SMU as Ammeter to Measure Current

[Jeffrey_Zola]

Update:  Measurements with 100 ms aperture at 50 Ohms Output Resistance were very well behaved, with no difference in the variance of current measurements from either instrument, nor any spikes in the SMU voltage- based on 51 samples taken once per minute for each of six DUTs.

Measurements with 100 ms aperture at 0 Ohms Output Resistance were also MUCH better than they had been without an aperture setting. There was one current reading from the SMU,however, where it spiked upward, with a corresponding drop in voltage. I wonder if it is unable to properly handle the "0" setpoint.

I did a quick calculation, and it appears that I am seeing about 22 Ohms impedance from the SMU. I have started an overnight run with my original 5-minute-per-sample external frequency at -22 Ohms Input Resistance, 1 mA Range, 100 ms Aperture Time, and SelfCal. Will sleep with my fingers crossed.  🙂

Jeff

[Brandon_G.]

Hey Jeff,

The default aperture time on the 4141 is 16.66 ms or 1 PLC @ 60 Hz.  The DMM at 6 1/2 digits is by default set to 100 ms or 6 PLCs @ 60 Hz.  

What was the voltage and current reading from the measurement point that was off?  How does it compare to the other points?  You may be hitting compliance during that one measurement point, which could cause your SMU to output a steady current and adjust its voltage so that the compliance current level is maintained.

How did you calculate the 22 ohms?  Was the 22 ohms the impedance seen during the spike or just in general?  I would expect if you're outputting 0V from the SMU your impedance would always be very very close to 0?  Maybe I'm misunderstanding something in your setup?  Though programming -22 ohms may null out the impedance, it would be worthwhile to find out why it's there in the first place.  If you have any graphs you can share, I'd love to see them.

Thanks!

Brandon G

[Brandon_G.]

Hey Jeff,

The default aperture time on the 4141 is 16.66 ms or 1 PLC @ 60 Hz.  The DMM at 6 1/2 digits is by default set to 100 ms or 6 PLCs @ 60 Hz.  

What was the voltage and current reading from the measurement point that was off?  How does it compare to the other points?  You may be hitting compliance during that one measurement point, which could cause your SMU to output a steady current and adjust its voltage so that the compliance current level is maintained.

How did you calculate the 22 ohms?  Was the 22 ohms the impedance seen during the spike or just in general?  I would expect if you're outputting 0V from the SMU your impedance would always be very very close to 0?  Maybe I'm misunderstanding something in your setup?  Though programming -22 ohms may null out the impedance, it would be worthwhile to find out why it's there in the first place.  If you have any graphs you can share, I'd love to see them.

Thanks!

Brandon G

[Jeffrey_Zola]

Thanks, Brandon. 16.7 ms is pretty close to 100 ms with respect to noise resolution performance, according to the graph in the spec manual.

I do not think compliance is an issue. I am currently measuring 66.7 uA on the PXI-4141, with the module in Voltage Output mode with Remote Sense. Voltage setpoint is 0 V with the voltage limit set to 10 V. The current limit and the current limit range are both set to 1 mA. The graphs use an Output Resistance of -21 Ohms.

There are four currents: (1) DMM alone (2) SMU alone (3) DMM in series with SMU simultaneous with (4) SMU in series with DMM:

It makes sense that the SMU-alone current is a bit higher than the others because this is the only measurement without the internal shunt resistor. I am concerned about the blips in the SMU-with-DMM current, even though they were not present in the SMU-alone configuration. If I do not know their origin, I have to assume that they could be seen without the DMM in series with it. One additional thing I haven't come to grips with yet was the need to invert the sign of the SMU current measurements.

There are two voltages, corresponding to the SMU current readings of (2) and (4):

The sign of the voltage is different for the two measurements, although the magnitude is similar. Qualitatively, there are two "sizes" of downward spike in the voltage- with only the larger of the two "sizes" impacting the current measurements.

The data from previous runs looks similar- but increasing the aperture time significantly reduced the number of spikes that are present in the current results. I have also not seen a current spike in an SMU-alone measurement with the 100 ms aperture time setting.

