|
|
01-21-2010, 10:43 PM | #45 |
Brigadier General
242
Rep 3,306
Posts |
__________________
"...it's not about the money and not about the brand of the car, it's about handling,performance and passion......And that, no other car has all together like an M3........when you talk about the most complete car the M is invincible." --Tony Kanaan. |
Appreciate
0
|
05-02-2010, 04:53 PM | #47 | |
Captain
80
Rep 961
Posts |
Quote:
I've had a load of fun figuring out how to use accelerometers to measure suspension motion. I've got a set of 8 accelerometer circuit boards (they're only $50 each from www.SparkFun.com) that communicate with my laptop over USB connections. They're about 3" square. In theory, that's two accelerometers for each corner (one for the chassis, one for the wheel), although in testing so far, I haven't actually hooked up more than four on one side of the car to see if I can get useable data. Here's a snapshot of the car with the wiring on the outside - the chassis sensor wiring is hidden but the wheel sensor wiring is just taped to the outside because that was the quickest and easiest way to get hooked up. My conclusion so far is that useable data awaits. That said, most of my learning so far is in the technology of accelerometers and laptops, and not in suspension motions and adjustments. But, I'm getting there. Some time this year I'll get this figured out. A few lessons learned: - the boards came pre-programmed. However, the data rate is too low and the demonstration software violates the ADXL345 sensor chip FIFO read timing parameters. So, I bought a programming dongle, downloaded the Atmel embedded development software, figured out how to program in "C" and then redid the data capture routines so they work properly. Along the way, I calibrated the sensor offsets to manageable levels and redesigned the sensor data transmission protocol (three times). - A round of email conversations with Future Technologies (the guys that make the USB chips on the sensor boards) allowed me to implement a work-around that made my Windows 7 laptop run its serial ports at 1 megabaud to keep up with the volume of sensor data. - There are no off-the-shelf programs that will take this kind of data and make it useful, so I bought a package called "Liberty Basic" and started writing Basic programs to capture and process the data. Even a few seconds of data overwhelms MS Excel, so it's got to be custom from end to end. - I gathered a bunch of data on my stock suspension. After days of trying to make the data make sense, I began to suspect that the data was fine, and it was a mechanical problem. Sure enough, the data was full of noise caused by the sensors wiggling around on the suspension. That day of collection became scrap at that point. - I noticed that rarely, but often enough to be noticed, my data sets had byte-alignment errors that presented themselves as out-of-range readings for the sensors I was using. At first, I ignored them, but the more I thought about them the more concerned I became. So, I once again rewrote the data capture protocol and added a serial number the data records I was collecting. Sure enough, there were three-second long "blank periods" in the data - it wan't just noise in the system, it was a wholesale drop-out. The combination of the protocol and my revised data-capture software made it clear that the sensors were fine - it was the laptop that was losing the data. It turned out that my Lenovo laptop's "electronic air bag" software was shutting off the hard drive when I hit a bump, which of course was when the most valuable data was coming in. I resolved the problem by replacing the OEM 250GB HDD with an Intel 160GB Solid State Drive. The SSD rocks, by the way, for lots of reasons that have nothing to do with its immunity to shock. My X200 now cold-boots about as fast as most machines recover from sleep mode. - After an extensive session of data gathering in early March, I finally have what I think will be usable data from the stock suspension. Sadly, it's also affected by the HDD protection dropouts, so before I can use it I have to write a clean-up routine to trim out any data that can't be verified as consistent and clean using the record serial numbers. The reason the March data is still waiting is because in March my day job started consuming my weekends. Second, the little time I did have was spent getting ready to install the KW Clubsport kit that was sitting on my garage floor. Through April, I found the time to do the install and after a bit of fiddling, I got the ride-height exactly where I wanted it. The next phase is to set the dampers. I don't have a working electronic system yet, so for the moment I'm going to follow the instructions Koni publishes for setting up double-adjustable dampers in their racing catalog. I'll keep plugging away at the sensor data, and once I get useable results, I'll design some new metal brackets to attach the sensors to the KW gear and gather more data. In the long run, the capture and display has to be real-time or near-real-time, but I'm still at the let's-walk-before-we-run stage. When I have data that allows me to produce a picture of the suspension motion, I'll post it. Last edited by JAJ; 05-02-2010 at 05:49 PM.. |
|
Appreciate
0
|
05-05-2010, 01:56 AM | #48 |
Lieutenant
31
Rep 533
Posts |
Great Write up...
surely i don't want to damage anything in the car. I got myself RD Sport Springs. What combo or alternative do you guys think i should go for ? coz honestly i love the write up and all...but i kinda dont understand it all...haha... let me know guys.
