Strain Gages Please!

Archive New Product Ideas
1

Those of us building robot armies are depending on you. Please SparkFun, sell us strain gages!

SparkFun has done a great job evangelizing do-it-yourself surface mount soldering. There may be a similar opportunity with force sensing applications. I’m not waiting on a tutorial though. I’m ready to start damaging parts right away!

I’ve been scouring the Internet for strain gages. Kyowa has an impressive product line. They sell strain gages with integrated temperature sensors and pre-soldered leads. And they sell kits and adhesives. Kyowa’s English-language documentation is remarkably good!

If you want to build giant robotic claws with grip sensitivity, you need strain gages. Strain gages are also good for weighing stuff. In other words, I need strain gages!

Unfortunately, there don’t seem to be many US suppliers. Perhaps strain gages would be a good addition to SparkFun’s catalog?

2

Digi-Key probably stocks them. Load cells are widely available, which consist of strain gauges mounted on hardware for weighing and force measurement.

Leon

3

As well as the gauges you also need the circuits. Currently I’m using AD8555 programmable amps in a single sided supply configuration as well as digipots to do automatic zero levelling.

I wouldn’t mind a bit of strain gauge action on Sparkfun as I’ve sort of muddled through using them and would like to see other peoples results and opinions on improving circuits.

For robotics I imagine raw gauges might be more useful than prebuilt load cells to give flexability in designing the grips or other robotic parts.

Any suggestions on forums where work on strain gauge circuitry and discussion are active would be welcome.

Phil

4

have you looked at what SFE carries? http://www.sparkfun.com/commerce/produc … ts_id=8713

Kind of pricey but it sounds exactly what you asked for.

5

by the way, you can make pressure sensitive gauges with the conductive foam that they ship DIL chips stuck into. The more you press, the lower the resistance. I don’t know about the repeatability, though.

6

Philba:
have you looked at what SFE carries? http://www.sparkfun.com/commerce/produc … ts_id=8713

Kind of pricey but it sounds exactly what you asked for.

That’s the right idea, and the price is great. In terms of robotic end effectors, I can see how that sensor could be integrated to measure compressive force along a single access.

I was thinking of multi-axis force sensing, based on beam deformation in the structural elements of the robot. (For example, to both grip and weigh, independent of grip orientation.)

I don’t see how to accomplish something similar using the SFE sensor, since the SFE sensor works only when it is in line with a compressive force. I’m open to suggestions. I haven’t done much with force sensing, so I may be suffering a poverty of vision.

Unrelated to robotics, I’m interested in building a scale. At first glance, the SFE sensor you point to looks promising. However, I’m interested in making repeatable measurements up to a couple kilograms, with accuracy and precision of 100 milligrams or less.

I’m not sure how to incorporate the SFE sensor into a scale without having imprecise/unrepeatable readings due to trivially off-center loading of the weighing platform. The weighing platform must, necessarily, be much wider than the sensor itself.

Any structure designed to channel the force straight down onto the SFE sensor could introduce friction or itself absorb force in the case of off-center loading? At least that’s what my intuition suggests, not having built a test rig with the SFE sensor. I fear that those factors would preclude accurate and repeatable readings in the milligram range with the SFE sensor?

I’m eager to read suggestions on how to use the SFE sensor in a scale, but I had tentatively resigned myself to using beam deformation sensors for the scale, as well, thinking those would be less susceptible to off-center loading, and would need no auxiliary structure to channel weight from the platform to the sensor?

The scale I’m building must be no more than 20 mm thick including housing, leveling feet, and the weighing platform itself. That’s the design goal.

I can tolerate relatively frequent (re-)calibration.

What have others done?

Maybe I should just go ahead and build a couple test rigs with the SFE sensor? I’ll think about it.

7

The SFE sensor looks interesting, though I recommend looking carefully at the recommended driver circuit due to the non linear resistive response. For you application I’m a little worried about the 3% accuracy however, particularly with the weigh scale application.

For more accuracy you’ll probably need to use a strain gauge plus instrumentation amplifier. This brings in a bunch of other problems like zero offsetting as well as self heating of the gauge.

Some circuits are very elaborate and have thermistors to compensate for self heating of the gauge. Ive found in practice that given sufficient time the device will achieve equilibrium with its mounting and then accurate measurements can be made…this depends on the structure and material the gauge is bonded to, sorry I can’t help out in giving guidelines there. For carbon fiber circular shafts, I’ve found equilibrium take about 2 minutes.

