Datalogging

[Belatti]

Originally Posted by Rubinho

I should have explained better, sorry. When you use CONST() Toolbox expects to find a name in the () which evaluates to a constant value in Tools|Constants. For example where I have said CONST(Wheelbase) you would set up a constant named Wheelbase with a value of 2.55 in Constants Management. If you want to use a literal value you can simply use it in the maths channel without the CONST() function. So:

(2.48 / pow([Speed],2) / [Lateral G]) * 0.777

Would be your channel using literals, or you could use my original version with constants named Wheelbase and Steering_Ratio set up in the workbook.

There's an assumption in this channel that the car follows the "bicycle model" (i.e. has only two wheels) so complications like toe and bump effects on the steered wheel angle is ignored.

P.S. I've just noticed in your channel you have speed as kph, you probably want this in m and Lateral G in m/s^2.

Thank you again!

I corrected my mistakes in the formula and the result was a flat line at +/-0 deg (blue line)

[Rubinho]

Originally Posted by forestdweller

Since this thread appears to be centered around understeer angle at them moment, I was wondering if any of the experts could comment on an observation I made. I've been comparing two different drivers, looking for differences in style that could potentially help one or the other improve. The application is SCCA autocross. The two driver's are sharing the same car.

I'd been looking at steering angle, Ackerman steering, and understeer angle. As there seemed to be a fair amount of "artificial" understeer at initial turn-in, which is a large amount of time in autocross, I filtered the understeer channel for lat g values greater than 0.9 g and plotted the frequency distribution. Some differences between the two drivers started to show up. Looking at datalogs from various events through the season, the patterns were pretty consistent regardless of which driver was faster on that particular day. (There are MANY other factors that could have contributed to that though.)

My take on it is that the second driver (black) is doing a better job of feeling the the limits of the tires on a consistant basis. OR the first driver (red) is just more aggressive. I'm leaning towards the former. Am I reading too much into a channel with a low signal to noise ratio?

Have a look and tell me what you think:

I am not familiar with the way RT Analysis calculates understeer. As I mentioned above, the majority of these channels show pretty poor signal-to-noise ratio. Let's assume for the purposes of this discussion that the calculation is a realistic representation of over/understeer. It is also not clear which direction is understeer, is -ve angle understeer or negative understeer (i.e. oversteer)? I will assume, for want of evidence to the contrary, that +ve is understeer, -ve is oversteer and 0 is neutral balance although this does not seem logical to me.

I would expect these histogram charts to display a normal distribution centrered around the mean cornering balance. That is to say, the bar/bars with the highest frequency are where the cornering balance sits in the majority of cases and that the probability of more extreme under or oversteer moments decreases with the shape of a bell curve. To some degree and certainly within the boundaries of a relatively small number of samples both drivers display this characteristic. An ideal curve would be very slim and centered on a small amount of oversteer as this would indicate a natually fast car driven consistently.

Red driver appears to have a smaller peak understeer and also less variance (narrower bell curve) than the black driver. Both display a similar amount of oversteer (shape of curve to the left) so I presume this must be a characteristic of the car. Black driver's graph has less of a bell curve pattern so appears more random meaning the behaviour is more likely to be a driver characteristic than one inherent to the car.

With only these charts to go on, the assumptions I have made above and not being biased by your original hypothesis I would say that the red driver is the more consistent and quicker. The black driver would appear to be forcing the car into unnatural situations more than the other.

If I were to have some more information (type of car, relative performance of the drivers, actual direction of the understeer channel and how it is calcuated) I may change my view.

[forestdweller]

Rubinho,
Thank you for your reply. I should say that I had "red" and "black" switched around in my original hypothesis. Unfortunately it's too late to go back and edit the post.

I should have provided a little more background information:

The car is a Mazda MX-5 with lightly tuned suspension. The front weight distribution is 52.6%. The tires, just the high performance street variety, are sized the same 225/45-15 front and rear.

At the time this set of data was recorded both drivers were satisfied that the car felt neutral and predictable.

