necessary to acquire Ie because the emitter resistor value is known, and the voltage across it is known. However, acquiring Ic may be useful to make sure the transistor capacitances have finished charging and it has settled before the acquisition.
Data is accumulated as the transistor cools down, and then it is plotted. So we get all the usual curves, at all temperatures.
The difficulty here is settling time of Vce and especially Ie. If Vce settles slowly, it's not too bad, because it is applied before the Ie pulse, when Ie is only 10-100µA. So if it's very slow and takes 5µs to settle, so be it. But the Ie pulse must have fast rise time, and settle to a flat top in under 1µs, because then the transistor is heating very quickly.
I used resistors and switched MOSFETs because I didn't want to bother with designing a fast settling time current source for the emitter. If you want to test both small signal transistors and big ones, then you need a current source with fast settling across a very wide range of currents. At low current this is not compatible with a big MOSFET pass transistor with large capacitance. So several current sources would be needed, with a way to switch between them to get rid of the capacitance. It's just a mess. It was much simpler to use a voltage source to charge a capacitor to the requires voltage, then switch resistors with MOSFETs. Since the MOSFETs doesn't have to work in linear mode, fast low gate charge low capacitance trench FETs can be used.
To keep voltages steady during the pulse, I just used capacitors. This slows down the change in voltage between measurements, but it's not too much of a problem.
https://www.diyaudio.com/community/threads/assorted-pulsed-transistor-measurements.373764/