New project?

[MurrayPeterson]

As most of you know, I have quite a technical "bent", and like to apply science to most things.

I just finished reading Neal Tovsen's blog article http://www.conecoach.com/2017/03/17/when-to-put-on-the-cape/, and it brings to mind some deficiencies in the current data logging and analysis tools.  In particular, how close to the limit are you really driving?  Or in other words, are you under-driving or over-driving?  GPS and OBDII sensors just don't come close to answering this question, and I think it is possible (although very difficult) to do better.

So, what I am proposing is to mount an accelerometer and gyroscope on the front or rear of a car, and log some high speed measurements (perhaps 1000 samples per second?).

That's the easy part.  The hard part is going to be doing the math to deal with all of the sensor's readings and then decide what constitutes "at the limit" on any given run.  If I am very lucky, I can discover patterns that indicate tire break-away.  At worst, I guess I will know a lot more about matrix math and Euler angles than I do now

Does this sound like a worthwhile endeavour?  Or do the current GPS logging and analysis tools give you everything you need?

[Cagare]

Murray, I would agree it is very difficult to do better.  I would state that there are so many variables that go into maximum available grip/contact patch that you could chase those variables to insanity. 

The goal may be to start with simple data logging techniques and analysis and intimately understand them before seeking more data.  In my experience managing large amounts of data sometimes it has to be approached simply first to draw more straight forward conclusions, which then drives in the right direction of what you should be collecting more detailed data on.  If you start to the level you may be thinking you will be so overwhelmed with information that it may be hard to achieve what you are after.  Call the starting portion of your project a bit of a gap analysis on existing techniques.

I think the project would be fascinating, and enjoyable, but I don't know how much more you will achieve outside of existing tools that currently exist.

[MurrayPeterson]

Murray, I would agree it is very difficult to do better.  I would state that there are so many variables that go into maximum available grip/contact patch that you could chase those variables to insanity. 

I really don't plan on chasing those variables -- many of them are almost impossible to measure in-situ anyway. 

I think the project would be fascinating, and enjoyable, but I don't know how much more you will achieve outside of existing tools that currently exist.

I think that I can get/use data that just isn't available via existing loggers.  Almost all of them use GPS to infer acceleration and speed, and even then, they miss the "little stuff".  I want to capture that moment when your front wheels start slipping sideways, while existing loggers capture the moment when the entire car is sliding.

[zero10]

You mention patterns that indicate tire break-away, I bet that to an accelerometer (and probably even to a gyroscope) tire break-away looks a heck of a lot like the driver backing off in certain drive configurations (mainly FWD or understeerey AWD) - perhaps a steering angle sensor would be useful as well?  When I toyed with this idea I wanted access to individual wheel speed sensors as well, and trying to poll these at the necessary frequency necessitates dedicated monitoring hardware - even on cars where you can read this on the CAN bus I don't feel like you can get updates fast enough.

I also dislike the use of GPS to infer acceleration, even the best commercially available GPS is vastly inadequate for measuring the detail required on an autocross course.


On the up side, there's well over 20 events this year in the Calgary area for data collection!

[MurrayPeterson]

I would *love* to have access to the raw data stream from the car's sensors, but that just isn't feasible   I'll be going the route suggested by Cagare -- collect the data first and then see what might be interesting out of all that noise.

This project isn't likely to go very far this year, since I have a brand new car that will need some serious attention to get it set up and tuned for racing.  I am spoiled from driving the same car for many years, and never had to spend any effort on setup.

[zero10]

I've got a canbadger that isn't getting any use

[MurrayPeterson]

I've got a canbadger that isn't getting any use

LOL
How complete is the info you get out using it?  And perhaps more important, how fast can it feed you data?

[zero10]

As for how complete the info is, the device is basically a raw bridge onto the CAN bus and since part of the challenge is building your own cable (or just buying one) you can pick which CAN bus you want to place it on.  The device is supposed to be very flexible when it comes to data rates and CAN bus parameters.  Javier said that it would grab everything on the bus and it supports both USB and ethernet capture so I'll venture a guess it's capable.  Unfortunately the internet is very... light... on documentation so I've been a bit shy of doing anything with it.

I think I have the right cable to plug it into my van - I should try that at some point and see what I can get.  I wanted to plug it into somebody else's van with EVIC and use the EVIC to disable the horn honk on remote lock so I could capture the message for that and replay it on my non-EVIC van   The lengths I go to for such stupid things....

[MurrayPeterson]

Sadly, the signals I really want are the raw feeds into the ECU, long before they make it to the bus.  I don't think all of the sensor signals even make it onto the bus (wheel speed data?).

I did find a cool little board that will save me a boat load of effort: https://www.sparkfun.com/products/14001
It even comes with a micro SD card socket for logging the data.

[MurrayPeterson]

Sadly, the signals I really want are the raw feeds into the ECU, long before they make it to the bus.  I don't think all of the sensor signals even make it onto the bus (wheel speed data?).

I might be wrong.  Found this reverse engineered info about the high speed bus on a Mazda 3:
https://docs.google.com/spreadsheets/d/1SKfXAyo6fbAfMUENw1KR3w4Fvx_Ihj6sTPSVXBdOXKk/edit#gid=0

Looks like individual wheel speeds are sent out every 15ms.  And a bunch of other good info is there too.

[CRX Don]

Does this sound like a worthwhile endeavour?  Or do the current GPS logging and analysis tools give you everything you need?

I don't have a definitive answer to your questions but in my opinion the hardest two things to do are: 1. Remember just what the heck you did on that last run, and 2: was it faster or slower? Data aquisition should help with both those especially if it incorporates video. The system I am considering is the VBox lite because it allows comparisons between laps based on GPS position so its easy to see how you drove the section differently and what the consequenses were. However the VBox does not allow you to input the position of gates so it is a little limited for autocross. I would expect that between video, a friction circle and words spoken into a mike and recorded you could gather a lot of useful data about how the car and driver are behaving.

But since you like techy stuff you might want to take a look at this: https://www.youtube.com/watch?v=uUyvyUthHIc

[MurrayPeterson]

So, I got the sensor board and tried it out this week.  I said that I wanted to capture the "little stuff", but that turns out to be a deluge of data.  The accelerometer is so sensitive that it captures road and engine vibrations, enough to pretty well swamp out any actual driving data.

Here is an example of 20 seconds of idling and then letting out the clutch and driving forward:
https://drive.google.com/open?id=0ByeFbiNdvLbGS1hlSDV0WlN5S28

I'll be playing with data smoothing and data acquisition rates, since this is not usable as it stands

[Reijo]

Maybe even a better, more vibration-free mount?  There are some special materials we have used to mount MRI's to prevent vibration interference from the building's vibrations (mechanical equipment, people walking, etc.).

R

[MurrayPeterson]

I am not sure one would help enough.  Those mounts act as a low pass filter, but I can do that with software as well.  The problem is that road and engine vibrations are too close to the frequencies that I wish to capture.  I am beginning to see why most data collection systems for racing top out at 10-20 Hz.

Of course, the GPS-based systems are immune, since they don't even see those low level movements.  Which makes them genius or very crude -- take your pick

[Reijo]

I wonder what DSL cameras and lens use for vibration reduction?   Might be software as well?

R