Your assumptions on sign convention were all correct: positive angle = understeer, negative angle = oversteer.

The understeer calculations are done in math channels that I inputed myself. I used the classic Ackerman steering angle formula for the required steering based on speed, lat g, and wheel base. The actual steering angle comes from a sensor calibrated to the average steer angle of the left and right front wheels. My logic behind using the average was that the GPS antenna (for speed) and lat g sensor were both located on the centerline of the car and that using the average angle of the left and right wheels would conform to a true bicycle model. There's some logical operations to account for left and right turns, but I've confirmed that the understeer angle works in both cases.

Between the two drivers, either could be quicker on a given day, but "red" seems to be pulling ahead in the stats.

Rookie DAQ questions: Is looking at an understeer distribution plot like this a known technique to help train developing or amateur drivers? Is it too susceptible to error to be a reliable tool?

-Chris

[Rubinho]

Originally Posted by forestdweller

Rubinho,
Rookie DAQ questions: Is looking at an understeer distribution plot like this a known technique to help train developing or amateur drivers? Is it too susceptible to error to be a reliable tool?

To be honest, I've not used it before but thought the plot was interesting. Assuming once again that the understeer equation is giving you accurate results I think it could be useful. The temptation with these channels is to plot them on a normal chart with a distance base and in those cases they tell you very little I think. A statistical approach like this does make sense. I've also plotted them around a track map in red (oversteer), green (understeer) and white (neutral within certain boundaries) which gives you a good overview of what the car's doing around each lap as a whole. Certainly, as we've discussed these channels aren't typically suitable for very close scrutiny.

If you were using axle accelerometers, yaw rate gyros or slip angle sensors to estimate cornering behaviour you could do it to a high degree of accuracy.

[Belatti]

Does anyone knows wich systems does F1, Indy, GP2, F2, etc teams have?

[wheadon]

I'm sure F1 teams have to run the same system which is made by mclaren

[Belatti]

McLaren makes the ECU, but the data logger is a different system:





See the "TEAM DATA ACQUISITION SYSTEM" in the diagram

[Belatti]

Just a couple of quiz here:

1) this is a graph of the damper sensors (they are linear and calibrated to show wheel movement and not damper movement). Can anybody tell me what may be wrong with the Fr Left one? I have checked the conection with the car standing still in the shop and looks OK, now in the next track test I´ll mount the sensors inversely to see what happens.




2) Can somebody help me to create a Math channel in PiToolbox to measure full throttle % arround a lap?

[1200Datto27]

Making a presumption that 0 on the scale is a static position, and that the travel shown is the wheel drooping, then I would hazard a guess that as the diagonally opposite corner is showing the least amount of that your issue is in the RR area. I have seen a similar trace on a car that had a deflated tyre on that corner, caused by the tyre taking the suspension deflection, and unweighting the diagonally opposite wheel.

[Belatti]

Oh sorry I was not clear enough... the car is perfectly well, there are no damper/suspension/tyres issues.

Believe me, the problem is in the FL sensor. I wish to know if I can repair/clean the sensor or if its just damaged and I´ll have to replace it.

[greenamex2]

Really does look like a sensor/wiring issue. What the sensor read when disconnected?

Also, can you either either zoom in on the spikes, or just supply a bit of the toolbox file (I use it as well, fantastic product)?

[phoenix]

Originally Posted by Belatti

Oh sorry I was not clear enough... the car is perfectly well, there are no damper/suspension/tyres issues.

Believe me, the problem is in the FL sensor. I wish to know if I can repair/clean the sensor or if its just damaged and I´ll have to replace it.

Is the sampling frequency set the same on all four damper sensor channels?

The overall shape of the curve looks ok - there is just a lot more going on, which could mean more data from a higher sampling rate.

If not, then the sensor is 'noisy' in some way. With the trace going to -40 and then back to the the nominal suspension travel I would suspect that the potentiometer is intermittantly going open circuit or short circuit. Some good quality 'switch cleaner' may fix that - worth a squirt if the pot isn't sealed. If it is sealed then I guess the only optin is to swap it for a new one, or a known good one from another damper just to prove that it is the pot causing the problem.

