Brake pressure sensor based LOAD CELL

[Tommy Cookers]

load cell manufacturers mainly like what we call shear because it gives less of what they call hysteresis
reminds me that magnesium alloys have a (potentially useful) higher ratio of shear strain to direct strain than do more usual metals
(the relationship between Young's modulus and shear modulus being significantly different)

this application could be satisfied with a much simpler device than those DW mentions

[DaveW]

this application could be satisfied with a much simpler device than those DW mentions

That is remarkable statement to make in the absence of a specification. I'm sure that Belatti (and I) would be happy to be told more....

A reference would help & comments on cross axis sensitivity, temperature coefficient, manufacturing tolerances, failure modes, gauge types, etc. Personally, I would not be happy about using Mg alloys for a load cell structure for several reasons (Is it possible to buy gauges optimized for Mg alloys?).

[Tommy Cookers]

Magnesium alloy I used in a 6 component balance because of the large dead load (a model oil rig about 8' high)
with shear gauge bridges and direct gauge bridges and analogue signal arithmetic
necessarily, the cross-axis sensitivities were trivial
the natural frequencies of this system loaded were enough for measurement fidelity (eg of wave loads), about 7 Hz minimum ?
the balance had a non-intrusive top force frame which realised good stiffness only when bolted to the (non-precision) test structure
that Mag alloy has a long elastic range, but at this stage was taken beyond it, without any effect on calibration etc
mystique only holds you back
btw with Magnesium you should avoid the usual acid wash

gauges are available with the 15 STC for this (15 ppm/degF linear expn coeff) material (used for aerospace and engine castings)
MicroMeasurements had then 68000 gauges in their master catalogue (this was available to relatively few customers)

experience showed me that most customers wanted and needed to be guided in specification
one internal customer refused to give any specification, and we went forward according to my judgement
with good results
this was a small scale model of breakwater block wave slamming loads, so having iirc around 80 Hz natural frequencies (in air)

[Belatti]

Thank you all for the inputs.

Lets go in order:

Damper loads at the ground are tiny, and can be safely disregarded if you are doing straightline aero tests.

I have seen the same thing with noise in the linear pots. What you need to do is use soft springs for the aero tests to increase your signal/noise ratio. You need some experimentation to set the static ride height to the correct height such that at the test speed you are running at the target height but its possible.

Another way to clean up the signal (which I haven't tried yet) is to measure vertical and pitch accelerations and use them to cancel out the oscillations in the damper signals.

Damper loads are never tiny in a 1300Kg tin top strait line testing at 250km/h, take into account that I dont have a NHRA smooth like mile to do the tests.

I cant use soft springs and then correct height either, I have only 1 test day per semester (tests banned is a world wide politic) and have to do my measurements at the same time Im testing other engine and suspension stuff, using a long straight track for the first and simultaneously set a partial laptime for the second.

[Belatti]

The trouble with trying to set your car up as an inertial base is that you are multiplying all your errors together, and if you start looking for small differences in large numbers then it all gets very ugly. So I think measuring the force directly has a lot of upside.

Exactly. I do like direct measurements. My idea is that damper pots are for measuring damper position and speed, period (not even ride height). I do work with cars that have been developed for years, even before I graduate, the rulebook is tight, so performance increments are rather small.

[Belatti]

loads can be quite easy to measure
the cost of 4 suitable load cells and a conditioning unit seems to be around $2000

or the amateur can design 'cells' around his own application's structural situation and make them, for a trivial expenditure
(this was my job for 20 years)
there is a lot of false mystique in this area, colleges are 50 years behind in their coverage of this technology
they seem unaware that gauges are essentially self-temperature-compensating (STC)
so eg for most amateur purposes including this one, temperature compensation needs no attention beyond the design stage
and anyway temperature compensation services will only compensate for (temporal) variations of spatially uniform temperatures
they will run a mile from anything else

if anyone is seriously interested in making these 'cells' I can contribute to the design process etc at no charge

btw
DWs need for load measurement following the choice of force as the nominal controlled variable in the active ride system ....
would have been alleviated if eg position or velocity had been chosen as the nominal controlled variable ?
which would be possible with some alteration to the control signals computed ?
eg if starting again today ? this would have practical advantages ?

Tommy, I live in a South American country governed by morons that set economic protectionist politics but with no incentives for companies that could develop replacements of the stuff that is missing. So importing $2000 load cells is a pain in the *ss, the leadtime is huge and the cost four times that.

[Tim.Wright]

Damper loads are never tiny in a 1300Kg tin top strait line testing at 250km/h, take into account that I dont have a NHRA smooth like mile to do the tests.

They really are small enough to be forgotton. I have done wheel force estimations based on suspension stroke and got a very reasonable correlation to the dynamometric wheels.

Given that:
1. You should be taking the readings in a single straight line section averaged over a time period of at least 1sec
2. The damper force is AC in style with an average force of zero

The damper forces should not be present in your results anyway, so it doesn't make much sense to go out of your way to measure them.

[Belatti]

Thanks Tim, as I said, I have been doing that but I need better tools. I cant measure a milimeter fraction with a ruler.

The context may be difficult for all you guys to nuderstand because, even if it is the major series in the country, we have usd 15.000 per race budgets (half or less of what big teams do have) and 70% of that is engine and tires, while 29% is wages and travel expenses...

Still, we are all the less amateur we can be.

