|By Aric Datesman |
Jul 5, 2012
Jeez, Sayar... Quite the pickle you're putting me in here. On one hand I promised to be nice, but on the other hand one of my issues is that you don't actually know how to use your equipment, and this falls under the "why is the VCR flashing 12:00" in terms of how base level a question it is.
Ok, from the beginning once again (as all of this was covered above in what I thought was enough detail for everyone to follow along)...
1. Per the spec sheet for your conditioner, at a scan rate of 500Hz has a resolution of 18,000 counts/divisions. Additionally at 100Hz it does 50,000 counts/divs, at 10Hz 120,000 counts/divs, and at 1Hz 200,000 counts/divs. In a nutshell this means that the Analog to Digital Converter (ADC) in the device can only handle so much data coming in at a time, so it begins dropping resolution and dumping it into "close enough" buckets to keep up (kinda like rounding...). So in this case, your testing at 500Hz means that the incoming analog -5 -> +5V signal from the load cell is split into one of the 18,000 available digital buckets at a rate of 500Hz.
2. From here we add in your load cell, which is rated to 5,000 pounds. Since it works in both tension and compression, you have a full range of 10,000 pounds (represented by the -5 -> +5V range on the conditioner).
3. Per step 1 above, this 10,000 pound range, when sampled at 500Hz, is divided into 18,000 equal buckets for the ADC to round the measured signal into. This is where the .556lb/div comes from... 10,000/18000. Think of it like standard decimal rounding... .5 rounds up, .49 rounds down. Then stand on your head and picture the cutoffs being different such that the multiples aren't whole numbers, but .556, 1.112, 1.668, 2.224, etc with the device lumping everything between .556 and 1.112 together, everything from 1.112 to 1.668 together, etc. Basically you're trading off resolution to get a higher scan rate.
3. Complicating this further is the accuracy of your load cell itself... In this case it's a 101NH with a repeatability error of 0.017% full scale. This works out to +/-0.85 pounds force, by way of 5000*.017/100.
4. Once we have the repeatability error, we can determine the error of measurement at a given force. So for a 500lbf force, the error induced by repeatability issues is .17% of the measurement (.85*100/500). As I mentioned earlier, the degree of this error goes down as the load goes up. Hopefully that bit makes more sense now.
5. Anyway, following from there we get back to the scan rate issue, whereby the ADC can only resolve readings in .556lb increments due to the 500Hz scan rate reducing the number of available divisions to 18,000. As you can imagine, stuffing a reading with an uncertainty of +/-0.85 pounds force into buckets incremented every .556 pounds is going to play hell with the distribution of which bucket the reading is going to fall into, and has a not-small effect on resulting accuracy. This is the bit toggle that Eric Krantz mentioned above, btw, and where I got the 1.112 figure, since any one measurement has a goodly chance of falling into 2 different buckets (2*.556=1.12).
With me so far? I ask because that's just the simplified version.... it gets more complicated once you start looking at the other error sources listed on the data sheet for the load cell. Sure, the repeatability error is .017%, but so are the hysteresis and the non-linearity errors. To get the total error you have to take the root sum square of the three of them. This turns what we thought was a .017% error into what is really an error of .0294% once you take all three sources into account (sqrt(.017^2+.017^2+.017^2)). If you like, feel free to run back through this number in place of what we got with the .017% figure. Kinda puts a hole in that generally within a pound figure you got from the manufacturer.
Of course, this is ignoring things like temperature change, heating of the strain gage, zero balance, wiring noise.... Like Eric said, it's a complicated subject. And I'm sorry if it comes across poorly, but if you're lost on step 1 perhaps this isn't something you really should be playing with (let alone posting results and analysis...).