High end data logging (F1?) question. Do they use GPS?

[Chris Wilson]

Do the F1 teams use GPS track mapping as part of their data logging strategy, or just G force sensors and a beacon? The few F1 traces I have seen in the public arena for Pi stuff have no GPS data, is this something that's been removed, or was never there in the first place?

Just curious... Thanks.

[tristancliffe]

Yes, all the F1 teams use GPS, partly for on-track postioning as required by the FIA, but also as part of their datalogging strategy. However, just like us, they are stuck with the limitations and accuracies (albeit supplimented with additional sattelite data to improve the error percentage) of GPS, and so do not use it for analysing driving lines, braking points, turning points - those are all done using conventional charts.

[RPD Motorsport]

As above, GPS is of little use for aynalasis due to lack of accuracy even at National level. We take data aynalasis down to 'do this 0.3m later...' etc. GPS is useful for creating pretty track maps, but most experianced data engineers wont even pull up a map and recognise the point on the track from the trace.

Aynalasis is baced on didtance from start/finish, and GPS used for stratagy/FIA etc.

[greenamex2]

When I first approached PI they talked me out of needing GPS. I thought this was because they didn't sell one at the time but having used the system I'd agree it isn't needed.

As PI have a (BIG) beacon at most circuits you don't need it for lap marking and the lap 'sectoring' seems to work pretty well without it.

[Paceracing]

Cosworth Electronics, (formerly Pi Research) do sell GPS systems for different applications including a high-accuracy 20Hz GPS which uses clever maths algorithms for error compensation due to errors in the GPS network. You are quite right however in that generally, the old fashioned method of a wheelspeed sensor to give distance, a lateral accelermeter integrated with speed gives corner radius and a track beacon to give end point distance is fairly reliable for track mapping. GPS is fraught with errors (even many GPS guided weapons have a quoted accuracy of up to 30M from the centre of a circle). To reduce the error you can use a Differential GPS which uses a second GPS at a known location calculating the error between known and measured location and broadcasting it back to the on-car GPS, however this is comes at a high cost as you can imagine.

Having played around with GPS overlays on Google Earth in the past, using cheaper commercially available GPS receivers, in some instances I have seen traces dissapear beyond the circuit perimeter and come back on-track again in the same lap! You generally need extensive filtering and maths 'fixes' to compensate for some of these errors.

I hope that this helps.

[Belatti]

I have read an article of a test in an F1 car with a high sample rate GPS (100Hz if Im not wrong). It was an RT3000 if I recall well. The system combined GPS and accelerometers and the data was passed through matematical filters to have good quality outputs. The accuracy of speed, time and position is great.

[Paceracing]

One point to mention,all GPS works at 1Hz. Faster GPS are resolving through maths algorithms and interpolation which effectively means 'best guess'.

[Chris Wilson]

Originally Posted by Paceracing

One point to mention,all GPS works at 1Hz. Faster GPS are resolving through maths algorithms and interpolation which effectively means 'best guess'.

Hmm, I am no expert but I would feel obliged to question that.

[Belatti]

Originally Posted by Chris Wilson

Hmm, I am no expert but I would feel obliged to question that.

+1

[Paceracing]

GPS data from satellites comes down in a data packet structure which amongst other things contains date, time, estimated error and other information. The structure of the data packet is such that the data frame length is 1 second in duration, (hence in GPS technology you will come across the term '1PPS' - 1 Packet Per Second) therefore to resolve GPS at higher rates, (i.e. more than 1Hz) you need a secondary method of calculation. Sometimes this is an informed 'guess' from the GPS receiver and other times, secondary sensors such as accelerometers and gyros will 'fill in the gaps' but invariably ultimate accuracy depends on the quality of such systems.

All the research is available so don't take my word for it. It's all well documented on the web.

Cheers,

Jason.

[nicemd]

What if we have signals from 6 satellites?

