David_
Advanced Member level 2
Hello.
While I went through designing a DC-DC converter(not done yet) I became aware of SMUs or Source Measure Units and I immediately realized that I want one of those(minimum 2 channels), but the price tag of such a device is too prohibitive for me to be able to get one.
But I don't see a reason why I wouldn't be able to design my own, and I am somewhat surprised to find that there are pretty much no such design on the web.
The commercial SMUs seems to be in most cases real marvels of engineering with ranges that can span down to some 10's of femtoAmps, which I where told by someone on a blog or something is equivalent to 1 electron per 16uS, which sound insane.
Of course I have no such illusions as to think that I can accomplish anything near that, but then again I don't see why I would need nor do I see any point for me to have a unit with such insane resolution(since I feel it to be insane resolution obviously I have no need for it), if I had such a device I could surely find some cool thing to do with it but sine the use that sparked of my interest in SMUs is to characterize my DC-DC converter design or to analyse it's output response(I am not quite sure if a SMU can do both of those things or only one of them), so given the fact that it wouldn't have to be very fancy I think it is a very cool project to implement.
It would involve:
1. A control unit based on a microcontroller(µC).
2. A Matlab script that establishes communication with the µC through USB(from Matlab it looks like a ordinary COM port but from the µC it will look like an USB host through the µC's USB transceiver)
3. An switch mode power supply to convert a single positive input voltage into both positive and negative outputs(I haven't yet decided on what kind of voltages I what but I think at least ±15 or ±20V).
4. Followed by a linear regulator stage to clean up the supply rails.
5. A analog output stage that interfaces ADCs to measure both channels output voltages/currents and DACs which will steer the analog output power components to output a variety of waveforms(including steady DC voltages, sine-waves, square-waves and PWL derived custom waveforms) while being able to supply high currents(high being at least 1A). This stage would also contain a sort of large number of switches that is managing the output polarity, whether the constant current feedback loop is used or the constant voltage feedback loop is used, the range settings, and probably some more stuff that i can't think of right now.
The analog output stage is the biggest mystery for me at this point, although I have been reading and trying to figure out some patents regarding SMU circuits but that has not revealed anything conclusive yet but I am hoping that I will be able to derive a basic structure for the output stage from those patents.
As to the output capabilities, I haven't decided on what output current limit I want but I suppose I want between 1A and 3A which is really just a random number... and resolution will also have to be decide upon, as to the voltage I think I would like ±20V at the output terminals with a resolution that is the same as for the current.
There are surely a lot of compromises to be made, usually I feel that I get my self into trouble due to a desire for higher resolution even though the desire for that is disconnected from the actual application but in light of this project I feel that I will have no problem in making compromises which is very positive. Although I will not get away with using the µC's internal data converters, actually using an XMEGA A1U µC could work since it have 2 * 12bit 2MSPS ADCs and 2 * 12bit 2MSPS DACs but even though I don't remember why I don't want to use them right now I believe there are a good reason why to use external data converters in a design like this, it might have to do with maintaining isolation between the 2 channels of the SMU.
One concern I have though is the speed of the µC required to realize the sampling frequency needed to be able to properly work with a DC-DC converter which is switching at 100kHz, could someone give me some kind of estimate for how fast a sampling frequency you think I will need to be able to properly make measurements on a DC-DC converters who's internal switching frequency is 100kHz?
I don't know where I got the following idea from but I read some blog about some SMU circuit and the blogger wrote that he needed a µC with a 100MHz clock in order to realize a 100kHz sampling frequency, which sounds very weird to me, if I want a 1MSPS system then surely a µC running with a 32MHz clock must be perfectly sufficient?
Any and all suggestions/opinions/criticisms are very appreciated.
I put this thread in the analog section because I am going to focus quite a lot on achieving a low-noise output and in that aspect this is very much a analog design issue.
Regards
Oh, here is a quick suggestion for how such a device could be constructed, though I have to admit there are some possible flaws and not so elegant solutions. Nevertheless it would be functional and very flexible.
