THSCC Forums

I know I was in kind of sensory overload mode for most of the event, but I thought that the grip was pretty sensational in Greenville. I had no problems whatsoever - of course, 10.5" wide (yes, I measured them) Hoosiers may mask a lot of what others are experiencing...?

--Karl, who needs more loud pedal...

Forget Greenville. Let’s go back and look at Greensboro real quick. I wanted to further illustrate why you have to take everything into account when looking at data.

Course map from Greensboro:


I want to focus on the section between “loop in” and “loop out”. For those that weren’t there, the little loop was a small square instead of a pivot cone but it drove basically the same. The section leading into it crested a hill and then fell away from you in both altitude and camber. Pretty much the prefect recipe for spinning like a top no matter what you are driving. Let’s take a look at the longitudinal acceleration through that section as seen by the DL-1. Remember, this is supposed to represent the acceleration and braking of the car.

Longitudinal acceleration through Sector 2 (“loop in” to “loop out”):


Everything starts off great as you can see me accelerating briefly around the corner down the hill towards the loop. I had to lift just prior to entering this sector so the car would make the turn. Right around the 0.5 second mark, I get hard on the brakes trying to slow for the loop. Slightly past the 0.9 second mark, it looks like I get off the brakes and punch the gas. I highlighted the section from ~2.6 to ~4.0 seconds on the graph (darker line) so we can see where this lines up with the track map.

Zoomed-in Sector 2 highlighting the area marked in the longitudinal acceleration graph:


That seems pretty reasonable, doesn’t it? It looks like I may have slowed a bit too much and then sped back up again to get near the traction limit for the turn. Right after that point, it looks like I slowed again slightly (oops, too much gas too early?) and then sped off down the hill. If you were there, you would know that was a perfectly reasonable approach to that tricky section of the course.

This analysis stuff is easy, right? Guess what? That ain't even close to what actually happened.

Next time: Accelerometers? TRUTH? THEY CAN’T HANDLE THE TRUTH!

Jim

So what is the truth? This is a good time to take a look at some of the data that can be derived from the GPS positional data. This is very similar to what the GPS-only based products will show you and illustrates some of the differences and limitations of each approach. If you read the previous “course data rants”, you might remember that the calculated speed from the GPS positional data is quite accurate despite the best efforts of Bob. As a result, deriving acceleration from the calculated speed data can be accurate as well subject to the inherent limitations of the slow sampling speed. This next graph represents the longitudinal acceleration based on the GPS data for Sector 2.

GPS-based Longitudinal acceleration through Sector 2 (“loop in” to “loop out”):


Note that the highlighted section in the above graph exactly corresponds to the highlighted sections in the previous graphs. According to this graph, it looks like I’m still slowing down in that portion of the course! Which one is right? Let’s see what they look like on top of each other.

Accelerometer and GPS-based Longitudinal acceleration through Sector 2 (“loop in” to “loop out”):


Yikes! While they agree with each other at least in shape near the beginning of the graph, they tell wildly different stories after that. Starting from the beginning, the reason the first GPS-based peak (red line) isn’t as high the accelerometer-based (black line) one is simply sampling rate. Remember that the GPS data is only being sampled 10 times per seconds while the accelerometer data is being sampled 100 times per second. The quick right-hand transition at the start of the sector is too fast to be accurately portrayed by the GPS data. In this case, believe the accelerometer data.

The data for the braking zone from ~0.5 to ~0.9 seconds seems to agree as the GPS-based graph recovers from its false peak. From there, it seems that we are looking at the data from 2 different cars or something. According to the GPS-based data, I don’t start accelerating until roughly the 4.9 second mark as the graph finally manages to get above the zero mark again. According to the accelerometer-based data, I’m slowing again at the 4.9 second mark and don’t start accelerating until roughly the 6 second mark. Huh?

Is Bob playing with us again? Have the accelerometers gone haywire? Let’s step back. What can we trust again? Speed. It’s all about speed. We can always trust speed. It may not be accurate to more than 1 mph but it at least provides us with a reality check. As if the last graph wasn’t busy enough already, let’s toss speed into the mix.

Speed along with Accelerometer and GPS-based Longitudinal acceleration through Sector 2 (“loop in” to “loop out”):


Ah hah! It turns out that Bob was the one telling the truth this time. Wait, I thought we were using the accelerometers as a “lie detector” for Bob? Do we need a lie detector for our lie detector? It’s like when we get a fuzz-buster and the police in Virginia go and get a fuzz-buster-buster. A cookie for anybody who knows what obscure movie the “fuzz-buster-buster” came from.

According to the speed graph (green line), I was slowing down that entire time and didn’t start accelerating until roughly the 4.9 second mark. Wow, that is exactly what Bob tried to tell us earlier. Scary stuff! So what in the Sam Hill is going on?

Next time: “Initial D gots nuddin’ on me” or “Surprise, it’s a trunk monkey!”

Jim

So, if we assume data gathering devices don't lie, here's what we know. You were slowing down in that segment. Your DL-1's accelerometers were not. Could they be moving around in your elaborate "mounting system"? Was your DL-1 mounted in any way differently between the two events? Are you seeing peak G's more based on how far your car can sling your DL-1 around in a slightly loose mounting scenario?


--Donnie

I've had it mounted the same way in the red CRX each time.



It's the best 50-cent mount money can buy! Besides, with all the drinks and such spilled there over the years, that carpet is solid!