RaceLogic's performance box for instance, have no accelerometers or gyro but provide pretty good positioning. 10Hz they claim.

[Paceracing]

Good question! I don't know is the short answer however intuition suggests that all satellites are synchronised to transmit at exactly the same time but like I say, that's just my guess.

I have never used Race Logic but if it doesn't have accels or gyro's built in, it must be using maths to calculate and 'fill in' between data points.

[tristancliffe]

Originally Posted by nicemd

What if we have signals from 6 satellites?

RaceLogic's performance box for instance, have no accelerometers or gyro but provide pretty good positioning. 10Hz they claim.

Most affordable GPS receivers never actually use more than satellites, which is essentially the minimum required to determine position (the data from three is used to work out the position, whilst the 4th basically helps the receiver work out atmospheric delays). 5 or more satellites are ignored most of the time, although the data from both will be used to determine which 4 to use.

High end systems can use the data from more to reduce the error, combined with very complex maths, and supported by gyro/accelerometer/wheelspeed data too.

A really good system on a really good day can be accurate to within 3 meters. I have even heard of system that can get the accuracy to within millimeters, but I don't think it works with a moving target (e.g. a racing car). The military certainly don't get more than 3m accuracy. Enough for a cruise missile, but useless for determining driving lines.

GPS Drift, which is where the apparent position changes over time due to lots of small errors in the system means that often it is pointless comparing a morning session with an afternoon session, as you could be 20m out.

Someone in the Monoposto paddock tried to sell me a GPS datalogger - he claimed that the accuracy was better than 10cm, and he was quite cross and hurt when I told him that was impossible for the £250 cost.

[NJH]

Originally Posted by tristancliffe

GPS Drift, which is where the apparent position changes over time due to lots of small errors in the system means that often it is pointless comparing a morning session with an afternoon session, as you could be 20m out.

Seen it many times with one of the big name GPS data logger systems that a friend uses, the drift pretty much ruins attempting to do any meaningful sector comparison. I have seen it as bad as 10+ m just comparing some laps in the same race let alone comparing a morning to afternoon. In other words so bad that the line is clearly way off the actual track surface. Sometimes comparing to the official timing it may be as much as 2s a lap out other times none of the laps were more than 0.2s out. Complete and utter waste of money IMHO but I am sure someone will be along in a minute to tell me we were not using it properly.

Anyways when I was helping my mate I ended up looking at graphs of accel / decel a lot of the time and video footage.

[Chris Wilson]

My 5Hz Garmin unit has the occasional odd trait of seemingly to be very accurate around 3 "sides" of a circuit, assuming it to be roughly rectangular or square, yet the 4th side exhibits drift as if I were driving along the grass verge, or even along the other side of the Armco. Usually and thankfully this is rarely the case It is potentially very useful, and I hope the technology will improve, and prices diminish to make this line tracing better at a price we might all consider viable. I'd like to hear what someone with a high end GPS like the Motec 20 Hz system thinks and how it compares in reality with a basic 5Hz receiver.

[Paceracing]

[QUOTE=Chris Wilson;2965702]My 5Hz Garmin unit has the occasional odd trait of seemingly to be very accurate around 3 "sides" of a circuit, assuming it to be roughly rectangular or square, yet the 4th side exhibits drift as if I were driving along the grass verge, or even along the other side of the Armco. Usually and thankfully this is rarely the case It is potentially very useful, and I hope the technology will improve, and prices diminish to make this line tracing better at a price we might all consider viable. QUOTE]

It sounds like an issue with reflected signals bouncing of grandstands or other structures which adds a 'time delay' into the system, (i.e. the signal is recieved after it has bounced around a bit). I have seen a similar phenomena myself and this was the cause.

