GPS drift

[steha]

Hi,
I'm currently using a racelogic driftbox for logging endurance races (8 hours), and we have a problem with GPS low frequency drift. The accuracy of a lap is just fine, but when comparing two laps that have several hours of real time in between,. the GPS data may differ 2-3 m in absolute positioning. This has some unfortunate effects when comparing laps with each other. That the racelogic lacks a good way to adjust the lap start offsets of laps to compensate for this is one thing, but I'm interested to know if I can expect the same kind of issues if purchasing a DLX with DashX, and if the application has any tools to compensate for this. Basically:

- Can I expect sat positioning drift of the same magnitude in the DL products as in the racelogic products.

- and does the DLX software have any way to compensate for this problem?

In short, we want to have very accurate comparisons of laps recorded several hours in between each other, without having to add an earthbound signal source, with or without manual compensation. Will any of the DL products solve this for us?


Thanks in advance

[faraday]

You want, I daresay without the associated price, more accurate GPS than the consumer/recreational style modules that are at the physical heart of all budget GPS automotive loggers make possible. What you are observing is real systematic GPS drift that can only be corrected with differential assistance.
Some of the sytems claim to have WAAS/EGNOS capability. I have no experience or much knowledge of these "free" system enhancements and reside in a part of the globe that does not have the service available anyway. I think I recall Racelogic claiming their kit was so enabled. Maybe R-T gear is likewise; I will check the literature later. The Driftbox uses the same module as the DL1.
The accuracy of the products in the short term (ms - s order of magnitude) is in the favour of the DL1 because the Driftbox's inertial sensor package is not suitable/used for the type of navigation/tracking that R-T uses with its 100Hz sampled accelerometers. Of course, this does nothing to enhance the long term accuracy you desire.
Current DL1s, on the other hand, have "raw" capability, which enables post-processing. The R-T post-processing software (extra cost) does not fix your issue directly, but by making PPP possible, should. This requires an internet connection and is not available at the time your run ends.
The longer the time between this and the time of your PPP post-processing, the more accurate this Precision Point locational method becomes as more stations report back their GPS guff to NASA or whoever it is. We are talking about days.
I am no expert in either theory or practice with this, so further research is necessary for you to fully acquaint yourself with the real benefits that the new DL1s can offer. I had the facility in a development unit late last year but I could not get PPP to work on my PC/connection. R-T have made enhancements to their software in this regard recently - I have yet to try it on my old data. Stay tuned (but not for 24hrs or so).

There is also a facility to manually shift the individual run files.
This is not applicable to individual laps, unless you reduce a log into individual (or chunks of) laps in individual files. When overlaying data in R-T analysis, the current number of selections is five per graph. These can come from between one and five files.

Hope this helps.

[steha]

Thanks a lot for the answer. So basically I won't gain much from buying a DLX when it comes to this particular problem (some PPP processing aside).

Actually I started with a plan to build my own software that splits the log into laps, and then match them with each other to "guess" the GPS drift between each lap. As you don't drive identically this will not be entirely accurate from lap to lap, but some suitable low pass filtering of the drift estimates should leave us with a fairly reasonable guess. With this drift value, I can then create a new log where the positioning data is corrected over time, and then use the normal logger software to play with it.

Of course, this all depends on that people are driving reasonably alike so that the car is not all over the place between laps, but that's not currently a problem.

Now all I need is some more spare time to actually do this.

[Andreas Lidstr]

Hmmmm

Dont get it all but...

...how far off will one be, let say after a 4 houre race using a DL1 when comparing different laps. One from the beginning off the race and one from the end?

Anyone that do Endurance racing that can provide a test log covering some houres of driving?

Reallly hope the DL1 has some advantages over the "other product" in positioning accuracy. (The other advantages are not that hard to see... ) Othervise I have to say sorry to steha for last nights "lies".
Time to go multilogger in your car I guess and do som live tests?

[steha]

Well from what I understand, according to faraday the actual GPS components in the driftbox and the DL1 are basically the same, which means that we will have the same long term (hours) problems.

But dual boxing the next race should be very interesting.

[Andreas Lidstr]

[steha]

So that should mean that there is a real and widespread need for a drift compensation application similar to what I was mentioning, provided that it is possible to actually make a good estimate from the input data.

I've only gotten so far yet as to reading a log file and convert the data to more sensible format (who the **** came up with the idea of degrees and minutes for latitude and longitude?).

Next step is automatic lap detection which should be fairly easy, and then try to find a nice algorithm for matching laps.

[faraday]

While I do not do endurance races, I'm normally at three-day meetings during which the first and last session/race of the day are 5-8 hours apart. I am not sure what the natural cycle time of GPS is, but over the duration of these meetings, it would be unlikely that many of our sessions fall on the same point in this cycle.
Map overlays from almost all meetings have shown greater variation within a session than between.
This is not obviously because the driver is crap or traffic is forcing different lines, but because the positional accuracy is no better than the natural drift. Most of my experience is with very fast cars - obviosly not F1, but in terms of laptime normally similar to elite F3 due to greater acceleration and more mechanical grip. This is probaly a situation that least favours low update rate GPS, but I still consider the result at least satisfactory and excellent value for money.

