# A question about choosing, implementing and placing a strain-gauge amplifier

This will be a little long question. Since I was and still am alien to some concepts, I will go step by step. Basically it is about implementing a strain-gauge amplifier and wiring configurations for a scenario. Sorry for the detailed way of writing, it wasn't possible to ask it in a shorter way.

Above is the definition of differential signaling. The Wheatstone bridge is under the category balanced differential signaling since it has mirrored signaling output around a common mode voltage:

Vsig = (Sig+) - (Sig-) Vsig = Vexc * ΔR/R

The strain-gauge I will use basically uses a full Wheatstone bridge. Here is the data-sheet. And here is the basic specs:

On the definition of sensitivity: 500με/mV/V means if the sensor is excited using 1V input and reads 1mV output, it corresponds to 500με strain. If the input excitation is 2 Volts, then the same 500με strain would yield 2mV output.

The measurements will measure +-100με which means the output will be 0.2mV at 1V excitation. The datasheet says the max possible excitation is 10V. So I will use a 9V power supply for the excitation which means the output Vsig will be max around 1.8mV.

My first goal is to amplify this 0-2mV range of output by an amplifier which has a good CMRR with a good noise immunity.

I have two challenges.

The first one is about dealing with the practical wiring and reducing noise during the transmission line. In other words where to put the amplifier how the wire the transmission lines ect.

The second one circuit-wise and is about choosing and implementing a good topology and an opAmp model. So I will divide the question into two parts.

Here is the illustration of the scenario where the distance between the strain gauge and the DAQ is at least 10 meters:

As you see in the above figures there is two ways of wiring. In Figure1 the amplifier is as close as possible to the gauge and the amplified signal is delivered in a single-ended fashion by a BNC cable. So BNC is longer than the twisted wiring.

In Figure2 the amplifier is as close as possible to the data-acquisition board and the balanced differential signal is delivered to the amplifier but this time being the long path and BNC is the short one.

Here is my first question:

1-) Which configuration is better Figure1 or Figure2? Im asking because here is my dilemma: If I use Figure1 gauge mV level signals will not be prone to noise much because the distance to amplifier is short but the single ended path is long. And in Figure2 the mV level signal path is too long to travel(?) even-though differential pair is more immune to noise the voltage is tiny(worry of SNR). So I'm confused which would be a better choice for the signal transmission.

Then comes the second challenge about the amplifier.

I will here focus on two options/references I encountered during my search. Here is the first one:

https://focus.ti.com/lit/an/sloa034/sloa034.pdf and here is the second one: http://www.linear.com/solutions/1183

Here simulated the single opAmp amplifier from the first link reference paper:

As you see ratio of R3/R1 sets the gain and input voltage is amplified from 0...2mV to 0..2V with around gain of 1000. But since the impedance seen at the negative input varies a bit, the gain varies slightly which is the third plot. Same paper gives two or three opAmp topologies which dont have this issue.

And in the second link from LT, there is a schematic. But when I download their simulation, it is different than in their figure in the link and it is as below:

This above is different than their schematic:

In the first schematic R11 is 100 instead of 100k and the it is using +-15V power supply not a single supply.

Here is my second question considering these:

2-) I never built a strain-gauge amplifier before. All I know that the opAmp should have a good CMRR. It seems like my first plot reveals the fact that using simple one opAmp amplifier might cause linearity problems. But then there are different topologies which uses two or three opAmp. Which configuration from these two papers you would prefer? If I implement LT’s amplifier which resistor sets the gain? R11? Im also confused why in their simulation R11 is different than the one on their figure and why do they use single supply in one of them and +-15V supply in the other. Which one is correct? And I know I shouldn't ask for an opAmp but what is so special about this LT6081(for being used in strain-gage amplification) and is there another alternative I can use? And what is CMRR trim?

• There are many eval boards from different manufacturers Analog Devices, TI,.. Apr 2 '17 at 7:14
• Figure 52 is NOT what you want - you should remove it because it removes static levels from the bridge output and I believe this is not what you want. It adds confusion to anyone else answering. On a more general note, your question is really quite long and adding stuff to it isn't bringing anything else to the party. On a really general note, modifying your question to incorporate suggestions in answers confuses anyone reading this - I would urge you to roll back the question to what you previously had. Apr 2 '17 at 11:41
• okay I rolled back. And regarding Marko Bursic's comment those eval boards have too many components on them to be placed; instead of 1 inamp I see many ICs. Is that because they are made of single opamps instead of a single inamp? Apr 2 '17 at 11:44
• @user1234 Did you have a chance to use the above mentioned AD8221 amplifier for your strain gauge measurements? If so, what did you consider for the REF terminal (2.5 V or connected it to GND) and what was the corresponding voltage for zero strain? Mar 6 '18 at 0:44

Firstly, regarding the selection of a sensor amplifier, go straight for an instrumentation amplifier such as the AD8221 (that's what I use for bridge and single sensor amplifiers like strain gauges): -

The advantage of an InAmp is that its input impedance is very high so it won't load the bridge and there is just one gain-set resistor needed.

