Electronics > Projects, Designs, and Technical Stuff
Seeking advice on multiplexing measurement system
kronk:
Hi there folks,
I am tasked with building a system to test multiple on-wafer resistors (TLM structures). My post here has mostly to do with the measurement/switching part of the system. I have read a lot on the topic in recent weeks and gone through several iterations as I have learned more, but I'm new to the test side of things. I'm posting here to see if those more experienced than myself can check my work or suggest something I don't even know exists!
Here are the constraints I'm working with:
* Four wire measurements are a must
* Resistors will be as low as 0.5 Ohm and I want 3 significant figures
* I need enough channels for at least 4 resistors, so I need 4 channels of 4 pole switching minimum
* While I can spend a few thousand on this, I'm budget conscious and interested in effective options that are in the <=$1000 range even if they will take a bit more work
* I have a Keithley 2400 and a couple Agilent 34401As on hand; there are also a couple of 7001s with four pole switch cards in the lab but someone else is eyeing them too
* The equipment I use/get for this will be dedicated to this simple task, so it need not be a fancy whiz-bang-do-it-all magic boxI've formulated two general approaches to achieve this:
* 4 wire instrument with separate switch system (SMU or really any DMM paired with a low loss switch; ie Keithley 2400, Agilent 34401A for the instrument, Keithley 7001, 7002, or 702 as the switch.)
* DAQ unit that integrates DMM with multiplexing capability (ie Agilent 34970A, Keithley 199, Keithley DAQ6510)My first question is if an integrated system is a better approach? Less loss or opportunity for interference?
My second question is what other separate switch systems exist besides the Keithleys? I've seen some older HPs on ebay, but they are very cheap and I suspect that is for a reason.
Third, I am a little skeptical of older/used equipment for this application - accuracy and precision are pretty important and I am concerned that the low upfront cost of used instruments means spending more on calibration before being able to use the thing. Thoughts?
Fourth, my inclination is that if I'm going to buy something new I may as well get an integrated system, currently leaning towards the DAQ6510. Is there a current-gen system (separate or integrated) that will meet my needs at a better value than that particular box?
Finally, here are my candidate solutions, coming from different sides of things:
* Save Time, Pay For New: buy DAQ6510
* Use What I Have, Use Extra Budget To Buy More Wafers: use Agilent 34401A and either fight for a 7001 or buy one used I will greatly appreciate any feedback or suggestions!
moffy:
You could put the 4 resistors in series and run a precision current through them then use something like the LTC2442 with 4 multiplexed inputs to measure the voltages, then just subtract and divide to get the individual resistances. It's effectively 4 wire because the current injection is separate to the measurement wires.
Berni:
--- Quote from: moffy on July 09, 2020, 04:42:27 am ---You could put the 4 resistors in series and run a precision current through them then use something like the LTC2442 with 4 multiplexed inputs to measure the voltages, then just subtract and divide to get the individual resistances. It's effectively 4 wire because the current injection is separate to the measurement wires.
--- End quote ---
This is also my thought.
If cost is more important than effort then you can just get a high channel count precision ADC chip and run a precision current source trough all 4 resistors in series. Hook that up to a MCU that spits out the results via serial.
The reason that real test rigs tend to use a rack full of test equipment is that this tends to be the faster and cheaper option for industry. Its expensive to pay an engineer to spend a few months faffing about with designing, building, testing, programming...etc a device the size of a shoe box that does the same thing as a rack of equipment and costs less. They can instead just buy a bunch of test gear, throw it into a rack, hook it up, slap some labview at it and be up and running in a few days.
The reason that old HP switch matrix mainframes are cheap on ebay is that people don't often need one and hobbyists are not interested in them so much. If someone in industry needs a switch mainframe they will buy a brand new one, because they can't afford to have it fail and stop production, they might need a 2nd identical one for a spare in case it does fail, and they want the manufacturer to support them by fixing the failed unit quickly and hassle free. Once they don't need the test rig anymore they will toss it out and you will find that switch mainframe on ebay.
kronk:
--- Quote from: Berni on July 09, 2020, 05:18:38 am ---
If cost is more important than effort then you can just get a high channel count precision ADC chip and run a precision current source trough all 4 resistors in series. Hook that up to a MCU that spits out the results via serial.
--- End quote ---
This was my first thought, and like you said my boss was not too interested in waiting a couple of months for me to do it cheap and custom. We aren't that cash strapped and we have a lot of other projects. He comes from a big company where buying new is the MO, I come from a small-mid size company that is only just getting out of the "buy RIE tools on ebay" stage.
Dulus:
Dont know how advanced you're planning to get but,
--- Quote ---You could put the 4 resistors in series and run a precision current through them
--- End quote ---
you could also incorporate a fifth fixed-known-precision resistor that you can derive the current from.
Save some more bucks from a precision current source :D
Only thing is that the current must be noise/ripple free, which you can perform with a low noise LDO type IC.
Navigation
[0] Message Index
[#] Next page
Go to full version