After I thought about how I derived the -21 Ohms, I realized that is was mostly accounting for the observed difference in measurements between the DMM and the SMU channel. I was attempting to calculate the impedance of the SMU measuring current.

The other suggestion that the NI-POWER help offers is to use a second-order DC noise rejection setting. Do you have experience with this setting?

Jeff

[Brandon_G.]

Hey Jeff,

In the spec, the noise vs aperture time graphs relates to the measurement noise of the SMU itself.  Normally, to characterize noise performance of the SMU, we will short the terminals of the SMU (apply 0V) and look at how much noise there is in the measurements.  Even with a perfect short you will still have some pk-pk noise in your measurement due to the inherent noise in the SMU itself.  In the case of an external signal applied that is not perfectly DC, increasing the aperture time will average the voltage (or current) over that whole period of time.

Therefore, in order to do a more apples-to-apples comparison of the DMM vs the SMU measure current, I would set the SMU aperture time to match the DMM itself.  If the DMM has an aperture time of 6x the SMU, then the spikes in the DMM measurement will get averaged out, making the spike less noticeable or not noticeable at all.  Make since?  I would set both aperture times to 16.66 ms so that we can test if both instruments will show the current spike at this aperture time. 

Was the SMU alone measurement (graph 2) you provided taken with a 100 ms aperture time or 16.6ms?  What about graph 4?  I'm trying to understand why the spikes may have shown up in graph 4 and not graph 2?  Were the settings the same?

Although decreasing DMM aperture time will give us a better compare between devices, it does not necessarily explain why the spikes are showing up in the first place.  Would it be possible to setup a test with just your DUT, external power supply and an external shunt?  You could measure the voltage across this shunt with a scope and then determine if the spike is present without the DMM/SMU in the system.  Since there seems to be spikes when transitioning from one current level to the next, perhaps you can induce these spikes by changing the DUT somehow?  You maybe able to set the triggering on the scope to see if you can capture any spikes.  

On the voltage graphs you provided is this the voltage across the SMU itself or the DUT?   What is the nominal voltage, 800 uV?  It's difficult to make out the scale from the picture, but if it's the voltage across the SMU, the voltage should be closer to -1.4 mV (0V +  -21*66.7u).  Perhaps this is the voltage measured somewhere else?  Also, if you're using remote sense, be sure to set your sense mode to 'remote', for proper voltage sense.

Since you have a resistance programmed on the SMU, a change in current will result in a change of voltage.  Therefore the voltage spikes would be expected when there is a current spike.  For now, I would program it to 0 for until we've identified the source of these spikes.  The variations in the current levels in the 4 graphs is due to 1. The DMM burden voltage (as you already mentioned) 2. Offset accuracies of the SMU/DMM itself. Performing self-cal on the devices and using the lowest current range possible, I suspect will minimize these differences.  Let us know what you find.

Thanks!

[Jeffrey_Zola]

For the data I showed yesterday, the SMU aperture time was set to 100 ms. The DMM configuration was not changed from the defaults with the 6 1/2 digits and 1 mA range. The only difference between graph #2 and graph #4 was the switching configuration in the matrix. The LabVIEW application is set up as a state machine that passes through an Initialization state that sets switches for the current bypass for each source and runs through the initialization sequence for the measurement instruments- which are not changed once the Initialize state has completed. There are different states for each measurement type, with the switching configurations and measurement VIs being the main differences.

I may not have sufficiently explained what the data represents- as there are no changes in current or any other setpoints. The stress machine is applying a constant voltage across six resistors and is measuring the resulting current using internal meters. There is a socket that allows access to the current path for external instrumentation. I used empty DIPs to interface with that socket. I soldered a 51 Ohm resistor to the supply-side pin with a wire soldered on each side of the resistor. These are my "DUT Supply" and "DUT Supply w Shunt" connections. I soldered a wire to the return pin, which is my "DUT Return" connection.

The tester is applying its voltage setpoint and measures the current thru each on a continuous basis. The NI system is intended to be a "fly on the wall" to make external verifications of the tester's internal measurements. The LabVIEW program, as I mentioned above, cycles through various states to take a measurement of the applied voltage (I did not describe the voltage connections in the previous paragraph- but they are present!) and resulting current for each of the six sources. There are three current-measuring states for the three measurement configurations, the voltage measuring-state, and a Wait state that counts down the time until the next periodic measurement sample. The current measurements are performed one-at-a-time with about 2 seconds per measurement.

