I am installing 1746-ws load cell modules in my micrologix 1500. The connecting scales are used in a high speed batching process. The resulting weight readings from my scales are displayed down to .01 lbs. The problem I have is that I am unable to stabilize my weight measurements enough to pass calibration. (Excessive filtering and sampling slows the system down too much and I overshoot my high setpoints.) Calibration for trade requires an accuracy of +/- one division, which in my case, is +/- .01 lb. I can stabilize my reading with additional math and achieving a reading down to .10 lbs. This does not satisfy Q.A because my old (non plc based) scale systems displayed down to .05 lbs and Q.A feels that the accuracy of the new system is not the same when displayed in .10 lb divisions.
If I could convert my .01 divisions into .05 divisions my system would pass calibration and Q.A would be satisfied. Any suggestions on how to accomplish this?

Multiply the floating point result by 20 and make the destination an integer. Then divide that integer by 20 with a float destination.

Note that the original value times 20 must be less than or equal to 32767 for my oversimplified solution to work!

Let me see if I am reading the problem correctly. The readings from the scale are displayed in values of 0.01 lbs increment. At present when you pass a 0.10 lb weight you get a reading of no more than 0.11 and no less than 0.09 lbs. This is the trade requirementof +/- one division. Your reading of the 0.05 lb weight however does not fall in the 0.04 to 0.06 range and QA is not happy with this. If this is a correct restatment of your problem then read on.

First Question - what is the actual or raw resolution of your system? I could not find a manual for the 1746-ws module to provide this answer thru the ab literature library so I assume it is some third party product. What I mean is this if your load cell is rated 0 to 100 lbs and your output is in 12 bits. Then you have 4095 divisions at 0.02442 lbs per division. Your final accuracy cannot exceed your initial. In this case it would be impossible to gaurantee the results you want. Now 0 to 100 lbs with 16 bits would yield 32767 divisons and 0.00305 lbs per division. THis would give you few problems achieveing your final accuracy.

Is this the module you are using? http://www.hardyinstruments.com/process_we...gh+scale+module

Or is it this module? http://www.hardyinstruments.com/process_we...logix%AE+module

I'm wondering because you said it was a 1746-WS, which is for the SLC-500 PLC, but you said you were using a ML1500, which would use 1769 modules, not 1746 modules.

The 1746 module has a resolution of 985000 counts and the 1769 module has a 23 bit A/D for a whopping 8388607 counts, so I don't think this is an issue of resolution.

What range is your load cell and what voltage are you using to excite it? You also mentioned that the application is a high speed batching operation. How fast are you taking measurements? Is it faster than the conversion rates of the module?

I am sorry, I miss stated my load cell module. I am using a 1769-ws module by Hardy. I am using a 300 lb capacity load cell with 5 volt excitation. As for my resolution, the load cell module has an internal resolution of 1:8,388,608. My weight reading uses up to a total of 32 bits depending on the decimal point location. I am currently able to display my weight reading in .01 multiples (i.e .01, .02, .03, .04, etc.). I need to display in multiples of .05. (i.e .05, .10, .15, .20, .25, etc). If I can convert my display into .05 divisions then my calibration range is +/- .05 ( with a 10 lb weight could be 9.95 or 10.05) wich is acceptable for calibration. I am just unsure on how to achieve this conversion or if it is even possible.


I am sorry, I miss stated the load cell module. It is a 1769-ws by Hardy. The load cell has a 300 lb capacity and uses a 5v excitation. I am unsure of your question about how fast am I taking measurements. I am measuring the the weight continuously as it is batched. I have the number of averages set very low to a setting of 5 to provide enough reaction speed. Unfortunately, the low number of averages adds to my stability issues. From what I have seen however, displaying at .01 divisions, my weight is only fluctuating by .03 lbs. If I could achieve a division factor of .05 I do not believe I would have any issues.



Doesn't the result end up being back into .01 divisions. Lets say my weight reading is 50.03 lbs.
50.03 x 20 = 100.6. As an integer it is 10006. 10006/20 = 500.3 This is basically where we began. To scale this back down 500.3/10 = 50.03. Am I missing something here. I was hoping that we could convert this into, in this case, 50.05. This would be rounded out to make my .05 divisions.

I think your 50.03 x 20 will be rounded to 1001, then when divided by 20 will be 50.05.
With the rounding you will always read xx.00 or xx.05

I might be misunderstanding the problem, so lets check if I understand you correctly.

