-
exciting a strain gauge load cell with AC
Posted by
Gibson486
on 02 Jul, 2018 14:11
-
Anyone here ever excited load cells with AC instead of a standard DC reference? I have a few questions....
So, it goes into your differential amp....I need to bias it...that also would mean that I have to ensure the each sin wave cannot extend past the rails? if so, then I need to ensure the op amp is powered by a source that is well beyond teh signal amplitude? Or, do I only care about the common mode voltage?
What would be the reason to move to AC excitation?
-
#1 Reply
Posted by
CopperCone
on 02 Jul, 2018 16:28
-
You move to ac because you can get high noise rejection since you are capable of using a bandpass filter and you can eliminate 1/f noise which is low frequency and can dominate your measurement. You can also get rid of offset errors from silicon. Easier to design and catagorize then an ac system of equal stability imo
You get hit with ac problems, i.e. high drift coefficents of capacitors vs resistors used for any filters and nonlinearity from the inductance and capacitance of your gauge.if the frequency is high enough you might have problems from bending wires and stuff. And distortion means you lose signal power. So you get alot of weird errors that are difficult to track down.
-
#2 Reply
Posted by
JS
on 02 Jul, 2018 17:17
-
AC cant go past the rails, at least not in most amps, but what's your setup?
You could switch the DC with an H-bridge and average a few measurements in both polarities to get rid of offset without the trouble of dealing with AC. In any case it's a lock-in amplifier, multiplying for a logic signal is easier than a sine and can be done in software, and keep the analog setup almost the same, but the addition lf the H-bridge driving the load cell.
JS
-
#3 Reply
Posted by
Gibson486
on 02 Jul, 2018 17:28
-
Tried it, it sort of worked.
Problem is that the wave coming out of the output is being clipped. I thought it was a biasing issue, but that did not seem to be the case. Instead, you have to bias the signal to max voltage. So, if I bias my signal to a 2.5V midpoint, that would mean I am effectively giving it a 2.5V reference instead of a 5V reference, so I only get half the output scale. Once I did that I got both sides of the sin wave at the output. Not sure why that happens, so I am still trying to analyze it. I put the a low pass filter at the output, and it worked pretty well, although I have not measured it an ADC yet, so time will tell.
-
#4 Reply
Posted by
Gibson486
on 02 Jul, 2018 17:30
-
AC cant go past the rails, at least not in most amps, but what's your setup?
You could switch the DC with an H-bridge and average a few measurements in both polarities to get rid of offset without the trouble of dealing with AC. In any case it's a lock-in amplifier, multiplying for a logic signal is easier than a sine and can be done in software, and keep the analog setup almost the same, but the addition lf the H-bridge driving the load cell.
JS
I will look at that more....the last issue I am running in to is the 1/f noise when just doing standard DC load cell with stable reference. For 24 bits and above, it is a huge pain. Hopefully this can alleviate some pain.
-
#5 Reply
Posted by
JS
on 02 Jul, 2018 19:04
-
Why not a chopper amplifier, which does the lock in for you all together, to much noise?
There is an interesting circuit I've build and tested using the technique I mentioned, is an arduino based milliohmeter, using an ads1115 and many pins in parallel to toggle ~100mA to drive the DUT. I've measured with some consistency 2mΩ pieces of wires and 2 digitsbto spate so in the order of 10μΩ LSBs. ADC is only 16bits and even less useful bits after considering every error source, but to meassure few μV without much consideretion of thermals and still get consistent results isn't something you can do, with this setup, without toggling the polarity. It does work great to deal with the offset, in the code it takes 100 samples in each polarity, goes back and forth twice and averages all the samples. You could toggle faster and more times, depending on your adc sampling freq and your desired update rates. Kind of hard to deal with 1/f noise if you want several updates per sec other than a lock in amp.
JS
-
#6 Reply
Posted by
Gibson486
on 02 Jul, 2018 19:18
-
Is there a instrument/differential amp that is a chopper? I have used them before in a circuit to control pressure regulation, and they are rather nice because of the minimum offset voltage. Either that, or I could use it right before the ADC. Can you use chopper op amps as filters? I have only used it as a buffer.
-
#7 Reply
Posted by
JS
on 02 Jul, 2018 19:22
-
Is there a instrument/differential amp that is a chopper? I have used them before in a circuit to control pressure regulation, and they are rather nice because of the minimum offset voltage.
Yes, that's the idea, offset is very low and stable, 1/f noise is eliminated but noise is usually higher than low noise opamps. How much voltage do you have in your cell at full scale? That would tell a ballpark what you need in terms of noise to make use of the 24 bit ADC, and how many usable bits are you expecting...
