Strain gauges produce extremely low level signals, so one must use special care when using them.
One thing you should look at is "Kelvin connection" 4-wire sensor connections. You need this to remove the resistance of the current source wiring.
You will also need to learn about instrumentation amplifiers if you have not done so previously. An instrumentation amplifier is a particular configuration which is stable with high fixed but configurable gain. You can build up the instrumentation amplifier configuration from multiple (three) OP AMPs or buy them as devices. If you build them from OP AMPs be sure you get very low noise devices.
You will likely need sigma-delta ADCs to get a enough bit depth. You will likely find that you will be looking at very small differences of readings, so you need a lot of bits to give you resolution in the changes. Twenty bits or more may be needed to get eight or ten bits of actual signal.
Strain gauges are usually mounted in pairs with one gauge fastened firmly to sense the strain and the other not strained to provide temperature compensation. Depending on your application, you may not need this, but at very low strain levels lack of temperature compensation will become an issue.
Electrical noise is especially troublesome at the low signal levels of strain gauges. It is actually pretty hard to get stray electrical noise pickup below 5 or 10 millivolts on a system with a micro processor. Putting the instrumentation amps on a separate board separate from and shielded from the micro controller helps maintain the quality of the strain gauge system. There is also a lot of ambient electrical noise. Florescent lamps are a big problem. You may need to use shielded twisted wire for the sense pair. If you use shielded cable, make sure the shield is connected at only one end, the sense amplifier end. Currents in the shield induce signals in the signal wires. Do not overlook the sense current supply lines. Any noise on those wires couple directly in to your signal at the strain gauge.
Filtering the strain gauge signals will help with noise problems. You probably are not dealing with or able to handle much bandwidth with your system and thermal noise is proportional to bandwidth.
You can probably use MOSFET multiplexers instead of relays to select the different strain gauges. Make sure the supply and switch control wires are very well filtered, as noise on these connections can couple to the signal channels. Use double-pole configuration to maintain balance if you switch the low level signals. It is better to multiplex at the output of the instrumentation amplifiers. Multiplexing at the instrumentation amp output could be single-ended. This configuration would require a separate instrumentation amplifier for each strain gauge. You could use relays on the low level signals but you would need to switch both wires of each sense pair. This would need a lot of relays.
Some ADCs include multiplexing at the ADC. High end ADC hybrid modules offer options for either single ended or balance ADC channel input.
I would start small with a single channel.
If you do not have access to an oscilloscope you should not dismiss unsteady readings. They probably indicate noise problems.
Good luck! I have no doubt that you will learn a lot about strain gauges and signal integrity in general with this project. Few people deal with such low level signals.
Wow
. Thank you very much for your concise yet really informative explanation
. I have some understanding regarding instrumentation amplifiers and I was thinking of using a chip like
INA126, but I changed my mind because I thought that the ADC chip and the circuit arrangement that I was using did not need it (I will explain further on).
Moreover, I understand (and thanks for mentioning
) that I have to deal with a lot of noise rejection. This includes both thermal and electrical noise. (My sensors will be exposed to sun
).
I am quite new to the Kelvin connectors, I have been looking into them since reading your post any they seem fantastic for my application.
I think it is a good idea if I show you a schematic of the circuit that I am trying to use (which I have attached). I am using an
AD7730LBR bridge transducer chip for which I believe includes the functions of an ADC and an instrumentation amplifier. As you can see an
ADR431 chip has been used for voltage reference and power conditioning. I downloaded the schematic from
this website.
The general plan is to make a shield (a mountable circuit) with the circuit on the schematic implemented on it and then put it on an Arduino micro-controller module.
My main concern is to come up with a way to connect all the 24 strain gauges in turns to only one bridge circuit and connect the output of those bridges to my ADC, but here are the limitations that I have faced and my thoughts on them:
1. I really really do not want to have 24 bridges for every gauge. Also if I did, I am almost certain that the current draw from the ADR431 will be around 0.5Amps which is above what Arduino micro-controller and ADR431 can handle. I possible can change the chip, but even then there will be a lot of branching going on to reach all the bridges and it will be nasty
.
2. I am aware that there are multiplexes ADCs like ADS1243, but they are quite expensive compared to the ADC chip I have (I got it with discount too
). So, I understand that my choice of going with AD7730 will result in a compromise between expense and accuracy of results.
3. I have been thinking of using a multiplexer chip like a 4051 or 4067 MUX/DMUX chips ( Are these the sort that you had in mind?). However, I am not sure if that is possible and if it is, how or in what type of arrangement relative to the strain gauges? (Of course, I am keeping also in mind all the noise factors).
So do you think there is an arrangement where I can just use one bridge and use a MUX to switch between them?
Once again thank you very much for such a good post and appreciate any further ideas on this.
