THSCC Forums

Donnie saw right where I was headed and he’s already experienced the pain of being in the presence of a mean drunk like Bob. Let’s look at how Bob smacks us down with respect to the DL-1’s software. It should be noted that Bob really isn’t that picky about time frame and he can wander the same distance in 15 minutes as he can in a day. Go back and look at the image of the zoomed-in area around the starting line. The red and black lines were my afternoon runs and the green line was my first run of the day. The morning run is much closer to the first afternoon run than the two afternoon runs are to each other. Bob is an unpredictable bloke to be sure!

First and last runs of the day:


Here is Bob in action. The software doesn’t allow you to offset any of the runs relative to each other so things don’t always line up very well. You can now see why some of my markers are drawn at odd angles. I’ve found that the software is a little more accurate if you can make the markers cross roughly the same point in the paths after factoring in the offset. There is a limit to this as some angles just don’t work and the marker can only be so far from perpendicular to the path. How far? Who knows-- the software isn’t very friendly and just tells you that it can’t be done after a certain point.

Even with my best attempts at defining meaningful markers, I found that this much drift is a bit too much to deal with. The bad news is that this is going to be common with the data collected throughout the day. I have no idea how the other software packages deal with this but it was really frustrating when I first ran into it. What happens is that while the start and end markers are usually good enough, you can’t compare parts of the course very well which, to me, is the entire point!

Drift in the onion:


Look at the havoc Bob is wreaking in the Onion! Trying to analyze this section for these 2 runs at the same time is impossible. Bob must die! How far was he wandering between these runs?

Drift at the start line:


Look at the scale in the image above, I’d guess those paths start 8 meters apart. That’s over 26 feet! After reviewing the RallyX data a while back, I started trying to figure out if I could fix this problem before Race Technology did. The software package really isn’t designed for this type of analysis and while they are fairly responsive to support suggestions, it seems like they’ve been dragging their feet on this issue. I have a few other things in mind as well but first things first.

I spent the last few weeks slowly picking apart their raw run file format and managed to decode most of the file. The punch line is that supposedly they are now sharing *some* of that information on their forum but I can’t get to it yet. That sure would have saved some time! Hopefully I can get that early next week to speed the process.

Anyway, I managed to read the raw run files into some C code and manipulate the positional data. Take a look at the following image.

Newly Bob-adjusted course start path:


The green line is the path I adjusted with my code. At this point it is a simple offset but it could be anything in the future. Like, say, Bob 3.0 adjusted offsets! So the start looks good, right? I could have gotten it a little closer but it gets hard to see both lines at that point. What about the rest of the course?

Newly Bob-adjusted full course path:

That’s a lot better, isn’t it! What about the onion?

Newly Bob-adjusted course onion path:


Here we see that while it isn’t perfect, it is a lot better than it was before. It’s close enough that we can set reasonable markers and get useful data from a side-by-side comparison. I think this is just the start of kicking Bob to the curb!

Newly Bob-adjusted course finish path:


Even the finish looks good. Look how close that final left turn is back onto the taxiway. This is definitely something we can use! I’m still working out a couple of issues with my adjustments but I think I should have something solid within the next week or two. I’m sure this just means Race Technology will release an update with this feature in 2 weeks but, as I said before, this is just the beginning.

Next installment: Look Ma! Actual data!

Jim

Some of this is over my head but I'm still laughing thinking about Bob. Would it help if I could take each line and make a seperate "layer" for each one in photoshop, so we can exactly overlap each one given the same starting point? Would that help us to see/understand the data better regardless of Bob's wandering?

I wanna help!

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

Thanks Ash.

That's similar to how I started with this whole process. I exported the raw positional data from a few runs into Excel and graphed it using the built-in tools. By playing with a base offset for each run, it was easy to make the paths line up well despite Bob.

The problem is that the analysis software doesn't allow you to import positional data that I can find. If I could have loaded my adjusted Excel data back into the software, everything would have been peachy. Instead, the only way to load data into the software is using the ".run" binary files that the DL-1 produces.

BTW, I got a call from Bob after my last post and he wanted to set the record straight. While he is a drunk, he isn't a bastard. Now Bob 3.0, *he* is a bastard!

Jim

Jim and Donnie.

Great info (especially the half I understand ):

Questions that you might want to address in coming episodes:

Background:

Are the MQGPS (or other) GPS only (no accelerometers) representation of g's accurate from run to run? (i.e. are they an accurate representation of relative g's so that runs can be compared?