The point of this thread is not a comparison of the Driftbox and a DL1, but since I have had both fitted simultaneously in several cars, I feel obliged to state that the DL1 normally returns an apparrently slightly better velocity trace (in one of my client vehicles we have four wheelspeeds as well) than the Driftbox. The positional result as determined from an overlay of laps or complete files actually favours the Driftbox, but these smack of a smoothness that I expect is a result of software manipulation, rather than inherent system accuracy.
The additional display and current "performance" analyses and measures of the Driftbox are almost useless for a circuit racing application in a racing car. They are road car features. The drift measure is definetely lacking sufficient accuracy for handling analysis and is sometimes misleading as a result. Even if I was racing a road car, I would prefer the DL1 because the software offers more racing style features.
The greatest advantage of the DL1 is that it is truely a data logger. Even if you are racing a daily driver road car, you can't be very serious about analysis if you don't organise RPM, steering and throttle data recording. Appologies if this statement causes offence.
How much does it cost you in tyres, brake pads and fuel... let alone entry fees...
Nevertheless, I must say it is better (and more serious) to properly understand a small amount of information than be casual with multiple channels.

It may eventually cost a little more with R-T kit than some other brands to achieve the same capability 'on paper', but the ability of the DL1 to interface with R-T dash/displays/video and serial/CAN from other automotive systems offers excellent value.
That said, if I did a few sprints/time-trials in my road car each year, I might narrowly prefer a Driftbox.
If this adjudicator was actually me, rather than a hypothetical petrol head, the DL1 would win easily, because as well as using it as a vehicle logger, I can use it as a realtime monitor/interface and logger with laboratory/workshop measurement equipment.

I like your thinking, steha, and envy what I presume is a solid programming background.
R-T use a simple X,Y co-ordinate position discriptor for the user's benefit in "Analysis" software. I'm not sure what unit is encoded in their native log file.

Time is the enemy and I'm getting hungry...

[faraday]

Hunger fixed, but time is still the enemy and I'm getting impatient.

I have waiting for over an hour for PPP to run on a log of a dozen laps at a short circuit. I suppose I'm meant to be happy that it at least started.
I have my suspicions that it's not calculating normally, because the progress window is repeating three events, ad infinitum perhaps:

Uncompress file
Data already cached
Store SP3/Clock data

My CPU usage is ~50%, almost all to Analysis. Buggered if I know what the rest means.
I feel I might be at peril of using up all my Internet data download allowance, so I think I will terminate the application, albeit at the peril of missing out on the Holy Grail.

I think I'll reduce the file to two laps and see what happens.............

[faraday]

Reducing the file size may have improved matters, but does not appear to have started us on a viable solution.
At least there is hope that what started might terminate, notwithstanding that the same three actions keep rolling over in the progress window. Now, however, there is a negative number of seconds displayed that is decreasing in magnitude.
If the process was going to finish at time zero, it would have taken ~24 hours, as the negative time shown started at 87,000 odd

Not many regard me as impatient

[steha]

faraday, thanks for the DL1/driftbox comparison. At the time when I got the driftbox my alternatives weren't actually between the DL1 and driftbox, but rather the driftbox or nothing. It is true as you say that it is aimed very much towards the road car clientel - but we use it for 2 things - in-car lap times, and the GPS-log. Everything else in the display and in terms of functionality is pretty much useless.

Please let us know if you get PPP to work, would be very interesting to hear if it yields good results.

[faraday]

Hope I haven't offended you or other Driftbox users.
The thing is still a useful tool in a racing car and not bad value for money.
From the commercial view, we should compare VBox Mini with the added input module to a DL1 and a Dash1. Then, the value for money of the equivalent Racelogic product really suffers.

I think we are talking about a competitive and dynamic market. The consumer is increasingly spoilt for choice.
If I was currently driving, I'd appreciate a current and easily recalled laptime display/recorder too.

[Grant]

I too use a Racelogic product (PerformanceBox), and I'm very happy with it. Its the best portable system I've used.

However, I may be getting a car that requires a more comprehensive gauge package, and was looking at the DL1 and DASH2 products. One thing I'd really like is to get rid of GPS drift so I can really analyze my lines. Racelogic has stationary beacons which they say solve this problem (albiet within a limited range). As faraday says the other Racelogic products aren't as cost-effective as their PBox/DBox line.

The DL1/2's method of dealing with drift (post-processing) sounds very clever. What sort of accuracy can I expect with it?

[grendel]

I'm doing a test which requires some absolute position accuracy.
The water trough is ~4 x 15 m and I need to log lateral acceleration vs speed in it.
>99% of the time position is totally OK. Within some 10 cm and the results are perfect.
<1% of the tests the path is off the line (by meters) and I need to correct manually.
This all over several years and I find the accuracy unbelievably good.