JS
-
#8 Reply
Posted by
CopperCone
on 02 Jul, 2018 19:31
-
search for zero offset / chopper differential/instrumentation amplifier, there are a few iirc.
And you can build them out of discrete parts, or make a chopper amplifier, but the noise will be high compared to using a sine wave circuit if you work small signals.
Choppers have pretty bad low frequency noise. I think you are mistaken.
-
#9 Reply
Posted by
rhb
on 02 Jul, 2018 22:26
-
I've had a look through my library. Generally all the discussions of strain gauges are DC. This included a monograph solely devoted to strain gauges. However, I did find this in:
Mechanical Measurements
Beckwith & Buck
Addison-Wesley 1969 2nd ed.
As it points out, use of AC adds stray reactance to the problem. You might want to try a constant current source and see if that helps.
-
#10 Reply
Posted by
tecman
on 03 Jul, 2018 15:14
-
AC bridge excitation is common. Using a synchronous demodulator offers the best noise rejection and stability. Problem arises if you have longish wiring to the sensor. Phase shift due to wiring capacitance needs to be compensated. Higher excitation frequencies offer better bandwidth of the output, but cable capacitance issues are a greater factor.
Paul
-
#11 Reply
Posted by
N2IXK
on 03 Jul, 2018 15:47
-
Another reason for AC excitation is in torque sensors, where the strain gage bridge is on a rotating shaft, and the excitation and signals are coupled onto the rotating shaft by means of rotary transformers. This avoids the noise and mechanical wear inherent in slip rings.
-
#12 Reply
Posted by
JS
on 03 Jul, 2018 16:01
-
search for zero offset / chopper differential/instrumentation amplifier, there are a few iirc.
And you can build them out of discrete parts, or make a chopper amplifier, but the noise will be high compared to using a sine wave circuit if you work small signals.
Choppers have pretty bad low frequency noise. I think you are mistaken.
No, LF noise is reduced by the chopper, given the chopping freq is higher than the 1/f corner.
It is interesting to consider the effects of a chopper amplifier on low frequency 1/f noise. If the chopping frequency is considerably higher than the 1/f corner frequency of the input noise, the chopper-stabilized amplifier continuously nulls out the 1/f noise on a sample-by-sample basis. Theoretically, a chopper op amp therefore has no 1/f noise. However, the chopping action produces wideband noise which is generally much worse than that of a precision bipolar op amp.
From here:
http://www.analog.com/media/en/training-seminars/tutorials/MT-055.pdfAs you usually want them for DC and LF applications, you can limit the bandwidth making the wideband noise not a problem compared to the 1/f noise, and chopper takes care of the 1/f noise!
JS
-
#13 Reply
Posted by
Kleinstein
on 03 Jul, 2018 18:36
-
Chopper amplifiers can help with 1/f noise and amplifier drift / offsets. However good AC amplifiers are still lower noise. In addition AC excitation also eliminated thermal EMF from the bridge. With AC excitation there is also the option to use a transformer, especially for the drive side.
Parasitic capacitance and wire reactance are mainly a problem with higher AC frequency - so the required bandwidth for the strain signal is an important factor to decide between DC with chopper amplifier and AC excitation. If only low frequency signals are to be detected AC excitation can be better. With fast signals DC excitation gets better.
The normal way is to combine AC excitation with phase sensitive detection (lock-In technique). With modern, fast ADCs it is possible to do the demodulation with the ADC and software.
-
#14 Reply
Posted by
CopperCone
on 03 Jul, 2018 18:48
-
When I remember looking at datasheets I saw that the chopper amps had like 5-10 times higher noise in the 10Hz bandwidth then low noise amplifiers.
Something like 100nV between 0-10Hz and 1.5uV between 0-10 Hz. (chopper vs LNA)
I just considered it 1/f noise, maybe its a designation error on my part. Is considering the 0-10Hz bandwidth to be 1/F noise region wrong?
Does the noise profile look different between the two amplifiers some how? I never compared the curves. To me for sensors 10Hz is a pretty good area for alot of processes like temperature, flow, resistance, mass and pressure that I was concerned with. Seemed to be a appropriate general match for the time constants of the stuff that I worked on. More was unnecessary and less was annoying to look at.
-
#15 Reply
Posted by
SeanB
on 03 Jul, 2018 19:09
-
Aircraft instruments use 400Hz AC at 6V to drive transducers, including things like fluxgate magnetic field sensors. You need a reasonably low distortion drive, and can get 14 bits out with ease, and the amplifiers are nothing special, just your run of the mill jellybean opamp ( Mil spec rated version of the 741, but the regular commercial version was just as good there) and a synchronous detector to sample at the peak of the sine wave to drive the ADC input.