__________________
|
Appreciate
0
|
05-05-2010, 05:04 AM | #49 | |
Major
56
Rep 1,447
Posts |
Quote:
__________________
Khalid
'12 BMW E92 M3 '03 BMW E46 M3 '14 Mercedes-Benz S400 '18 Alfa Romeo Giulia Quadrifoglio |
|
Appreciate
0
|
05-25-2010, 11:05 PM | #52 |
Captain
80
Rep 961
Posts |
For those of you following this thread, ponder this:
I think it's kinda cool, actually. The context is that I ran the car at a modest speed over a speed bump. As you can see, the wheel (red line) ran over it, the body (green line) followed with a time delay, and the suspension (blue line) ran out of travel twice on compression at about 3cm upward travel with a 5cm extension in between. The first "run out of travel" event (the blue line rises to 3 cm or so and stops rising) is when the suspension absorbs the rising side of the speed bump. Once there's no travel left, the body has to follow the rise of the wheel over the bump. As the wheel goes over the top of the speed bump, it starts to drop, but the body continues rising. The suspension droops by 5 cm until the body starts to descend. The second "run out of travel" event is when the wheel stops dropping down the backside of the speed bump and starts to roll level again. The chassis keeps descending, compressing the suspension down on the bumpstops until it too stops moving. It bounces a couple of times before settling. The tire plays a critical role in the ride characteristics of the suspension. The brief descent of the "wheel" below ground level is evidence that the tire squashes (allowing the hub where the sensor is to drop below it's baseline position) absorbing some of the impact as the body drops onto the suspension. Chassis sensor - attached to the bodywork in the engine bay. I strapped it firmly to the strut brace so it moved with the bodywork. Wheel sensor - attached to the strut body behind the brake backing plate. Both sensors on the driver's side. This is a 1 second snippet of about 30 seconds of data, which in turn is one run out of about 15 segments in that specifci excursion accounting for about a half-hour of total data capture. The scale of the analysis is quite daunting! The vertical index lines are about 0.10 seconds apart. Last edited by JAJ; 05-26-2010 at 12:33 AM.. |
Appreciate
0
|
05-25-2010, 11:41 PM | #54 |
Captain
80
Rep 961
Posts |
For those of you who find this of interest, here's the raw acceleration data that allowed me to build the motion analysis above. The accel data is sampled at 800 samples a second. 980 cm/sec^2 is the force of gravity (one "G"). The max force at the wheel is around 4 G's upward and the max force at the chassis is around 1.8 G's upward and 0 G's (free fall) downward.
The position plot of the suspension is computed by integrating the acceleration to calculate velocity, then integrating the velocity to position. I'm theorizing that the roughly 80 ms (12 Hertz) oscillation in the differential and hub acceleration time-series is actually "hum" from the tire. It was a 235/40x18 Dunlop 3D snow tire when I captured the data back in March. Referring to the motion picture (the first post above) the red wiggly line in the wheel position picture (the "hub" position) drops below zero when the tire sidewall compresses and absorbs the fall of the car as it falls off the back side of the speed bump. It's amazing what you can see. Last edited by JAJ; 05-26-2010 at 12:17 AM.. |
Appreciate
0
|
05-26-2010, 12:13 AM | #55 | |
Captain
80
Rep 961
Posts |
Quote:
Second point - see my suspension calibration post from Friday May 21, 2010. Third point - see my other two posts from tonight, May 25, 2010. Sensor calibration is the biggest challenge. I've written (Basic) software to compute the timing of two independent but simlar sensor data streams precisely, resample (anti-alias) the datasets to a common timebase, physically calibrate and remove sensor offset at the sensor itself, then again in the post-processing, and finally do the geometry to project the raw 3D data onto a pure vertical axis. Each stage of integration, after all this care in getting clean data, still requires manual data selection and strictly controlled de-trending to provide useful and believable output. In the end, rather than measuring the damper or spring motions, I ended up actually measuring the hub as it's driven up-and-down by the tire and seeing how its motion drives through the suspension to move the chassis up and down. |
|
Appreciate
0
|
07-18-2010, 06:38 PM | #56 |
Banned
1982
Rep 1,847
Posts |
|
Appreciate
0
|
02-11-2011, 03:16 PM | #59 |
Kind of a big deal
303
Rep 1,674
Posts
Drives: an 1M not often enough
Join Date: Jan 2011
Location: between Unlimited and Hard to Get
|
Impressive, that you did quite some ground work on data acquisition, but more so that your approach of double integration of an accelerometer signal results in a clean and recognizable position curve (~ an under-damped chassis with dry/stick friction).