With the MQGPS nominal 15 hour battery life, would it help "drift" issues to keep it on all day starting well before the first run?

Cut to the Chase which is "How can I learn to be faster, especially as a relatively experienced and competitive driver in a competitive car using one of these neat toys?"

First, How can DL1, etc or MQGPS (which seems to facilitate between run views of relative g's) be used to make the "next run" faster? With or without a faster codriver?

Second, How do Jim, Donnie, etc. use DL-1, etc. data after an event to help make them faster for the next event?

Thanks!

_________________
Dick Rasmussen

FS 50 2018 Mustang GT

Jim,

More questions:

It appears that even with your complex adjustments that the position errors in the plots are greater than the likely actual position variations on any good run on a typical autocross run. Especially on "thread the needle" courses or sites with "off line marbles" (aka Heartland Park and Sanford).

If so, what is the real value of the position data, especially at an event if comparing runs or drivers?

It seems to me that for those few course segments where there is a "real" choice of line (some big corners and optional slaloms) that an accurate segment time combined with the driver knowing which line was used would provide the useful data. What say?

Edit to add: One thing I like about GEEZ (but it is only useful after the event) is the ability to get pretty good relative segment times between runs due to the clearly identifiable cornering peaks (assuming consistent lines in each run). Do the GPS only units (MQGPS?) show accurate peaks in relation to time?

Also, from looking at the MQGPS charts from the website, it appears that you can set markers based on distance traveled. If so, is this distance (from run to run) close enough to give good relative segment times and/or average speeds in a segment?

Thanks!

_________________
Dick Rasmussen

FS 50 2018 Mustang GT

Yikes. You ask very good questions Dick. I'll do my best to answer what I can and the rest I'll just gloss over completely.




This is something I'm interested in myself. I can only answer for the GPS data pertaining to the DL-1 but I have a feeling it is representative of what you'll see in the other products.

The short answer is that the GPS-based acceleration data is definitely a little off compared to the accelerometer-based data. The peaks were off by as much as 10% in some cases and the GPS-based curve tended to lag the accelerometer-based curve. It's funny you ask this because I was going to touch on this subject next as it leads directly into looking at the course data. I'll touch on the peak issue as well as what I mean by the curves lagging each other.



My current observations lead me to believe that as long as the unit is turned on a good 20 minutes before your runs and left on for the duration of the heat, you'll get the same results. Hopefully I can do some more testing to validate this but the bottom line is that is certainly isn't going to hurt to leave the thing on.



Here comes the part where I gloss over your question. Honestly, I think this depends entirely on the person and the product. In my case, I'm not looking to do this between runs at the moment so I haven't given it too much thought. In a tour-type 3-run format, the last thing I want to be doing is messing with electronic gadgets between runs given the current state of things.

However, this is the exact area I want to look at with regard to enhancing usability.



Stay tuned for a look at this past weekends data. I figure I should get to it by page 7 or so.

Jim

These issues are directly related and follow the base operation of setting proper course markers (such as "start of run", "end of run", etc). Without accurate positional data, we would be back to relying on the accelerometer data to determine waypoints in the data stream.

As you know by using the GEEZ, pure accelerometer data really isn't a reliable way to get positional data. Sure, you can find places where you were pulling peak g's (or not) and based on your knowledge of your runs, make assumptions as to where you were on course. Wouldn't it be nice to see a map and *know* where you were? You always back up that positional observation by checking the data associated with it and that winds up being your "mark".

It's already been shown that the _relative_ positional data we are getting from the GPS units is quite good. The accuracy of individual segments are dependent on length and speed but it is generally in the order of a couple hundredths of a second. Is that good enough? And while it won't show you the exact detail of the maneuver, it will show you that you chose a different line into an optional section. I'll show an example of that real soon.

The fundamental issue is that the software made it occasionally impossible to draw proper markers for *multiple* runs at the same time. The positional offset hack of the raw ".run" files should help with that in the short term.

Jim

If we had a separate GPS unit that would let us mark the cone positions *AND* upload them to a computer I could add that layer in with the laps to get a better perspective on the data.

I can mark all the things I want with my unit, but if I can't upload the data to a useable form on the ol' computer, it's useless unless your dataloggers give coordinates. If they have coordinate capability then I might be able to go back in and add cone position but they would be approximate because of Bob and whatnot.

