Dependent signal

[AndrewRF]

Dear NI,

I have a problem with pressure measurements. I have 16 vacuum gauges from 26PCCFA6G (http://sccatalog.honeywell.com/pdbdownload/images/26pc.series.chart.5.pdf) and PCIe NI-6363 DAQ. Since gauge have bridge scheme (Vcc, GND, Output_A, Output_B),  the only way to read signal from each gauge -- is to use the Differential regime of DAQ. Otherwise, gauges will have 3 common points (Vcc, GND, Output_A) and readings will be dependent.

So I connect two (for test, I need all 16) gauges to ai_0 and ai_1 input at DIFF regime via SCB-68 connector block as shown here http://www.ni.com/pdf/manuals/372551a.pdf p.4-6, floating signal sources. Balanced resistors are 13 kOhm each. Test conditions are next:

-DAQ continuous sample rate is 5kS/s (500kS/s), block size is 1k (100k)

-acquired data  block is averaged and mean value is shown on the screen.

-two pressure value at gauge ai_1: 0.01 atm and 1 atm, only one value of pressure at gauge connected to ai_0:1 atm

If the gauges outputs are independent, signal ai_0 must be constant. And it is in case of low sample rate. But in case of high sample rate (from 60 kS/s to 500kS/s) the signal of ai_0 is changing while pressure changes on gauge connected to ai_1.

Please, see attached files for clearance:

1.png: p0=1 atm, p1=1 atm, SR=500kS/s (at 5kS/s the mean values are the same, I can attach only 3 files to the post)

2.png: p0=1 atm, p1~0.01atm, SR=500kS/s

4.png: p0=1 atm, p1~0.01atm, SR=5kS/s

So, at low sample rate (SR) signals is almost independent (see mean-Dev1_ai0 value), while at high SR they are dependent. Is there any way to fix this problem? Is the problem at gauges, since they have 1 ms response time? But measured pressure is almost constant at time during facility test run...

Thanks for any advise and sorry for my engrish 🙂

[johnsold]

AndrewRF,

I can see several possibilities for explaining some of your results.

1. On the PCIe-6363 differential channel 0 is connected: AI 0+ = AI 0 = pin 68; AI 0- = AI 8 = pin 34. It may have been a language issue, but it was not clear that you heve things connected correctly.

2. What is the power source for the pressure sensors? Is the common or ground of that power source connected to AI GND of the DAQ device?

3. The resistors of the sensor should be adequate to provide the bias currents (assuming the power source is properly grounded as mentioned above). The 13 kohm resistors may add more error than they prevent.

4. Sampling faster than the response rate of the sensors does not get you any new information and may effectively increase your noise bandwidth.  If your pressures are changing faster that a few hundred hertz, you need a different sensor.

5. Look at the CMRR vs frequency graph for the 6363 (page 2 of the specifications). The CMRR is down to ~50 dB at 100 kHz. From the shape of the curve it will likely be lower than 40 dB at 500 kHz.  A 40 db CMRR results in a 50 mV error with a 5 V common mode signal.  Also look at the settling time. With the source impedance of the sensor (2500 ohms) the settling time is several miroseconds. This will contribute error to the measurements at higher sampling rates.

Lynn

[AndrewRF]

Dear johnsold,

thank you for your reply. I will answer to your questions below:

1. LabView SE has Connection diagram, so it is very difficult to make a mistake 🙂 Yes, ai_0 connected to 68 and 34.

2. Source is a AC/DC adapter (based on 50Hz transformer and LM317-like linear regulator). V- of power supply is not connected to AI_GND and this may be the issue...

3. I have tested gauges and DAQ, when R_b and R_d are absent (or their resistance is infinite). The result is the same. Output resistance of gauge is 1.5 kOhm, so I take Rb and Rd ~ one order greater...

4. Pressure is almost constant during test run. In described conditions the pressure is constant. The reason to use high sample rate -- is to get more data. I believe, that averaging of data can decrease the measuring error due to environmental noise. Please note, that in desirable test run I need to use 16 gauges (about 60 kS/s for each gauge). Of course, it is necessary to decrease that noise in other ways: decreasing the wire length, shielding, proper power source etc. I believe that I already done all possible to decrease noise. The averaging is a question.

