Synchronizing a switch advance to multiple SMU measurements

[AbandoningCausality]

Hi,

I have a 4138 SMU and a 2536 switch matrix. I'm trying to "know" when the switch is moving in hopes of later realigning the data to the switch in software.

The SMU will be set to a constant voltage the whole time so no need to worry about that triggering.

The switch should be itterating over a switch list.

Here's one way I can imagine doing this:

If I had this, it would be really simple to just decimate my SMU read array.

Another way I could imagine would be to delay n measurement periods between switches. As long as their base clocks were syncronized I wouldn't have to rely on triggering. BTW, if I do a scanned switch with no triggering, how fast does it run through the sequence? 

Another way is if I could get an accurate timestamp/base clock tick count on each measurement and switch action, I could try and later realign it in software.

Another way would be to write an advanced sequence or sequence source mode on the SMU to only output triggers every N measurements

Thoughts?

[TommyDunkz]

Hi AbandoningCausality,

This sounds like an interesting application! Is the switch necessarily triggered on each measurement? You could send a timestamp in software whenever the trigger is sent, although this may be too computationally intense if you're going to be running at high speeds.

I found a Community example which looks similar and may be a good reference. I have included a link to this example at the bottom of this post.

Also, the PXI-2536's maximum scan rate is 10,000 crosspoints/s, as noted on page 3 of the device's spec sheet (also linked below).

Community: SMU Switch Handshaking with Pulsing: https://decibel.ni.com/content/docs/DOC-43560

PXI-2536 Specifications: http://www.ni.com/pdf/manuals/374625d.pdf

[AbandoningCausality]

I'm not totally clear on how soft timestamping could work. For one, we want to sample really fast and trigger quite fast. If I follow the link I see 50k crosspoints/s:

The other issue is detecting when "the trigger is sent". Of couse, if we soft trigger it, that's an option but the diagram I made is hard triggered.

The community example is useful. It's good to know cold switching is an option. It seems to be taking a measument after sourcing though. What I described above is sampling at the 1.8Msample rate, not the 100 ksample sourcing rate.

One question I have with the example is the SMU seems to depend on a trigger and the Switch depends on a trigger so which instrument is the first mover? Why don't the instruments just timeout? Is there any way to make each instrument trigger on both a software and a hardware trigger? The goal of the software trigger to start everything?

Thanks for your thoughts. I have my application mostly working but I was interested if it was possible to do what I described in the diagram.

[TommyDunkz]

Hi AbandoningCausality,

You are correct about the scan rate. That was a typo on my part.

Just to confirm--are you using LabVIEW for this application, or a different development environment?

I'm still trying to clarify the problem statement at this point, so please correct me if I'm wrong in my interpretation. Presumably, the dataset that your application generates will have timestamps for each data point (I'm guessing you're looking at something like a waveform or array of data)--however, you are asking if there is a way to correlate the generated data points with the particular switch configuration which was in use by the 2536? In order to find the exact time at which the switch configuration was changed, according to your scan list?

Regarding timing, the hardware timing engine on your SMU will be your best bet for a timebase. Switch modules don't have onboard timing, so unfortunately we don't have a quick way of generating timestamps from that card whenever jumping from one switch configuration to the next.

[AbandoningCausality]

Yes, LabVIEW.

Timestamping was one possibility that occurred to me to solve my general problem of synchronizing the data with the switch positions. Hardware triggering was another idea (first idea). In that case, I wouldn't need a timestamp because I'd have a fixed decimation ratio (every 10 samples corresponds to one switch position.

[Agamemone]

Hi Abandoning Causality,

It would make sense to create a timestamp to notify the user when your configuration changes. It sounds as though your application is currently up and running? With that in mind, are there any outstanding questions you have or anything that needs clarification?

[AbandoningCausality]

My application is running but I don't have it hardware handshaking as shown in the top most picture. I don't know if that is possible. Tommy's summary of the problem was accurate. I also think timestamping is not possible because as tommy mentioned, the card does not support that and software would not have enough resolution for 50kSamp/sec.

[Agamemone]

Could you change the voltage level when the configuration changes to mark the change?

[AbandoningCausality]

Can you clarify what you mean? What is measuring the voltage? What is generating the voltage syncrounous to the switch advance? What would be triggering the switch to advance syncrounous to the SMU measuring 10 samples?