The voltages I showed represent the measurements from the SMU that were taken from the same VI as provided the current data. The X-axis of these Excel pivot graphs show the six DUTs, and the data point collection time. The trend lines I explained in a post earlier this week. I am reposting the raw graphs as attachments here.

[Jeffrey_Zola]

Sorry for the double post:

I have a set of data using the 2nd order noise cancellation, which is attached here. The scale on the y-axis is much smaller than the yesterday's graph. Now we're talking!

My ultimate goal is to replace each module pair in my current setup (one PXI Switch module and one PXI DMM) with a larger switch module and one channel of the SMU. John's comment in message #4 of this thread makes more sense now. I will not be able to perform simultaneous measurements on different channels. Tried it.  Didn't work.

Jeff

[Brandon_G.]

Hey Jeff,

I'm glad to see the noise reduction on your measurements when using higher ordering filtering.  I do want to emphasis though that even though the measurements are now steady, the current spikes could still physically be present on your system- they are just filtered out in your measurement.

As I was thinking about this more I was wondering if you might not be allowing enough settling time in your switches.  If I'm understanding your test setup correctly now, it sounds like you perform measurements 1, 2, 3, 4 once every 2 seconds and therefore are continually opening/closing switches to perform each of these measurements?  I was originally thinking you were doing the same test 4 times over, leaving the switches static for each of the 4 measurement schemes.  Can you confirm if my understanding is now correct?  

If these switches are not static and you are sometimes seeing these spikes, I wonder if in some cases the switches haven't fully settled yet?  Are you using hardware triggering or software timing to control the switch and DMM/SMU measuring cycle?  If you are using software timing, I could see the possibility of intermittent spikes as the delay between throwing the switch and beginning your measurement can be varied from run-to-run.  You can easily test out this theory by adding a software delay or adding niSwitch Wait For Debounce.vi, if you haven't already.

From the measurements perspective the higher order filtering has cleaned up the spikes, but if you'd like to identify the source of these spikes, please let me know and we can continue troubleshooting.

Thanks!

Brandon G

[Jeffrey_Zola]

Thanks, Brandon. I am reasonably certain that the switches are settled before I take measurements.

My LabVIEW program has two controls on the Front Panel (I call them Switching Delay and Measurement Delay) that feed into ms Delay VI's. With the DVM+ and SMU HI S columns always connected to their read row and a bypass current path provided thru each DUT by connecting its DUT Return row to its DUT Supply column (with or w/o external shunt!), the switching sequence goes like this, repeated for each DUT:

1. Connect the read row to the DUT supply column & connect the DUT return row to the DVM- or SMU LO S column.
2. Flat sequence delay using the Switching Delay as input. (There are two current paths thru the DUT- thru the meter and thru the bypass)
3. Disconnect the DUT Return row from its DUT Supply column.
4. Flat sequence delay using the Measurement Delay as input.
5. Take the current measurement.
6. Re-connect the current bypass
7. Switching Delay
8. Disconnect the read row and the DVM- or SMU LO S.
9. Switching Delay

I have been using 500 ms for both delay inputs, but these are likely longer than they need to be. These were "round numbers" that I threw into the controls when I created them. Because I have only been taking "real" data on five minute intervals, the timing of slightly longer than two seconds per DUT has not been an issue.

I understand that there may be noise still present, but that I am filtering it out of my data. I am in a mode where I need to finalize a test setup to make external V/I measurements for as many sources as possible, and need to demonstrate technique in order to identify what hardware will need to be procured and what cabling an interface apparatus needs to look like. Once I have parts, etc., on order, I can return to refine the measurement setup further.

Thanks,

Jeff

[Brandon_G.]

Hey Jeff,

I agree that 500 ms is more than enough time to let the switch settle.  It sounds like there's something else going on there.  If we can help troubleshoot this down the road please let us know and we'd be happy to assist.  

Cheers!

Brandon G