How (& Why) To Filter Motorsports Data – Example Using Race Technology GoPro Software.
The Race Technology GoPro software aims to give you a great synchronised video and data experience for a modest outlay. However, if you have used it you may have found that the video is ok but the data is not. Specifically you might have found the acceleration data is tricky to interpret. You know there are secrets hiding within but how can you trust what you are seeing? One method is for you to filter motorsports data to make it more useful to you. But how?
This article (and 2 bonuses) is a worked example using RT’s software but you can apply the concepts to any data system. If your data system software doesn’t allow filters like this (AiM for example) then I have these spreadsheets you are welcome to use instead.
Wait! – Use A GoPro As A Video Datalogger??
Yes! If the idea of using a GoPro as a motorsports data analysis tool is news to you, then have a quick scan of my previous article on the subject. You can do this 👇
I am (sadly) not on commission for RT software sales. I do however use it every time I’m at the track. If you use data for driver development then having data instantly in sync with video is super-powerful. The video gives you context. The data gives you the detail on where to improve.
It is not all straight forward though – the acceleration data you need to get your insights, requires attention before you can trust it. Here is how.
GoPro Data Quality Issue
GoPro’s can log GPS and accelerometer data in sync with video. Great until you realise that some of the data quality is not that good.
The GPS data is at 10Hz and is generally good quality – equivalent to most consumer level motorsports data analysis systems. When it works you can get some key motorsports data analysis information:
- Lap times,
- Splits times,
- Speed trace,
- Delta-t(called “time slip” in this software)
- Plus, Lateral and Longitudinal acceleration,
Sadly, without help, these accelerations are not that good – the data you get can be inconsistent, confusing and hard to work with.
Having questionable acceleration data means that doing …
…. becomes unreliable.
This is the last thing you need.
“Where Am I Going To Find 2 Seconds Per Lap?”
Are the fundamentals of race driver data analysis new or confusing for you? You’re not alone!
Start pinpointing precisely where you can lower your lap times all by this time next week with:
👉 The Complete Beginners Guide To Data Analysis 👈
Give your data logger the chance to become more than just an expensive lap timer to you.
Choosing The Best Source Of Data
My first tip to get the most from the RT GoPro software, is that you should use the GPS derived acceleration data not the sensor data:
I’ve spent hours (and hours) reviewing comparisons of these two forms of the same data. One uses the acceleration sensors (and would normally be your preference) and one uses the GPS data.
My conclusion is that the GPS acceleration data seems to be more consistently aligned to what the car appears to be doing in the video. Not always but more than the acceleration sensors.
The rule is use the best data you have and so I’d recommend you go with the GPS acceleration data.
Raw Acceleration Data Looks Like This (A Mess!)
Below you can see what a typical unfiltered longitudinal acceleration trace looks like for a single lap:
Unfortunately this is completely useless for your data analysis.
What you need, is to extract the “signal from the noise” in this data. Filters are the answer.
Finding The Signal In That Noise
When you filter motorsports data (or any data frankly) you want to uncover the signal i.e. only the information that is useful too you. In this case, your whole race car accelerations. These are slow compared to say engine vibrations, but at the moment you’ve everything together.
The standard RT settings includes a “smoothing” to help you do this.
This is a rolling average that smooths the data together. This approach does produce a smoother, more useful looking trace, but there are issues with moving averages. Namely that the noise is not actually removed … all your engine vibrations are lumped in together with your whole car movements.
Luckily you have alternative filtering options within the RT software. These are what I am going to suggest you choose.
You just have to find out how to set them up. Read on …
For deeper insights on choosing filters and the issue with moving averages have a read of this article for a filter I build in Excel for this and watch the video.
What You Are Aiming For …
Amazingly when you do apply a proper filter to that noisy acceleration data above, you end up with something like this:
I still think it is astonishing really that this line is hiding in that raw data mess above – but it is!
With this longitudinal acceleration data you can start to use the shapes of the traces to help your driver data analysis.
Several “aha!” moments will soon be at your finger tips …
For example, look where the cursor (vertical line) is highlighted in that image above. Hopefully you can read the number but if not it is showing you that the race car was braking at -0.97g.
You can compared this to other parts of the track. If you do then you can see that this is where the driver has braked the hardest around the lap. You can then identify other areas where the driver might be able to brake harder and later for other corners. That could save you a lot of lap time.
Therefore, you can see that these squiggly lines become practically very useful. Again, if interpreting data is new for you then check out the course 👈 but hopefully you see that to filter motorsports data is worth it.
Video And Data Automatically In-Sync
When you filter motorsports acceleration data you end up with the two traces below the onboard video.