Regarding platform off center offsets, you can just use software to obtain a ‘zero’ point, and then take measurements from this. Another alternative is to use digital potentiometers to adjust for zero before a weight is applied. Normal commercial weigh scales will zero adjust each time the device is turned on, and usually give an option to manually zero adjust.

Regarding your weigh scale platform, there are a number of options, generally you want to restrict sideways movement and any other compression/extension forces by hollowing out the internals where the gauge is attached, leaving only the force(usually a lever type) that is applicable. Have a google around for load cell construction to see what I’m talking about here. When measuring bending force in a carbon fiber shaft we found we needed a secondary structure like a little bridge to reduce measuring the ‘squashing’ of the shaft when load was applied.

I’m interested in what you are doing with multi-axis force sensing. I’m not sure you’ll get constant readings no matter what you do. The gauge will measure distortion, however I don’t imagine this would be constant if your robot arm is changing the orientation of the object being weighed. You might be able to build up a calibration scale that is dependant on your robots range of movement, which would be sort of cool. That is have it grip a known weight then move though its various degrees of freedom measuring the gauge at each…though I suspect this would only be practical if the number of degrees of freedom is limited to one or two axis.

Anyway, please keep us informed how things go.

Phil

8

The resolution you are seeking may be impractical. If you’ve ever tried to weigh anything in the milligram range, you very quickly begin to realize that the slightest things will throw off the measurement. Furthermore, the requirement that the housing be no more than 20mm high is another issue that even commercial scales have difficulty achieving.

I would suggest taking apart a commercial digital scale to see how it is constructed. Electronic postal scales are fairly cheap - and might have some hacking potential beyond just learning how it’s put together. The information you get from that would help you deal with off center loading and friction since those problems have already been solved.

9

rgbphil:
For more accuracy you’ll probably need to use a strain gauge plus instrumentation amplifier. This brings in a bunch of other problems like zero offsetting as well as self heating of the gauge.

Thanks for calling attention to self-heating of the sensor. I slapped my forehead. I’d been obsessing about ambient temperature fluctuations and thermal conduction between the sensor and weighing platform.

Actually, all your comments were very helpful.

The scale will be a computer peripheral. My intention is to include extra instrumentation, such as temperature sensors, and then use calibration profiles and gratuitous computation to make sense of everything.

I’d like to keep the parts cost down, but the reality is that the cost of my time will dwarf the cost of the parts by a couple orders of magnitude. So my general intention is to massively over-specify all of the electronic components.

My current thinking is that I’ll go ahead and buy a couple of the SFE sensors and build a crude scale to prove the micro-controller, computer interface, and software. Once the system is loosely demonstrated, I’ll revisit the force sensing subsystem and its mechanical design.

rgbphil:
I’m interested in what you are doing with multi-axis force sensing. I’m not sure you’ll get constant readings no matter what you do. The gauge will measure distortion, however I don’t imagine this would be constant if your robot arm is changing the orientation of the object being weighed. You might be able to build up a calibration scale that is dependant on your robots range of movement, which would be sort of cool. That is have it grip a known weight then move though its various degrees of freedom measuring the gauge at each…though I suspect this would only be practical if the number of degrees of freedom is limited to one or two axis.

Calibration will be interesting! I don’t expect to achieve precision with the robotic grip. There are too many variables. Plus, I’m coming from a place of ignorance.

From a gripping standpoint, it might actually be more important to detect if you’ve gripped the object away from its center of mass. In other words, to detect torsional forces as you attempt to lift/grasp. That might end up being more valuable than the ability to crudely weigh the item grasped. Although both would be helpful, depending on the task being performed.

I’d like to finish the scale project in the next few months. The robotic grip is a long term research project. A big problem with all of my projects is that I live in an apartment building. Despite good sound proofing, I can only perform noisy machining operations during weekdays, when most people are away and the noise won’t offend. With other demands on my time, projects go slowly.

The mechanical aspect of these projects is the most intimidating to me right now. My past experience has left me without an intuitive sense of force distribution, so I’m sure there will be a few embarrassing moments along the way.

I made a note on my dashboard to stop by and post updates.

10

signal7:
The resolution you are seeking may be impractical. If you’ve ever tried to weigh anything in the milligram range, you very quickly begin to realize that the slightest things will throw off the measurement. Furthermore, the requirement that the housing be no more than 20mm high is another issue that even commercial scales have difficulty achieving.

I agree. Actually, the reason for this scale project is precisely because I can’t find a commercial scale that satisfies my requirements. If I could buy such a scale, I likely wouldn’t build one from scratch. Although I certainly value the educational aspect of the project very highly.

