SSR for low-current needs

[chronos04]

Hi all,

We are looking for a switch matrix for low current measurements. According to the "Switching Low-Level Signals to a DMM" technical note of NI, Reed relays are the best for that because of the low path resistance and lack of minimum current. It is said that SSR switches show a path resistance quite high. However, the PXI-2534 specifications report a path resistance of 1 Ohm, with not minimum current, and low thermal EMF (2uV). Is that correct?

Then, my question is whether this PXI-2534 switch matrix based on SSR relays would be as appropriate as the Reed PXI-2532b for low current measurements; or am I miss-understanding any specification? In general, would the PXI-2534 have any disadvantage with respect to the Reed counter-part (apart from the speed/bandwidth, which is not an issue for us)?

Many thanks,
Juan

[Now_With_Underscores]

Hey Juan,

What's the lowest current you want to measure?  You're not missing anything on the 2534 spec sheet... the path resistance row-to-column is 1.4Ω max.  Note the terminal block's path resistance is not included in this figure, but is likely less than another Ohm or so.  Our technical note generalizes relay types and thus doesn't apply across the board.  FYI: The 2533 is the same type of module as the 2534, but is 4x64 instead of 8x32.  Underneath the hood, both the 2533/2534 use Avago's ASSR-1510 SSR, which has low leakage, low parasitic capacitance, and is generally well suited for low current signals.  Let us know more about your signals and we can comment further.

[chronos04]

Thanks for the reply John.

We plan to use some row-column connections to measure current and others to measure voltage (measurements are to be done with the 4141 SMU). When measuring currents, we want to be sensitive to variations in the order of 10-100 nA, with the current to be measured flowing through a 1-2 kOhm resistor. In the case of voltages, measuring variations of about 50 uV should be enough.

Juan

[Now_With_Underscores]

My primary concern with using the 2533/2534 is that the SSR relay (ASSR-1510) has a rated leakage current of 100nA max at room temperature (500pA typical), but spikes up to 1uA across temperature...  per relay; multiple relays are tied in parallel, so expect leakage to multiply.  Note that said leakage is measured with a 60V signal applied across the opened relay; lower voltage means less leakage.  Alas, lower voltages increase capacitance, which will slow your measurement time. 

In general, switch matrices aren't the best choice when measuring 10nA sensitivity.  There's simply too much leakage, even with reed relays.  For example, an empirical worst case leakage test we performed on the 2530B yielded 254nA worst case leakage.  You would likely achieve better results, but NI can't warrant that performance.  You can make matrices work by adding guard circuits, offset compensation, etc, but really the best method is to directly connect the instrument to the DUT... unfortunately, this requires more instruments. 

If adding more instruments isn't an option, you can achieve less leakage with multiplexer configurations.  In our empirical tests - and while using guarding, offset nulling, and other low signal optimizations - we achieved sensitivity in the 10-100pA range using a PXI-2575 module.  This wouldn't be my first recommendation, but it is an option if you simply can't add more SMUs.

Tell us a bit more about your application (# of signals, # of simultaneous measurements required, sample rate, etc) and we'll see what we can come up with.

[chronos04]

It seems that we should find a trade off between flexibility and sensitivity, maybe relaxing one of our requirements.

The idea of our application is sequentially characterizing a set of devices that share the same chip. For each device we should apply three voltages and measure a current (which we can do with a single 4-ch SMU). Typically we can have about 32 devices per chip and we would like to measure the devices in an automated way, i.e. without changing any wire connection. For instance, we go to device #1, do a set of measurements, then switch to #2, etc. Speed is not a big concern because we may spend seconds/minutes measuring the same device; and the sampling rate for voltage/current measurements will not be higher than 100-150 ksamp/s.

Replicating the number of SMUs does not seem to be feasible...most likely we should employ the switch matrix for a pre-characterization with lower sensitivity, and then measure specific devices with higher sensitivity via directly connecting them to the SMU. From your reply, I understand that even going for a reed relay (e.g. PXI-2530B / 2532B) would not make it much better compared to the SSR PXI-2533/2534. Using multiplexers would reduce the connection flexibility, but maybe we should also consider them and reduce the number of parallel devices to be measured.

Let us know if you come up with a better idea.

Thanks,
Juan

[Now_With_Underscores]

Hey Juan,

The specified noise accuracy on our FlexSMUs (PXIe-414x) is only valid when the aperture time is 2 PLCs (33ms/S@60Hz).  Thus, sample rate is going to top out at well under 30S/s.  For example, on the PXIe-4141, an aperture time of 5us per sample is going to add ~2000ppm range error (~20nA additional noise uncertainty).    You'll need to slow down your measurement speed to reduce system noise.  See Figure 2 in the 4141 specifications.

Another option is to build dedicated current sense circuits into your test PCB (e.g. with a current controlled voltage transconductance amplifier) and then buffer each output voltage into your matrix.  You could calibrate your test PCB's circuitry periodically (e.g. with a PXI-4071, SMU, etc, depending on required sensitivity). 

Barring a custom design, I agree that the best method to keep the initial sensitivity requirements is to matrix all the less sensitive rails you can, and then have a dedicated SMU channel per sensitive rail.  If you can loosen the specs, then a matrix becomes more feasible.