Tuesday, November 26, 2013

From bathroom scale to scientific test apparatus

It all started with thinking about how I could instrument my drop tower. More to the point: how I could instrument it without resorting to the thousand dollar setups from places like Omega Engineering. During my search, I came across the affordable button load cells sold at RobotShop.com for robotics and hobby electronics. This one in particular caught my eye, with its 1000 kg-f (9800 N) rated limit. Certainly enough for any materials I may want to test, and even high enough to make a thrusting target for measuring thrust loads.

I kept digging, and learned that the 1000 kg-f load cell I'd found was only rated for 100-1000 kg-f, so any low end force data couldn't be measured. I also found some bathroom scales advertising that they used load cells. One advertised being able to measure a load up to 400 lbs-f (1780 N): for less than half the cost of a big button load cell. So, like any good engineer would, I bought the cheap thing and started figuring out how to make it useful.
This scale has no idea what's about to happen to it.



The first thing I did was convince myself the scale used load cells. After a bit of work with a screwdriver I had my answer. Yep, it did. 4 of them.
Top two load cells in the scale. Can you hear the evil laughter?
 Recall from the statics primer that a load gets distributed amongst all of its supports, so 4 moderate-range (0-100 lb-f) sensors could provide a measurement up to 400 lb-f. I still needed to convince myself that the scale could read low weights, however. So I slapped some batteries in it and put a small weight on it. Indeed it could measure fairly small loads reasonably well. From this initial glance, things were looking pretty good.

Feeling confident, I stripped the glass face and plastic tubing off the bits of the scale I actually cared about:
The important bits.
Close-up of a load cell (Under that white goo is the strain gauge)
The load cells in this scale are shaped like an 'M'. On the middle bar of the 'M', a strain gauge is placed so that if the bar bends, it deforms the gauge. To the end of the bar is riveted a plate that covers the strain gauge and is where the load is applied. The load on the top of the plate gets transferred to the end of the middle bar, and actuates it like a bending cantilever beam.

I established pretty quickly that I wouldn't be able to re-use the electronics in the scale. First, it had to be activated before it would start measuring. Second, it timed out if no weight was put on it. Finally, it had a stabilization time before it would register a final weight. Sadly, the electronics were too smart for their own good. Out they went.

Time for some new electronics

I was satisfied that the load cells were promising, but had no electronics to read them. I'd have to figure out how to make my own. I came across several helpful sites, but the most useful was this one from the Enrico Fermi Institute's Electronics Design Group. They provided a really nice tutorial on attaching a button load cell to an Arduino board for data acquisition. They also mentioned exactly where they got their parts, what those parts were, and how much they cost.

At first, it sounded like I would just need an amplifier, a DAQ and I'd be in business... But I realized pretty quickly that my circuit wasn't looking like the ones I'd found. I didn't have 4 wires (one for each node of a Wheatstone Bridge), I only had 3 (Doh!). And of course, the connections on the scale's circuit board weren't labeled in a useful way. It was time to figure out what these wires were.

Out came the digital multimeter, and I measured the resistances between each of the 3 wires. As it ended up, between two of the wires (I'll just call them wire 1 and wire 3) I got approximately twice the resistance value that I found between wires 1 & 2 (~850 Ohm) and wires 2 & 3 (~850 Ohm). But what did it mean?

Thanks to the wonders of Google, I came across several people posting about the same sort of load cells. From this thread, it appeared I had what is referred to as a half bridge arrangement. I was missing two resistors and and one voltage output in my circuit. Originally, the resistors and voltage lead were probably on the electronics board, not that they would be much help to me. From the thread, the circuit diagram for what I needed to build looked like this:
In words: there was likely 2 strain gauges in my load cell, which jived with some of the comments in the Omega technical information for measuring bending strain. But I'd have to wire up the other two resistors and the missing output voltage lead.

The source voltage (\(V_{in}\)) could be provided by some power source (like a set of batteries or DAQ), and the output voltage (\(V_{out}\)) would be the load measurement (as a voltage). I could monitor the output voltage with a digital multimeter to see if things were working. But first I needed to construct the rest of the Wheatstone Bridge: easily accomplished after a trip to a RadioShack that wasn't a glorified cell phone kiosk.