[MaxxUK]

Originally Posted by Paceracing

GPS data from satellites comes down in a data packet structure which amongst other things contains date, time, estimated error and other information. The structure of the data packet is such that the data frame length is 1 second in duration, (hence in GPS technology you will come across the term '1PPS' - 1 Packet Per Second) therefore to resolve GPS at higher rates, (i.e. more than 1Hz) you need a secondary method of calculation. Sometimes this is an informed 'guess' from the GPS receiver and other times, secondary sensors such as accelerometers and gyros will 'fill in the gaps' but invariably ultimate accuracy depends on the quality of such systems.

All the research is available so don't take my word for it. It's all well documented on the web.

Cheers,

Jason.

I've been using GPS based data-logging for 6yrs+ and data-logging in general for the best part of 20yrs and for the most part I am able to use it to determin sector time and 'general' racing line. I have had sessions where the line is spot on for the whole session (compared to each other and video) but occasionally there is a small glitch which throws a sector out. Some circuits are quite poor for getting very accurace GPS traces, I find Brands and Rockingham quite bad but have gotten very good results at times.

A lot of it is probbaly down to mounting of the antenna, if your only able to see 1/3 of the sky from it's loaction you are only going to get 1/3 the potential signal quality. Don't forget that as you turn you will be referencing different satellites and this can be a 'completely' new set if sky vision is restricted.

I know some stuff about GPS technology but not a great deal ... but I know a man who does. I put the above statement to him and he replied.

"The GPS signal has a number of components, which does include data andsignals that are used for calculating how far away a satellite is from the receiver. These signals are "continuous" and the GPS receiver keeps a constant lock on them - typically the correlator which is the bit of electronics and software which tracks these signals is updated every 20ms (50Hz) as this is convenient - but 20mS isn't a limitation. In fact in theory you could output GPS position/velocity/time at any rate you like, 1kHz, 100Hz, 0.1Hz... however in practice because the GPS signal is relatively weak and "noisy" there is little practical point in going above 20Hz. Calculating a solution faster than this means that you use lots of CPU power to calculate to solutions more frequently, then filter or smooth it to improve the noise performance - so you've done lots of work for almost no useful benefit."

I also asked about this question and answer

"What if we have signals from 6 satellites?"


"Good question! I don't know is the short answer however intuition
suggests that all satellites are synchronised to transmit at exactly
the same time but like I say, that's just my guess."

"Satellite transmissions are not synchronised - it's one of the receiver's jobs to track all the signals and then just interpolate to get the transmission time at a single point in time. The most signals the better, 4 being the minimum as the receiver needs to work out position in 3 axis + time. We need 4 equations', to find 4 unknowns. If there are more than 4 signals available then all the data is combined using weighted average techniques - so the "best" signals are used more than the others."

I also asked about accuracy. I'm not sure about the 1m quote for GPS 'alone' but when coupled with 'actual' accelerometer readings I would suggest this is about right.

Standard GPS positioning techniques are typically accurate to about 1m. If you've got the budget you can get to about 0.2m using a standalone receiver and subscription based satellite based differential correction services. Using local differential "RTK" techniques you can get to about 2cm.

The quotes are from the technical director of the company who many people regard as making the best/most accurate gps equipped data-logger.

From my point of view I use the system 'on track' maybe 120 days a year in my work as a driver coach/instructor. For the last 4 years with sync'd video. For the vast majority of the 1,000s of sessions that entails I have had data accurate enough to do proper sector timing down to the 100th, some I have even checked with frame based video comparison. It is quite adequate for examining the 'general' racing line but for really precise stuff you need to use in conjunction with video. Using the same companies 20Hz option I have achieved 'very' accurate racing line comparisons.

When I have gone to the trouble of making the lap marker exactly at the start finish line I have never seen lap time variance between the logger and official time of more than a tenth and often it is spot on the the hundredth.

There have been GPS Drift issues and GPS glitches but these are the exception rather than the rule and are generally obvious enough to discount.

Maxx

[Paceracing]

This could go round and round in circles. There are a number of excellent books written by Professor Jean-Marie Zogg, a lecturer at the University of Applied Sciences, Switzerland who specialises in GPS technology. For anyone interested, they are worth a read but a bit heavy going at times.

