Hey all,

This isn't really new, but I thought I'd put it out there. I've written a python script that pulls all the meta data out of the PDR recordings. All the data and the Metadata tags (keys to decoding the data) are in the mp4 files recorded by the PDR. There are other tools that can pull the data, but I wanted to write one in Python. When the script is done running, it outputs a csv file with all tags on the top (67 parameters), time at the left and the rest of the spreadsheet filled with the data sampled during the mp4 recording.

It's interesting to see this information all in one place. One thing I've done with it so far is a lap analyzer. This is my first session of the day at Portland International Raceway.

Lap Analysis:
================================================================================
Lap Clock Time Lap Time Speed @ S/F GPS Location
--------------------------------------------------------------------------------
1 2.22 min N/A N/A 45.594689, -122.693509
2 4.13 min 114.58s N/A 45.594682, -122.693499
3 5.98 min 111.10s N/A 45.594681, -122.693471
4 7.81 min 110.00s N/A 45.594683, -122.693511
5 9.61 min 108.10s N/A 45.594677, -122.693490
6 11.50 min 113.10s N/A 45.594681, -122.693507
7 13.33 min 110.10s N/A 45.594709, -122.693513
8 15.02 min 101.29s N/A 45.594681, -122.693499
9 16.83 min 108.50s N/A 45.594687, -122.693503

Lap Time Statistics:
--------------------------------------------------------------------------------
Fastest lap: 101.29s (Lap 8)
Slowest lap: 114.58s (Lap 2)
Average lap time: 109.60s
Total session time: 876.77s (14.61 min)

If anyone has any suggestions for this tool, let me know. It's a public repo on github if you want the code.

ACV

 

Can you see what channels are available in the data?

 

Here are the channels in the files I recorded. I think this will be relatively stable, but it could change for different versions of the PDR. All this was extracted from the MP4 file itself.

Channel 01: Boost Pressure Ind
Channel 02: Coolant Temp
Channel 03: Intake Boost Pressure
Channel 04: Oil Pressure
Channel 05: Oil Temp
Channel 06: RPM
Channel 07: Accelerator
Channel 08: Clutch Pos
Channel 09: Brake Pos
Channel 10: LF Tyre Pressure
Channel 11: RF Tyre Pressure
Channel 12: LR Tyre Pressure
Channel 13: RR Tyre Pressure
Channel 14: LF Tyre Pressure Status
Channel 15: RF Tyre Pressure Status
Channel 16: LR Tyre Pressure Status
Channel 17: RR Tyre Pressure Status
Channel 18: LF Tyre Temp
Channel 19: RF Tyre Temp
Channel 20: LR Tyre Temp
Channel 21: RR Tyre Temp
Channel 22: Gear
Channel 23: Trans Oil Temp
Channel 24: Fuel Level
Channel 25: Fuel Capacity
Channel 26: Steering Angle
Channel 27: Distance
Channel 28: Speed
Channel 29: ABS Active
Channel 30: Traction Control Active
Channel 31: Vehicle Stability Active
Channel 32: Performance Traction Management
Channel 33: Driver Performance Mode
Channel 34: Battery Voltage
Channel 35: Lateral Acceleration
Channel 36: Longitudinal Acceleration
Channel 37: Vertical Acceleration
Channel 38: Latitude
Channel 39: Longitude
Channel 40: Altitude
Channel 41: Heading
Channel 42: GPS Fix
Channel 43: GPS Precision
Channel 44: Number of Satellites
Channel 45: Beacon
Channel 46: Temperature Multimedia Processor
Channel 47: CPU User
Channel 48: CPU System
Channel 49: CPU Idle
Channel 50: CPU IO Wait
Channel 51: CPU IRQ
Channel 52: SD Card Read Data Rate
Channel 53: SD Card Read Operations Rate
Channel 54: SD Card Read Time
Channel 55: SD Card Write Data Rate
Channel 56: SD Card Write Operations Rate
Channel 57: SD Card Write Time
Channel 58: Recording Event Odometer
Channel 59: Wheel Speed Left Front
Channel 60: Wheel Speed Right Front
Channel 61: Wheel Speed Left Rear
Channel 62: Wheel Speed Right Rear
Channel 63: Outside Air Temperature
Channel 64: Intake Air Temperature
Channel 65: Temperature Internal Board
Channel 66: Temperature Camera Module
Channel 67: Yaw Rate

 

Nice to be able to extract the data that the car provides. Cosworth’s Pi Toolbox (free with a registration) can graph these channels for subsequent analysis. Check out HiPo Driver on YouTube to understand how to use the tool. It’s not intuitive but he has developed and offers (free) files for analysis. It’s very comprehensive. It’s my go-to when analyzing data post sessions.

