I am working on a project that requires ~.01g sensitivity to detect very very small changes in weight, but the actual measurement is irrelevant, it just needs to be consistent. For example, if I place a toothpick on the apparatus, I'd like to see a measurable difference. Based on the work I've done so far and what I've read, this is not easy (or cheap).
So far I have used a 5kg load cell with an HX711, which appeared to work until I realized my data was being affected by temperature and humidity changes. I don't know how to account for these changes, so I put this method on hold. I also think the 5kg load cell is designed to only .05% accuracy.
My next attempt was to use some of that conductive rubber string to see if it would give me relative measurements, but it was not sensitive enough. I would have to add .5kg or so for the returned values to show any significant change.
Another idea I had was to use a Raspberry Pi and a camera. I'd adjust the focus so that it became a macro lens, attach it to a fixed point on a spring facing a dithered/textured printed image then have it detect changes between each frame using OpenCV. Though, I think this is just a much more abstract and complicated way of mimicking what digital calipers do.
I have also considered using a spring and a range sensor (probably not accurate enough), and I think for it to be accurate enough I would need to use a laser displacement sensor (out of my price range). Contact measurement is OK, so my final trial would probably be to pick up some digital calipers and hook that up.
Am I missing any other method to do this or am I just plain out of luck?
With that resolution, you have basically two choices: cut the large weight into small pieces, weigh individually on a cheap jeweler's scale and add the bits, or pay $$$ for a suitable 5+ kg scale.
If the actual weight measurement does not matter, and only a relative difference between one measurement and another matter, would it be possible to use a linear encoder with a spring?
There are some 33 micron strips that are within my price range which should get me close to the accuracy I'd need, but then I think at that point it would be easier to do the caliper hack.
Marktonium:
So far I have used a 5kg load cell with an HX711, which appeared to work until I realized my data was being affected by temperature and humidity changes. I don't know how to account for these changes, so I put this method on hold. I also think the 5kg load cell is designed to only .05% accuracy.
I'm not an engineer, so some math help would be great!
Any idea where I'd start to look to compensate for the temperature and humidity changes? I can generate both humidity, temperature, and raw readings from the HX711 + load cell but I'm lost after that.
Any idea where I'd start to look to compensate for the temperature and humidity changes?
Possible, but the calibration process is difficult and time consuming, and the result error prone.
You would need a 2D table (or two-variable function fit) of corrections versus temperature and humidity, for a series of standard weights. Overview of calibration here.
A controlled environment, as mentioned above, is easier.
Delta_G:
What you should do is keep the scale in a temperature and humidity controlled environment (as best you can) and calibrate the scale when those values change.
jremington:
A controlled environment, as mentioned above, is easier.
I'll start doing more research on sensor calibration.
Precision detection of mass difference is much easier to do with scales. Have a counterweight
matching the mass under test, and measure the difference. Temperature changes should cancel
out in the balance arm if its kept all at the same temperature, and given precision ruby knife-edge
bearings you should be able to get towards that 20ppm resolution figure. Its a very tall order, note.
The standard way to deal with a precision scale is to null the balance with some mechanism capable
of only very small force (and which thus doens't need high accuracy itself). An actuator and a thin
spring is possible, small linear solenoid, various methods can be dreamt up.
There is no cheap route to precision measurement of force that I've heard of - good loadcells can
get you so far, then it gets tricky.