I also suggested he put the g-cube aligned backwards(front facing rear) in the
car - (recalibrate of course) effectively reversing left/right. Then if the
left-right difference "stays" its the cube, and if it "switches", its the car
handling/or mounting issues (external to the cube).
Byron, i did find an asymmetry in the g-cube, but it was at the .02g level -
and manifested itself as a higher rms in measurements aroung 1g left, vs 1g
right. Ill email you more details - as soon as i find the time :)
dan
Byron Short wrote:
> Hey Ron,
>
> I've attempted to answer this a couple of times, and keep
> getting dragged off into other tasks, so my apologies for
> the late reply.
>
> Well, heck, you've got me a little bit stumped. I'm going
> to hope that it's a problem with the G-Cube, because hey, we
> can certainly replace that easily enough. However in
> another case a user told me about something similar, but
> when the G-Cube was put into another car the problem
> completely disappeared, indicating it wasn't the G-Cube.
>
> Here's the case for why it probably isn't the G-Cube...
>
> When you do your calibration (the one where you put your
> G-Cube on each side in turn) what you are doing is setting
> adjustment levels for how the G-Cube perceives 1g in each
> direction. Once that calibration is done (and especially
> when it's repeated with very similar results) we know that
> at 1g the G-Cube is extremely precise and repeatable.
>
> There is a tiny "moment arm" inside the G-Cube that deflects
> as g's are applied to the G-Cube, and that flexing is
> precisely measured and reported back to the recording
> computer as a duty cycle. We then convert that number to a
> g number by applying what we learned in the calibration
> process.
>
> The design of the moment arm is such that it's extremely
> consistent and linear across the entire rated g range,
> (which by the way is from -2g to +2g). The spec sheets on
> the accelerometer package show deviations beyond linear of
> only negligible amounts (in the 1%-2% range), with the bulk
> of this deviation in the low side of the scale, say at 0 +-
> 0.25g. So in these instances the deviation should be below
> what we even measure. At 1g we are calibrated to the exact
> number, so that should be our most accurate point. The
> typical values that we reach out on the course in the
> lateral direction also happen to be right around our test
> value of 1g, so our accuracy is really maximized in the
> precise area where we need it most.
>
> All of this leads me to conclude that it would be *extremely
> rare* for such a large deviation, the 0.15g you indicate in
> your post, to be in the G-Cube, especially so close to the
> test point of 1g.
>
> So can I imagine any scenario where the G-Cube could be
> causing this? Well, yeah, but it's pretty unlikely. If
> there were a flaw in the moment arm that caused it to
> essentially hinge at a precise g point, that g level being
> somewhere over 1.00g, but below about 1.25 g, we could get a
> reading like you are describing from the G-Cube. I wish we
> had a good repeatable test to measure in the 1.00 - 1.50g
> range, but we don't. If we did that would prove or disprove
> our case one way or the other. Without that, all I can say
> is the scenario I described above the for the problem to
> occur is one that I made up, and not something we've ever
> seen before.
>
> So, let's talk about the car a little bit.
>
> First, don't worry about where the G-Cube sits in relation
> to the CG of the car, unless you feel that you must have
> readings from that location. The car does get slightly
> different readings at different locations, but not enough so
> to be worrisome. The accuracy isn't effected by the
> location of the G-Cube, btw, it's just that you get the
> readings as they occur in that particular location. Not a
> big deal.
>
> What is a big deal is the alignment to straight ahead.
> Let's assume the alignment is off by 5 degrees, pointing
> slightly left instead of straight forward, and compare two
> turns, left and right, in which we are pulling 1.20 g's
> laterally, while simultaneously accelerating at 0.30g.
> (This is a total g of 1.24 -- within a reasonable range for
> the test car.) In both the left and right turns, the
> lateral g's are lowered (but just a little) by the fact that
> the lateral force isn't running perfectly down the lateral
> pole. Instead of the lateral force being at 90degrees, it's
> at 85 degrees on the left side, and 95 degrees on the right
> side. The difference this makes is negligible. Instead of
> 1.2g's, we'll only see 1.195g. (sin(85) * 1.20). So in
> fact after rounding, there's no measurable error there. But
> wait! We have 0.30 of acceleration there that is also going
> to spill over into our lateral readings. In the case of
> the left hand turn we have 0.30 * sin(5) that will be added
> to our lateral g's. This brings us back up to 1.23 g's.
> Conversely, the right hand g's go down by the same amount,
> down to about 1.17g's.