I should ask what you mean by 'run out of travel' events because these would cause sharp acceleration spikes, whereas your position curve looks to be within harmonic boundaries. Did you feel/hear the hard bump stops yourself during this measurement? Perhaps you do mean highest/lowest position on the harmonic curve? Just started a study on a custom EDC controller, but would rather use a u-Ctrl for that. Your PC setup is of course perfect for measurement and calibration, although I had hoped that more turn-key acquisition setups for accelerometers were available (at higher cost no doubt). subscribed! Last edited by eeghie; 02-11-2011 at 04:38 PM.. |
Appreciate
0
|
02-12-2011, 12:31 AM | #60 |
E60_to_E90
36
Rep 1,201
Posts |
Awesome write-up JAJ. A very informative read!
__________________
2011.5 E90 Dinan M3, SS II, M-DCT, EDC, M Drive. 20" MORR VS8.2 wheels
Performance Mods: Dinan St.1 suspension, exhaust, intake; TTPs; BPM St.II, DCT, servo tunes Cosmetic Mods: OEM CF splitters, mirrors, center console, spoiler; Arkym CF diffuser; LUX; tint |
Appreciate
0
|
02-16-2011, 09:11 PM | #61 | |
Captain
80
Rep 961
Posts |
Quote:
The run out of travel events are the flattening of the top of the suspension compression curve. I didn't go over the speed bump very quickly and it only just touches, so the "impact" is only really visible in the processed output - in the raw data it's buried in the noise and rumble. I don't recall a "thump", but then it was a year ago. Best of luck on the EDC controller - the challenge will be knowing how to select the settings, since you don't know where you are, only where you've been. The problem with damper tuning is that it's road-surface dependent, so a rough road benefits from a softer setting than a smooth road. Lateral G's don't really affect how much damping you need - it's all about texture. |
|
Appreciate
0
|
02-16-2011, 10:08 PM | #63 | ||
Kind of a big deal
303
Rep 1,674
Posts
Drives: an 1M not often enough
Join Date: Jan 2011
Location: between Unlimited and Hard to Get
|
Quote:
Quote:
I started diving into old Matlab stuff and found that it lets you record directly from the PC's line-in, but I would need an accelerometer with the right output signal. Did you consider accelerometers with analog output? |
||
Appreciate
0
|
02-18-2011, 12:18 AM | #64 | ||
Captain
80
Rep 961
Posts |
Quote:
For sets longer than a second, you don't get consistent trends for detrending. The accelerometer signal has (what I presume is) low frequency noise components that make the integrated result wander around by 10's of cm over a few seconds. Quote:
I thought a lot about analog accelerometers because I'm an analog circuits guy from decades ago. That would have been vastly easier for me to set up and calibrate than digital processing, but vastly harder to post process. I'd have been able to bandpass filter it, integrate and detrend it and get meaningful numbers without breaking a sweat. Digitally, it's not so easy for me. If I was starting now, though, I'd take a close look at Bosch BMA180 digital accelerometers - they have built-in bandpass filtering and data cleanup. They might work better. |
||
Appreciate
0
|
02-19-2011, 01:35 PM | #65 |
Major
159
Rep 1,288
Posts
Drives: 2014 435iX, FBO
Join Date: Mar 2009
Location: Ontario, Canada
|
JAJ, thanks a ton.... I am really enjoying this thread.... subscribed.
I just can't believe the great work your doing on all the data gathering - fabulous. As you said, you are doing this for you - but we are all loving it. I hope you will post the wheel bounce pics on the KW stuff and translate what that shows into real life feel.... IE: does it work? Can you show better/faster reaction of suspension travel and does that translate into better feel, better handling, lower track times. I guess what I am asking is: What is your success criteria and how are you going to measure it?
__________________
2020 M2 Competition HS, DCT, 763s, Carbon: splitter, side skirt, grill, diffuser, wing
Previous: 2014 BMW F32 435iX - JB4, MHD Flash, BMS Meth Kit, ER Intercooler, intake, catless DPs, KW Streets, 437M Reps with 245/35 and 275/30 Michelin PSS 4S |
Appreciate
0
|
04-08-2011, 04:37 AM | #66 | |
Lieutenant General
951
Rep 15,818
Posts |
Quote:
Would the Eibach route be counter effective? I do want a drop as the vert is really high in the front, but not at the cost of losing handling... and I do not want to change the shocks as I want to keep EDC. I was also looking at the KW Adjustable Spring Sleeve-Over Kit as well as the GC kit Though both of those kits cost 3 times more than a set of Eibach or H&R springs... I would like Dinan, but it's cost is too much to spend on a lease. Thank you for the amazing write up
__________________
|
|
Appreciate
0
|
Post Reply |
Bookmarks |
|
|