 

I use the Cosworth toolbox as well. It's really neat for what it is. It gives the data as graphs, but it doesn't really analyze it. Questions come to mind. How did I do lap-to-lap? What's my weakest corner? How do I compare to similar cars? Don't get me wrong, I think it's great to have the PDR and the Cosworth Toolbox. I think we can do more. For example, with this data real analysis can be done for Corvette on tracks around the nation. When I went to Sonoma or Laguna Seca, I could have a virtual coach that uses data from a dozen other cars to coach me to the fast line on these tracks. This line is based on the data of experienced drivers. It's something that could easily be done for any Corvette with a PDR. I don't think there's another manufacturer that could do this. I just hope the Corvette boys in Bowling Green are thinking along the same line.

For me this journey started when the ZR1 did not crush the overall Nürburgring record. I believe it has to do with active aero. I decided to see what the PDR had in terms of data to decide if it was enough to control an active aero wing and dive planes. In short: it is. My thought was I could make a demo of this using small servo motors and a microcontroller. Imagine watching the PRD video and having a small wing / dive planes on the desktop reacting to the what's happening on screen. Also, with the speed of these cars, I would add a pitot tube to the system to get static and dynamic pressure, so the wing is working like a true airfoil. It's reacting to the driver input and the air around the car -- just like a modern fighter jet. All of this is not only possible but cost effective.

 

Nice work, thanks.

It took me a minute to figure out what to do with everything, but some Google and AI queries got me squared away.
I'm able to extract the meta data and create the .csv file. Unfortunately I'm a google sheets person and not a MS Office person. Google can't handle more than about 147,000 rows with the full data (limited to 10M cells). Excel can handle a LOT more, but I don't have it.

Is it possible to be able to tack a single PDR input file and break out the .csv of the meta data by lap? So the output would be PDR_####-Lap-1.csv, PDR_####-Lap-2.csv, etc.?
If that's possible then Google Sheets could handle the resulting number of output cells required by lap for a large session.

I use a Garmin Catalyst for the coaching part, so far I've loved it.

Using the files from HiPo driver on his website I've been able to do lap to lap comparisons just fine. (https://www.hipodriver.com/resources)
You'll want to scroll down and download this: PDR2-Lite-2022-11

If other drivers would share their files one could simply import their data and compare it to their own data and overlay the two in Cosworth Pi.

One thing I would LOVE is to be able to edit the files and keep the meta data. I haven't found a video editor (yet) that doesn't destroy the meta data. Is it possible to scrape the data into a file, edit the video file in your editor of choice, and then write back the correct data? I could understand if the trim was only made at the end this might be possible, but if I trim 10s from the start and 20 mins. from the end how do you sync the data back up?

 

Very cool that you have the data out. Was looking for something to do to run some models on it.

Care to share the Github link?

Keep in mind that Cosworth Toolbox (a limited version for GM) and Pi Toolbox (the full Cosworth product) have very different capabilities.

The GM Cosworth one is very limited. Pi toolbox lets you do lap analysis, turn analysis and comparisons and such, so has most of the capabilities being referenced.

 

It was easy to find with a Google search:
https://github.com/a7v7/gm_pdr_analyzer/

 

Unfortunately @a7v7 hasn't logged on since this post.

I'm trying to use the PDR data on a site called Firelaps, but talking with the creator he says the PDR GPS information isn't correct. I was hoping a7v7 could chime in with some details from what he has discovered.

Essentially, in order to use Firelaps it requires the absolute GPS time, and according to the sample from your script I sent them, what is extracted is only the session time (how long since start, not when start - well, started).

Hoping this is something we could get from the MP4 still.

 

The PDR file also works on Cosworth PI, Hipo driver has some handy worksheets to get you going.