Ultimately I don't really care if understanding the GPS data makes me faster. I am a tech weenie so I like the GPS stuff, and I like problem solving and thinking about a drunk guy in a circle. It's fun just messing with him.

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

Put two GPS units in the car, one in front and one in rear, then use a third, stationary, unit to triangulate your location.

It's amazing to think that I've been accused of being a geek. Obviously, the accusers didn't associate with any other THSCC members!

No Karl is the bastard.

_________________
02 Focus SVT
STF 9

I can speak directly to this one on the MaxQ. DONT CUT IT OFF DURING THE DAY!!! Its a pain to constantly sync up to the satellites then to sync up the bluetooth.

I have only used mine once, but I just turned it on and put it in the cup holder of the Miata. With no top, it reads perfect, now I only need to work on mounting of the pocket pc so I make sure I get all my runs.

_________________
02 Focus SVT
STF 9

Yikes, that book is $110 used. May I borrow yours for some light reading?

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

hmmm. Ok so I should leave my unit running all day. No problem there, the "power outlet" (ne cigarette lighter plug) on the Soli is live even with the key off. Should I leave it in record mode also? I have a 1 GB card which should hold more than a day's worth of continuous data. Should be easy to see the drift if the car does a "Bob" while parked.
I noticed tonight that my software doesn't match the help files screen shots in many ways, so maybe I have a function for overlaying or offsetting paths that isn't in yours?
I'll know more once I have multiple run data to play with. so far I have only recorded a street trip to verify it was working, no repetitive data.

Donnie, the "thing in the corner" is another Blue Man peeking into the frame. If it entertains someone you know to see it as a penis so be it, I don't see it as one. Then again I don't usually go around looking for pictures of penises, not that there is anything wrong with that...

_________________
SPIN or WIN!
there's no glory for going slow.

Nope, DL-1 software is too buggy to work with huge files, and definitely doesn't have a "convenient" way to split files (you can do it, but it's a pita). Would also make analysis between runs impossible.



Not really, no. No easier than overlaying two runs where you start at the same place only to see the lines shifted by several meters.



I've never seen that.



Oh, the old "I don't have a problem, *you* have a problem" shtick! I get it!

Whatever. Genie trumps blue man anyway.

Dick, to address one of your points, I personally don't care to look at data between runs. I don't believe I could decompress enough *and* do that to be able to drive like I need to. Plus I feel like I'm pretty good at replaying in my head to know big things I could improve. As for after the day is over, it depends on my mood and how much sleep I need. If I have no codriver, I often don't use the data at all (I feel like I usually know pretty well what I did wrong -or- if I don't the data probably wouldn't tell me easily what I need to know anyway). If I have a codriver, then I almost always look at it. What's funny is I've never had a time where I didn't learn something I could have done better from the data, even when I've been the "better" driver in the car. I've also never had a time where I was the slower driver where the "better" one didn't learn something from me, too (even times where that better driver beat me by 1.5 total seconds at a Pro in the same car!). The point there is I think that almost every time you can learn something from your codriver and vice versa from looking at the data.

The key steps are:
. setup start and stop markers
. setup reasonable segment markers
. overlay two good runs
. compare and see who did better in what segments and the data will show why

One thing to watch out for is the dreaded "I beat you here, but only because I gave it up in the segment before by MORE." Just because you win a segment doesn't mean you did better in total. It may be that you *had* to lose a previous segment by more just to win this one by a little (and thus your net is a loss). Jonathan kicked my butt on both courses at Nationals, but there was at least one segment on each course I legitimately beat him, too. Same thing against Braun at the Atwater Pro in the Miata. I usually beat Danielle at Pros in the Spyder, but in most cases I learned something from her data to pick up time.


--Donnie

Donnie,

Your experience with comparing drivers with the DL1 is essentially exactly the same as what Miles Beam and I experienced when we codrove his Vette in 2005 and compared GEEZ runs after the events. We then had ideas for driver improvement for the next event.

Ditto for what I expect I could realistically do between runs and that is with just one driver. Other than a quick check of revs on my recording tach to help with borderline gear shifting situations, I do as you do.

That said, I wonder if the MQGPS display would allow a quick identification of relatively low g's in a particular corner where I "whimped out" but wasn't sure it was necessary to do so? If so, for me that might be the biggest potential value at a major event.

I'm looking for an excuse to spend some of my tight budget on a MQGPS but so far it looks like more seat time and/or more sets of new tires would be a better investment . . .

_________________
Dick Rasmussen

FS 50 2018 Mustang GT