Data Logging for bhp calculation

Hi, I'm new here. I've have just started to use Rachchrono with a QStarz 818x 5Hz with a view to calculating approx. bhp from runs as well as the 0-60, 1/4 mile etc. I have done a few runs and have noticed a few things that I don't understand. Any help would be great, thanks in advance.

Firstly the speed data and distance data don't seem to match. There is up to a 10% discrepency on the data I have! For example:-

On a run of 11.6s to do a 0-60 my average speed is 14.57m/s
Thus, in 11.6s I have covered approx. 169m
According to the distance data I have covered 181m and so my average speed is 15.65m/s.

Secondly, why does the distance data not use the altitude in its calculation? I have been looking at the data to understand how it's calculated and it correllates with pythagoras of x and y position. Is the altitude data not accurate enough or is it assumed that you find some flat road?

Finally, to calculate power I require acceleration, kindly provided in the .csv download. However, this requires a lot of massaging to remove some big values and get it to a stage where I am confident that it's realistic. My car in 2nd can pull 0.5g (on a GTech) so I've set this as a maximum figure. What sort of averaging should I use on the acceleration data to get the best results? I've started to search but as ever it takes a while to sift through the pages and get to a result that is relevant.


  • aolaol
    edited January 2009
    I'm not sure what is wrong with the average speed calculation you're seeing. But RaceChrono uses GPS speed and calculates the distance by it self. RaceChrono does not take in account of altitude changes as GPS altitude information is not very accurate, especially in race track conditions. The distance calculation should be pretty accurate on a level surface, within 2% of the real distance. Also there is some zig zag effect in GPS plot which cause the distance to be higher than in reality. On the other hand, GPS calculates the speed from doppler effect which probably takes altitude change in account...

    The acceleration values are there mainly for video use, and are estimates calculated from difference in speed and bearing. In my opinion, to calculate power it would be better to have actual G force data from real accelerometers (not pseudo data calculated from GPS data). Unfortunately RaceChrono does not support accelerometers yet.

    I have had real accelerometer support in the works really long, but it seems I just have too many things to do :(
  • edited January 2009
    Cheers aol. What's wrong is that the distance that is quoted and the distance that you can calculate from the speed are different and by some margin. Would zig zag effects make an error of 10%? Is the GPS speed data the measure variable and is thus the one to use? How does Racechrono calculate distance for 60ft times etc? Does it use speed data or distance data?

    In terms of power caluclation it's more of an exercise than for any proper purpose. I would have hoped that the data was accurate enough to enable calculation of average acceleration every second for example and then use an interpolation of that data for the guestimate of power. The actual acceleration data in the .csv has such a large amount of noise at 5Hz that it's not suitable for the task at the moment.

    Fundamentally accelration is the change in speed with time so calculating it is not an issue, the error of the GPS position is the issue so that the transition of the speed data is not as smooth as you'd like to calculate acceleration at the same rate. Over an ageraged period of time you should be able to get data that has a low enough error to give some results that can be used.

    I'll have to do a few more runs and see how I get on...
  • Right, did some more runs yesterday in a mate's Audi S4 and made sure that the GPS signal was spot on. The results are really good. We did 7 second gear runs from around 20 mph to red line and of those 1 was lost due to GPS signal and 1 had a glitch. Of the remaining 5 the spread of times was 2.88s to 2.96s for a 30mph to 60mph run.with the average at 2.90s. 4 of the runs were within 0.02s of each other. Great!

    From these runs I was then able to calculate power and air resistance to give approx. power at the wheels. To improve accuracy I averaged the power over a second for each point. A better method will be to manually smooth the acceleration data to remove noise, although on these runs it was a lot better to begin with.

    The final stage was to take the calculated power at the wheels and apply a losses equation taken from rolling road data for a quattro car. The data points I had were to convert power ATW to fly power, 1.15 at 3k and 1.35 at 7k, there. This yielded great results and a power curve that was a good approximation of what had been measured on the dyno. I think my weight was a little low for the car but the figures compared to a recent dyno like this:

    Power at the wheels, GPS 252bhp, Dyno 257bhp.

    Max power, GPS 310bhp, Dyno 318bhp.

    Max torque GPS 330ftlbs, Dyno 350ftlbs.

    The possible weight discrepancy would explain the majority of the error IMO. Will check the weight again online today.

    So overall a good day, got to do the same for my Audi S2 and BMW now...
  • aolaol
    edited January 2009
    Wow, that sounds very good accuracy! Can you send me the NMEA (from \RaceChrono\Sessions\*-nmea.txt on your memory card) of these runs to me? Also the calculations you have done, if you wouldn't mind giving them to me. I could use your research for improving the long neglected Performance Testing section a bit with your data... You can email the stuff to tracks(at) .
  • edited January 2009
    The reason your calculated 169m is different to the GPS data of 118m is that your average speed probably isn't 14.57m/s. Unless you've worked out the average speed by totting up the instantaneous speeds at each data point and averaging them, I suspect you've assumed linear acceleration, which the car will certainly not give you. The car will vary its acceleration at certain points due to grip, power available and changing gear.

    As for estimating power, I suspect there's too many possible sources of error to get flywheel power from this method. Power at the wheels will be more accurate. As you've determined, weight is an important factor and sadly, you can't get it from the Internet. It's rare that a car will weigh close to the stated weight. Manufacturers don't vary the stated weight of a car based on the options it comes with. Normally, just the different engine sizes are catered for. Factory fitted options can weigh quite a lot, especially things like air conditioning. You need to get the car on some scales or a weighbridge with the driver onboard. Head down to your local recycling centre as they all have a weighbridge and ask the guys nicely!