While I went through designing a DC-DC converter(not done yet) I became aware of SMUs or Source Measure Units and I immediately realized that I want one of those(minimum 2 channels), but the price tag of such a device is too prohibitive for me to be able to get one.
But I don't see a reason why I wouldn't be able to design my own, and I am somewhat surprised to find that there are pretty much no such design on the web.
The commercial SMUs seems to be in most cases real marvels of engineering with ranges that can span down to some 10's of femtoAmps, which I where told by someone on a blog or something is equivalent to 1 electron per 16uS, which sound insane.
Of course I have no such illusions as to think that I can accomplish anything near that, but then again I don't see why I would need nor do I see any point for me to have a unit with such insane resolution(since I feel it to be insane resolution obviously I have no need for it), if I had such a device I could surely find some cool thing to do with it but sine the use that sparked of my interest in SMUs is to characterize my DC-DC converter design or to analyse it's output response(I am not quite sure if a SMU can do both of those things or only one of them), so given the fact that it wouldn't have to be very fancy I think it is a very cool project to implement.
It would involve:
1. A control unit based on a microcontroller(µC).
2. A Matlab script that establishes communication with the µC through USB(from Matlab it looks like a ordinary COM port but from the µC it will look like an USB host through the µC's USB transceiver)
3. An switch mode power supply to convert a single positive input voltage into both positive and negative outputs(I haven't yet decided on what kind of voltages I what but I think at least ±15 or ±20V).
4. Followed by a linear regulator stage to clean up the supply rails.
5. A analog output stage that interfaces ADCs to measure both channels output voltages/currents and DACs which will steer the analog output power components to output a variety of waveforms(including steady DC voltages, sine-waves, square-waves and PWL derived custom waveforms) while being able to supply high currents(high being at least 1A). This stage would also contain a sort of large number of switches that is managing the output polarity, whether the constant current feedback loop is used or the constant voltage feedback loop is used, the range settings, and probably some more stuff that i can't think of right now.
The analog output stage is the biggest mystery for me at this point, although I have been reading and trying to figure out some patents regarding SMU circuits but that has not revealed anything conclusive yet but I am hoping that I will be able to derive a basic structure for the output stage from those patents.
As to the output capabilities, I haven't decided on what output current limit I want but I suppose I want between 1A and 3A which is really just a random number... and resolution will also have to be decide upon, as to the voltage I think I would like ±20V at the output terminals with a resolution that is the same as for the current.
There are surely a lot of compromises to be made, usually I feel that I get my self into trouble due to a desire for higher resolution even though the desire for that is disconnected from the actual application but in light of this project I feel that I will have no problem in making compromises which is very positive. Although I will not get away with using the µC's internal data converters, actually using an XMEGA A1U µC could work since it have 2 * 12bit 2MSPS ADCs and 2 * 12bit 2MSPS DACs but even though I don't remember why I don't want to use them right now I believe there are a good reason why to use external data converters in a design like this, it might have to do with maintaining isolation between the 2 channels of the SMU.
One concern I have though is the speed of the µC required to realize the sampling frequency needed to be able to properly work with a DC-DC converter which is switching at 100kHz, could someone give me some kind of estimate for how fast a sampling frequency you think I will need to be able to properly make measurements on a DC-DC converters who's internal switching frequency is 100kHz?
I don't know where I got the following idea from but I read some blog about some SMU circuit and the blogger wrote that he needed a µC with a 100MHz clock in order to realize a 100kHz sampling frequency, which sounds very weird to me, if I want a 1MSPS system then surely a µC running with a 32MHz clock must be perfectly sufficient?
Any and all suggestions/opinions/criticisms are very appreciated.
I put this thread in the analog section because I am going to focus quite a lot on achieving a low-noise output and in that aspect this is very much a analog design issue.
Regards
Oh, here is a quick suggestion for how such a device could be constructed, though I have to admit there are some possible flaws and not so elegant solutions. Nevertheless it would be functional and very flexible.