924RACR

I'd like to get some external feedback on my data here, as I see that the community is starting to develop some notable level of expertise analyzing the stuff. My goal is to improve driver calibration; I'm sure there's more work to do on car setup, and I have a plan, but I do expect that there are a few corners I'm not 100% optimized on yet.

That plus all the thoughts I've been expounding on others' data gets me thinking perhaps it's time to put my own up there!

For those already using an AIM product, here's a link to the full race data:
http://vaughanscott.com/AIM/Data/Rac...un_PM_Race.drk

It's my last race of the season, many of the "off" laps are confused with traffic - this is the same race I posted that exciting 3-lap video from. It also contains my best lap ever, only 7/10ths off the lap record. But both I and the BMW driver I was chasing know our cars are capable of more. I'm anxious to hear what the experts out there think (track is Waterford Hills, MI, for any who've been there).

I know one of the preferred formats is just a screen print - how does this look? I tried to separate out the channels of use - lateral in yellow, steer angle in green, yaw rate in red, Ackerman target yaw rate in light blue (calculation may not be perfect, so take that one with a grain of salt), and speed in purple:


Thanks!

SundayDriver

I have been staring at this for a bit, trying to see what the patterns show and have some questions, as I have not thought much about yaw (and rate) data. Can you explain what Ackerman Target Yaw is and how you calculate that?

It also seems to me that there is an important item that is missing - what is the car doing at any point in time? Can you add inverse corner radius? For example, if corner radius is steady but yaw rate is increasing, we know the car is rotating relative to it's path. Without corner radius, we don't know if you car turning faster (around the intended path) or rotating the car around it's axis.

FYI - In my Motec, they use the formula of G-Lat (m/s/s) / Speed ^2 (m/s) to calculate inverse corner radius. (For those who have not thought dealt with corner radius, the reason to use inverse radius is because it goes from zero to some value, while radius goes to infinity and screws up math and plots.)

cbracer

First off, where is your speed data from? It looks like a rear tire that loses grip on an open diff?

I assume target yaw rate is a calculated channel from speed and steering? You need to zero your sensors. The calculated target yaw is offset from the yaw sensor. It contantly reads more negative in the negative direction, and reads less positive in the positive directly. You would have been better off just inverting your steering wheel angle sensor, strangely enough. I'm guessing by the scale that steer angle is steering wheel angle, not steered tire angle. Your A_Y which I assume is lateral G, appears too filtered when trying to compare it to the yaw rate.

924RACR

Mark - yes, I'll go add those calcs/math channels in and post another pic, thanks. If I think long and hard enough, I should be able to re-derive the steering corner radius also from similar calcs as the Ackerman yaw rate.

FWIW - Ackerman target yaw rate is a simplified term for expressing the yaw rate that would be achieved at that speed, given that steering input, using the classic 2-wheel bicycle model (think Gillespie or other such vehicle dynamics texts) and assuming no loss of tire adhesion - pure linear tire response. The Ackerman model should match up in reasonably steady-state maneuvers with lower lateral loading (<0.5g). So simply put, it tells what the driver wants under those conditions. If the actual matches the target, the driver is getting what he/she wants - whether or not the tires are at the limit. If the actual is above the target, the car is yawing too much - oversteer; if it's below, the car is understeering.

Of course, the model incorporates certain factors which may or may not be right. I know my steering rack ratio (assuming it it reasonably accurate/linear for the -90-0-90 range) and wheelbase; this calculation also uses "characteristic speed" which is measured experimentally - it's basically a fudge factor. In practice, one drives a lower lateral and gets the ackerman model to line up with the measured values. Of course, that's tricky on a racecar; how often do you drive at such low laterals?

CB: correct, the speed sensor is on a rear wheel that does sometimes lose grip - though I do run a worn factory 951 LSD. Nice to know my little motor can still put out that much power!

Correct, steering angle listed is handwheel angle, not road wheel angle. I'll double-check the calcs when I go back in to add inverse corner radius as Mark asked - probably later today. It's a pretty straightforward matter to re-work all the signals in the analysis SW, one of the big reasons I like the AIM product line! Yes, A_Y is lateral. Unfortunately I cannot reduce the filtering on that, as it's from the internal AIM lateral sensor, and they have the level of filtering already set - though I agree, it does seem to be over-filtered!

924RACR

OK, just looked at what you were talking about with respect to yaw target offset. There's a straight, pretty much the only ontable straight on the track, between corners labeled 7 and 8. There you can see that in a straight line, the car is tracking pretty straight (lateral's hovering around 0) and steering, yaw rate, and target yaw are all around 0. Perhaps you're noticing that most of the time, target is higher than measured yaw? Yeah, the car does push a bit - I like a bit of what the NASCAR boys would call a "tight" car, but some of that needs to be dialed out - planning some setup tuning for the spring, starting with re-leveling the car... right now it's 48/52% front/rear split, and I'm going to try re-zeroing it, see how that drives. Puts the power down very well, but doesn't like to turn in. I was thinking about adjusting shocks, till I realized I still haven't exactly leveled the car.

You will also notice in a few spots - 6, 9, 11, and 12 - that I do achieve oversteer, with the actual yaw exceeding the target. Besides, as mentioned in the previous post, I'd take the target yaw with a grain of salt, since I'm not 100% convinced that it's where it should be (fudge factors and all), though I know the basic formula's good.

Sorry, I should've added the track map:

kary993

Mark,

Since I am no math wiz I tried to build a channel as you described in the above calculation. I think I am not building the G m/s/s correctly. Can you provide more detail on how that is built from what sensors in your system?

Thanks!