[Tim.Wright]

What are your wheel rates?

[DaveW]

Apologies, but I feel I must respond to Tommy Cookers' last post.

... with shear gauge bridges and direct gauge bridges and analogue signal arithmetic ...

That sounds as if you were following the thought posted by Greg, which he referred to as "SIMO" (I actually thought he meant "MISO", but in your case it would be "MIMO"). If that is true, then I think you have moved the complication from gauging to signal conditioning, and exhaustive calibration. I guess each copy will have its own transformation array (or analogue signal arithmetic gains).

necessarily, the cross-axis sensitivities were trivial

I guess you mean gauge sensitivities, but I was concerned about structural cross coupling, which you will have calibrated out (If you've guessed lucky, according to Greg).

... which realised good stiffness only when bolted to the (non-precision) test structure

That idea bothers me. It suggests that the attachment details, and the way it is torqued up, will affect the transformation matrix...

More seriously, I have in the past done crazy things like strain gauging an aircraft wing, and then spent an age - first finding the sensitive gauges, then wiring those up to amplifiers to estimate individual gains, before completing the wiring and SCU build, and finally calibrating the finished channel(s).

The technique works, eventually, after a couple of iterations. In my case, with both wings & tailplane gauged, I was able to identify the fuel state of the aircraft with the help of a pull through maneuver (That wasn't the objective, but it did supply evidence that the idea actually worked). But it not something I would advise for a "production" load cell. In the case of the Lotus F1 car there were four off per car, three cars, and then spares to build....

[Greg Locock]

by SIMO I was referring to calibration process, one well defined force or moment, look at the result on many strain gauges. It is directly equivalent to traditional modal analysis with a single shaker, or multiplane dynamic balancing.

The analysis of a run on the track would be MIMO. This'll only work with a linear system, and redundancy would be a great aid to accuracy.

[DaveW]

Apologies, Greg, I misrepresented you.

In the case I described, keeping it simple, I festooned the wing with loading points, which allowed me to apply a dead weight at a range of positions, both span-wise & chord-wise. Each application of the load generated a set of output voltages from the array of strain gauge bridges (Multiple Input). The task was to generate a scaling vector to reduce these to a single output that represented the applied load. Hence MISO.

[Tommy Cookers]

Apologies, but I feel I must respond to Tommy Cookers' last post.

... with shear gauge bridges and direct gauge bridges and analogue signal arithmetic ...

That sounds as if you were following the thought posted by Greg, which he referred to as "SIMO" (I actually thought he meant "MISO", but in your case it would be "MIMO"). If that is true, then I think you have moved the complication from gauging to signal conditioning, and exhaustive calibration. I guess each copy will have its own transformation array (or analogue signal arithmetic gains).

necessarily, the cross-axis sensitivities were trivial

I guess you mean gauge sensitivities, but I was concerned about structural cross coupling, which you will have calibrated out (If you've guessed lucky, according to Greg).

... which realised good stiffness only when bolted to the (non-precision) test structure

That idea bothers me. It suggests that the attachment details, and the way it is torqued up, will affect the transformation matrix...

this (30 years ago) was a 6 component 'dynamometer' balance I designed to have negligibly small cross-axis sensitivities
the 'negligibly small' criterion was probably eased by really only 4 ? components being significant to this project
some economies of activity were needed, in principle each test could have involved measurements from about 200000000000 waves
in short, Magnesium alloy allowed better sensitivity both by the mechanism previously stated and
due to its low moduli allowing a thicker section (helping 'column strength' needed for large roll and pitch moments)
4 members mechanically in parallel, each gauged with 2 bridges each representing in output its side force or vertical force
so 8 outputs then arithmetically analogue-combined at unity gain into 6 channels of data representing the 6 components of load

yes, you couldn't want to make yourself a slave to anything indicated pre-arithmetic
and I worried about IMO what you are calling structural cross-coupling, but experience always showed it was an irrelevance
the through calibration made after the trauma of integration with the test structure agreed with the earlier calibration
and supported the policy of neglecting the cross-axis terms
conveniently, later 6 component balances were smaller and so could all be 1 piece (then helped by access to the master gauge catalogue
to reduce milling I designed the gauged sections to be curved, iirc in part because anticlastic bending helped the strain field shape

one essential task for the designer is minimise cross-axis sensitivities (eg in a 6 component balance)
otherwise the necessary calibration will become impractically extensive, and balance performance in use will be degraded
and much of so-called drift or hysteresis perceived in balance output is actually an artifact of calibration methods
time-related internal thermodynamic effects are significant in larger balances, as some eg NPL and Schenk are aware
these factors should be addressed eg in (semi) automated calibration systems, but are ignored as inconvenient in some offered

[Caito]

What's the sampling frequency?

Filtering the signal should considerably reduce your noise (like using Hi-Res with an oscilloscope). At least if we can safely assume that the time-average of the pot noise is zero .

[Caito]

https://www.sparkfun.com/products/12728

Optical proximity sensor. You get a 2.7V variation in 26cm (useful from 4cm to 30cm).

That's ~108mV/cm. A basic 5V 8 bit ADC (256 levels) will get you ~20mV/bit.

You could amplify your signal, given that you would only need (let's say) from 4cm to 10cm and improve your resolution a little bit.

Of course, I'm saying you should connect the sensor to the chassis of the car. It will still have noise, but you could average them out if you run at constant speed.

Cheers,
Caito.-