Regarding the placement of the amplifier, it's a little bit open. I sometimes have up to 20 metres of cable between sensor and amplifier because the environment that my sensors are used in is usually well above 200 degC so there aint no decent silicon that can sit close by! In this scenario my DAQ is very close to my amplifier (inches away) and I use a single-ended connection to a single-ended ADC.

So, in practise, both circuits are used.

As for the theoretically best implementation for reducing noise I stand by my answer to your earlier question; a balanced signal does not need to be a differential signal - it only has to present a decent impedance balance to earth.

• Thanks a lot, please see my edit. There are some minor concerns: 1-)First of all for negative -5V I want to use a ICL7660 Voltage Converter. Is that okay? 2-)This article analog.com/media/en/technical-documentation/application-notes/… implements RFI filter at the input, do you recommend? 3-) How can I make that 2.5V offset adjustable 4-)Should I use a buffer or an RC filter at the output to the ADC, please see this discussion: electronics.stackexchange.com/questions/227645/… Apr 2 '17 at 11:39
• You don't need a -5V rail for what I've shown. RFI filters can be useful but are they always needed, no, certainly not. Why would you want to make the 2.5 volts adjustable? An ADC should usually have anti-alias filters so yes, an RC at the ADC input is usually required (more so than RFI filters at the InAmp input). Apr 2 '17 at 11:45
• One last thing, oh so this can be used with single supply as in your schematic? And what can I use for a 2.5V reference? Not a voltage divider right? thanks (I think I will start implementing this as a prototype but because of surface mount too hard to solder on breadboard or perfboard) Apr 2 '17 at 11:50
• Use a precision voltage reference - it can be shared amongst multiple channels too. Sometimes I feed the REF pin with a DAC output so I can auto-balance the inputs (more of a customer requirement of course). The 2.5 volt ref can also be used to implement a precision 5 volt excitation voltage too. You can get little PCBs that convert SMD to a 0.1" pitch but you'll have to search for these. Apr 2 '17 at 11:55
• theres a similar question on voltage reference here: electronics.stackexchange.com/questions/48757/… seems that I can trust this IC: ti.com/lit/ds/symlink/lm336-2.5.pdf (?) Apr 2 '17 at 12:17

1-) Which configuration is better Figure1 or Figure2?

figure 1 is better than figure 2, but neither is great.

for high impedance source + over long distance transmission, you want the amplifier to be as close to the source as you can, and the transmission be done in terms of current, and preferably digital and with isolation.

so the best would be to put the acq as close to the sensor as you can. and transmit digitally; many solutions based on ADC/MCUs with onboard high resolution ADC would be my choice.

All I know that the opAmp should have a good CMRR.

yes. go with a truly balanced solution - aka instrumentation amplifier. if it cannot be done, consider a 2-opamp solution so you minimize the requirement on the gain-setting resistors.

• thanks but any of the two links I provided is a truly balanced solution as you mentioned? Apr 1 '17 at 21:19

To answer your question about the LTC6081, I would say it is used for the following reasons:

1. Low noise
2. High CMRR (100dB)
3. High PSRR (98 dB)
4. Rail to rail operation

Besides all that, it can operate on a single 3.3 or 5V rail. This makes easy to use in typical circuits where you have those voltages available for the other circuitry (such as an ADC and/or microprocessor.) Easy to use in that you don't need a separate supply for the opamp.

• Thanks- The problem is I need a DIP package. This amplifier sold as surface mount uk.rs-online.com/web/p/operational-amplifiers/0423680. I cannot solder surface mount. Do you know any similar one I can use instead? Apr 1 '17 at 21:36
• Nope, but soldering that kind of thing isn't that difficult. I solder sot23 parts with 6 pins by hand, as well as 0402 sized resistors/capacitors/inductors. Perfomance wise, the LTC2050 or LTC2051 would probably be better - but that's the little sot23-6 bugger. Anything modern with good performance will surface mount. Might be a good idea to learn how to work with them. It really isn't hard.
– JRE
Apr 1 '17 at 21:44
• But I have to make a prototype so I will use perfboard not PCB. Do you think I can use these on perfboard? Apr 1 '17 at 21:46
• There are adapters that you can use. Small board with the smd footprint, somes traces, and pins for the perfboard. Or have a pcb made. Given the signal levels, that'd probably be better. Design, simulate, layout, order, assemble. Depending on how fast you the board made, the prices range from cheap (weeks of delivery time) to outrageous (overnight manufacture and delivery.)
– JRE
Apr 1 '17 at 21:51