Reading Load Cells with BRX

[pbw]

I'm automating a cement Batching Plant, that has 3 individual scales, with 4 Load Cells each.
I decided to try and read the load cells with a BX-08UT. I set it up in the office with a small weigh bar (10,000 Lbs) and was blown away by the results. It was accurate and pretty solid, so I decided to build the system.
Today I tried to hook it up to actual load cells with the supporting frame, (4x500,000 load cells) and the readings were all over the place, and no amount of filtering will take it out.
This is a noise free system (no motors running or valves switching yet) and my power supply is rock solid.
I hooked the load cells up to and old readout and it was very stable, to rule out the Load Cells or outside conditions.
What's weird is that there are occasional spikes which can be easily filtered out, but about 10-20 seconds after startup I get a huge spike that last for 5 seconds then it will occasionally fall back and the cycle continues.
Am I missing something?
Shouldn't this setup work theoretically?

[Controls Guy]

I have successfully taken load cells into a Direct Logic (not BRX) thermocouple module, but it was a really really bargain-price project.  For a normal project, I'd use a scale head with analog or Ethernet output, or else a combiner with 4-20ma transmitter.  Fact, I'm doing just that on a current BRX project.

[pbw]

Thanks for your reply,
What Brand and Model are you using?

[Controls Guy]

Interface.  Will get models on Monday.

[brucek]

I've done this both ways. Serial in from a weight scale and 3 load cells to a summing card and then to a signal conditioner. There is a program I got somewhere that had examples I believe that takes you thru the serial part to strip just the weight out. I think you just have to figure out how many words to strip to get to the weight. The signal conditioner just goes to an analog 0-10v or 4-20ma. The serial is probably the easiest here is a txt of it you just have to import.

[Garyhlucas]

If it were me I'd be looking at solving the apparent noise issues first because adding more hardware to mask the problem is likely to bite you later.  That big spike has to be caused by something.  I'd be looking at shielding and grounding issues. Are all your shields grounded at only one end? If the load cells are connected to the cable shields and they are mounted on metal then you have ground loops. 

Safety grounding for power is all about creating as many ground loops as possible so no two adjacent pieces of metal have a voltage potential big enough to harm you.  Because every piece of electrical equipment in your plant leaks a little into the grounded metal there are circulating currents always present that you are not aware of and they change as things start and stop.

[pbw]

Thanks for your feedback guys, I tried all kinds of filtering techniques to no avail, as a last resort I placed a 220uF bipolar cap on the input terminals and that fixed my issue.
I calibrated it with a 3,480Lbs weight and it's accurate all across the range and very stable.
I'm thinking maybe switch to a tantalum cap for reliability.
I left provisions in the panel to add a indicator and use Ethernet to interface if this doesn't work out.
Am I missing something? Should I fear the hammer dropping down the line? Is this a acceptable workaround? We're not selling this machinery so it's a open and flexible project.
Attached is a Trend View of the raw input, and the input after I added the Capacitor.

[RBPLC]

Not to hijack this topic but I'd be cautious about using the BX-08UT. I'm finding it to be problematic and am curious of other users are having issues with both the 4/8 channel variants.

As a brief synopsis, these modules seem to be especially prone to issues involving grounding/ground loops when using temperature devices and I suspect the same issues would show up with other types of low signal level sensors. I bought several of the BX-08UT modules for a larger project I had planned. A few days ago I started pre-testing my panel before fully installing it in its enclosure while on the bench. I've got a BXDM1E-18ED23-D, BX-08AD-3, BX-08UT installed in this particular panel. I was using a Fluke 726 calibrator to simulate a Type T TC to try to set calibration values. When doing this I started noticing that depending on where I physically set the calibrator on my table I was seeing different input values on the modules. Long story, these modules seem highly sensitive to grounding issues. For example, I was using an AD THMT-T06L06-01 TC in the BX-08UT and when I touched the probe to the sub-panel I was seeing temperature changes of 10F or more. If the outer braiding accidentally touched the sub-panel I was seeing the same issue. If both the probe and outer braiding touched another value. I did a bunch of troubleshooting and the bottom line is I don't know how you would implement a sensor like this on a piece of equipment and get reliable and stable temperatures if there are multiple points in a system that can give spurious readings. I've troubleshot this on an EBC100 with the BX-08UT and see similar ground related issues. For reference, I've got NI cRIO's with 9213 modules next to the BRX systems and see none of these issues with the same sensors. I'm pretty bummed right now and am looking for alternatives to the BX-08UT. I'm wondering if the BX-08THM or a signal conditioner might be a better option for temperature measurement.   

