Handling noisy DUT with SMU (PXIE-4139)

[alexklrc]

Hello.

I am trying to use a PXIE-4139 Precision SMU, a switch matrix, and an isolated power supply (in a PXIe chassis), to build a calibration/test fixture for a noisy DUT. I need both differential signals and single ended signals, and a high amount of flexibility as it goes through a switch matrix (which prevents me from directly hooking up SMU LO to earth/gnd). Unfortunately the DUT presents a large noise source as it has several noisy switching supplies. The noise appears to be in the 100 kilohertz range up into the single megahertz. Also, the device has ferrite beads on the ground return paths and some signal paths, and moderate filter capacitance (0.1uF with no series R) on several signal lines.

One important fact is that all the signals of interest are essentially DC, so aggressive filtering is possible if it does not disrupt the SMU.

The ferrite beads in addition to the noise and the capacitance creates large DC offsets both for current and voltage ranges with the PXIE-4139 SMU. A small resistor to isolate the SMU and tying LO/LOs to a noisier but lower inductance ground also helps.

What I want to figure out though, is what is the best way to leave my SMU essentially floating against the DUT, and selectively connect LO/LOS, HI/HIS to this noisy DUT through a switch matrix while doing my grounds and shields correctly. I am concerned that adding an LC filter or a common mode choke to the SMU will disrupt the control loop, and I am not certain where to attach filter and isolation components, and where to join up sense and force lines depending on these filters.

My first rough idea would be to have an intermediary PCB with split ground planes, put common mode chokes /differential RC filters on every critical signal pair to bridge the gap, and add isolation resistors where possible:

The differential signal would also have bleeder resistors to the noisy ground plane so that the common mode limitations of the DUT and SMU are respected.

Thanks.

[santo_13]

A great feature of 4139 is the source adapt, where you can adjust the control loop settling by setting the tuning parameters to suit your exact load, so, ideally, you should be able to tweak the SMU control loop to work the best with your external load (effort to find the best setting may take time).

Connecting the LO to DUT GND is the best way to keep the supply floating as long as you're not exceeding the Earth-LO differential limits.

Overall, your simplified schematic should work, as backup, develop your first PCB with additional paths to allow you to play with components to get the best performance.

[alexklrc]

Unfortunately the two main ways I have to design this are fairly incompatible so each has to be their own design. 

1. Split grounds, every single signal individually filtered with common mode choke. (in screenshot). Lots of space, but most aggressive filtering and potential for issues. Also, the noise is a super wide range of frequencies which chokes struggle with on the low end.

2. Solid uninterrupted ground planes and fills connecting all grounds. Shields (which are connected to earth ground by the SMU chassis) connected at a single star point with a large low inductance bond to ground. Some physical separation between the noisy side and quiet side, though all the signals route to the "quiet" side anyway, and even through a matrix. Basic RC filtering on positive ends and differential lines, but all the LO connections are connected directly to the ground planes/fills.

I am doing 2. for now as it's much simpler if the SMU does not misbehave. The offsets when breadboarding at least are very large, and it does not appear that SourceAdapt tuning removes them. Unfortunately - until I design the board I am not sure I could accurately simulate the effects of having a solid ground plane and shielding for noise reduction.