5. I believe that CMRR is not an issue. May be I wrong, but the test gives the same mean values of gauge signal at 5 kHz and 500 kHz (of course, when the pressure is constant at gauge 1 and gauge 2). Settling time can be the issue. If so, there is no any way to make the gauges independent at high sample rates...

What are you think about voltage range? In my tests it was [-10;10] V. Gauge signal full scale span is 100 mV, so the proper range is [-100;100] mV. If the described at 1st post effect is instrumental, will proper voltage range make the results better (I mean decreasing the influence of one gauge on another)?

[johnsold]

AdrewRF,

2. It is absolutely essential to have all the devices connected to a common ground.  If they are not connected properly, then currents may flow through parasitic impedances resulting in strange and sometimes destructive behavior.

3. With one order of magnitude between source and bias resistances errors can be on the order of 10%.  They will likely be smaller if the resistances are well matched, but if you need them go for two or three orders larger (up to about a megohm).

4. High sample rate means wide bandwidth.  Wide bandwidth means larger noise (increases as square root of bandwidth).  I am not sure you gain anything by samping faster than the sensor bandwidth which is about 1 kHz.  What is the nature of the noise? Is it broadband white noise or some kind of interference such as from power line or switching power supply/PWM motor drive?  There may be better ways than just averaging, depending on the kind of noise.

5. High source impedance and rapid channel switching can produce the appearance of coupling between channels.  This is often called "ghosting."  Waiting for the signal to settle before conversion or driving the DAQ device with a low impedance external amplifier are the only fixes.

Voltage range: You have two things to consider: Common mode voltage and differential voltage.  You did not specify the excitation voltage you are using for the gauges.  Let me show the effects with two examples.

Example 1. Excitation = +10 V and ground.  Common mode voltage = 5 V. Differential voltage = +/-100 mV. This means that the voltage at the AI inputs is in the range [4.95;5.05] V. So you need to use the [10;-10] V range.  You cannot use  a 5 V range because the maximum input voltage is 5.05 V.

Example 2. Excitation = +5 V to -5 V. Total voltage across gauge is 10 V so the full scale signal is still 100 mV. Common mode voltage = 0 V. Differential voltage = +/- 100 mV. The voltage at the AI inputs is in the range [-50;50] mV. Now you can use the 100 mV range.  This is equivalent to gaining 7 bits of resolution on the conversion.  Note that if you use standard fixed +5 V and -5 V regulators the output tolerance is about 0.5 V so you may have a common mode voltage up to about 0.5 V in the worst case.  You may need to trim the voltages with adjustable or tracking regulators or use a slightly larger input range on the A/D converter.

Lynn

[AndrewRF]

Dear johnsold,

thank you for answer, it is very helpfull for me!

[AndrewRF]

Dear NI Community,

do you have an expirience in Honewell pressure gauges 40PC series?

What type of gauges is better to use for NI-6363 device (concerning "ghosting")?

26PC - http://sensing.honeywell.com/index.php?ci_id=49833 (we already use it)

40PC - http://sensing.honeywell.com/index.php?ci_id=138832 (we are thinking to buy it)

As it has been written above, we use 16 gauges with NI-6363 and cannot get readings faster than ~0.1 kS/s due to "ghosting"

[johnsold]

I have used the 26PC series devices. They are nice sensors but work well only with a suitable signal conditioning circuit. Notice that the signal is in millivolts and the source impedance is ~2500 ohms. Both factors make these devices unsuitable for connection directly to the NI-6363.

The 40PC devices have internal amplifiers. The output signal is several volts at low output impedance. These will connect nicely to the NI-6363. Note that each sensor requires up to 10 mA of power supply current.  Make sure that the source can supply enough current for all the devices you plan to use. Note also that the output is single ended so you do not need differential connections to the DAQ device. You can get twice as many channels than with the 26PC series. You should not have much, if any, ghosting with the 40PC devices.

Lynn

[AndrewRF]

Thanks!

1. Does anyone know impedance value for 40PC? Maybe, someone has estimated it...

2. Hm.... My expirience in NI-6363 usage: at LV SE 2010 for this board 32 channels can be selected both for differential and single mode. I cannot prove this since i have only 16 gauges. Maybe its just software "bug". Nevermind...

[johnsold]

The output impedance does not appear to be specified but typically amplified outputs have relatively low impedances.

The NI 6363 has 32 analog inputs which can be configured as 32 single-ended or 16 differential channels.

Lynn