[Casper]

Swap the sensor with another one, if the problem does not follow the swap then look for an issue in the wiring. Does the wiring for that sensor run near any other electronic item on the car, this can lead to issues due to interference, none of which I understand. If the issue is the sensor I guess a new one is the answer.

[greenamex2]

Originally Posted by Casper

Swap the sensor with another one, if the problem does not follow the swap then look for an issue in the wiring. Does the wiring for that sensor run near any other electronic item on the car, this can lead to issues due to interference, none of which I understand. If the issue is the sensor I guess a new one is the answer.

Top suggestion, do it at work all the time, usually works.

[Belatti]

thanks for the answers guys, I guess I can open and use some switch cleaner and swap the sensors as well to see what happens... Ill tell you later...

@phoenix: the sampling rates are 200Hz for all the sensors

[Casper]

Originally Posted by Belatti

thanks for the answers guys, I guess I can open and use some switch cleaner and swap the sensors as well to see what happens... Ill tell you later...

@phoenix: the sampling rates are 200Hz for all the sensors

You will always nearly always get a noisy signal at those rates and I would suggest it costs a lot of money to avoid or minimise it. Filtering is the name of the game, read Templeton's book to help here.

[Belatti]

can you give me a link or tell me the full name of the book? I googled but couldnt find it

[forestdweller]

Originally Posted by Belatti

can you give me a link or tell me the full name of the book? I googled but couldnt find it

I'm not Casper, but I think this is the book he was referring to:

http://www.amazon.com/Competition-Da...0806606&sr=8-2

[Belatti]

Oh, thanks forest!

Templeman, Templeton... probably it is

[speedsense]

Originally Posted by forestdweller

Since this thread appears to be centered around understeer angle at them moment, I was wondering if any of the experts could comment on an observation I made. I've been comparing two different drivers, looking for differences in style that could potentially help one or the other improve. The application is SCCA autocross. The two driver's are sharing the same car.

I'd been looking at steering angle, Ackerman steering, and understeer angle. As there seemed to be a fair amount of "artificial" understeer at initial turn-in, which is a large amount of time in autocross, I filtered the understeer channel for lat g values greater than 0.9 g and plotted the frequency distribution. Some differences between the two drivers started to show up. Looking at datalogs from various events through the season, the patterns were pretty consistent regardless of which driver was faster on that particular day. (There are MANY other factors that could have contributed to that though.)

My take on it is that the second driver (black) is doing a better job of feeling the the limits of the tires on a consistant basis. OR the first driver (red) is just more aggressive. I'm leaning towards the former. Am I reading too much into a channel with a low signal to noise ratio?

Have a look and tell me what you think:>

Is there a speed difference between the two from turn in to apex? Just because the laptimes are close, one driver may be faster on entry than the other,. Is the line the same? You can see this in the data by the lateral G curve, compare them together for each corner to see if the cornering "lines" are the same. For instance, a G peak closer to the turn in indicates a late apex, or closer to the exit will indicate an early apex. You could also add a math channel to the steering signal of DERV(Steering), to see the steering velocities of the two drivers, and compare this velocity on entry. It will show how abrupt one driver's steering maybe over the other. One may be a trail braking the other may not While there are many types of driving techniques that will separate many a driver from another , overall there are two types of driver's in the world, there are those that can rotate a car around any handling problem, commanding the car to his will and there are those that need the car to rotate for them, the car does the commanding of the technique. The setup that will follow either driver will be very different Lastly be careful how you use a calculated understeer (handling) graph, though it points out understeer and the places on the track it occurs, Understeer varies in value between drivers. Some drivers will not like any understeer at all, others will think a small to med understeer is neutral to them. While "handling" math channel charts are useful and a great reference for analysis, they do have a downside, there isn't an absolute number to understeer based on the driver's insight/experience and a really good driver that can drive around a problem car can make your handling graph look perfect even when the car is way off....

