ADL3 Internal G sensor calibration

[David Ferguson]

I was working with my first ADL3, and noticed that the calibration for the internal G-sensor didn't seem quite right. My observations were based on bench top-use where after zeroing the internal G channels, I set the dash on it's side to measure +/- 1.0 g. I found that I had to set the lateral G-sensor to 125 mV/G, while the longitudinal and vertical G-sensor was best at 120 mV/G.

Lastly, this dash is mounted into the car tipped back at a 25 degree angle (so long G & vert G are offset when static). Does the dash compensate for that if I zero the long G sensor when mounted? Or should I zero it in the vertical position, then use an I2 Math Channel to generate a corrected Long G (and VertG)? something like: CorrLongG = (SensorLongG - offset) * Sine(25 deg)

The online help is unusually silent (for MoTeC) on this topic, the inputs topic saying that the only internal sensors are battery voltage and device temperature. Any advice or information would be appreciated.

[Holmz]

I was working with my first ADL3, and noticed that the calibration for the internal G-sensor didn't seem quite right. My observations were based on bench top-use where after zeroing the internal G channels, I set the dash on it's side to measure +/- 1.0 g. I found that I had to set the lateral G-sensor to 125 mV/G, while the longitudinal and vertical G-sensor was best at 120 mV/G.

Lastly, this dash is mounted into the car tipped back at a 25 degree angle (so long G & vert G are offset when static). Does the dash compensate for that if I zero the long G sensor when mounted? Or should I zero it in the vertical position, then use an I2 Math Channel to generate a corrected Long G (and VertG)? something like: CorrLongG = (SensorLongG - offset) * Sine(25 deg)

The online help is unusually silent (for MoTeC) on this topic, the inputs topic saying that the only internal sensors are battery voltage and device temperature. Any advice or information would be appreciated.

Assuming that the car pitches and that the roads are NOT level adds complexity.
If you move the unit on the bench, and adjust the gain to show 1 G then it sounds like you are doing it right.
You'll want to have it on its LHE and RHS to make sure that the offset is correct, and ditto for top an bottom sides for longitudinal.

I think that at the very least the equation should be:
CorrLongG = (SensorLongG - offset) / Sine(25 deg)
but that is clearly not right as 25 degrees goes to zero... So this looks more like it:
CorrLongG = (SensorLongG - offset) /Cosine(25 deg)

This is not the most accurate.
If the unit is not too far off with respect to roll and pitch (i.e. the sensors are close to alignment to begin with), then all is pretty good as cos (small number) is close to 1.
But the 25 degrees, means that the pitch and longitudinal G will be off in a dynamic sense.
So the values will not be good in an absolute sense, but will just fine for relative comparison on the same circuit from lap to lap.

[Blu302]

Assuming all sensors are calibrated correctly, the direction will be a dynamic couple between 2 or more outputs. That said, if the ADL is leaning back (say 45Deg) and a forward motion of 1G is applied, there will be a ratio of 0.707G on the 2 outputs. If the forward motion was roughly 1.414G, there would be 1G output on each. Each output will change in a linear fashion to each other relative to angle. To be accurate you would need to have a 3d map to compensate given that the corrected output from the longitudinal and vertical will need to used with the lateral. Essentially as you lean the ADL back, the ratios will change in proportion to pythagoreas until it is laying on its' back where there will be only one output for forward motion. Should remember pythagoreas from school.

A squared + B squared = C squared

Sides of a triangle where A is longitudinal and B is vertical, giving C as the ratio between the both seeing as they act at right angles to each other. A table should be sufficient for correction but there are probably easier ways.

I hope it's not too much jargon and that it helps

[Holmz]

Assuming all sensors are calibrated correctly, the direction will be a dynamic couple between 2 or more outputs. That said, if the ADL is leaning back (say 45Deg) and a forward motion of 1G is applied, there will be a ratio of 0.707G on the 2 outputs. If the forward motion was roughly 1.414G, there would be 1G output on each. Each output will change in a linear fashion to each other relative to angle. To be accurate you would need to have a 3d map to compensate given that the corrected output from the longitudinal and vertical will need to used with the lateral. Essentially as you lean the ADL back, the ratios will change in proportion to pythagoreas until it is laying on its' back where there will be only one output for forward motion. Should remember pythagoreas from school.

A squared + B squared = C squared

Sides of a triangle where A is longitudinal and B is vertical, giving C as the ratio between the both seeing as they act at right angles to each other. A table should be sufficient for correction but there are probably easier ways.

I hope it's not too much jargon and that it helps

Yeah but A is this case is longitudinal and B would have to be vertical-G... To you have a 3 axis accelerometer or 2-axis?

so if you only have long-G you need to divide by cosine(lean-back).

[Blu302]

I probably got the sensors mixed a bit but the side to side G force will be unaffected, so doesn't matter if a 2 or 3 axis sensor is being used depending on orientation. Using Cosine doesn't account for a bump in the road or angular velocity around a track as one can stay constant while the other changes. Both sensors need to be taken into consideration as the force output is a compound of them both.

[Martin]

Motec,

Is it possible to add this into the ADL3 softare to input a tilt angle of the ADL3 for Gforce reading puposes?

[JamieA]

yes, it does need to be done, and is on the list of things to do, but hasn't yet made it to the top of the list.

[Martin]

id like to see this list.......

[David Ferguson]

I've been working with another ADL3 on the bench, and noticed that it gets really warm if left powered on and connected to Ethernet for say 30 minutes (not sure if Ethernet had anything to do with it, but wanted to note that) The dash was not supplying power to anything external. The G-sensor calibration changed over that time (at 70degF, I zeroed, and sent the configuration; 30 minutes later the dash was 133 degF, and the G-sensors were reading like .17 - .19).

Is it expected that the G-sensor calibration will change with temperature? Can Motec build in a compensation for this?

Any update on support for dash tilt?

[DarrenR]

See the attached ADL3 config.
In the Advanced Maths calculations you will see the calculations for Corrected Long and Vert G and also Pitch and Roll calculation. The Dash mounting angle is a 'Constant' value.
I also push the G sensor values through channel maths to filter them, just for nice stable readings while testing.

The calculation fully compensates the G sensors for the mounting angle in the Pitch direction only. So if the dash was pitched 45deg forward, longitudinal G force (vehicle acceleration) would affect the G long and G vert sensor equally by half. The corrected values shows the correct G long, and no G vert, for example.
The Dash still need to be squarely mounted laterally. Lateral G could be compensated also in the same way but figuring this out hurt my head enough.

Setting the dash mounting angle at 0 keeps the G long corrected and G Vert corrected the same as the filtered values, but increasing the mounting angle (as a remote control in monitor channels) puts a phase offset on the sensors. This also carries through to the pitch and roll calculation.

Pitch and Roll gives you 0 deg when the dash is at its mounting angle and goes +/-180 deg in all directions. Note there is a small glitch at 90 deg... Also it is only correct when acceleration due to gravity is the only acceleration applied, so it won’t work for lean angle on a bike...

We have found all G sensors tend to drift a little bit with temperature, some worse than others. It is always best to calibrate the G sensor Zero point with the Dash at normal operating temperature. Span should also be calibrated every so often. Using the 'custom' calibration and calibrating the dash on a level surface using 'read value' is better that entering the zero and span voltages.

Two things to note is you can't zero the corrected values, you have to zero the sensor values with the dash square to the ground. And vertical G must read 1G when static, not 0.

Any Questions?