When you set a 10 pound weight on the load cell the weight that it reads is not steady but it is fluctuaitng by .03 lbs. At first glance it seems that you have some signal noise or vibration problems. I am concerned that adjusting the display to a 1-1/2 digit decimal portion instead of a 2 digit decimal is only a band-aid, and that the problem will come back to haunt you in vexing ways. Wouldn't it be better to try and correct whatever is causing the fluctuations?

Generally the accuracy of a weighing system is expressed as a percentage of the full load calibration thus taking simple numbers if the weighing system accuracy is +/- 0.1% then using a 100kg set of weights calibrated to international standards the weight should be between 99.9kg and 100.1kg. If I choose my resolution to be 100.1000 or 100 it makes no difference to the accuracy just the amount of decimals I wish to display and the aggro I get from QA people telling me the 4th decimal point is bouncing so as Alaric says if the weight display is not stable there must be a system problem whether mechanical or electrical and simply frigging the display to keep the QA people at bay is not answering the problem.Forgot to say your figure of 0.01 in 10lbs would be to me an acceptable accuracy, 0.05lbs in 10 would not especially if 10lb is the full scale deflection of the weigher but maybe is for your process

?

50.03 x 20 = 1000.6...If the Destination in the MULtiply is an integer it will be truncated to 1000. Then the final divide will yield 50.00. So, instead, make the Destination of the MULtiply a float. Then MOVe that float to an integer and it will be rounded to 1001. Divided by 20 and stored in a float it will be 50.05. So do the MULtuply by 20 with a FLOAT destination. Move that float to an INteger to get it rounded, and then divide that by 20 with a float destiantion to get your .05 (1/20th) desired resolution.



Actually Ken, with the rounding you will always read xx.x0 or xx.x5. Isn't that exactly what he asked for?

Here is some test logic that I tried. Seems to work just like he wants. It will help stabilize the display too if it is bouncing around less than .05 pounds.

EDIT: I forgot how to put images directly into posts here...
[attachmentid=4177]

Okay, I understand now. I don't work much with the floating point math and assumed that all digits in the floating register would transfer over. I did not realize only the digits to the left of the decimal would be the ones to transfer. This looks as if it could work for me. Thanks.



You are right and I do not want to just hide this problem. The thing is though that there is a lot of floor vibration from heavy mixing equipment and there are roller conveyors mounted to the top side of the scale bed. I can eliminate the fluctuation through filtering and sampling but the excessive filtering and samples slows down the scale response and I will overshoot my upper setpoint. As it is, We have seven scales in the process and this is the first of those that we have switched to be a plc based system. The remaining six scales read in divisions of .05lbs as well.

just out of interest really, do your quality people ever come along take a selection of your widgets that are supposed to weigh 10lbs retire to their hovel where they have a nice stable weighing system which they can check weigh on and then emerge with an arm load of statistics proving the process scales are rubbish or do they simply rely on the fact that when they put a 10lb weight on the scales it shows 10lbs or thereabouts

Our Q.A people only care that the scales are calibrated (by outside sources) and have a sticker indicating calibration date and next due date. The have never come out and tested any packages. As far as they are concerned, the system is working if it passes calibration and they do not receive any complaints as far as weights go. They are however, the first ones to dictate in what kind of divisions we can measure in. Pretty good quality system eh.

Been there, might be worth you having a go, I usually do that sort of thing at night or weekends when no one is liable to catch me just so that I am ready for when some one starts asking awkward questions

0.01 lbs is so small that air currents will cause the scale to "drift". The solution is to mount your scale on a large 2 foot by 3 foot by 4 foot block of concrete embedded on ground with at least a 5000 PSI compressive strength to begin with (no second or third floor stuff here). Then install a compressed air system that blows off the scale before each weighing. Then you need to build a large plexiglass box around it that has to be closed before the scale will even attempt a reading. Finally, tap off the compressed air system and bleed a little bit of air into the box and put a micromanometer on the box so that you can put a very slight but repeatable draft on the box for each weighing, along with a control valve to do the metering of the air.

If you think that's bad, try working with a scale with the same percentage range accuracy rating but the scale is 60 feet long and 14 feet wide (ie, truck scale). Every time the wind blows, it rains, snows, or a truck drives by, the readings will fluctuate. You can stand and watch it literally fluttering every time you get a little breeze. Fortunately it will at least auto-zero to avoid most buildups of dust and dirt.

300lb capacity x 0.01 increments = 30,000 increments. That is too much. Many years ago when I worked in the Scale Industry, we used 5,000 increments as the standard to determine resolution.

Ian

Just goes to prove that just because the technology can do it doesn't make it smart or right.

It is better to have excessive resolution than not enough. You can always reduce the resolution through software, but you can't increase it...