924RACR

OK, yeah, I've got it built correctly - units should be 1/distance (1/m or 1/ft, whatever your prefer)... in my case, with the AIM notation, the calculation is as follows:
(A_Y*9.81)/((Speed*MI2KM*KMH2MS)^2)

A_Y is lateral in g's, time 9.81 m/s^2 per g

Speed is obvious, measured here in mph, times built-in constants in the AIM SW to convert from mph to kph, then kph to m/s, then the whole thing squared.

I still have to crunch through a derivation (my vehicle dynamics book's still at work) for other corner radii based on steering angle or yaw rate...

924RACR

OK, if I'm not mistaken in my derivation, the inverse radius that would correspond to the driver's steering input should be basically:
1/R = tan(road wheel angle)/wheelbase

Right? Not horribly off-base?

SundayDriver

I comes directly from the Lat g sensor. You just need to convert from g's to m/s/s. As Vaughn indicated, 1g is 9.81 m/s/s. Multiply yoiur lat g reading by that number and you have it.

kary993

Thank you Mark!

924RACR

OK, I've added/updated the plot to include the inverse corner radius, as requested; in the pic below, it's the second/lower red trace. Corresponding to that is a blue line which should be inverse corner radius based on steering, assuming my formula above is correct. Obviously, the two aren't even close, so if the formula used is correct, either my steering angle has to be off, or my rack ratio way off. This would also get to the root of why my target yaw rate calculation seems so wrong.

Mark or others, can you confirm that formula?

Here's the updated pic:

SundayDriver

If you invert what is plotted for corner radius, looking at T1, the blue trace says the corner is around 100m radius, while the red indicates about 20m. I looks to me like the blue is pretty good. I have not had a chance to think about the calculation you posted or to really look at the latest plots yet.

Again, the blue trace for inv corner radius is shifted vs. the red. I think there is either something wrong with your steering calibration or the car has a weird set up. Steering angle should really be measured at the outside wheel. What you might be seeing is some sort of variable ratio in the steering (could be by design or could be bump steer) so your steering wheel angle is not linear vs. wheel angle.

924RACR

Bump steer is quite probable, as low as the car is - thanks, I didn't think of that. Front spring rates are high to minimize, 660#, but not THAT high. Too bad we can't eliminate the bump steer (not legal in IT).

It is VERY possible that the analog inputs like steer angle and yaw rate are shifted vs. the internal AIM channels of speed and lat. I'd noticed that once before and complained about it... OK, just checked, looks like that bug was fixed, not impacting anything here. (It was associated with changing beacon point manually to get different sessions to line up on-track).

Then perhaps the shift you're seeing is an artifact of the built-in filtering of the lat signal.

I think I can also derive a corner radius from yaw rate; let me run that through, post another pic from that. That should provide another sanity check vs. the steer angle derived radius.

Also - I think you're thinking backwards on the inv corner radius? Using the sector 1 example, I see a peak inv corner radius of 0.02 /m for the red, 0.06 for the blue - which gives me 50m for the red, 17m for the blue. Not so much the values, but the important point being that the blue trace should correspond to a smaller radius than red when higher - perhaps you got them backwards in reference? Should be an inverse relationship... if the blue is higher, I'm understeering; if the blue is lower, I'm oversteering (at least, relative to my lateral)...

924RACR

OK, here's an attempt at adding on the inverse corner radius as calculated from yaw rate. I think it seems plausible/meaningful; in part, I like the looks of it since the characteristic speed (fudge factor) required to make it work as shown is more plausible than what's required to get the target yaw rate to work - higher rather than lower than a street car. One way to directly measure characteristic speed is listed below, only works if you have access to a skidpad, but it's basically a measure of understeer characteristic of the car; the lower the speed, the easier the car goes into tire-punishing understeer. With a car that just goes straight to oversteer, the characteristic speed is essentially infinite.

Anyway, here's the added signals; I've already drawn some conclusions, ubt I'll leave you to draw your own:


To measure Char Speed, you define a radius (dry pavement, obviously), then figure out the steering input required to maintain that corner with linear tire response (less than 0.4g). Then increase vehicle speed till the steering input required is doubled; the speed at which this occurs is the characteristic speed. Street cars will typically be in the 20-30m/s range. The speed required to make my ackerman target work (match the yaw response of the car) was around 12m/s; I know it pushes more than I'd like, but not that much! The speed used for the above picture to match up inverse corner radii is 55m/s, more as I'd expect for a racecar (particularly directionally correct).

Thanks...

SundayDriver

Yes, sorry I reversed the colors in my description. Now if I get it right this time, the calculated radius appears reasonable while the steer based does not. Since they disagree so much, and you can not trust the assumed wheel angle due to bump steer, etc, you really can't make a conclusion about OS/US from that. The Yaw based looks pretty good.

I think if you put all this together, you have a pretty decent picture of where the car is pointed, what kind of side load you are getting at that moment, how much it is rotating relative to the corner radius as well as steering/corrections.

Now bear with me, as I have never really considered all this with Yaw data, so I am taking shots at what I "think" I am seeing. There is a lot to consider and it would be easy for me to get some of this backwards, so I encourage everyone to jump in and challenge my thoughts here...

Seems to me that you can now see that the car is generally understeering in mid corner (I am looking mostly at T1) as the yaw rate drops off, and corner radius increases as the car slides outside the intended line. Then you often get some oversteer - probably intentional to get the car back on track.

The plotted corner radius indicates that you could be faster if you had a smoother radius - BUT, the filtered lat g may be hiding the speed of your corrections. If the plot were unfiltered g's then I think we could assume that either you need to speed up your corrections or the car needs to be better tuned to a more neutral attitude.