why can't I run load cells directly into a PLC?

R

Thread Starter

Rodger Furey

I've contacted a few load cell suppliers and none of them appear to be particularly enthusiastic about running load cells directly into a PLC. I can do it, they say, but they all recommend that I first go into a summing box/weight controller and then analog out to the PLC. None of them have given me a good reason for doing this. It may be that they just want to sell me the electronics, but I want to make sure there isn’t a valid performance related reason to avoid going directly into a PLC. I like the simplicity of the concept, I avoid dueling digital displays, and I can see each individual load cell. App accuracy is of paramount importance. Any help would be greatly appreciated!

App info

(3) 500 lb load cells, S type. 1500 lbs total
2 mV/V
0-15 VDC excitation
App accuracy requirement of the load cell sum is +/- 1 lb

The three load cells will be supporting a tank. I would like to run all three load cells into a thermocouple/mV module (Allen-Bradley Flex Logix 1794-IT8). I’ll sum the load cells in the PLC. I would like to monitor each individual load cell as well as the sum of all three. I will be datalogging the load cell outputs and the sum will be used in a program with multiple/changing setpoints that controls a process.

The load cells are to be in a class 1 div 1 room.
The PLC is not.
The cables from the load cells to the PLC will be 25-50 feet long.

I imagine I'll need an intrinsically safe barrier between the power supply/PLC and the load cells.
 
Siemens makes a loadcell input module called siwarex.

Why would you want to get involved in reinventing the wheel, of course you can connect a loadcell to the PLC analog input, the question is what will you do next and how will you do it. Go buy a proven off the shelf solution, after the price of the loadcell an amplifier costs nothing and if you try do it yourself it is going to be a long and lonely road.

Donald P
 
C

Curt Wuollet

You can run them directly into a PLC and sum them.
A summing box lets you normalize them and this cancels
certain errors. And you get the instantanious sum which
you won't get with 3 analog inputs if there is any
vibration or movement. Either way will work, but resistive
summing and a single conversion will be more accurate.
Since you want accuracy better than .1%, I would heed their advice and also seriously consider 6 wire cells
(Kelvin Sensing). With 10V excitation and a 100mv input
scale you will only be using 1/5 of the available counts
and your error budget doesn't look good. I would use a
good IA (Instrumentation Amp) at the summer with a gain
of 5 with the mV card or a larger gain and use a standard
analog card. With 20 mV full scale, you are resolving down
to < 20 uV. Even without thermal effects and noise, that's
a tall order. Certainly doable, but much easier to do wrong than right. That's why scale people can still sell scale hardware for these "simple" applications.

Regards

cww
 
I'm using 4 similar loadcells on a belt conveyor, and the summing cards we use is nothing more than parallel terminals with balance pots in between them. I have successfully bypassed the summing card by just wirenutting the loadcells to the cable feeding the weigh scale module card (a 1771-WSM I think).

I think you will be better off using a weigh scale module card. It has built in settings (sampling rate, calibration stuff, etc.) that would be a hug e pain to reproduce in house.

Good luck
 
M

Michael Griffin

You can't connect a load cell "directly" to a PLC very easily, because you need excitation, amplification, and calibration. You really need a signal conditioner in between. There are however plenty of basic load cell signal conditioners on the market without displays. They are available as DIN rail mount packages (to be mounted in an enclosure), or as field mountable individual boxes. The signal output is typically 0-10V, +/- 10V, or 4-20ma. Power supply is typically 24VDC or +/- 15VDC.

We have used quite a few RDP S7-DC load cell signal conditioners. If you are using three of these pay close attention to the installation instructions, as they are only two way isolated.

Any instrument vendor that sells load cells, LVDTs, etc. should have a line of load cell signal conditioners. Also, try a Google search for "load cell signal conditioner".
 
J
The "valid performance reason" for summing load cells signals external to the PLC is one of the costs of ownership of load cells - all load cells drift (except the matched, sputtered Sartorius GWT cells).

It is typical to have either a local scale service or a plant tech recalibrate the system quarterly, or more often, depending on how badly your particular load cells drift. If it's wash down area, drift is even worse, because the hermetic seals on the load cells eventually fail.

The calibration process is iterative. It's tweak, tweak, tweak, tweak. I've witnessed techs taking anywhere from an hour to half a day to tweak the load cells on a vessel. In one plant where I worked, the contract scale guy had his own desk and parts rack in the plant. The point being - periodic, routine re-calibration is essential.

Do you want to have the either your plant tech or the local scale service guy messing around in your PLC tweaking the zeroes and spans?

Let the cal/service tech tweak the electronic box all he wants or needs to and just take the weight signal to your PLC. It's a lot cleaner that way.