[nicemd]

Very interesting, thanks!

[Chris Wilson]

MaxxUK: Thanks for the detailed reply. Good info. So you feel the 20Hz receivers give a noticeable and worthwhile accuracy increase? One issue I have recently found is if you are inputting GPS direct into an engine ecu for logging is that the higher resolution ones (20 Hz and above) can apparently demand so much processing from the ecu processor that it can detract from its main job of running the engine. Motec, for example, have now declared that their current 20 hz receiver only be used to feed direct to one of their dash boards or standalone loggers, and not to feed their ecu's with the data down the serial port. Previously they sold a different 20 hz logger with no mention of this. I can only assume their tests have shown an uncomfortable amount of time being spent by the engine ecu processor messing with the additional data from the high end GPS receivers. I myself have been feeding 5Hz GPS into my Motec ecu direct, as I don't like digital dashes, and Motec's standalone loggers are too pricey for me at the moment. The ecu has oodles of spare processing at 5 Hz with my fairly basic demands for it on the engine side of things. I tried feeding a different, none Motec, and stand alone logger with GPS, again at 5 Hz, but Motec do clever stuff with filtering and maths, under the hood, so to speak, in their i2 analysis software for GPS channels. Trying to get anything near as good a result in the other logger for analysis with that loggers analysis software is proving a bit ropey. It won't be the logger, that just logs numbers, it's just matching the maths filtering i2 uses. Someone cleverer than myself might manage this, to me it's very daunting. As I said earlier, I think GPS is here to stay and as it evolves and hopefully gets better and cheaper for motorsport speciality usage, it is, for me, a useful visualisation tool.

Thanks for all the replies.

[Belatti]

Check this out:

Why do F1 teams still prefer to use optical sensor technology rather than GPS to measure accurate body velocity and slip angle?

Success in Formula 1 racing is determined by the smallest of margins. A split second shaved off a lap time, wheel traction and speed through the corners are all critical factors that decide who’ll eventually sit on top of the podium. Since the FIA standardised the specification of F1 cars this has become even more crucial. It’s now up to the mechanics and technicians to analyse and optimise the car’s performance so that it always delivers its maximum potential. The use of GPS and telemetry, has given racing engineers the power to interpret vast amounts of data collected either during a race or practice session, and use that information to tune the car for optimum performance. However, even though GPS is widely used for many applications, it is not reliable or accurate enough to be used in F1: the FIA banning of the use of 2-way telemetry and remote engine mapping probably didn’t help either. Instead, the industry has enhanced its focus on optical speed sensors in order to ensure that the cars always deliver their best on the circuits. Importantly, the teams rely on the optical sensors to be able to understand any new tyre developments.
So why does the industry prefer optical speed to GPS?

The problem

Car technology today is very complex. Every automotive company does its utmost to ensure that its cars perform as well as they can. This means that each of them invests heavily to constantly improve ride, handling, stability and performance. The same principles apply to F1 and LMP teams, current road cars have more control systems than any current F1 car! These tests have given us such benefits as ABS, ASR and ESP. In order to refine areas like velocity, acceleration, side slip and angles, further testing is necessary not just to improve performance but also to help to design the systems that will solve the problems. Optical speed, distance & slip-angle sensors play a predominant role in this area.

How do optical speed sensors help?

Optical speed sensors are used throughout the automotive industry for:

Distance measuring in all manner of operating states like standstill, braking, high-speed ride navigation and position.
Measuring the slip-free acquisition of vehicle speed.
The determination of slip between rail and track wheel.
The acquisition of speed changes with direct reference to the track.
Why does F1 prefer speed sensors to GPS for measuring body velocity and slip angle?

