8 replies to this topic

[Lukin]

Hi,

I was wondering about a way to check the calibration on a g-sensor. I heard someone say that you double check it's correct by turning to car on it's side to check the lateral sensor (if it's perpendicular to the ground the lateral sensor should read 1g) and the same can be done with the longitudinal sensor, by lifting the rear to make it perpendicular to the surface.

Is this correct?

If so, then it's not really that practical, but can a decent approximation be made by rotating the car to a known angle and seeing what the acceleration is detected.

Since an angle of 0 degrees to the ground would give a zero acceleration, and at 90 degrees the acceleration would be 1g, can the lateral and longitudinal acceleration (in terms of g) be approximated by the sine of the angle?

Eg at 30 degrees acceleration would be 0.5g, at 45 degrees it would be .707g and so on and so forth?

If this isnt the case, is there a simple method for checking the calibration on a g-sensor?

Cheers

[paulogman]

turn the g-meter itself on it's "side" not the whole car?

[Greg Locock]

The calibration method depends on the type of sensor, but for a DC sensor (which is the only sensible one for handling) you are right.

Another way would be to drive at known speed around a circle of known diameter, which will be slightly confusing due to vehicle roll, but you'll get that on the track anyway.

[BRIAN GLOVER]

I think the basic mechanism is mechanical. You deflect a pendulem to a predertimied value and calibrate the instrument accordingly. I gotta check that out tomorrow. I would imagine it is the same as an aircraft instrument except one is for vertical Gs and the other for lateral. The aircraft instrument is set at one G when the needle is at rest and the car is set at 0 G at rest. Both have 3 needles. The two red ones are slaved to the green needle and record max loads with a reset button. You reset the meter for each manouvre and it is recorded electronically for debrief. Some passenger cars have them now but race cars Gs are recorded via telemetary. I don't think its a good idea to roll the car on its side. There has got to be an easier way. How do you get to 2 or 3 gs that way. Turn the sensor sideways and crash the car into a wall. Nah.

Originally posted by Lukin
Hi,

I was wondering about a way to check the calibration on a g-sensor. I heard someone say that you double check it's correct by turning to car on it's side to check the lateral sensor (if it's perpendicular to the ground the lateral sensor should read 1g) and the same can be done with the longitudinal sensor, by lifting the rear to make it perpendicular to the surface.

Is this correct?

If so, then it's not really that practical, but can a decent approximation be made by rotating the car to a known angle and seeing what the acceleration is detected.

Since an angle of 0 degrees to the ground would give a zero acceleration, and at 90 degrees the acceleration would be 1g, can the lateral and longitudinal acceleration (in terms of g) be approximated by the sine of the angle?

Eg at 30 degrees acceleration would be 0.5g, at 45 degrees it would be .707g and so on and so forth?

If this isnt the case, is there a simple method for checking the calibration on a g-sensor?

Cheers

[Fat Boy]

Lukin,

You're really overthinking this, buddy.

1. Put the sensor (not the whole damn car!) on it's side in one direction, take a reading. That's +1 g lateral. Racing convention in the States is that this is in a left hand turn...I guess it's an oval thing.
2. Put the sensor on it's side in the other direction, take reading. That's -1 g lateral.
3. Take a reading with the sensor in it's normal position and the car as flat as possible (a setup pad if you have one). That's your zero point.

Repeat the process in the appropriate directions for longitudinal acceleration.

Verify all calibration points with a simple bubble level (gravity is rarely wrong!).

All accelerometers that are meant to be used on a car for lat. and long. accelerations are quite linear in their output. A 3-point calibration is fine. More than likely there will be some shift in the zero point due to temperature, so try to put it in a place that has a relatively constant temperature. If you know that where you have it gets warm, then use a heat gun to approximate the operating temperature of the accelerometer on the track. Don't get too brave with the heat gun, 40-50 deg. C should be about the max you heat it to.

Hope that helps.

[Lukin]

Originally posted by BRIAN GLOVER
Some passenger cars have them now but race cars Gs are recorded via telemetary. I don't think its a good idea to roll the car on its side. There has got to be an easier way. How do you get to 2 or 3 gs that way. Turn the sensor sideways and crash the car into a wall. Nah.

I guess I should of explained it more clearly. It's for a Formula Ford and the data is accurate in terms of what is expected from the g-g plots for shape. Ie on corner entry data reflects braking, then as steering angle increases so does lateral g etc.

Someone pointed out though that the overall values might be a little high, as if the actual values are being multiplied by a scalar. Personally I think they are accurate (around 2 g max) but I would still like to confirm the values themselves, hence the need to check. I just want to use gravity as a benchmark.

If it was wrong and the actual value was being multiplied by a scaler of, say 1.2, then rotating the car 45 degrees would yield an acceleration value of 0.849 g as opposed to the real value of .707g.

[Fat Boy]

Originally posted by Lukin


I guess I should of explained it more clearly. It's for a Formula Ford and the data is accurate in terms of what is expected from the g-g plots for shape. Ie on corner entry data reflects braking, then as steering angle increases so does lateral g etc.

Someone pointed out though that the overall values might be a little high, as if the actual values are being multiplied by a scalar. Personally I think they are accurate (around 2 g max) but I would still like to confirm the values themselves, hence the need to check. I just want to use gravity as a benchmark.

If it was wrong and the actual value was being multiplied by a scaler of, say 1.2, then rotating the car 45 degrees would yield an acceleration value of 0.849 g as opposed to the real value of .707g.

Why on God's green earth are you interested in tilting the car at some crazy angle when you can just move around the sensor? Race engineers get in bad enough trouble from mechanics without going out of your way to draw flak. You're being silly.

Without touching anything and the car level, check the output of the G sensor. Is it "0"? Pull the sensor off the little velcro patch that it's sitting on and put it on it's side. Does it read "1"? Now put it on it's side in the other direction. Does it read "-1"?

If at any point in time, it doesn't read the above numbers, then recalibrate as I told you before and be done with it.

Here's a little more on the subject. It doesn't really matter if the sensor is reading perfectly accurate or not. Heresy, I know. What you are much more interested in is the shape of the curves in the data rather than the exact numbers. If all your numbers were multiplied by 1.2 it wouldn't change a single thing that you would tell you driver in terms of what he was or what he was not doing on the track. In general, precision is much more important than accuracy when doing data on a racecar. We're much more interested in trends rather than scientific absolutes.

As you get to higher levels of racing, the accuracy demands go up. Certainly, if you're at the level of using the data in simulations, the accuracy of every number is quite important. At the Formula Ford level, just make sure it's consistent.

[Lukin]

Cheers,
I didnt think of moving the sensor by itself. I was writing that second bit as you posted. Will give that a burl. At the moment all the data logging is being used for is to compare driving styles etc, but I would like to utilize for better effect. Anyway, thanks for ya help everyone.

[Fat Boy]

Well, it sounds like you're on the right track. Now you just have to spend the next 5-10 years looking at squiggly lines so you know what you're looking at. Have fun!