For a while now i have been thinking about building an automated Transistor matching device. So i thought i would bounce a few ideas off people and see what you think would be the easiest/slikest setup.
My first thoughts were to use a MCU like a basic stamp etc for the heart of the thing.
I have some 2' long heatsinks that i can outfit with a couple of poer terminals and a ribbon cable that should fit quite a few devices. For TO-3 devices i could outfit sockets on the sinks pre wired with a large hold down bar. that way many sets of transistors could be pre mounted and allowed to acclimate to room temp.
Then the MCU would trigger on one transistor at a time, for a predetermined time. I could then use one of those RS-232 multimeters to read a value and store it in memory, even print it.
Something to that effect. that way a person(s) could run through a batch of a 1,000 transistors pretty quickly
Two people working side by side could run through them in one continuous shot, One person loads the heatsinks up and plugs one sink into the tester, the other person removes the sink from the tester and catalogs or marks each device, then removes the transistors from the sink and puts it back into the que.
Any thoughts or ideas on a set up like this? I would think something like this should yield pretty consistant results per batch.
Zc
My first thoughts were to use a MCU like a basic stamp etc for the heart of the thing.
I have some 2' long heatsinks that i can outfit with a couple of poer terminals and a ribbon cable that should fit quite a few devices. For TO-3 devices i could outfit sockets on the sinks pre wired with a large hold down bar. that way many sets of transistors could be pre mounted and allowed to acclimate to room temp.
Then the MCU would trigger on one transistor at a time, for a predetermined time. I could then use one of those RS-232 multimeters to read a value and store it in memory, even print it.
Something to that effect. that way a person(s) could run through a batch of a 1,000 transistors pretty quickly
Two people working side by side could run through them in one continuous shot, One person loads the heatsinks up and plugs one sink into the tester, the other person removes the sink from the tester and catalogs or marks each device, then removes the transistors from the sink and puts it back into the que.
Any thoughts or ideas on a set up like this? I would think something like this should yield pretty consistant results per batch.
Zc
I think this is a very interesting idea... I have been working on something similar to it, in the form of a computerized transistor tester.
I would suggest getting a desktop/laptop involved instead of (or added to) the stamp/PIC if at all possible. Think about it... you haven't specified how you want these to be matched. Do you want to only find a match to a single transistor from the remaining parts? Or would you want to find every matching pair from a large pile of transistors? And if so, you need to consider the trade-offs... maximum number of pairs or minimum error between pairs, etc, etc... a desktop makes this much easier to manage and automate, I think.
Also, what parameters do you want to match? Just hFe/gm, or others like capacitances, breakdown V, etc. And BJT, Mosfet, or both?
Sorry for the millions of questions, but I'm really curious about this now
I would suggest getting a desktop/laptop involved instead of (or added to) the stamp/PIC if at all possible. Think about it... you haven't specified how you want these to be matched. Do you want to only find a match to a single transistor from the remaining parts? Or would you want to find every matching pair from a large pile of transistors? And if so, you need to consider the trade-offs... maximum number of pairs or minimum error between pairs, etc, etc... a desktop makes this much easier to manage and automate, I think.
Also, what parameters do you want to match? Just hFe/gm, or others like capacitances, breakdown V, etc. And BJT, Mosfet, or both?
Sorry for the millions of questions, but I'm really curious about this now
Those are all good questions. and i guess the answers depend on how hard each test would be to implement.
My initial thought was just to match hFe/gm or Vds. Some way of sorting devices into at least a couple of ranges to sort pairs from.
This idea was sparked by reading the Pass paper on Mosfet matching and that it is better to match the devices at the temp and current they are going to be used. and that the Vds varies with temp. So my thoughts were to mount banks of devices onto many sets of heatsinks. let them acclimate, then automate the tests so that the current running through each device was on for the for the same amount of time for each device.
So, lets talk about how this would be done for the various tests, and what test's are really important for device matching.
Measuring for Vds Per pass's paper looks pretty simple, and i would imagine that the same sort of test for Bjt's would be about as easy. it should be pretty simple to setup an automated system for testing devices as such. But is that enough?
I work on many types of amplifiers. everything from cheap and crappy PA system amps that have zero device matching and then the mega buck high end amps where my goal is to repair them (or build them) to at least the same spec as the factory or perhaps better.