[pbw]

I noticed when I connect the shield to ground I get a different reading, most likely due to a ground loop like Gary stated, but none the less stable and accurate with a 220uF bi-polar Cap wired on the inputs.

[Garyhlucas]

When installing analog sensors with metal cases I always check whether the ground wire or shield wire connects to the case of the sensor, some do some don't.  If it is and the sensor is mounted to metal or could touch metal you have to be very aware of the ground loop potential.  If mounted to grounded metal do not connect the shield at the PLC end, insulate it so it can't touch ground.  The shield connection at the sensor is what grounds the shield.

Another VERY important point. When the sensor wires must pass through a junction box you need to treat the shields just like conductors and use insulated terminal blocks, NOT a common ground block in the junction box.

[pbw]

Finally it makes sense.
You're a good teacher Gary, maybe you should do a "Industrial Controls for Dummies" for guys like me.

[ADC Product Engineer]

When installing analog sensors with metal cases I always check whether the ground wire or shield wire connects to the case of the sensor, some do some don't.  If it is and the sensor is mounted to metal or could touch metal you have to be very aware of the ground loop potential.  If mounted to grounded metal do not connect the shield at the PLC end, insulate it so it can't touch ground.  The shield connection at the sensor is what grounds the shield.

Another VERY important point. When the sensor wires must pass through a junction box you need to treat the shields just like conductors and use insulated terminal blocks, NOT a common ground block in the junction box.

As an addendum to what Gary said, never daisy chain grounds for the same reason.  You will create ground loops.  Grounds should always be single source, single destination.

[RBPLC]

These are all good installation practices but doesn't resolve the fundamental issue with the use of this card. For example, when using the THMT-T06L06-01 sensor, in order to get consistent measurements, one would have to ground the the sensor at only one point (at the process measurement end). The sensor lead would then have to be insulated along its entire length in order to prevent inadvertent touching of the lead wire to another grounded point on the piece of equipment. Not to mention if you were trying to use a TC to control a process where the sensor might be moveable and at times might/might not be grounded. If you can't guarantee that the sensor will/will not be grounded, you can't guarantee a specific offset for the calibration value and can't guarantee absolute temperature value. This also creates issues when trying to calibrate the input (for example with a process calibrator) because as soon as the sensor is hooked up (if it is grounded) the offset caused by the grounding is different and the absolute value of the reading is no longer correct. That being said, I have installed a 47 nF capacitor in parallel with one TC input (similar to what pbw suggested) and this seems to resolve the issues with grounding in my bench setup. Using NI 9213 modules, I have never seen this type of issue with thermocouples. Also, testing on the P1k P1-04THM module, I don't see these issues either.     

[BobO]

These are all good installation practices but doesn't resolve the fundamental issue with the use of this card. For example, when using the THMT-T06L06-01 sensor, in order to get consistent measurements, one would have to ground the the sensor at only one point (at the process measurement end). The sensor lead would then have to be insulated along its entire length in order to prevent inadvertent touching of the lead wire to another grounded point on the piece of equipment. Not to mention if you were trying to use a TC to control a process where the sensor might be moveable and at times might/might not be grounded. If you can't guarantee that the sensor will/will not be grounded, you can't guarantee a specific offset for the calibration value and can't guarantee absolute temperature value. This also creates issues when trying to calibrate the input (for example with a process calibrator) because as soon as the sensor is hooked up (if it is grounded) the offset caused by the grounding is different and the absolute value of the reading is no longer correct. That being said, I have installed a 47 nF capacitor in parallel with one TC input (similar to what pbw suggested) and this seems to resolve the issues with grounding in my bench setup. Using NI 9213 modules, I have never seen this type of issue with thermocouples. Also, testing on the P1k P1-04THM module, I don't see these issues either.   

We're looking into this. There may be things we can do to mitigate it. Would you be willing to test a modified module?

[ZoNiE]

We just had a similar issue. We are using a BR-08THM and type-T TC's and started having random noise issues. It turns out that when the cart holding metal pans full of food (conductive) that the TC was in was touching the grounded wall of an oven, it was causing ground loop induced noise. Our work around is to put stops to prevent the cart from being pushed in too far, but this may not always work. Hope that there is some sort of fix for this to allow for ground contact, as the possibility of one of the TC's that is not in use may be left on the floor of the oven by the user. We may try the 47nf cap solution... I may be interested in testing a modified module...