John
 
It must be possible to do that, but why? Why do you want to monitor each loadcell? To see the loading configuration i the tank? To se the windforce? Or what?

I have been using ABB weight transmitters that you put in the S100 I/O system. If you use that you could use their programblocks for weighting as well. You save time and money in my opinion.

The diffrence betwen this an your discription is that all three loadcells are put together in a junctionbox so you cant monitor them separate. This have been a very good way to monitor tank and silos every time I have been using it.
About the cable it is imprtant as always when you are haneling mV signals that the cable i shieleded in every pair or tripple. And that they are separetly twinned (?). And if you are having distance in a cabinet its important to use the cable insaide that as well.

MTL are having barriers well suited for weghing in hazardous areas.

Good Luck!
Björn
 
J

Joe Jansen/ENGR/HQ/KEMET/US

According to the website, the IT8 is an RTD input module. However Thermocouples will have the same problem in that the response curve is non-linear. You will be trying to fit the response curve of the load cell against the response curve that the module is expecting for the RTD /
Thermocouple. This leads to some math that is not pleasant to do in ladder logic. Not impossible, mind you, but very unpleasant.

You will most likely be money, and time, ahead by simply purchasing the appropriate controller, and using the 4-20mA signal which will give you a pleasantly linear response, which can be scaled in the PLC using a simple
SCP function block to give you the true weight.

HTH!

--Joe Jansen
 
The loadcell output is very small and needs pre-amplification to scale it to 0-10V levels. Use a supplied amplifier from the vendor or similar product, or a module like Donald suggested that is made for the application.

~Ken
 
B
I concur with this. Load cells can be the most painful thing to calibrate. It seems like something simple at first until you watch someone who actually knows what he is doing calibrate them properly.

Most times I suspect the tech gets it "close enough" and moves on. Getting a really accurate reading can take a long time.
 
Joe,

Don't know if that is strictly true of a 1794 series mv/TC module (although, if the IT8 is an RTD input card as you've researched, then I'd suppose the TC card is a different part number, or I'm just plain wrong).

I'm guessing A-B has a 1794 module roughly equivalent to the INT4 or NT4 module for the SLC500 rack. If so, then that module can be configured for TC input (a variety of types), to read the CJC thermistor (for ambient PLC temperature), or for linear 0-50 mV or 0-100 mV inputs.

Rodger,

That said, the distance concerns me if trying to push a signal directly into the PLC. Voltage drop in the S+/S- signal lines will shoot accuracy in the foot. The way around this is to use a 6 lead load cell (signal, excitation, and 'sense' lines). The sense lines are connected to the load cell excitation terminals at the load cell, and brought back to what is essentially a regulated power supply. This uses the 'sense' leads for
voltage feedback and automatically compensates the excitation power supply. In effect, if the load cell excitation is desired to be 15.000 VDC at the LC it'll pump up the excitation supply appropriately (say, to 15.250V, or whatever) to compensate for voltage drop in the S+/S- lines. Also, long runs of low mV signal lines will tend to pick-up electrical noise which also degrades accuracy).

In any case, at 15 VDC excitation a 2 mV/V - 500# LC puts of 30 mv at 500#, so (if driving the signal directly into a 50 mV full scale input) you'll be wasting 20 mV of potential range (and subsequently, accuracy). To throw
another complexification factor into the mix, I'm guessing the 15 VDC excitation is a maximum rating. Very often better accuracy is obtainable using 10 VDC, or even 5 VDC excitation sources - self-heating effects are greater at higher excitations, and cause more pronounced thermal drift effects on accuracy. 10 VDC is a good compromise, and makes the math easier.

These are partial reasons one might want to use a load cell meter as a front end instead of raw mV load cell output.

The other way is to keep the load cell leads very short (so voltage drop is negligible), run them to a bridge type signal conditioner located right at the LC (Action Instruments, API, Calex (sp?), and other make these), and
send weight info to the PLC as a 4-20 ma signal..

If investigating this option you'll want to look over their spec sheets to see if they provide the required accuracy. Keep in mind, too, how much dead weight the tank represents - in order to get good accuracy it might be
necessary to span the milliamp signal for 1-20 ma, or 0.5 ma to 20 ma.

I'm not hawking their product (but it comes to mind) - Hardy Instruments makes both Devicenet and, if memory serves, A-B rack module weigh scale front ends (at least for the SLC rack; don't know about anything else, and it's been a couple of years since looking at this stuff). They are a bit on the pricy side, but probably on par (once 3 signal conditioners, additional wiring, and the convenience factor are figured into the mix) with going the ma signal conditioner route, they do up to 4 LC channels per module, and have other amenities built-in which should make programming easier.

Bob
 
R
I've just discovered this website recently and this is the first time I’ve used it. I appreciate the responses.

