8 replies to this topic

[Greg Locock]

A long time ago some people here were kind enough to send me some GGV data for their cars. I did less with it than I should, but did at least start to understand the difference between the nice little diagrams in Peter Wright's book, and the grotty data you get in real life (that was before my day job involved said grotty data, now I spend a lot of time degrotting).

My experience is that real tires, in real events, can easily develop cornering stiffnesses 30% greater than the Flattrac machine gives, and for limit handling they can easily be 20% different in limit mu. Maybe our track is very grippy, but that is a crazy difference to have to work around. Ah, thinking about it we've got a milled concrete surface and a grippy asphalt one, it'd be nice to see the cornering stiffness on both surfaces.

Mark Ortiz discusses some of that in his Jan 2010 newsletter, I wondered if anyone wanted to take it further?

[Ben]

My honest answer from a practical standpoint is to work in delta.

Example; We took a number of rear constructions to Calspan then did moment method work using Pacejka models fitted the Calspan data . This work retrospectively predicted the driver's main feedback and allowed us to develop a new setup to exploit the advantages of a particular tyre. We did this by assuming the moment method results with the standard tyre showed a reasonable balance, control moment, etc and then tweaked things to get back to this point using the new tyre with more overall grip.

I suspect the need for absolute values may be more important in the world of OEM road cars, but in racing trends are more important. Incidentally the step we made was >0.5% which is sizable given the class we're racing in.

Ben

[Greg Locock]

One area where corrections to Flattrac data is crucial is the Fishhook/rollover stuff. I don't do much limit handling work other than that, as things like brake in turn and so on have very poor repeatability, even with a robot driving. I did some useful work recently on the difference betweeen a good tire and a bad tire for limit handling, but that may have been luck since I was just modelling the manouevre based on the driver's perceptions rather than very much data (I knew the speed, and the radius of the turn, and so the approximate latacc).

As such I'm mostly working in the linear range, which is good, as you'd expect the Flattrac data to be more trustworthy, but is bad becasue we seem to get these 30% differences between Flattrac and track.

[Tony Matthews]

Greg Locock, on Jan 4 2010, 10:17, said:

A long time ago some people here were kind enough to send me some GGV data for their cars. I did less with it than I should, but did at least start to understand the difference between the nice little diagrams in Peter Wright's book, and the grotty data you get in real life...

To be fair to Peter Wright, Greg, he did include some real-life grot, the 'nice little diagrams' were, after all, the envelopes, not the contents!

Edited by Tony Matthews, 05 January 2010 - 23:34.

[Greg Locock]

Ha, yes that looks more like it. One of my favourite hairball plots is shock velocity vs force from a real lap, compared with the nice curve we get from the dyno.

[DaveW]

Greg Locock, on Jan 6 2010, 00:48, said:

Ha, yes that looks more like it. One of my favourite hairball plots is shock velocity vs force from a real lap, compared with the nice curve we get from the dyno.

Amongst other things, suspension set-up & driving style affects realizable individual wheel GGV diagrams, I think.

Damper force/velocity diagrams from track data may be affected by:

1. measurement deficiencies (particularly when potentiometers are used to estimate velocity)
2. measurement corruption (it is rare for damper force & velocity to be measured unambiguously)
3. frequency response characteristics of both transducers & associated signal conditioning
4. frequency response characteristics of the dampers themselves
5. variations in other parameters that can affect damper characteristics (e.g. temperature)
6. damper performance may depend on the source impedence.

Potentiometers are not really robust enough to work in a hostile environment & dedicated analysis software often (by default) manipulates recorded measurements to "hide" obvious corruptions caused by wiper "bounce", etc.

Suspension movement on-track is "random-like", so unfiltered plots of load vs velocity will collapse onto a trajectory only if the transducers, associated signal conditioning & the recording equipment have matched frequency response characteristics and, of course, the dampers themselves have characteristics that are independent of frequency.

The last (point 6) is my somewhat heretical conclusion after testing many vehicles in the fairly controlled environment provided by a four post rig and having access to comparative results measured on an damper "dyno". Put crudely, if a damper wants to stop moving, it will do so when it is installed in a vehicle, but not when it is installed in a dyno (unless the dyno also wants it to stop).

Edited by DaveW, 07 January 2010 - 11:43.

[Ben]

DaveW, on Jan 7 2010, 11:41, said:

Amongst other things, suspension set-up & driving style affects realizable individual wheel GGV diagrams, I think.

Damper force/velocity diagrams from track data may be affected by:

1. measurement deficiencies (particularly when potentiometers are used to estimate velocity)
2. measurement corruption (it is rare for damper force & velocity to be measured unambiguously)
3. frequency response characteristics of both transducers & associated signal conditioning
4. frequency response characteristics of the dampers themselves
5. variations in other parameters that can affect damper characteristics (e.g. temperature)
6. damper performance may depend on the source impedence.

Potentiometers are not really robust enough to work in a hostile environment & dedicated analysis software often (by default) manipulates recorded measurements to "hide" obvious corruptions caused by wiper "bounce", etc.

Suspension movement on-track is "random-like", so unfiltered plots of load vs velocity will collapse onto a trajectory only if the transducers, associated signal conditioning & the recording equipment have matched frequency response characteristics and, of course, the dampers themselves have characteristics that are independent of frequency.

The last (point 6) is my somewhat heretical conclusion after testing many vehicles in the fairly controlled environment provided by a four post rig and having access to comparative results measured on an damper "dyno". Put crudely, if a damper wants to stop moving, it will do so when it is installed in a vehicle, but not when it is installed in a dyno (unless the dyno also wants it to stop).

Excellent post Dave. You really know when someone "knows stuff" when they're comfortable in admitting what they/we don't know. And as you point out those type of people are often labled heretics.

Ben

[DaveW]

Ben, on Jan 7 2010, 12:01, said:

Excellent post Dave. You really know when someone "knows stuff" when they're comfortable in admitting what they/we don't know. And as you point out those type of people are often labled heretics.

Ben

Thanks, Ben, though I'm sure I don't deserve it.

I should have mentioned that the lines or curves connecting measured co-ordinates to generate "hairball" plots provide absolutely no additional meaningful information in either case and may, indeed, be misleading. You can test that thought by plotting the co-ordinates with no interconnections.... I'm sure the plot will still be "noisy", but your eyes might find the result rather less distracting.

Edited by DaveW, 07 January 2010 - 20:29.

[Greg Locock]

Something I've been meaning to do for years is to use the recorded time history of shock length to drive the shock-dyno and see how closely the hairball plots match between the road and the rig. Won't be happening any time soon, sadly.