Different voltage measurements from various NI devices, why?

[querty999]

I'm getting stable, noise free, voltage readings from an NI-9219 but noisy readings of the same signal using PCI-6251 and USB-6221. Been fighting this problem for days. I'm about out of ideas. Please help.

The noise is not small. The measured signal has a normal range of +/- 10 volts. The noise level is several volts for both the 6251 and 6221. Noise is virtually non-existent with the 9210. The measured signal is coming from a BNC output on a material testing machine. (The signal represents force.) 

I have attached a tee connector to the BNC output of the machine. The tee divides the signal so I can attach two NI devices at once. The 9219 readings are rock steady but the other two devices are all over the place while measuring the same signal at the same time.

One difference in the NI devices is the multiplexer. The 9219 has no multiplexer. The 6221 has a multiplexer. Not sure about the 6251 but I feel sure it does.

I need to get stable readings from the 6251 because it's part of a third party system. I can't use the 9219 instead.

Any ideas?

[querty999]

Here's some additional information.

All three NI devices give virtually identical noise free readings when measuring the voltage on a 1.5 volt battery.

Noise is present with the 6221 even when connected to a laptop computer with no external power supply. The 6221 is powered from the wall outlet however.

[querty999]

Documentation for the 9219 says is has built-in filtering. Documentation for the other, more expensive, devices do not mention filtering. Could the 9219 filters make the difference in the noisy readings? Should I use be using an external filter with the other devices? Where do filters come from?

Documentation for the 9219 shows a mux on every channel, according to the block diagram. Since the 9219 has an ADC on every channel why does it need muxes?

9219 documentation also says the ADC's are "isolated". I wonder if this means optical isolation? Documentation for the other devices doesn't mention isolation of any kind.

The only drawback of the 9219 appears to be sampling speed. But it looks like the other devices sacrifice a lot of useful features for speed. I don't even see how the 6251 and 6221 are useful with all the noise they transmit.

[crossrulz]

What other signals do you have in your task?  This is sounding like a ghosting issue, where you are getting parts of another signal in this reading due to a high source impedence.  You may need a simple voltage follower circuit to act as a buffer and give the DAQ a low impedence source.

The 9219 has ADCs that are isolated from each other, so you will not see the affects of ghosting.  And those MUXes are to handle the many types of measurements that DAQ can measure.

[Henrik_Volkers]

If you look into the spec of the multiplexed DAQs usually the 2. or 3. figure will show the settletime error.

The error is a function of source impedance and effective sample time and has a log scale on the error.

The noise specs are usually measured with a single channel  only , without multiplexing 😉

Common effect on unbuffered multiplexed DAQs.

To reduce these effects:

• avoid multiplexing or
• reduce the sample rate and/or
• insert a buffer (amplifier)  with low output impedance

The 'trick' to read a GND line inbetween channels, to reduce ghosting/channel crosstalk is questionable, lowering the samplerate has in my experience the same effect.

[querty999]

---

wrote:

What other signals do you have in your task?  This is sounding like a ghosting issue, where you are getting parts of another signal in this reading due to a high source impedence.  You may need a simple voltage follower circuit to act as a buffer and give the DAQ a low impedence source.

Querty999:  I only have one signal in my task. I tried reducing the sample rate to just a few samples per second. It didn't help. Are unity gain amps all constructed from components, or can I a buy one ready to go?

The 9219 has ADCs that are isolated from each other, so you will not see the affects of ghosting.  And those MUXes are to handle the many types of measurements that DAQ can measure.

---

[querty999]

---

wrote:

If you look into the spec of the multiplexed DAQs usually the 2. or 3. figure will show the settletime error.

The error is a function of source impedance and effective sample time and has a log scale on the error.

The noise specs are usually measured with a single channel  only , without multiplexing 😉

Common effect on unbuffered multiplexed DAQs.

To reduce these effects:

• avoid multiplexing or     Querty999: How to do this? Is there a way to bypass the mux?
• reduce the sample rate and/or  Querty999: I tried reducing sample rate. No joy!
• insert a buffer (amplifier)  with low output impedance  Querty999: Are these available fully assembled? Or just made from components?

The 'trick' to read a GND line inbetween channels, to reduce ghosting/channel crosstalk is questionable, lowering the samplerate has in my experience the same effect.

Querty999: "read a GND line inbetween channels"?? I don't understand.

---

[adena.l]

Hi querty999 - 

What exactly does your physical setup look like? How many channels are you using/how many voltages are you trying to measure? How do you have everything wired into your different devices?

[querty999]

Adena,

I am trying to read a voltage that represents force. It signal comes from a material testing machine, called a tensile testing machine or UTM. The machine pulls on a sample until it breaks. The amount of "stretching" (strain) and force determine the characteristics of the material.

The UTM has a BNC jack on the back. The force produced by the machine appears on the BNC jack as a +/- 10 volt signal. I need to measure this signal precisely to record small changes in force.

I have attached a tee adaptor to the machine's BNC port. The tee has BNC connectors on all three legs. To one side of the connector I attach the NI-9219. The other two NI devices mentioned previously are attached to the other leg of the tee, one at a time.

The 9219 doesn't have a BNC connector. It uses spring terminals, so I have attached a BNC splitter. The splitter is is a BNC connector with two wires that I attach to the 9219.

The other devices have BNC terminals so all I need is a cable with BNC connectors at both ends.

I need to make this measurement using the PCI-6251 device instead of the 9219 because the 6251 is part of a third part application. However I find the noise level of the measurement using the 6251 is too high (20mV). I can't read small force changes with that much noise.

I measured the signal with the 9219 just for comparison. The noise amplitude is only 1mV. I need that kind of low noise for the third party application work properly. The 9219 has shown that low noise measurement is possible, but maybe not with the 6251.

In summary, I need a low noise reading from the 6251 like I'm getting from the 9219. But apparently that's not possible, even though the 6251 is supposed to be a more sophisticated device. The 6251 has a much higher sample rate. But having more bad readings at a high sample rate is not useful.

I am measuring only one voltage.

[Henrik_Volkers]

SO you need to measure withthe 6251 , that read a isolated  1.5V battery fine.

The 9219 is well isolated and filters and reads fine.

Seems that all points to a EMC challenge 🙂

General:

• Use a differential setup for the 6251.
• Provide a symmetric bias current path . (a 1Mohm resistor from AI+ and AI- to AGND) (I assume a low impedance source, however it doesn't hurt 😉 )
• Read 300ms of data at 1MSPS , do a FFT , how does the spectra look like?

How is the Force sensor connected?

Type of sensor and signal conditioner?

How are they powered? 

Isolated mounting of the sensor?

Housing connected to maschine??

Every single contact/wire can be important. You can post some data 🙂

Try the DAQmx  example 'Voltage - Finite input.vi'  , samplerate 1M, 300k sample,  save the TDMS file, and make a copy of the frontpanel for the settings.