Bridging it up

My very first Wheatstone Bridge. Isn't it cute?
I used a breadboard to mock up my homebrew Wheatstone Bridge (yes, there is a beer called this), as seen above. This first try used two 100 Ohm (1/4 amp power limit) resistors for the bridge resistors. I had actually grabbed the wrong size, and didn't realize it until I'd gotten home. But I was impatient, and set it up anyway. I managed to get a functional circuit, but as the sites led me to expect: the output voltage was hardly large enough to measure on my multimeter.

Satisfied that I had some idea of what I was doing, I ordered a set of Texas Instruments INA125P instrumentation amplifiers from Mouser Electronics, and an 8 channel USB data acquisition (DAQ) module. Once those arrived, I rewired my circuit to include the amplifier, an LED and a switch to make testing easier. I decided to use a 100 Ohm gain resistor to get approximately a 500x signal gain in the amplifier, which I anticipated would bring the output signal into the 1-5V range. I also replaced the bridge resistors with 1 kOhm resistors, since the load cell branch resistances were about 850 Ohm each, so that the resistances all the way around the bridge were closer.

Then it was time for more testing:

It's aliiiiive!
To my pleasure, the whole thing appeared to be working. Next on the list was to see if I could get a reasonable calibration curve for it. This would mean putting known weights onto the load cell and establishing a relationship between the applied load and the measured voltage. If it wasn't possible to calibrate, then going any further with these load cells couldn't happen.

Calibration time 

The load cell cover has a small foot for centering the applied load, a feature I intend to use to make sure the cell is loaded properly. But this meant I had to figure out how to balance weights on the thing. As it ended up, I had a bunch of 16 oz cans of black beans (organic even), and a length of the PVC pipe I used to make my drop tower. A can of beans just fit into the tube, so I used the tube to steady a stack of cans.

There was a snag though: the cans were only about a pound a pop and I could only fit 4 in the tube. I wanted to see if I could push this thing to 100 lbs, so drastic measures were necessary. First, I grabbed weights from my garage gym: a 10 lb-f and a 35 lb-f kettlebell. Those I carefully balanced on the load cell to get intermediate loads. But I still lacked something close to 100 lbs that I could easily put on the load cell.

Then I remembered: I have a stool and another bathroom scale. I also happen to possess a rather well calibrated 226 lb-f weight. Going back to the statics lesson, I set one foot of the stool on the load cell, one foot on the center of the scale, and made sure they were level. Then I hopped on and shimmied my butt back and forth to get some higher weights. Not perfect, but it was a start. Ah, the glamorous life of an engineer.
Dramatic recreation of high load calibration. No engineers were harmed.
I plotted my recorded loads and voltages in Excel, and fitted a trend-line to the data. Much to my surprise, the data fell in an almost perfect line, well within the measurement error. I had a functional calibration curve for a range from 0-100 lbs. There was definitely a fistpump.
My dignity was totally worth this.

What's next

The next step I have planned is to wire up the DAQ, re-calibrate (since the input voltage will change) and try to do some low-load measurements with my drop tower. After that, I'll build a holder for all 4 cells and wire it up to be usable in the drop tower. I'm also hoping I can use it to measure some thrust loads (these load cells don't care about orientation as long as the applied load is normal to the beam). I'll post updates when they are ready.

4 comments:

  1. hi. can i have the schematic for the circuit. tq!!!

    ReplyDelete
  2. If you want the final one I used, check out this post, about 3/4 of the way down for the image. You should be able to open it and save it directly.:
    http://sparkyswordscience.blogspot.com/2014/02/bathroom-scale-apparatus-gets-upgrade.html

    ReplyDelete
    Replies
    1. thanksss but its quite messy because of 4 load cell. can i have schematic circuit for 1 load cell only. pleaseee. this is my email sitiirani@yahoo.com

      Delete
    2. Sorry, but it sounds like you need to draw one up based on whatever it is you're putting together. Mine was specific to my arrangement.

      The links in this post and the wheatstone bridge arrangement image here are good places to start.

      Delete