Here are some features that I feel should be in the kind of device needed:
- Testing (comparing/matching) power devices (bjt/fet)
- Power devices should be heatsinked
- A temperature sensor should be on the heatsink, and reading available with the data to the software
- Devices should be tested by pairs, at the same time, with a common temperature. if no heatsink, for the non-power devices, then preferably those should be in close contact to each other and time should be given to them to equalize their temperatures
- voltage and current for power devices should allow for the high power types, so at least 100V, possibly even more, and the highest current possible, and be able to test their hfe against current, same with vbe, while temperature is known and monitored
- for small devices, 2 parts at the same time in a socket arrangement to place them physically in contact for best temperature exchange
- software should allow saving each tested part's data, for comparison by software, in addition to the visual from curves
- plotting 2 devices curves at the same time on the same plot can very easily and quickly identify matches, visually. differences can be analyzed by software, with a % given for deviations and the match quality be quantified
- for the smallest devices, perhaps 4 of them could be tested simultaneously, with half of each type NPN/PNP, so good pairs of pairs could be made for dual diff amps. not easy, with the large differences between types, but would go a long way to make the best matches possible
- Testing (comparing/matching) power devices (bjt/fet)
- Power devices should be heatsinked
- A temperature sensor should be on the heatsink, and reading available with the data to the software
- Devices should be tested by pairs, at the same time, with a common temperature. if no heatsink, for the non-power devices, then preferably those should be in close contact to each other and time should be given to them to equalize their temperatures
- voltage and current for power devices should allow for the high power types, so at least 100V, possibly even more, and the highest current possible, and be able to test their hfe against current, same with vbe, while temperature is known and monitored
- for small devices, 2 parts at the same time in a socket arrangement to place them physically in contact for best temperature exchange
- software should allow saving each tested part's data, for comparison by software, in addition to the visual from curves
- plotting 2 devices curves at the same time on the same plot can very easily and quickly identify matches, visually. differences can be analyzed by software, with a % given for deviations and the match quality be quantified
- for the smallest devices, perhaps 4 of them could be tested simultaneously, with half of each type NPN/PNP, so good pairs of pairs could be made for dual diff amps. not easy, with the large differences between types, but would go a long way to make the best matches possible
Tek just sent me a link to their SMU's here: http://www.tek.com/sites/tek.com/fi...1KW-57501-1_SimplifyFET-Testing_AppNote_0.pdf and I though it would be interesting to share since the device discussed here is very similar in capabilities to the SMU in the app note. SMU's are not cheap, but less than $30k. This explains why Tek stopped making curve tracers.
I do think a DIY dual SMU with sensor interface for a temp probe would be great. Even better if someone created software for it and it was network interfaced (no drivers or PC dependencies) like the Tek/Keithley SMU.
The hardware would be essentially 2 12 bit DAC's, two 16 bit ADC's, one temperature probe interface, one high current power amp, possibly a high voltage power amp and a low voltage power amp for base/gate drive.
Ideally you would want to extract parameters that can be pulled into models automagically. Like Locky's box its really all about the software.
I do think a DIY dual SMU with sensor interface for a temp probe would be great. Even better if someone created software for it and it was network interfaced (no drivers or PC dependencies) like the Tek/Keithley SMU.
The hardware would be essentially 2 12 bit DAC's, two 16 bit ADC's, one temperature probe interface, one high current power amp, possibly a high voltage power amp and a low voltage power amp for base/gate drive.
Ideally you would want to extract parameters that can be pulled into models automagically. Like Locky's box its really all about the software.
I designed a LC USB meter and that was fun.
I got the USB software from Home - WFFwiki
He provides PIC18F4550 circuit and code and the PC code in C#.
I just modified it to interface to the LC circuit.
I got the USB software from Home - WFFwiki
He provides PIC18F4550 circuit and code and the PC code in C#.
I just modified it to interface to the LC circuit.
This stuff can't be cheap anyway, and it's very specialized, not really multi-purpose like what we need.
