Strain Gauge 8.3 × 10⁻⁶ Impossibleness - How?

[I wanted a rude username]

An all-purpose ape asks an interesting question, replicated here in full:

> While disposing of my old concrete patio at the local landfill, the attendant signaled me to pull my SUV/trailer forward onto the 60 ton truck scale indicating 7304 lbs (3320kg). I walked off the scale to visit the cashier, noticing the display value dropped by 185lbs (84kg) reflecting my body-weight and accurately, to within 1/2% - in the context of the scale's capacity that division is 8.3e-6. A strain gauge and circuit configured to supply an 5vd ADC input over a variation of 60 tons would go from 5v down to .000004v. 24bit A/D converters have that resolving power, but the analog signal conditioning / amplifiers etc... seem impossible to me; wind was blowing, my engine was running, resonance, microwaves from nearby phones, nearby behemoth landfill tractors scraping the earth. Is this a feat of DSP or perhaps optics instead of resistance based strain gauges? Perhaps a highly segmented under-weldment with 100 individual strain gauges? I am looking forward to comments!

Discuss.

[Someone]

0.1% accurate or better plus resolution beyond so that sort of short term measurement is entirely plausible. Probably electronics tightly integrated in the (sealed and shielded) load cell + heavily damped and slow system.

[Dundarave]

With precisely one single datapoint, I’m not sure how anyone could generalize about the measuring system.

It could have been an unreproducible fluke that his “measured weight” was so close to that of his bathroom’s scale.

[amyk]

Here's the specifications of one 60-ton truck scale:

https://acmescales.com/catalog/industrial-products-mettler-toledo/industrial-products-mettler-toledo-truck-rail-scales-mt/mettler-toledo-vtc190-truck-scale-2/

10000 increments of 12lb, so basically 4-digit accuracy -  easily reachable with a 16-bit ADC.

[CatalinaWOW]

I have noted the same thing.  Two partial explanations come to mind but don't fully explain it.

1.  Many of these segment the platform into two or more segments, weighing the front and rear axle(s) separately and then adding the results.  Splitting it left-right is the other obvious move.  Cuts the dynamic range by a factor of two or more.

2.  Low bandwidth/long time constant.  Watching the display settle it isn't too long, on the order of a second, but definitely enough to eliminate engine vibration and many other high frequency noises.

[Kleinstein]

Weight scales are quite demanding applications and they do use quite a bit of effort to make them stable and low noise.
The better load cells / strain gages usually use some low frequency AC excitation and a careful thermal design. Having the scales outside makes the thing quite demanding. For the short term the slow thermal response helps.

What was observed was the low noise and short time stability, not really the accuracy. The performance of many commercial scales is still impressive. Not all of them use strain ganges - some of the lab scale ones also use electromagnetic force compensation.

[Echo88]

https://www.planetanalog.com/wp-content/uploads/1976-10-05_30ppm_scale.pdf

[armandine2]

https://www.analog.com/en/education/education-library/transducer-interfacing-handbook.html

see chapter 12 page 194 for analog strain-gage design example - Precision Platform

[I wanted a rude username]

The title of that Planet Analog article dated OCTOBER 5 1976 ...

> This 30-ppm scale proves that analog designs aren't dead yet

It's clickbait from before there were clicks. Except the content is actually good.

[Alex Eisenhut]

https://www.analog.com/en/education/education-library/transducer-interfacing-handbook.html

see chapter 12 page 194 for analog strain-gage design example - Precision Platform

Beautiful, engineering gold right there

[David Hess]

see chapter 12 page 194 for analog strain-gage design example - Precision Platform

That is a fast settling design.  The analog filter time constant is slowed after the initial change in weight to remove additional noise, so the slow time constant of the filter does not extend the settling time.

Here's the specifications of one 60-ton truck scale:

https://acmescales.com/catalog/industrial-products-mettler-toledo/industrial-products-mettler-toledo-truck-rail-scales-mt/mettler-toledo-vtc190-truck-scale-2/

10000 increments of 12lb, so basically 4-digit accuracy -  easily reachable with a 16-bit ADC.

16 bits or slightly less is realistic because in practice accuracy is limited to perhaps 1 in 20,000 because of mechanical considerations with the load cell and strain gauge mounting.  In the past it was also tough to minimize drift in the signal conditioning and achieve 16 bit *accuracy* in the ADC, but that may be considered an inexpensively solved problem today, albeit still not intuitive.  The mechanical aspects of the load cells and strain gauges still have a lot of art to them.