DC crosstalk between PXI 6220 channels due to high source impedance

[raptrick]

Dear all,

I’m trying to digitize 10 single ended signals simultaneously with the PXI 6220. The source impedance for each signal is 1.5meg which should not be problem since the input impedance is >1G. Bias currents are taken in account and give only an offset error (assuming that they are constant). Crosstalk according datasheet is -75db at 100khz which is fairly okay for my app however the source impedance is not indicated for that specification. When I acquire data (Dc voltages) from the PXI6220 i see unexpected high crosstalks (again source impedance 1.5meg). The sample rate is 200Hz and I connected only one AIGND pin to the reference of the system. I verified whether the GND is okay by connecting one input pin directly to 5V (Rout=0) . This channel indicates 5V while the others (Z=1.5meg and grouded) give non -zero value of voltages (2V). DC leakage between channels in the cabling is not verified yet.  

1)      Does anyone have a clue why I observe so much DC crosstalk at high source impedances.

2)      Does the PXI6220 digitizer have 16 parallel channel or does it make use of an multiplexer.

3)      Do I have to connect all AIGND’s to my system’s reference or is just 1 okay.

Patrick

[raptrick]

See measurement results. You can see cross talk between adjacent channels due to an unbuffered multiplexer.

[6.97 1.27 0.459 0.33 0.203 0.147 0.115 0.0652 0.645 0.289]

[-0.00125 6.97 1.98 0.69 0.293 0.335 0.27 0.185 6.97 2.63]

[0.000369 -0.00774 6.97 1.97 0.728 0.305 0.211 0.0869 0.152 0.0788]

[-0.00125 -0.00936 -0.0272 6.97 1.96 0.599 0.36 0.157 0.143 0.122]

[-0.000604 -0.00871 -0.0262 -0.0252 6.97 1.96 0.554 0.303 0.123 0.0928]

[0.000369 -0.00709 -0.0239 -0.0243 -0.0408 6.97 1.9 0.57 0.174 0.188]

[-0.000928 -0.00903 -0.0265 -0.0256 -0.0405 0.0247 6.97 1.9 0.978 0.544]

[0.000369 -0.00839 -0.0262 -0.0252 -0.0415 -0.0233 -0.0288 6.97 0.603 0.35]

[0.00102 6.97 1.98 0.689 0.292 0.336 0.271 0.186 6.97 2.62]

[-0.000604 0.0289 -0.0129 -0.0162 -0.0356 0.0075 0.797 0.221 0.176 6.97]

Conclusion: the specs are a bit misleading to my opninion. There is no recommended source impedance spec or any equivalent input circuit. No source impedance is specified for the crosstalk performace.

Patrick 

[johnsold]

Patrick,

I think the culprit is probably the multiplexer and not the amplifier input impedance. Look at the settling time graph on page 2 of the speicifcations. At 100 ppm error (which several source impedance curves cross) the settling time increases exponentially with the source impedance. 4 us at 1kohm, 6 us at 2 kohm, 13 us at 5 kohm, and 25 us at 10 kohm.  It is probably not a reasonable to extrapolate that data more than two oreders of magnitude, but it is not unreaonable to expect some problems even at your slow sampling rate.

The user manual has this to say (pg 4-7):

---

 M Series devices are designed to have fast settling times. However, several

factors can increase the settling time which decreases the accu racy of you r

measu rements. To ensu re fast settling times, you  shou ld do the following

(in order of importance):

1. Use Low Impedance Sources —To ensu re fast settling times, you r

signal sou rces shou ld have an impedance of <1 kΩ . Large sou rce

impedances increase the settling time of the NI-PGIA, and so decrease

the accu racy at fast scanning rates.

Settling times increase when scanning high-impedance signals du e to

a phenomenon called charge injection. Mu ltiplexers contain switches,

u su ally made of switched capacitors. When one of the channels, for

example channel 0, is selected in a mu ltiplexer, those capacitors

accu mu late charge. When the next channel, for example channel 1, is

selected, the accu mu lated charge leaks backward throu gh channel 1. If

the ou tpu t impedance of the sou rce connected to channel 1 is high

enou gh, the resu lting reading of channel 1 can be partially affected by

the voltage on channel 0. This effect is referred to as ghosting.

If you r sou rce impedance is high, you  can decrease the scan rate to

allow the NI-PGIA more time to settle. Another option is to u se a

voltage follower circu it external to you r DAQ device to decrease the

impedance seen by the DAQ device. Refer to the KnowledgeBase

docu ment, Decreasing the Source Impedance of an Analog Input

Signal , by going to ni.com/info  and entering the info code rdbbis .

---

A buffer per sensor is the best fix.

Lynn