You do bring up a good point. While I have no doubt there is some small effect, I can't seem to measure it in the small amount of testing I've done. I mounted it using tape, velcro and a clamp at different spots in a Miata and went up the street and back. I was able to turn into my street fairly consistently and afterwards, when looking at the data, I couldn't tell which way the unit was mounted other than a bad angle. I finally had to quit when the church across the street started letting out.

Jim

I saw much higher G spikes in the Spyder on my Datacam2 when I had it mounted with *just* velcro than when I added a really tight bungy over it to cinch it down. That's why I threw that out.

Not that I doubt your mounting job is solid or anything.

Hey, maybe it's not the carpet that's an issue...maybe the entire floorpan is moving!

--Donnie

Now *that*, I'll buy!

We will know soon enough as I'm having a bracket made which will bolt into the floor. Or at least the closest thing to a solid object we can find.

Either way, the trends remain the same even if the extremes are off by a smidgen.

Jim

I think either there was a small earthquake right as you were setting up for the loop, or you were drifting.

Perhaps the suspension is messing with the accelerometer data, in the sense that the car will pitch and yaw to some degree when accelerating and braking(and turning)? While you may be decelerating in the forward direction, you could be accelerating in a sideways direction...

Or, your inner alien is confusing the sensors.

_________________
~ RallyX VP 2009 ~
- - - - - - - - - - - - - - - - - - - - - - - -
'92 Prelude
'76 Celica

I should have clarified my statements about my mounting experiments. I never mounted the thing with just velcro. I tried 3 different methods with the first one "involving" velcro. For that one, I had 2 strips of velcro on the dash as well as the DL-1 but I also had a bunch of tape holding it down as well. The velcro alone seemed a little wobbly which is what prompted a trip to Home Depot for the tape. Maybe that is why I haven't seen any big spikes?

Jim

It was a whole lot of the latter but I'm inclined to believe the former started it.

It is also the perfect example of what happens when you try to use a trunk monkey as an accelerometer. Believe it! Trust me...

Jim
- hey, gotta keep the Subaru guys in the back awake...

Tell me more about Greenville...Since I was there, I can relate more to that course.

Please

_________________
~ RallyX VP 2009 ~
- - - - - - - - - - - - - - - - - - - - - - - -
'92 Prelude
'76 Celica

I will, I promise. I'm going to finish up my story about the Trunk Monkey disguised as an accelerometer and we'll head back east.

Here is a video from yesterday's CCR event at CMP in Kershaw. They had an auto-x event on part of the road course similar to the Nasa events.

It was *cold* and windy which, even though I froze my petunias off, was the perfect to follow up to Greensboro. I get one more shot at the cold at VMP today and hopefully whatever I learn will still be valid in the fall (read: Topeka).

Jim

Here is a video from my best run at VMP with NASA yesterday. This was run 6 (I think, out of 12!) and even though the tires got plenty of heat in them, you can see the car rotates pretty well.

There was a big rotation near the start of the course and it will be interesting to compare that section with a couple of runs where I didn't let the car rotate quite that much and took more of a traditional approach.

The offsets after the first slalom were brutal. As anybody that's been to VMP can tell you, that section of the pavement is *very* bumpy and they had us going through there at nearly full tilt which was a touch exciting to say the least. I had a few moments in there on other runs where it felt like the entire car moved a couple of feet laterally when I hit the bumps wrong. It was real tough to be consistent through there and is another great spot to check out how much time was lost (or gained?) on other runs.

Jim

Jim,

My two cents worth of observations...

Regarding the GPS longitudinal data, does it calculate this as if you were on a flat plane (spherical earth) with no elevation changes? For example if you were falling off a cliff, what would your longitudinal speed be (zero or something much higher?). I don’t think this has any (or much) impact on the discrepancy that you are seeing, but in the back of my mind I keep thinking that both systems assume you are on a flat surface (with maybe the GPS values being actually calculated as if you was on a flat surface).

With your accelerometers only capturing X and Y axis (which is all I assume it does) that any elevation changes (Z axis) and pitch, roll and yaw of the car (due to course or body roll) affect the X and Y axis data. I know it does, but I don’t know how much. That is why I wonder if those systems that DO include Z axis, pitch, roll and yaw sensors would produce different longitudinal acceleration data? Maybe even data that better matches the GPS data? If it does affect it more than I think it does, I wonder if this might be part of the explanation. I was not at the event, but you describe that section with terms like “crest of hill” and “off camber” which suggests to me that this was a section that maybe was more “not flat” than the rest of the course?

If it is not due to the above then I wonder if it is a mounting problem (as Donnie mentions) or if there is some other flaw or broken assumption in how the X and Y acceleration data is generated from the accelerometers.

Richard

_________________
Richard Casto
1972 Porsche 914
2013 Honda Fit Sport
2015 Honda Fit EX
http://motorsport.zyyz.com
Money can't buy happiness, but somehow it's more comfortable to cry in a Porsche than a Kia.

You touched on some of the issues Richard.

Incidentally, I used a DBA (that's Donnie Barnes Approved) method to mount the DL-1 on Sunday for the VMP event. I wound up taking the rear storage panel thingy out (not exactly legal but it only weighs a few pounds) and used some scrap aluminum, thin rubber padding and sheet metal screws to secure the DL-1 directly to the car's body. It wasn't pretty nor anything I'll be using in the future but it was definitely secure. The result? The data looks the same as everything else I've been collecting. Draw your own conclusions...

The bottom line is that the accelerometers were not telling a lie. They were functioning *exactly* as designed and the readings are as valid as the calculated GPS-based data. Huh? It's true! Seriously. The problem is that the data has been skewed by a little monkey business. I'll explain in a bit...

Jim