Actually I don't know of anything out there that is strictly aimed at the diy power amp builders. What I'd like to have is an efficient way to test and match all of the critical parts used in preamps and power amps, to make sets, with more than 2 parts in them, like for example as I mentioned the dual diff amp stages used in many preamps and input stages of power amps.
Being able to find pnp and npn pairs that best match would go a long way to make those input stages really well balanced, as they should be.
I did quite a bit of simulations that show how critical the proper balance can be.
Not only the output offset gets reduced, but many other things like distortion get much better.
We know how devices can be different when we buy them, and with the right test tool we can minimize those differences where it counts.
I have a big project with a bunch of amps and other things, and I won't build them without first selecting the parts properly.
This project I''ve had for a very long time and I'm still working on it after so many years. It may take some time, but I want to do it right.
And I don't believe this can cost as much as some think, especially with most of the functions done in software.
The temperature is a critical parameter, and should be included in the testing, so if it's to be done right, it has to be there.
Right, and this doesn't have to cost $1000s. We can do it.
That would also be awesome! Great idea!
And this wouldn't have to be at more expense in hardware, it can all be software.
And software is also something that can evolve, so more features can even be added later as we discover things we can do and didn't think of earlier.
He's done a great job and it's a nice tool, but for me it doesn't go far enough and it's still too restricted, because it's a windoze only software, and that's a show stopper for me.
His tracer software doesn't do the part matching, but this could be added.
He didn't make it to test power devices far enough.
I think we need a good heatsink, with temperature sensing, and dual sockets, on the heatsink, with a fast system to mount and unmount the devices, so they can be tested fast and in pairs.
Doing pairs can help even if we want to make matched sets of several devices, as one device can remain on the test tool to match it with more parts.
It's obvious that using this requires having a large enough batch of parts to match, so the chances are better to actually find good matches.
This is probably the costly part, for the power devices, but I think the results would be worth the effort.
We have a diy community, and there are possibilities of doing part exchanges, when many people do the same builds, using the same parts, there could be a pooling of parts to increase the batch size and share the results.
This design will be a bit complex, and it would be great to have the work shared among several people.
It would not only get done faster, but it's also good to have more pairs of eyeballs verifying things done by others, to reduce mistakes.
It's mostly modular, so this whole thing can be split up in sections that can be assembled into a final design.
Even the software could be done like the open source stuff is, and grow over time.
One thing is imperative for me, is a mac native software, no emulation!
Cool!
What does the software do? I bet it's aimed again at windoze only. Being C# pretty much ties it to that.
I've been getting interested in the pic platform lately, and just got a few little things to be able to play with it and get familiar.
I've been thinking about using pic in some design I'm working on, or perhaps the attiny.
I saw that bryston is using pic in some of their power amps. Obviously we won't see their software, but this is a source of ideas of how it can be done.
There is one rather difficult thing to do for something that should be buildable by any diyer anywhere on this planet, which is to make use of parts that can be obtained anywhere.
I doubt this to be feasible universally, but there are ways to get passed that.
For example for a useful design that has wide interest among diyers, there is the store on this site and those who can't find things could find it here.
I doubt this to be feasible universally, but there are ways to get passed that.
For example for a useful design that has wide interest among diyers, there is the store on this site and those who can't find things could find it here.
Right now I'm thinking about the design to handle 2 TO3 parts on the heatsink.
This should be on sockets, for easy and quick release to swap parts out fast and repeatedly.
Of course the device cases should be isolated from the heatsink, so the high voltages wouldn't be an issue for the lowly humans who would be stupid enough to touch this while in operation.
The thing is, the TO3 sockets are not made for many part swappings, they're only for final mounting and probably would have a problem with too many part insertions.
But what choice is there?
Anyone knows of good usable sockets that could sustain such abuse?
The power transistors aren't all using the same type of case, which complicates the design.
It requires several methods of attachment to accommodate as many types of parts as possible.
I was thinking of a way to hold the TO3 cases in place without using screws, while at the same time offering a little protection against busy hands that would want to touch the device cases while high voltage is applied.
This could be done with some kind of clamp on top of both devices in the rig, which would have a conducting part that would be pressed on top of the TO3 case, and some insulation on top to prevent hands from reaching in.