[speedsense]

Originally Posted by Rubinho

There are two Pi suites for data analysis. Club Expert Analysis which is supplied with the club logging products and Pi Toolbox which comes with the professional level products. Both Club Expert Analysis and Pi Toolbox Lite come included in the price of their respective data products. To get the full version of Pi Toolbox a hardware licence key must be purchased for £1000. The Lite version of Toolbox contains all features except real-time telemetry viewing, power spectral density display, and multiple worksheets/tasks. Full maths channel capability is in the Lite version.

The user interface is very similar to I2 and was designed first. Some people even say that the I2 interface is a rip-off of Toolbox's.

Well, actually CDS (Competition Data Systems) was first and Toolbox looks very much like CDS software,especially in the map plotting and graphical interface. And cost is half as much as either Motec or Pi, at least here in the states it is even though the software isn't free either, but is also included in the system package either as a Lite or Full verison. CDS was the first windows version (not DOS crammed into windows) and was also the first toolbox oriented software. While the Pi toolbox and I2 have some features that CDS doesn't, CDS contains a lot that the others don't have or even thought of. Overall the analysis package is extremely flexible and what you can analyze in 20 minutes in CDS, would take an hour in Toolbox and an hour and half in Motec.... The biggest difference is that CDS was designed and programmed by people who actually have touched and driven a race car and understand what it is that a racer needs to know and have created tools to make it possible. Probably the reason it's copied....

[EfiOz]

Originally Posted by Belatti

Does anyone knows wich systems does F1, Indy, GP2, F2, etc teams have?

F1 - A selection of Magneti Marelli, some Pi and a lot of custom built stuff.
Indy - Mostly Pi
GP2 - Magneti Marelli
F2 - Pi
F3 - Spec Bosch system

[EfiOz]

Originally Posted by Casper

You will always nearly always get a noisy signal at those rates and I would suggest it costs a lot of money to avoid or minimise it. Filtering is the name of the game, read Templeton's book to help here.

It look like a dodgy shaft inside the position sensor. They either get dirt or corrosion on them or the wiper loses tension and starts to jump and the ignal drops out. Active sensors are particluarly known for this.

I've logged suspension position at up to 1000hz with no noise problems. The amount of filtering required to clean up that signal would make it almost unusable or misleading in the least.

[Belatti]

Hi all, me again.

After a couple years working with Pi inside a team, I will do some freelance and planning to change ship for cost reasons... I bought a used AIM EVO4 with a MyCron3 dash.

The EVO4 has internal triaxial acelerometer, 2 wheel speeds, GPS, RPM cable, etc... and 5 analog channels. I plan to use two brake sensors, throttle and steering wheel there.

For the channel thats left I want to buy a channel expander so I can use 4 suspension potentiometers.

The thing where the AIM webpage seems to fail is to explain to me whats the difference between the Hub and the Ch. Exp.

http://www.aim-sportline.com/pages/c...-expansion.htm
http://www.aim-sportline.com/pages/c...r_data-hub.htm

Could someone give me a hand telling me how to connect and what to buy for it? Thanks!

[RF_Racer]

Originally Posted by Belatti

Hi all, me again.

After a couple years working with Pi inside a team, I will do some freelance and planning to change ship for cost reasons... I bought a used AIM EVO4 with a MyCron3 dash.

The EVO4 has internal triaxial acelerometer, 2 wheel speeds, GPS, RPM cable, etc... and 5 analog channels. I plan to use two brake sensors, throttle and steering wheel there.

For the channel thats left I want to buy a channel expander so I can use 4 suspension potentiometers.

The thing where the AIM webpage seems to fail is to explain to me whats the difference between the Hub and the Ch. Exp.

http://www.aim-sportline.com/pages/c...-expansion.htm
http://www.aim-sportline.com/pages/c...r_data-hub.htm

Could someone give me a hand telling me how to connect and what to buy for it? Thanks!

I been using AIM for a few years, the MXL Pista and have an Evo3 as well. The hub allows additional peripherals to be added where as the channel expansion creates additional digital/analogue channels.

I personally think the AIM system is great and easy to use. I tried Pi system 2 and was too fiddly.