A little history: I was involved in a load cell app a few years ago. After watching the service tech endlessly diddle around, trying to calibrate the system I finally thanked him for his efforts, ripped out the summing box and controller and ran the signal wires directly into the thermocouple card of the SLC500 that was controlling the process. The zero and span calibration was done in the PLC with a scaling (SCP) instruction, very easy. The primary problem I had with this system was the accuracy. I never managed to get it better than +/-2% before I had to abandon it for other things, and I was never able to determine how much of the error was due to electrical or mechanical issues. It’s been working for 5 years now because +/- 2% was fine for that particular application. My current app requires an accuracy of +/- .06%. I am cautiously thinking about doing the same thing but I need to address the deficiencies of my previous effort first and I sure would like to not be flying solo on this project.

Mechanically, I am hanging the new tank with three load cells (instead of two cells and a balancing linkage like on the old system), and I am making every effort to minimize to near zero the effect the connecting piping can exert on the system. I also do not have the significant temperature variations (-25 to 80 F) and vibration issues I had in the old system. So I'm feeling pretty good about this.

Electrically, I am less confident, primarily because of the reluctance of the load cell people to get on board with my approach - and most of the responses here it seems. If anyone is interested in pursuing this topic, I have a few follow up questions/comments. Right now I'm not sure what to do…maybe have a beer.

Sample rate: The PLC scan time should be around 10-20 msec. Would a weighing instrument sample the load cell inputs any faster? And even if it did, I would be shooting the summed value out to the PLC which would control the filling and draining of the tank, subject to the limitations of the PLC scan time.

I’ve never heard of a six wire load cell. Can someone explain what the extra two wires are for?

Calibration via an OIT on my old system is infinitely easy. Unload the load cell, press a button to zero it. Load it up to full scale, press a button to save the full scale value. (I also record each calibration value and use that data to see if there is a trend that might indicate an impending failure.) Drop the empty tank back on the load cells and zero out the tank weight. Am I missing something that hasn’t reared its ugly head in my old low-accuracy app?

I’m not sure if it would be better to convert the mV signal to 4-20 mA/0-10 VDC and run it into an analog module or keep the original mV signal and go into a thermocouple/mV module. I imagine there will be some small error introduced with the signal conversion.

I would like to see each individual load cell so if one of them starts to act up it will be immediately obvious. I can set up an alarm if one of the cells output goes out of range. If I’m only reading the sum then I would think that a misbehaving cell would be less obvious.
 
M

Michael Griffin

<clip>
> The zero and span
> calibration was done in the PLC with a scaling (SCP) instruction, very
> easy. The primary problem I had with this system was the accuracy. I never
> managed to get it better than +/-2%
<clip>

The PLC was scaling the reading, not calibrating the signal. You still need to make sure the analogue input signal is physically calibrated to the input range of the PLC.

> My current app requires an accuracy of +/- .06%.
<clip>

I don't have any experience with tanks, but 0.06% accuracy is very ambitious in any industrial application. Resolution is easy - just use an A/D converter with enough bits. Accuracy is the difficult part. The entire system has to be capable of this, including the sensors, amplifiers, and A/D converters. The measurement errors from each of these add together. Noise on low level signals will also reduce accuracy. There seems to be some contradiction in your requirements for high accuracy and cheap instrumentation.

> Sample rate: The PLC scan time should be around 10-20 msec. Would a
> weighing instrument sample the load cell inputs any faster? And even if it
> did, I would be shooting the summed value out to the PLC which would
> control the filling and draining of the tank, subject to the limitations of
> the PLC scan time.
<clip>

How fast will the signal vary? One thing to keep in mind with three separate signals to the PLC is that the analogue card (of whatever type) will typically have only one A/D converter, and it will multiplex the inputs. This
means that the three readings will be taken at slightly different times (the specs on your card should tell you what this is). I don't know if this makes any difference to your application.

You also need to be sampling the signal at several times the rate at which it can vary. If vibration is possible, then you need to filter the signal to avoid erroneous readings.
 
One thing you might want to do is put together an error budget spreadsheet. This will include specifications from the LC manufacturer's data sheet, whatever converter(s) you are investigating (if any), load cell power supply, and temperature ranges at the LC location and within the electrical enclosure.

Many devices will have a temperature coefficient (tempco) specification stated in terms like "150 ppm/°C", and can show how much measurement variation can be expected by operating at one temperature versus another. To end up with 0.06% accuracy I'd suspect you'll need to be very circumspect in regard to these things.

You didn't mention what is used for the LC power supply in the previous project, but, for instance, if it is capable of 0.2% regulation at any particular load, then that's about the best accuracy the measurement system could be expected to deliver.