Maybe nearby a LED showing high voltage is applied, so as to warn not to touch. When the LED is out, the clamp could be released and parts removed.
This would be less of an issue with the plastic cases like the TOP3 or similar, but those need a different type of socketing and clamping.
Anyone has any ideas about this?
btw: For the software, there may be a way to write one version that works on many platform, by using Tcl/Tk... That would be fairly universal wouldn't it be?
This should be on sockets, for easy and quick release to swap parts out fast and repeatedly.
Of course the device cases should be isolated from the heatsink, so the high voltages wouldn't be an issue for the lowly humans who would be stupid enough to touch this while in operation.
The thing is, the TO3 sockets are not made for many part swappings, they're only for final mounting and probably would have a problem with too many part insertions.
But what choice is there?
Anyone knows of good usable sockets that could sustain such abuse?
The power transistors aren't all using the same type of case, which complicates the design.
It requires several methods of attachment to accommodate as many types of parts as possible.
I was thinking of a way to hold the TO3 cases in place without using screws, while at the same time offering a little protection against busy hands that would want to touch the device cases while high voltage is applied.
This could be done with some kind of clamp on top of both devices in the rig, which would have a conducting part that would be pressed on top of the TO3 case, and some insulation on top to prevent hands from reaching in.
Maybe nearby a LED showing high voltage is applied, so as to warn not to touch. When the LED is out, the clamp could be released and parts removed.
This would be less of an issue with the plastic cases like the TOP3 or similar, but those need a different type of socketing and clamping.
Anyone has any ideas about this?
btw: For the software, there may be a way to write one version that works on many platform, by using Tcl/Tk... That would be fairly universal wouldn't it be?
The tektronix sockets for the curve tracers are quite interesting with kelvin connections, but really pricey.
I have built matching ficures for transistors in the past. They don't need to be all that complex actually. A setup at the target operating points usually will do as good or better than matching curves. Input pairs your lloking for base/gate bias voltage for a particular collector/drain current. If the fixture is set up right you can check AC gain and noise on the same fixture. This is essentially what the Quan-tech transistor noise analyzers do. Their circuits can be redone today pretty easily.
Output transistors are more involved but essentially very similar.
That would be less of a design tool than a dedicated selection system.
I can post the relevant schematics of the Quan-tech if that's of interest.
I have built matching ficures for transistors in the past. They don't need to be all that complex actually. A setup at the target operating points usually will do as good or better than matching curves. Input pairs your lloking for base/gate bias voltage for a particular collector/drain current. If the fixture is set up right you can check AC gain and noise on the same fixture. This is essentially what the Quan-tech transistor noise analyzers do. Their circuits can be redone today pretty easily.
Output transistors are more involved but essentially very similar.
That would be less of a design tool than a dedicated selection system.
I can post the relevant schematics of the Quan-tech if that's of interest.
Can you show some links?
For convenience, sockets should allow quick and easy part swapping.
The sockets can't be the same for all of the parts, so obviously each type will require different setups.
I think for the big TO3 types, that a socket under a heatsink would be best and some type of clamp above to make contact to the collector while holding the parts down tightly not only for the collector contact, but also the thermal one.
This would allow easy and quick swap of those big parts.
There should be 2 parts tested at the same time, for immediate comparison.
One part could be quickly changed while keeping the other in place, so matches could be easily found.
I think the curves are just an extra accessory for visual comparison, but the data points should be better compared by software. Much quicker, can be saved, and closer matching can be done.
I think I'd want more than just a single operating point test.
Sounds interesting.
Since the software can handle most of the tests, why not make as many as possible, for best comparisons.
It might be a bit difficult when matching a large number of parts, to keep them organized to match the parts with their saved data sets, but I think this would be a definite plus, to be able to make best matches, and go beyond just matched pairs.
Not easy to match npn and pnp, but with a large enough sample, this should be more feasible, and having a lot of parts tested and their data sets saved, can allow feeding this to a software matching routine. This would require some system to mark tested parts to remember which ones correspond to a data set. (numbering system)
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.