PCIe-6341 Adding a DC Blocking Capacitor Troubles

[KeithB]

LEt's get to the point...

I have a 0.6V differential signal from a low impedance, electrochemical device.....we want to look at some noise measurements up to 10 kS/s......to increase the resolution, I was thinking of a simple RC high pass filter added to the circuit to remove the DC component and use the +/- 0.2V range instead of the +/-1V range; see attached picture for basic layout of channels. I started w/ differential channel measurements, even adding a grounded channel in between the channels with different ranges if I was going to multiplex hoping to improve settle time.

The circuit simplified.... high current power supply @ 2.25V in series w/ the fuel cell (to boost V for load bank) and a low ohm custom shunt resistor, attached to the load bank. The load bank is in remote sense (S+,S-), not sure if this is cause of all my problems, i.e. the inputs are ground(ed) referenced...

1. Multiplexed mode...scanning channels 1, 2, 3, 4 repeat, up to 10 kS/s.

1a. I started with a 400k resistor on Ch4, but in multiplexed mode the channel output started at 0, but then in the first 1-2 second would drift up 60mV and almost stabilize out.....and then match the Ch1 signal pretty well (minus the bias).

1b. I then figured, I drop the RC resistor a bit to increase the current through the cap...but not really understanding the influence of what is occuring inside the DAQ card. So at the end of the day I have a 90k in the circuit.....a little drift up but only recording 10 second windows, so i can record for 15 and delete 5 seconds.

But, I have this friggen 0.05 mV p/p perfect 60Hz sine wave now showing up....which possibly was dampened out at 400k, but more visible at 90k (see attached files, sorry made this quick).....also messed with the AI- to GND resistor a bit...lowering that made a bigger 60Hz signal added.

2. Realizing the multiplexed measurement is screwing with the capcitor...I ran a single channel scan....nice and flat; no drift up, but still have the 60Hz component from somewhere.....only place I can think is if the inputs into the Load bank S+ and S- are ground referenced?.....engineer, but not electrical so am a bit stumped.

A colleague helping out was on the try and see, teaching...even recommended trying some NSRE with the AI sense connected to the fuel cell negative terminal...but I thought I'd post first.

So is using this layout even possible, just to drop the signal above 10Hz from the +/-1V range to a +/- 0.2V range using an RC high pass filter? I can live without the multiplexed measurement, but this 60Hz signal boggles me....its a ceramic capacitor, no polarity.

Shoots, any help would be appreciated....maybe 60Hz always there, I'm just pullin it up...

Keith

[johnsold]

KeithB,

Does the load tolerate 2+ V between the sense and power terminals? Electronic loads typically use internal circuitry which is similar to regulated power supplies except for the direction of current flow.  Most power supplies with remote sensing only allow a few tenths of a volt between the sense and power terminals.  The spec sheets for the RBL488 do not indicate any limit.  The behavior outside the limit may be undefined.  The manual does emphasize that all signals and measurements in the BRL488 are referenced to S-.

Grounding intermediate channels in a multiplexed system does not reduce settling time.  It does tend to reduce ghosting, which results from charging capacitors in the multiplexer through high impedance sources.  That should not be an issue in your system.

You are making three measurements and all can be referenced to the negative end of the cell (and S- on the load).  I would probably go to single ended measurements.  The resistor of the high pass filter will be the one from AI4+ to AI Gnd.  Also connect the chassis ground of the 2.25 V power supply to AI Gnd (after verifying that the load can float above ground).

Then I would start looking for the source of the 60 Hz.  Does its presence change with the current? What happens if you operate at low current without the DC power supply? Are there other devices near the test setup which use a lot of power? Are the computer/Daq system, load, and power supply all connected to the same AC branch circuit? What happens if you use a resistor for a load rather than the electonic load?  Your diagram indicates that the internal resistance of the cell is 2 milliohms. You are measuring about 0.5 mV of 60 Hz which suggests 250 mA flowing through the cell or inductive pickup in the loop of wires containing the high pass filter.  Only 3 nA need flow thorugh through the two 90.2 kilohm resistors to generate 0.5 mV.  

I do not understand why you are seeing the high pass filter transient decay to a steady state value which is not zero (DC). Ground loop is a likely suspect since you have mV of DC and high currents elsewhere.

Lynn

[KeithB]

Thanks for the tips; I am going to go try the NSRE referenced to S- location today; this isn't a huge priority as we really don't need uV resolution which is essentially system noise anyway.

I can't say for certain, but we've been using these loads in this mode for years, with no problems; The V boost from the powersupply is just necessary for the high current/low voltage operation.  I do have power supplys which specify very low differences between the remote leads and the lugs at the unit itself, but as you said I never saw this in the load bank manual also.

After looking at the tech's setup, there was an additional cable after adding the shunt, that happened to pass 2 inches from a fan attached to the cell for cooling.....These fans are not always running, so that might be the intermittent source. Even further, considering the entire run of cables in the station, there's really no way to avoid some inductive pickup off the AC power...of which all units are connected to the same AC lines.

Either way...I appreciate you taking a look in detail...I'm having to break out of the 'black box' mode; frustating, but enlightening as well. I'll post an update after trying the NSRE approach.

[KeithB]

Oh yeah, DAQ wires are shielded.

NRSE referenced to S- removed the transient decay and the value settles around zero.

Interestingly, I'm still picking up the 60Hz on the DC blocked channel...but we're running power spectrums so this should be easily recognized.

Thanks again.

[johnsold]

KeithB,

60 Hz interference can be difficult to eliminate completely.  If the pick up is inductive, shielding will not help much.  Magnetic fields penetrate the shields and couple to the wires inside.  Superconducting shields might work but the liquid helium budget probably will not get approved by Accounting.

Tightly twist the wires in the high pass filter circuit.  Magnetic coupling is related to the area of the loop enclosing the flux.  Twisting reduces the area.

If you can separate the 60 Hz in the analysis and it does not cause dynamic range or non-linearity problems, that may be easier than trying to eliminate it.

Lynn

[KeithB]

Thanks again for the advice.

K