>
> So after a little math, even a relatively large alignment
> error of 5 degrees only produces a spread of 0.06g's. We'd
> need 10 degrees of error to produce a 0.10g spread. This is
> a lot of error. So again, I'm not convinced this is it.
> But do be careful and make sure that if you aligning inside
> a console for instance, that the console faces perfectly
> forward. My Camaro was always difficult to align within the
> console because the console was off by about 10 degrees of
> straight forward.
>
> One reason the car could be showing more g's one way than
> the other is lean. The car leans out of the turns, and we
> provide a straight line compensation factor for that lean.
> But if the car leans for instance more in left hand turns
> than in right we could see this effect on the trace.
> However, there's not enough lean in any chassis to make up
> 0.15g. So I doubt that would be the cause, at least not
> entirely. Besides, the most likely difference in lean is
> the driver weight being on the down side of the vehicle in
> right hand turns, making right turns higher than lefts--the
> opposite of your problem.
>
> Next would be the car's actual handling. Since you are
> pulling more g's left, this is when the driver is on the
> high side. This follows suit with the general theory that
> when the weight is on the inside edge the car will handle
> better, and on the outside edge the car will handle worse.
> But still, 0.15 is a lot.
>
> I can't think of any one thing that could get us a
> difference of 0.15g. But are you really sure it's that
> much? You said your adjustments always take "more right" to
> make the maps look right...how much more? The full scale
> only goes to 0.100. If the difference is really 0.100 the
> full scale shouldn't be enough. Are you sure that 0.15 is a
> fair representation of the amount of the problem? On
> average, how much adjustment have you been putting in? If
> you find that 0.020 or 0.030 is enough to correct the
> problem, I'd re-open the alignment issue. It's the most
> likely of the physical explanations.
>
> Next, I'd try to duplicate the error on a different G-Cube,
> or in a different car. If you swap G-Cubes with another
> user (be sure to re-calibrate!) does he now have the high
> left problem, while your problem goes away? Or put a second
> G-Cube in the car alongside your G-Cube. Be sure both are
> aligned exactly the same, preferably with the edges against
> each other to assure that. Then record the runs. There
> should be very little difference between the runs. If you
> see some sort of evidence of a "hinge" as we discussed
> above, that would be a clincher. Or if you could develop a
> simple course with a long sustained left followed by a long
> sustained right, we could run the course several times the
> regular way, then flip the G-Cube over (straight upside
> down, but still pointing forward). This swaps the two
> lateral poles, and should produce a mirror image map. If
> that mirror-image map now shows the problem to the right
> instead of the the left, and it does it consistently, that
> might also be a good clincher. Again, my HOPE would be that
> it's the G-Cube...it's the easiest fix of the possibilities.
>
> In summary, I don't have any sure things. The easiest thing
> to do is to just replace the Cube, but I doubt that that's
> the actual problem. Call me if you'd like to discuss some
> possible tests further. I definitely want to address this
> during this off-season time!
>
> --Byron
> 800-775-9511
>
> Ron Katona wrote:
> >
> > Hi all, newbie list member question here...
> >
> > 100% my GEEZ data from this season showed left G forces (peak and
> > sustained) about .15 G higher than right Gs. The car subjectively does
> > not feel like it turns better to the left. It has coil-overs and has
> > been corner balanced, carefully aligned, and weighs only 60 lb more (out
> > of 3130) on the left side with the driver in place. Other people have
> > driven the car and haven't noticed that it won't turn right (maybe I've
> > found the perfect NASCAR setup).
> >
> > I'm using a G-Cube with a Palm-IIIxe and the latest versions of GEEZ
> > downloaded from the website. I've calibrated the G-Cube several times
> > using the Palm and the same batteries used for the runs. The G-Cube is
> > velcroed placed in a little bin between the front seats and is quite
> > snug in there. It's about as close to the CG of the car as I can get it
> > and appears to be level and aligned with the centerline of the car.
> >
> > FWIW, I've seen peaks of up to 1.40 G left, but only about 1.25 right.
> > Sustained is about 1.25 Left and 1.10 right. All other data seems normal
> > for this car (ESP M3). In order to get my maps to look like the courses,
> > I usually have to add "more right" on the calibration window. To me, the
> > right turn data seems much more realistic.
> >
> > .15 G seems significant to me. Could the car really be that much better
> > to the left, or is it more likely that the G-Cube is
> > misaligned/calibrated? If it's the G-Cube/GEEZ software, what's the most
> > likely problem? Any ideas, opinions, advice greatly appreciated.
> >
> > Ron Katona
|