Drive your bridge with AC, and use an AC amplifier to get the differential signal out, and add enough gain to get it high level before the demodulator. AC coupled stages do not have to have too much DC stability, and the offset voltages are removed by the high pass coupling capacitors between stages. Only time you need stable DC and low offset is the final DC amplifier before the ADC, after the synchronous rectifier, and here the signal is easily much larger than the drift or noise. Even the lowly JFET transistor does a good enough job of making the demodulator, and you can get much better analogue switches these days.
Look up aircraft signal conditioning, it all runs on AC, and over long wires in a pretty RF hostile environment as well.
-
#16 Reply
Posted by
CopperCone
on 03 Jul, 2018 19:13
-
beware capacitor drift
-
#17 Reply
Posted by
JS
on 03 Jul, 2018 19:36
-
If going for sinewave instead of square wave driving you should use a simple generator and a multiplier and not to worry about drift, only phase uf you are working at high frequency
JS
-
#18 Reply
Posted by
CopperCone
on 03 Jul, 2018 19:40
-
you can always measure RMS you don't need to do a lockin amplifier
-
#19 Reply
Posted by
rhb
on 04 Jul, 2018 16:39
-
What are you trying to measure? And what are you using for the ADC?
-
#20 Reply
Posted by
Kleinstein
on 04 Jul, 2018 17:35
-
you can always measure RMS you don't need to do a lockin amplifier
Measuring RMS is not a good idea with AC excitation of a bridge: first there is additional bandwidth and thus more noise even if used with a good band-pass filter. The second point is that one looses the "sign" - with a phase sensitive detector one also gets the phase and this way a sign for the output voltage. Finally RMS converters are usually not very linear/temperature and stable and may be more complicated than Lockin like detection.
The noise spectrum of AZ amplifiers and normal OPs looks quite different. AZ OPs have essentially no 1/f noise - some (chopper type) have a larger noise somewhere in the 10s-100s of kHz range and others have a larger noise below some 1 kHz than above (sampling type AZ).
The 0.1-10 Hz range is often used for the low frequency noise that is often dominated by 1/f noise, but not always. When it comes to the 0.1 -10 Hz range, there is not that much difference between low noise AZ and classical OP. Below some 1 Hz the AZ OPs can be lower nose.
-
#21 Reply
Posted by
CopperCone
on 04 Jul, 2018 17:39
-
a PSD has charge injection and phase drift if you need to filter it to prevent a amplifier from saturating. Digital one is hard if it needs to be fast and may have other downsides like control loop stability. One of those modern RMS/DC converters is actually a really nice easy to use chip. It will be easier to design then a lockin amplifier and may work well. You will also not lose signal power to distortion (well I guess you might not really lose it to a simple lock in, like one that uses a single gate as a switch, since you will get the harmonics power used). It will also not require any kind of phase trimming like a analog lock in would.
But yes you will lose polarity information. I imagined it being used on a scale or something that does not reverse current.
It is easier because you can just look at the oscillator, filter attenuation and your RMS/DC chip specs to get a good design number.
-
#22 Reply
Posted by
Kleinstein
on 04 Jul, 2018 19:13
-
The usual analog RMS/DC converters have poor stability specs - you start to pay a premium if it needs to be better than 0.5%. The rather accurate LTC19xx chips are no longer made. A simple PSD build around an HC4053 and a simple OP is usually much better. Phase shifts are a minor problem as they go with a square law - so it is only the square of the phase error that enters and one can usually keep the phase error small enough so that the square will make is no a real issue. At a relatively low frequency phase trimming may not even be needed as the phase error can be small be design. Even with higher frequencies phase adjustments tend to be not that difficult and usually does not change much over time.
At the high signal level of the detecting PSD charge injection is not such a big deal either if the signal quality is good (like in a DMS bridge). Charge injection is more like a problem for the input chopper of an AZ OP - but even these work well today.
Loosing the sign means one would lose half the dynamic range and have to stay away from the balanced bridge. As many RMS converters don't work well at less than 10% of the nominal signal the loss is even larger. Starting at around 60% full scale makes gain drift also act like an offset drift. Phase shifts and capacitor drift gets a big issue in a steep band-pass filter needed with an RMS converter. So a RMS converter is more like a poor way and I can hardly imagine any real advantage compared to a DC bridge with using AZ OPs.