How do I get rid of cross talk between 2 channels on the PCI E series DAQ?

[DonS]

I have a PCI E series with a BNC-2090. When I connect my signal to ACH0 and ACH1 in SE mode and read both channels in continuos mode I get cross talk between the channels. If I read each channel seperatly I have no problem. What can I be doing wrong, please help.

[Filipe A.]

Hello;

There are a couple of thing that can be the reason for that behavior:
1) Both signals have amplitudes that vary a lot, one related to the other, so the instrumentation amplifier of your DAQ device is having problems to settle to its values. The solution for that problem is to increase the interchannel delay in between the channels. You can do that both in Labview and NI-DAQ function calls.

2) There is some ground loop in between both signal sources that is changing the behavior of the analog inputs. You can try to hook up both signals in differential mode to double check if that is the reason for the problem.

Regards
Filipe A.
Applications Engineer
National Instruments

[linm]

How do I go about changing the delay between input channels?

[Salvador_Santolucito]

Hi,

The analog input acquisitions run off of two clocks:

1) The sample clock is what you specify as the sample rate. This is the clock period between each data point on each channel.

2) The convert clock is what actually runs the ADC. The convert clock rate is what sets the limit for analog input acquisitions. For 4 channles running at 100 kHz, the convert clock rate is 400 kHz (4 convert pulses for each sample clock pulse). For each pulse of the convert clock, the ADC converts the voltage to digital data. The distance between each convert clock pulse is the time delay between channels.

By default, DAQmx uses the slowest possible convert clock rate. The convert clock pulses are evenly dispersed between sample clock pulses. This means that for a 1 kHz sample clock on 4 channels, the convert clock rate would be 250 Hz. However, you can specify a faster convert clock rate if you want less inter-channel delay. The max convert clock rate is the max sampling rate of the board. If you choose a faster convert rate, you will have four quick convert pulses followed by a pause until the next sample clock pulse occurs. You can NOT specify a slower than default convert clock rate. Sample clock period must be greater than or equal to the number of channels times the convert clock period.

I have attached a picture of how to specify a convert clock rate.

-Sal

[linm]

Filipe suggests to increase the interchannel delay to reduce noise from settling problems. If DAQmx already defaults the delay to be as long as possible (sampling rate times #channels), it seems that doing anything else to the convert clock would make things look worse. Is this true?

My specs:
16 channels
continuous sampling at 500Hz
this corresponds to convert clock of 8KHz (16 x 500) is this correct??
PCI-6025E
BNC-2090 used to interface analog inputs with 6025
latest LV & Daqmx on XP

My specific problem:
the BNC-2090 has 2 rows of 8 BNC inputs. If I input signals from rows right above another, there is significant crosstalk. For example, if I have a 5V TTl on input 1 and 0.5V TTl on input 9, the 0.5V gets pulled to become 2V.

The switches on the BNC-2090 are set to Single Ended (SE) with Non-Ref Single Ended (NRSE).

Operating at RSE does the same thing and there is no signal degradation operating using only the top or bottom row.

Any ideas? Thanks.

Message Edited by linm on 03-04-2005 03:28 PM

[Salvador_Santolucito]

Hi,

You are correct that you can only make the convert rate faster than default. However, when it comes to signal noise, convert rate is usually not the problem.

I would look at your wiring and termination. We have a great online tutorial that appears to be temporarily down. I will look into it and see if I can post it.

-Sal

[linm]

---

I would look at your wiring and termination. We have a great online tutorial that appears to be temporarily down. I will look into it and see if I can post it.

-Sal

---

Sal, that would be great...I'm looking forward to your reference.

[terpsho]

I have the same problem at my PCI-6025E.

I have modified my input configuration to see if I can improve it.

However, the change does not improve too much about the cross talk.

In my previous setup, my measurement source is a grounded signal source; but my input configuration is a single-ended ground reference connection.

It should have some random noise from the common ground according to the document: (Is it the reference you guys talking about?)

http://zone.ni.com/devzone/conceptd.nsf/webmain/177A8B29FEDC0F5886256FA90083C0F8.

Before I make my change, the power spectrum of the noise is below -70db, which is around 1-bit error, in my old input configuration.

I did my measurement for 45 seconds (100 samples per second).

Message Edited by terpsho on 03-21-2006 08:24 PM

[Michael.Phillips]

Hi Terpsho,
I recommend taking a look at the recommendations in this KnowledgeBase article to reduce ghosting in your measurements.