When I was young, I was interested in electronics as a hobby. Now that I’m older, I find myself drawn to building useful instruments because commercial off-the-shelf solutions don’t satisfy my requirements.

signal7:
I would suggest taking apart a commercial digital scale to see how it is constructed. Electronic postal scales are fairly cheap - and might have some hacking potential beyond just learning how it’s put together. The information you get from that would help you deal with off center loading and friction since those problems have already been solved.

I recently did just what you propose. This whole scale project was motivated by my dissatisfaction with commercial kitchen scales. I prepare food from scratch, in small quantities. In terms of your health, there is a big difference between 500mg of salt in your daily bread and 1500mg of salt in your daily bread. I already use a scale to measure bulk ingredients like flour and water, but most scales are not accurate or precise enough to measure the minor ingredients (salt, oil, yeast, herbs and spices).

Anyway, I disassembled my old electronic kitchen scale (1 gram precision, unknown accuracy).

The scale sensor is a cantilever (beam) load cell. The load cell dimensions are 80mm long by 13mm wide by 13 mm high. In other words, 0.5" square, metal bar stock.

A sticker on the side of the load cell reads either “6060” or “0909”, depending on orientation. This could be an allusion to a 6060 aluminum alloy, or just a coincidence?

Two, co-incident holes have been drilled horizontally, in the middle of the load cell, to form a cavity in the customary style. The cavity cross-section is reminiscent of the number “8”, turned sideways like an infinity symbol. If you perform a Google Image search for beam-style load cells, you’ll understand what I’m talking about regarding the cavity. Presumably drilling two, co-incident holes is cheaper than milling a proper cavity in the beam.

The beam load cell is mounted horizontally, as you would expect. One end of the load cell is screwed to the plastic floor of the scale with two screws. The other end of the load cell floats in space. Two holes are drilled vertically in the floating end of the load cell.

The floating end of the load cell has an “E” engraved on the side.

An injection molded plastic weighing platform sits atop the floating end of the load cell. Two vertical pin formations on the bottom of the weighing platform mate with the vertical holes drilled in the floating end of the load cell. This configuration prevents the weighing platform from shifting or rotating on the load cell.

A drawn aluminum plate hangs over the top of the plastic weighing platform. The aluminum plate is not fixed and may be lifted off the scale for cleaning.

Strain gages are mounted on both the top and bottom surfaces of the beam load cell. Three varnished conductors emerge from the cement on each of the top and bottom surfaces of the beam. These conductors are soldered to a small circuit board on the fixed end of the load cell. Four conductors issue from this PCB on the load cell: red, black, white, and green; the standard colors for a four wire load cell.

Presumably, the strain gages on the load cell number four, two on the top of the load cell and two on the bottom.

I haven’t analyzed the scale’s PCB in detail (not to be confused with the load cell PCB). Near the sensor solder point on the scale PCB is a Texas Instruments LMV324I op amp (http://focus.ti.com/lit/ds/symlink/lmv324.pdf). The particular IC is the 14-pin SOIC package that the data sheet designates “D (R-PDSO-G14)”.

At first, I presumed the strain gage was configured in a Wheatstone bridge, but the LMV324I is a quad op amp. So I guess I’d have to analyze the traces and other passive components to be sure.

The scale has a simple LCD display. The scale’s micro-controller cannot be readily identified because it uses that packaging technique common in 1980s era digital watches: the semiconductor is affixed directly to the PCB, the conductors soldered with fine wire, and the whole thing covered with a mound of unmarked, black, melted plastic.

A picture is worth a thousand words. If I can get my web hosting sorted, I may try to post some tear-down photos of the scale.

11

Re the inside of your commercial scale, quad op amps are a traditional way of making instrumentation amplifiers from scratch…special purpose instrumentation amps are just quads connected internally. Look at the datasheet for the AD8555.

Cantilever…that’s the word I was looking for before. Sounds like a load cell to me. Spreading a number of single axis gauges is also common. I must admit to thinking of a strain gauge with a full bridge already on the foil as that’s what I’m using, but they’re probably cheaper with single axis gauges. Note you also may have only one or two gauges with the other arms of the bridge being fixed with resistors.

It’s interesting your kitchen scale actually uses a load cell…most I’ve heard of will use some sort of slider moving a pot or in the better ones some sort of optical sensor. The tiny jewelery scales might give a better idea of a high precision setup.

Alternately you might get away with keeping the load cell as is, and connecting up a modern instrumentation amp, with your own micro and display for giggles. It’s quite reasonable to assume with a different amp/sensor circuit you’ll get very good accuracy. Again its the issues of self heating, conduction (which you’ve noted) and calibration that will be a pain. However they’re not impossible.

Phil

Phil