*Forgive the driving … Uber to the apex anyone! 🤦♂️🤣
In that small clip above, you can actually tell how well the driver (me!) is performing.
You can see:
- I had some power-on oversteer at corner exit.
- That made me open up the steering and lift off the throttle.
- The impact of this driving mistake you can see in the little dips on both the lateral and longitudinal acceleration traces.
- You can also see that I was off line – having just completed an overtake 😊 – and why having the video for context with your data is so valuable.
That is why you want to filter motorsports data in this case. But how?
How To Filter Motorsports Data In Race Technology
You have a number of options for filtering channel data.
To get to these options, open the “Variable Manager” from within the RT software:
The dialogue box for the Variable Manager looks like this:
Very Windows95 interface I’m afraid (sorry RT 🤷♂️) but you get used to it.
If you click on the little + signs, this will open up more options.
Click on the one next to “GPS” for the logged GPS channels. It should look like this:
This is beginning to be an assault on you eyes but push through.
Click on the + for “GPS lat accel (GPS lateral acceleration)”
Then click on the + for “Advanced options” and you should see this:
There is a lot going on here.
You can come back and have a play with all these settings. For now, I am going to focus just on the filtering.
You can see in the above image that next to “Filtering” is “Butterworth Lowpass, 0.4Hz”
This is what the settings look like after they have been set. Next to “Filter” for you I would expect it to say “None”
To change that you need to click on the word “Filtering” … Yes the actual word.
You should now see this:
The Filter Design Dialogue
Again there are a lot of options for you to try.
Select the radio button options and fill in the numbers you want here. Then click ok and see how this affects the results in the charts.
Bonus 1 – Going Deeper With Filter Selection
At this point I need to confess. I am no signal processing expert. There is a lot to know here. If you are reading this and want to give me some pointers, I am keen to learn so please sign-up to the newsletter and email me!
Here though you have two primary choices of filter. Butterworth and Chebyshev. These are compared in the image below that I got from this article with more information you might find interesting.
From what I can work out, the Butterworth provides a smoother trace but struggles when there are step changes. The Chebyshev filter is great at following sharp changes, however it has a ripple you can see in the image.
As I do not really know much about what is best, I created several versions with various settings and compared them.
Below is the same raw acceleration data filtered from before, compared to a number of different filtered versions:
The raw data is at the top. The filtered data is in the below 4 traces.
At first glance all the results for the filters all look fairly similar. Unfortunately the resolution of the image is a low so it is hard to analyse a proper comparison here for you.
Safe to say though I did a lot of trial and error trying different settings. I then watch several laps of onboard video. Then looking at the traces to see if that was I was seeing in the video reflected well in the data.
This took ages (!) as I used a number of different data sets from different cars and even different versions of GoPro.
Filter Settings For You To Try
In the end, I would have liked to have done this more scientifically. What I did was by eye but I’m reasonably happy that the settings I’ve chosen will also give you good results too. Perhaps if you try them let me know.
One thing I kept having to remind myself of is that this was all coming from a GoPro so not to get too fussy.
The settings you can try are in the “Filter design” dialogue box image above and repeated below.
I found the follow to work the best for both lateral and longitudinal acceleration:
- Upper frequency: 0.4 Hz
- Filter order: 4
Again if you know more than me on this, please let me know. Also if you try some settings and find an alternative works better, again please let me know (including your race car type), and I can update this article.
Finally, if you want a quicker way to iterate through Butterworth filter settings in Excel, please grab one of my spreadsheets here.
Summing Up – It Is Worth It !
Whilst this might seem something of an added pain to setup, I think you will find that the control it gives you over your data is worth it.
Hopefully I will no not have to change these filter settings again.
If you save your layout using the dialogue below then every time you open a new video you will have the same settings:
This clearly saves the whole layout you have for all your windows.
Bonus 2 – The Analytics Layout I Am Running
The tip here is to slow build out your own so that it makes sense to you.
Ultimately, the ability too rapidly compare video and data in sync like this has really helped me during (not just after!) testing days. This is the biggest bonus for me. Oh and the cost and convenience relative to other options.
Hopefully if you apply the same approach and filter motorsports acceleration data, you will be also able to get more from the software and your GoPro.
Best of luck and be sure to sign up for free to the newsletter below to be the first to know about my new articles, podcasts, tools and courses.
Where do you go after looking at split times? Try the delta-t channel. Learn about it here.
Are you using all your racecars grip? Know how to get this info from your logger? Here’s the article.
No clue where to start with motorsports data analysis? Could this be the course for you?