The advantages of Correvit optical sensors

Optical sensors have numerous advantages over GPS and offer the following benefits:

250 Hz output rate (4ms updates)
The slip angle is measured referencing the road surface and not the vehicle roof – this is not the same.
Has low latency.
Non-contact optical sensor
Measurement uncertainty of final value better than 0.1% because of precise optical grating technology.
Can be used under extreme environmental conditions.
Has a good linear output.
Easy to use.
Low maintenance and service demands following many years of refinement.
The disadvantages of GPS

GPS was seen as a “magic bullet” and, for some applications, it is very useful but GPS cannot compete with optical sensors because of its disadvantages:

Only 20 Hz output rate.
The antennas have to be mounted on the car roof, which is not good for transient testing.
Noise is high ( averaging can help the GPS but there are long term errors, lasting 10 minutes normally, which is about 0.3 degrees peak-to-peak: the averaging process can not remove this.)
To get maximum accuracy you need a 2 meter separation of the antennas

[Casper]

I have a close friend who is into GPS design for military purposes and he laughs out loud at anyone who tells him GPS is a good thing for DA on race cars. Me, as an observer and having no technical background but who has used and sworn at GPS for DA over the last five years I have to agree with him. Doubtless that the system we use is a low end device built to a budget and in the end just about all it is good for is lap timing compared to physical sensors and that might be debatable. I would like to get two GPS timers into a car and run it for about twenty laps and compare the results, my intuition tells me the times on each would vary quite surprisingly.

[RPD Motorsport]

Must agree. We have done this to... We ran a car with 2 complete logger systems including 2 seperate GPS units, 2 IR timing units and we were also supported by TSL with magnetic timing as in race meetings.

The IR matched the TSL timing down to at least 0.001 each lap, and each GPS time was +- 0.6 each lap.

Yes it may be POSSIBLE to get GPS to work down to the cm from an engineering point of view but in real life we dont, the systems we use in motorsport (and the MOD in most cases) can't, and most of all its just not needed. Timing can be done from IR/magnetic, loggind data and telemitry can be done very accurately by distance from the IR trap, and everything else done from there.

Its intresting to get a track map from it as sometimes road camber etc can skew an acceleration based one, but an imperfect map has no bearing on data aynalasis or on any results/calculations. 99% of the time our data engineers dont even look at a track map and only pull one up to go through things with drivers.

And just as info, the FIA timing screens (as discussed earllier in the thread for F1/LMS) are done by a unified distance telemetry feed and the timing marker. All teams must have a channel called 'calc_speed' which uses a pre-defined wheel cirumferance and takes the higest speed of the non driven wheels (to eliminate lock up errors).

[Icarus_nz]

In my experience a lot of depends on what your are trying to achieve.

For those of us bought up on accelerometers and squiggles the GPS stuff seems full of problems.

In terms of trying to establish 'absolutes' GPS does have problems for all the reasons above.

However, if your focus is on identifying trends and 'relatives' GPS is very visual, and that is good for drivers to perform self analysis.
Squiggles are abstract and it can take time to learn to read them well. Many drivers struggle to focus long enough to learn.

Bear in mind that I am taking about lower levels of motorsport here.

Many engineers these days are spoiled by the quality of info they have. Having to use a different or even a really basic logger is a good disapline. Identifying effectively what needs to happen by crystal ball gazing will test your understanding of the fundelmentals.

As always the question is driver or car?

I"m interested to hear the comment that a team is telling its drivers that (for example) to brake 0.3m later. Thats a big call. At the end of the day the driver is the first point of feedback and they need to be encouraged to develop 'feel', not an absolute reliance on the data.
Mark Webber used to say "Did it understeer? No. Then go a foot deeper. Yes? Pull back" The track surface is dynamic and only the driver can feel that. The suggestion to go deeper from data may be correct at that point in time (the past) and may point to a trend but is never an absolute.

[cliveagyoung]

I'm late picking this up!!
OK so they are not too accurate but for a quick in car reference they help.

Anybody recommend a low cost make I can use for comparisons on track.??