  • aolaol
    edited January 2009
    Nope, I calculate the distance from the coordinates. Sum of length each vector between each plot. Only way to calculate average speed is the traveled distance divided by time it took, getting speed of each plot will give you just rough estimate... True about the car weights.
  • Hi both, thanks for the replies. The S4 runs were great and I've got some data from my Audi S2 also.

    The avearge speed was the average of the speed data points (every 0.2s) over the total time so there's no problems there. The distance data was the finish distance - start distance. The avearage speed was calculated from total distance / total time.

    The differences are an area of concern although the speeds measured tie in with my expected results and some reecnt 0-60 tests gave results around the 5.5s mark and this ties in with GTech data.

    I will look at some of the improved data and see what the distance vs. speed data errors are (using the methods described above).

    For power the calculations were:-

    (1) P=FV
    (2) F=ma

    Substiture (2) into (1) for F

    (3) P=mav

    The power at the wheel calculation is:-

    (4) Air Restance = 1/2.rho.Cd.A.v^3


    P = Power (W)
    m = Mass (kg)
    a = Acceleration (m/s^2)
    v = Velocity (m/s)
    Air resistance (W)
    rho = Air ensity (kg/m^3)
    Cd = Drag coefficient
    A = frontal area (m^2)

    Conversion kW to bhp = 1.341 kW to bhp.

    The data for both cars is widely available and I also have RR plots and run-down losses to have a look at, the power curve was just for fun but was the right shape and within about !0% from the data provided.

    The acceleration data provided is very noisy and requires some massaging to get smooth results, of course this can dilute some of the data also and so the error on the power at the fly is large.

    As for power at the wheels, that's pretty simple to work into the RaceChrono package with the input of car weight and making the following assumptions. Most modern cars have a Cd around the 0.3 mark and A of around 2m^2 and an assumption of rho = 1.2 isn't far out, this is a small amount in second gear compared to the 220+bhp ATW calculated and measured on the RR. The acceleration data would have to be improved using some data checking to smooth the data using previous results.

    I will happily send you some data. I'll make sure it's all OK ;-) and get a copy off to you.

    Can I post pictures on here? Have got some nice Ecel charts of the data described.
  • Another method for power measurement could be to run the car up to the rev limiter in second, select neutral and allow the car to coast down thus measuring in real time air resistance, rolling resistance and transmission losses at the smae time, will see how it goes when I get time.
  • You can posts pictures with BBCode, or rather links to pictures in internet.
  • edited January 2009
    I've created a private album for this at

    pw is race_chrono
  • edited January 2009
    Email sent.

    I've just looked at the .csv file and the discrepency Inoticed between avearge speed x time vs. distance quoted has all but gone. I'll check this on future runs but it seems that the original runs I did had poor GPS. Live and learn! Still, for consistency I'll use the speed data if I need to calculate distance. Great!

    Got some number crunching to do on my S2 data and see if I can get some 35 to 65mph runs (turbo lag in 2nd means that 30mph is too low). I'll also see what sort of power curve I can create.

    The only thing outstanding at the moment is the acceleration data noise, if I get a satisfctory result I'll post up, likewise if anyone has any ideas...
  • I've had a few goes at this and it's getting a lot better. The key is to get the acceleration data smoothed out. To do this (on a 5Hz receiver) I make each point the average of the previous, current and next points (3 total). This smooths out the major peaks and troughs and can be used to give reasonable results. To further improve the data I've been plotting the acceleration graph in excel and then manually alter any points that are way off.

    You need to analyse the graph of the speed data to ensure that it is smooth, I have had runs where the GPS has suddenly shifted or a satelite has dropped out and this means that a section of hte data cannot be used.

    The no.1 rule is to apply common sense and if something looks wrong or different to the majority of other runs then it more than likely is.

    There seems to be a cyclic error on the acceleration data of about a 1s period. The magnitude of this is reduced with the averaging but a manual adjustment is still required to remove the really silly points.

    For losses I have found that the best method is to assume a straight line distribution for the Patw to Pfly figure. I have taken the value for max RPM of 1.35 at 7000rpm and assume that losses at 0RPM are 0!

    This gives:-

    P(atw to fly) = 1+(RPM/7000*0.35)

    TO calculate RPM I have maxed the car in that gear on the limiter (7400RPM) and then traken the peak speed from that run. Another approximation would be to hold say 3000RPM and use that. From this measurement I can then get the RPM/mph data (or other speed measurement/ power measurement) and then calculate RPM from the speed output from RC.

    To finish, I have compared the caluclated power at the flywheel and RR data for my S2 and it's getting close, it's certainly the right shape, it's just inertia effects and the change in Cd with speed that give errors now...
  • aolaol
    edited February 2009
    That is quite impressive. I'm quite surprised with the fact that the results seem so convincing with the mere 80 euro receiver! I wonder if the variation stays as that small with more runs?
  • edited February 2009
    I've done quite a few runs now and they're all within 10%, I have used rolling road data to get the losses and the car accelerates faster in second so there are also inertia effects to take into account, c10bhp.

    As it's a turbo car there is also turbo lag and ECU load to take into account when comparing gears. This is most visible with the sub 5k differences in the power curve.

    I think the best method of comparison for my car is a 35-65 test in second (3500 to 7000rpm), the power curve generation is for fun more than anything else but consistent tests should show differences due to modifications.

    I've also done 5 0-60 tests and got some interesting results. The launch is the hardest part to replicate, with the gearchange a close second. The results vary by up to 1s but comparing the 40 to 60 times they were all within 0.2s of each other...
  • Best is (as on a dyno) a gear with 1:1 ratio (4th usually on a 5spd) to reduce transmission losses so they fall out of the equation.
Sign In or Register to comment.