I hear what you say about curve tracers. Expensive, and even then the information is graphically delivered, and necessarily an averaging 'visual' function will be necessary to choose matches anyway.
Why not construct a very simple tester which measures the two significant parameters; Vbe, and Hfe?
I have done this with great success for bipolar output device testing. The rig cost me about $US20. I use a small 12V gelcel, and first test for Vbe at a prescribed current (I use 50mA) and Vce (I choose 6V). This gives me a measurement for Vbe with a Fluke DMM and I then go through about 100 devices, putting each into little piles with the same Vbe to within a mV.
When I've done all of them, I then pick up a pile of identical Vbes, and measure them in the same rig for Vbe. I do this by measuring the voltage drop across a base stopper, and dividing the mA into the known 50mA flowing through the collector. Within each batch I can then match to within 2% on beta, and better than 0.2% on Vbe. This says nothing about Hfe linearity over the full range of collector current, of course, only concentrating on the bias current of around 50mA, but in truth with largish emitter resistors this is mere bagatelle. All things considered, it's probably as good a way as any AND IT'S CHEAP!
Cheers,
Hugh
Why not construct a very simple tester which measures the two significant parameters; Vbe, and Hfe?
I have done this with great success for bipolar output device testing. The rig cost me about $US20. I use a small 12V gelcel, and first test for Vbe at a prescribed current (I use 50mA) and Vce (I choose 6V). This gives me a measurement for Vbe with a Fluke DMM and I then go through about 100 devices, putting each into little piles with the same Vbe to within a mV.
When I've done all of them, I then pick up a pile of identical Vbes, and measure them in the same rig for Vbe. I do this by measuring the voltage drop across a base stopper, and dividing the mA into the known 50mA flowing through the collector. Within each batch I can then match to within 2% on beta, and better than 0.2% on Vbe. This says nothing about Hfe linearity over the full range of collector current, of course, only concentrating on the bias current of around 50mA, but in truth with largish emitter resistors this is mere bagatelle. All things considered, it's probably as good a way as any AND IT'S CHEAP!
Cheers,
Hugh
jackinnj said:
I have been thinking a little bit, but not too much, about
building curve tracers. I think it wouldn't be much more
work to actually build one that is PC controlled. A decent
sound card would probably do as an interface, giving us
two channeles out and two in. The curve tracer would
basically need some amplifiers and a PSU (or use an external
one). Soundcards usually aren't DC coupled, but at least
mine has a decent LF response and we are doing a curve
trace after all. The added benefit of this would be that part
of the curve tracer moves into your PC in the form of software
(which assumes you can program, though) and you can capture
the data and do all sorts of fun things with it, apart from just
showing the trace. You could automatically compute parameters
from the trace, like Early voltage. A little database (could be
just a file) to save the data of each transistor, and you can
compute statistics. And why not number the transistors when
you measure them. Then you can just search your database
for matched pairs, a certain individual with some particular
data etc.
Well, personally I probably won't bother to take the time to
do it, but for those who really want a curve tracer, think
about it.
Hello Christer,
yes, a computer-based curve tracer is certainly an attractive idea, and makes the resulting data much more accesible. To drive the collector (or drain, or anode) you need an amplifier capable of providing high and low voltages at high current. To reduce dissipation in this amplifier, most curve tracers apply short pulses to the Device Under Test, which also reduces dissipation in the DUT. With a good quality soundcard, there will be enough bits available that gain switching of the amplifier isn't necessary, greatly easing its design (gain switching can cause stability problems), and perhaps an old DC-coupled audio amplifier could be used. Designing and building the sensing amplifiers wouldn't be difficult, so this seems like the way to go.
Nevertheless, Hugh's method is much cheaper...
Jackinnj: Agreed, there have been lots of designs that use an oscilloscope in XY mode to plot the results, and with digital cameras being so cheap, one could easily mount a digital camera in front of an analogue oscilloscope to enable screen capture.
yes, a computer-based curve tracer is certainly an attractive idea, and makes the resulting data much more accesible. To drive the collector (or drain, or anode) you need an amplifier capable of providing high and low voltages at high current. To reduce dissipation in this amplifier, most curve tracers apply short pulses to the Device Under Test, which also reduces dissipation in the DUT. With a good quality soundcard, there will be enough bits available that gain switching of the amplifier isn't necessary, greatly easing its design (gain switching can cause stability problems), and perhaps an old DC-coupled audio amplifier could be used. Designing and building the sensing amplifiers wouldn't be difficult, so this seems like the way to go.
Nevertheless, Hugh's method is much cheaper...
Jackinnj: Agreed, there have been lots of designs that use an oscilloscope in XY mode to plot the results, and with digital cameras being so cheap, one could easily mount a digital camera in front of an analogue oscilloscope to enable screen capture.
here - take a look at this
since I have been doing some consulting in NYC, I haven't had time to finish it up:
http://tech-diy.com/mcutracer_apr03.htm
since I have been doing some consulting in NYC, I haven't had time to finish it up:
http://tech-diy.com/mcutracer_apr03.htm
http://www.maxim-ic.com/appnotes.cfm/appnote_number/253
an app note about how to build a I-V tracer, with software(Qbasic)
/rickard
an app note about how to build a I-V tracer, with software(Qbasic)
/rickard
The right stuff
Thanks for that, it looks just like the right thing. With data in Qbasic it's easy to manipulate, and a 12-bit DAC should reveal DUT non-linearities rather than its own. All it needs is some amplifiers to suit whatever testing is required.
Edit: The Qbasic listing needs its extension changing from txt to bas for Qbasic to recognise it as a program.
rikkitikkitavi said:
Thanks for that, it looks just like the right thing. With data in Qbasic it's easy to manipulate, and a 12-bit DAC should reveal DUT non-linearities rather than its own. All it needs is some amplifiers to suit whatever testing is required.
Edit: The Qbasic listing needs its extension changing from txt to bas for Qbasic to recognise it as a program.
Transistor Matching Q's
Wow -- I had not idea so much dialogue would be stimulated!
Maybe my questions need to be restated a bit:
1. How close a match is "close enough"? I realize this will depend on the topology of the circuit, and that output BJT's may have different requirements than say, small signal BJT's for differential input sections. i.e., 20%, 10%, 1%, etc?
2. What about matching to 3 different points for Hfe? For example, Hfe at Ic points of say 5 amps, 0.5 amps, and 0.05 amps? Then pick the ones that are closest based on least errors.
3. For those of you that purchased 100 or so transistors in order to find the number you wanted, did you calculate variance or standard deviation? Are some makers better at this than others?
4. What other parameters should be measured/matched?
Thanks and regards!
Rob
Wow -- I had not idea so much dialogue would be stimulated!
Maybe my questions need to be restated a bit:
1. How close a match is "close enough"? I realize this will depend on the topology of the circuit, and that output BJT's may have different requirements than say, small signal BJT's for differential input sections. i.e., 20%, 10%, 1%, etc?
2. What about matching to 3 different points for Hfe? For example, Hfe at Ic points of say 5 amps, 0.5 amps, and 0.05 amps? Then pick the ones that are closest based on least errors.
3. For those of you that purchased 100 or so transistors in order to find the number you wanted, did you calculate variance or standard deviation? Are some makers better at this than others?
4. What other parameters should be measured/matched?
Thanks and regards!
Rob
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.