So lets say i have a mega buck high end amp sitting here. one channel is blown and i need to replace all the output devices. not far off actually as i do have one sitting right here in just that condition.
Now then i am going to have to buy a lot of devices to get enough to replace all 10 in this amp. I might have to buy 1000 to get the tight match needed? thats ok, but then i have 1000 devices to go through. If i can sort them into logical groups, i could then use those devices for other projects, or share them with other DIY'ers.
So lets start with that. Where do we go from here? what test's are critical? Do we need to so some sort of curver tracer?
Zc
My initial thought was just to match hFe/gm or Vds. Some way of sorting devices into at least a couple of ranges to sort pairs from.
This idea was sparked by reading the Pass paper on Mosfet matching and that it is better to match the devices at the temp and current they are going to be used. and that the Vds varies with temp. So my thoughts were to mount banks of devices onto many sets of heatsinks. let them acclimate, then automate the tests so that the current running through each device was on for the for the same amount of time for each device.
So, lets talk about how this would be done for the various tests, and what test's are really important for device matching.
Measuring for Vds Per pass's paper looks pretty simple, and i would imagine that the same sort of test for Bjt's would be about as easy. it should be pretty simple to setup an automated system for testing devices as such. But is that enough?
I work on many types of amplifiers. everything from cheap and crappy PA system amps that have zero device matching and then the mega buck high end amps where my goal is to repair them (or build them) to at least the same spec as the factory or perhaps better.
So lets say i have a mega buck high end amp sitting here. one channel is blown and i need to replace all the output devices. not far off actually as i do have one sitting right here in just that condition.
Now then i am going to have to buy a lot of devices to get enough to replace all 10 in this amp. I might have to buy 1000 to get the tight match needed? thats ok, but then i have 1000 devices to go through. If i can sort them into logical groups, i could then use those devices for other projects, or share them with other DIY'ers.
So lets start with that. Where do we go from here? what test's are critical? Do we need to so some sort of curver tracer?
Zc
Neither of those links work anymore so I don't know what is in them but...
....let me know if this is a pipe dream...
Ideally, the transistor "tester" would be automated and connected to an excel database. Little plastic baggies with paper labels numbered 1 to whatever hold the transistors after testing. The baggies would have holes in them to hang on hooks like at the store to keep them organized.
The test sequence would be to insert the DUT into a socket, click on a button on the screen, the test would run, and then an indicator would turn on to show when the test is complete. Take the DUT out of the socket and place in the labeled baggie and hang on the hook. Other Excel macros could sort the transistor per whatever criteria you choose.
The person performing the test would of course select the particular transistor type on the screen to limit voltages so that the DUT would remain unharmed. If a MLSSA type of signal were used like in speaker testing, I don't think a heatsink would be necessary since it would be finished quickly.
Twelve bit ADC would allow 4096 different steps or if someone were ambitious they could use a programmable DSP...
...like I said, this may be a wild haired scheme....
....let me know if this is a pipe dream...
Ideally, the transistor "tester" would be automated and connected to an excel database. Little plastic baggies with paper labels numbered 1 to whatever hold the transistors after testing. The baggies would have holes in them to hang on hooks like at the store to keep them organized.
The test sequence would be to insert the DUT into a socket, click on a button on the screen, the test would run, and then an indicator would turn on to show when the test is complete. Take the DUT out of the socket and place in the labeled baggie and hang on the hook. Other Excel macros could sort the transistor per whatever criteria you choose.
The person performing the test would of course select the particular transistor type on the screen to limit voltages so that the DUT would remain unharmed. If a MLSSA type of signal were used like in speaker testing, I don't think a heatsink would be necessary since it would be finished quickly.
Twelve bit ADC would allow 4096 different steps or if someone were ambitious they could use a programmable DSP...
...like I said, this may be a wild haired scheme....
yeah something like that! you could mark the devices with a sharpie or whatever method you choose to sort. but you should be able to have a way to select voltage and current through the DUT for matching.
I dreamnt of being able to load up a dozen or more devices at a time and have the tester quickly test all of them sequentially and then mark each device with a sharpie, and reload.
I need a way to grade 1000 Mosfets at a time!
I dreamnt of being able to load up a dozen or more devices at a time and have the tester quickly test all of them sequentially and then mark each device with a sharpie, and reload.
I need a way to grade 1000 Mosfets at a time!
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.