 

I was using HiPo Drivers worksheets on Pi Toolbox last year. When I attempted to obtain a new license this year it point me to a subscription plan on version 14. AliveDrive as it's called works very closly to what HiPo Driver set up in the Lite version. I've subscriped at $66/yr for the Plus version which provides me a lot of info to digest. They have an automatic license key process that gets dumb and requires running a repair executable file (which they'll provide you) on occasion. If you reach out to them they'll get you the executable file if you have issues. Just keep it and run it if they tool doesn't recognize your license key.

 

Thanks, but this whole thread is about taking the data in the MP4 file and using that data elsewhere.

I already use the Cosworth Pi software, and not only is it cumbersome, it lacks some basic functionality. HiPo has done a great job working with the limits, but frankly it's not worth the effort (time) for the output. I get far more insight out of my Garmin Catalyst in minutes, then I do having to pull the SD card, connect my laptop, start up the software, load the file, and then start.

Firelaps is a great alternative, and as it's web based it's super easy to gain insight from. Sure, you have to upload the data from the SD card still if you went that route with the PDR, but Firelaps has a device that collects and submits data automatically.

For now I'm using a $149 Racebox Mini (price has gone up since the initial launch). It collects data, I download it on my phone from the website, and I upload it back to Firelaps. I get analysis and results back in about 60s. I can run it on a tablet vs. hauling around a laptop at the track as well.

 

Im building off the work from the OP but adding plots and GPS track natively into my tool. If I successfully package this into an executable I will share it.

Screenshot attached

 

Wow! That's looking nice already.

I really, really, really wish that the PDR system properly captured and encoded the GPS time. It seems that it logs time based on when the recording starts, so you get a relative time to your start, but not an absolute GPS time. With absolute GPS time the data is far more portable to other systems, but perhaps that's exactly why they did this.

What I've learned is the PDR talks a great game, but it's functionality as a track tool is near zero.

 

Hmm, Ill have to look into the absolute GPS time. I dont see why it wouldnt log GPS time. But yes the track log starts when you hit go on the PDR screen, if thats what youre referring to.

I have been able to output a log file which works with the free version of Motec I2, but you dont get sector times with the free version.

I disagree about the lack of track functionality. All the key pieces are there, but the existing tools on the market are terrible. Hence my Claude project =-D

 

I see what you mean. The timing is not exposed on the CAN bus to the telemetry logger. It may be logged in the infotainment system, though.

Can you explain how not having the time breaks compatibility with other systems? I haven't worked in other systems enough to have that cause an issue for me.

 

Hey guys,

Sorry I didn't look earlier. This is awesome. I would really like to see the code for the mapping project! BTW I use Claude code. The original work I posted on github was done largely with Cc.

My thinking is to create a website where you can upload PDR data. Claude could analyze it and give you an report of how your doing. Where you can find speed, etc. The idea is to open it to multiple people and multiple tracks. You could compare your data to friends. The Claude analysis would be humorous. I'm impressed with how funny Claude can be.

I think of this as a hyper-focused social network: Corvette / PDR / Track Day analysis.

ACV

 

I'm sorry I've only just seen this now. If you are interested, PM me and we could discuss details of what you are trying to do.

 

Per usual, I quickly went way deeper on this project than I had planned. (and spent way more money in Claude tokens)

Video is now part of the dashboard and is time synced with the plots and the icon on the map.
Map sectors can be added/removed, and configured
Notes page added
Loading PDR data is native to the dashboard, not in a separate UI
Added brake/throttle color gradient map overlay
Added lap visibility shortcut key to toggle between adjacent laps

I tested this at a track day and still find it hard to get easy and clear insight into how a sector was better than another. The brake/throttle overlay is of high value, as are the brake/throttle traces. I wish there was a logged channel to tell me what the ediff is doing as that could be an indicator of total traction limit. I have some ideas for better insights, but havent implemented yet.

 

I did find an interesting plot which shows a driving error.

This moment is some high speed oversteer which was induced by early throttle application. The green line in steering angle, orange is brake percentage, and blue is throttle percentage. The sharp wave in the green trace shows me counter steering the car to correct the slide, but you can also see that I both lift off the throttle and apply brake, which could have induced a further slide/spin, and I really should have just lifted a bit on the throttle and not applied any brake. Its fun to look through the data and find these things, and once you see them its easier to spot next time.