> I've never heard of a six wire load cell. Can someone explain what the
extra two wires are for? <

Its a 4 wire load cell (2 excitation wires, and 2 signal wires), plus another pair (sense lines) connected where the excitation wires are tied to the LC bridge. Power supplies specifically designed for LC (and other strain gauge type bridges) will have a set of terminations to connect the sense lines to.

The excitation wires must carry a fair amount of current, and so voltage drops will exist across them. The sense wires, on the other hand, are connected to a fairly high impedance measurement circuit in the supply, carry very little current, and thus drop almost no voltage. They measure the voltage at the load cell, and bias the power supply to compensate for excitation wiring voltage drops. In practice, the power supply will put out slightly more voltage so voltage at the load cell will be at the desired
setpoint. If 10.000 VDC at the LC is desired, then (depending on wire length, terminal block resistances, et al) the supply will actually develop10 volts plus whatever is necessary to compensate for these losses.

> I'm not sure if it would be better to convert the mV signal to 4-20
mA/0-10 VDC and run it into an analog module or keep the original mV signal
and go into a thermocouple/mV module. I imagine there will be some small
error introduced with the signal conversion. <

You'll certainly gain in resolution by throwing a signal conditioner into the mix.. Lets take a simplier example with a 100# load cell rated 3 mV/V and a 10 VDC excitation supply. Lets put this raw into a 50 mV PLC input that gives 32767 bits at the 50 mV full scale (655 bits/mV).100# on the LC will develop a 30 mV signal (3.333 #/mV, or 197 bits/#). If, however, you put a signal conditioner in line so the 30 mV LC signal translated to 50 mV to match the card's full range, and ran the numbers, you'll get 50 mV at 100#, which is 2#/mV, and gives 328 bits per pound. This is a 66% increase in resolution. Now, it's better to convert to milliamps (less noise pickup, et al), but the same math holds ... the idea is to use the full range the PLC card is capable of.
 
M

Marcos Bartkevicius

Dear colleague:

> ...Calibration via an OIT on my old system is infinitely easy. Unload the load cell, press a button to zero it. Load it up to full scale, press a button to save the full scale value... <

You must have in mind that recalibrating is not the same thing as rescaling. As you described, what you do in your system is a rescaling of the measurement. Everytime you do this, you end up with a different resolution. The cause is that you are not using the full range scale of the analog card. This can only be acomplished by the use of any analog signal conditioner that will brig the analog zero of your system to the analog zero of your card, and the analog maximum of your system to the analog maximum of your card.

> ...I’m not sure if it would be better to convert the mV signal to 4-20 mA/0-10 VDC and run it into an analog module or keep the original mV signal and go into a thermocouple/mV module. I imagine there will be some small error introduced with the signal conversion... <

In the case of some error introduction, the error calculated at the end (program reading) can be mathematically proven to be under the resolution loss described above. Indeed, the reading range you are "throwing away" by connecting the LC without signal conditioning is far more relevant constraining to errors (difference between the real analog value and the STEP value the analog card delivers) than the error the IA will introduce into your system.

> ...I would like to see each individual load cell so if one of them starts to act up it will be immediately obvious. I can set up an alarm if one of the cells output goes out of range. If I’m only reading the sum then I would think that a misbehaving cell would be less obvious... <

Some good IA have digital outputs to signal the PLC that one individual input has become unusable. Wire these inputs to the PLC and you will be able to keep the system running even in the event of a bad LC. In this case, you will need an analog system that will adapt each input independently concerning to 0 and RANGE and deliver all three signals to the PLC, plus the error digital inputs.

> ...I’ve never heard of a six wire load cell. Can someone explain what the extra two wires are for? ...My current app requires an accuracy of +/- .06%... <

It seems to me that if you never heard about six wire LC and you are running after a 0.06% accuracy you haven´t done a good reserch job. Like other colegues already said, the accuracy you will meet at the end is the sum of all accuracy/errors/drifts in your system. The amount of accuracy lost with the connection of a 4 wire LC directly to the PLC is by itself the prohibitive concern to achieve the 0.06% accuracy.
 
Six wire is used to "track" the excitation voltage. It is useful if you have longer cable lengths.

When you have a long cable (which is made of copper wires), your excitation at the "load cell bridge circuit" varies with temperature (because copper resistance change with temperature).

So out of 6 wires (3 pairs), one pair each is used for signal, excitation and sense. The sense wire acts like a feedeback and corrects the temperature effects. This is somewhat similar to a "2-wire" and "4-wire" resistance measurement on a multimeter.

If your application is Not in a temperature controlled environment and you have a long load cell cable, you are bound to have a very large with change in temperature!

Regards
Saji
[email protected]
 
Top