924Board.org

ptheskil · Joined: 03 Aug 2004 Posts: 128 Location: Essex, UK

Spent a couple of hours with my buddy and his iphone yesterday. He's downloaded an app that exports all of the iphone internal accelerometer/GPS/gyro data once every 0.05sec into a data file. With a couple of hours mucking about in a spreadsheet you can calculate instantaneous power from the acceleration data and vehicle mass and use this to derive torque. You can work out the road load from coast-down data and correct your measurements to give gross output at the wheels. As you know the vehicle speed you can calculate rpm and there you are - enough data to plot a power/torque curve.

If I got this right then there should be an image attached somewhere for my '81 931ROW. If you assume an 85% transmission efficiency (is that right? I can't remember where I got that figure from but it is in my head for some reason) that puts me at 147kW (200PS) and 241Nm (178lbs.ft). Pretty damn happy with that.
[img]https://picasaweb.google.com/118033499298158047258/931#5604459289362978002[/img]
_________________
1981 931 series2 Euro spec

morghen · Joined: 21 Jan 2005 Posts: 9095 Location: Romania

IT IS WORTHLESS CRAP !

i got 260 or 270 hp on that thing in second gear with good traction.(and yes i've carefully calibrated it and provided accurate info about the car/driver/etc..)
in first i got nothing...in third i got 160 or something like this.

that simply cannot be

_________________
Supercharger and EFI kits
https://www.the924.com

ptheskil · Joined: 03 Aug 2004 Posts: 128 Location: Essex, UK

I'm sorry if your experience of performance measurement apps/gadgets is "worthless crap" but I can tell you now that if that is the case you either:
1. don't know how to calculate performance data correctly from 3-axis accelerometer and GPS data or
2. used a worthless crap app/gadget.

My analysis uses raw 3-axis accelerometer and GPS data from the iphone and the calculations are done by two professional engineers with over 20 years experience in powertrain component development and engine performance measurement. We are co-owners of a joint patent for an advanced boosing concept which is being evaluated by several European and Japanese automotive OEMs as well as several off-highway vehicle OEMs. This establishes our engineering credentials.

The data from the iphone we used agrees within 1% with the accepted value of "g". The data over the 30sec run was assessed for error by comparing calculated velocity at the end of the run with that at the beginning (start and end of run from stationary) and also double-checked against GPS data which is an independently measured parameter. The biggest unknown and potential source of error in the results is the vehicle mass which we did not measure. The vehicle is 100% stock and so published curb weights were used in the mass calculation. Overall I would not expect more than about 5% error in the reported results.

There is a big difference between the integrity of numbers spewed out by an app/gadget of unknown quality designed and marketed to amateur tuners for a profit, and a robust calculation by qualified engineers using 3-axis accelerometer and GPS data under carefully controlled conditions.

If anyone is interested I would be happy to run through the process we used but if the results are going to be dismissed as "worthless crap" on the basis of a disappointing previous experience then I won't bother.

Is anyone interested in a sensible discussion/debate about this?
_________________
1981 931 series2 Euro spec

the_mad_electrician · Joined: 16 Nov 2009 Posts: 1073 Location: Central Georgia

Hp and torque numbers aren't the ones that matter anyway. The time it takes to get around the track is the important number.
_________________
81 924 N/A

2004 Ranger "Edge"

2005 Mazda 6

Harm · Joined: 02 Apr 2009 Posts: 1376 Location: Holland

morghen · Joined: 21 Jan 2005 Posts: 9095 Location: Romania

dont get me wrong...i'm not here to rain on your software/calculations or your credentials.
i'm no bum when it comes to education....no biggie...but i know a thing or two.

my experience is indeed with a downloaded application for the iphone..that also uses the GPS and the accelerometer in the iphone.

i said it is crap because i've tested it and it produced STUPID results.
lets have a talk about errors and influential factors:

- g meter accuracy (for what the phone is used a cheap accelerometer can be used and i doubth that apple used the most expensive unit they could find)

- g meter calibration are you sure that ALL accelerometers are calibrated the same? do you have a quality spread sheet from apple? do they all produce the same set of numbers when submited to the same conditions?

- your calculation, yea...no dissrespect to you or your engineers...but maybe the method you use to calculate is not quite the best..i'm not saying it is not...but this is always factor.

-grip...ok...this might be a stretch but the engine makes power...power has nothing to do with grip...yet you have grip as a factor...if you spin the wheels a bit...the power is there but the g is not.

i think i can find some more factors that can influence your estimation/calculation.

the simple thing is that you are trying to calculate something that others measure...so even if the factors have minimum influence on the results...you still cant compare them with the results the others got from a dyno...because their results are altered by other factors.
yea..they can be in the same ballpark...but then again...they might not be

and as long as you dont measure the amount that each factor has on the result...you cant calculate the result....you can ESTIMATE it.
_________________
Supercharger and EFI kits
https://www.the924.com

ptheskil · Joined: 03 Aug 2004 Posts: 128 Location: Essex, UK

@the_mad_electrician, @Harm,

I agree that the track is where it counts and a dyno will usually (but not always) give you a more accurate result. One day I'll do a dyno run and compare results. In the meantime I'll use the results I've got to give me a warm feeling that things are generally OK and in the right ballpark. I'm certainly not claiming any great accuracy or trying to make some sort of my_car's_more_powerful_than_yours statement. Actually, the shape of the torque curve is not right so I suspect that the ignition timing is not dropping to the boosted advance curve as it should. If the crud in the WUR is anything to go by I suspect that the pressure sensor in the DITC is not reading right. I can get 100RON fuel locally (for a price) so I'll take a chance on not having a detonation problem. Time will tell.

@morghen,

You make some very good points. I can't claim that I measured anything with great accuracy and I agree that a dyno would usually give more accurate results. A vehicle dyno is still potentially subject to grip/tyre issues though but I would expect it to be less significant and more consistent between runs. I didn't have a dyno to hand last Saturday so this is all I've got right now. That said, I don't think my numbers are too far wrong. I agree that the accuracy of the iphone data is unknown so that is potentially a big problem. You can adjust for any zero offset errors (in this case it was 2.5%) but you have to use the values it gives you for acceleration. I don't know how accurate Apple's sensors are but the z-axis data is within 1% of "g". I have no idea how linear the response is or if the y-axis is as accurate as the z-axis so it's only an untested assumption that they are. I'm pretty happy with the method and calculations though

Of course I would be!

I only disagree with you about the difference between calculating and estimating values. Every parameter ever measured or reported has been calculated from something. Even a vehicle dyno calculates an estimate for torque and power from rotational speed, roller reaction torque (usually measured using a load cell) and some system losses which are also calculated. Granted, a decent dyno set up will be regularly calibrated so you would be right to have greater confidence in dyno figures. But a dyno still only reports power and torque based on calculations. The only difference between that and my recent experiment is that I use different parameters to calculate the result. Every dyno you ever saw calculates an estimate of the values.

I still maintain that my method is very valuable. I can repeat the measurement every month, week, day if I want to and more importantly before and after any modifications I make. Even if the absolute numbers are wrong the differences are significant. It will take me one minute to collect the data, 15 minutes to analyse it and will cost me a couple of spoons of petrol. I think that's pretty useful.

If you have an app (or something else) that can measure raw acceleration data please send it to me and I would be very happy to run the numbers through my spreadsheet and see what comes out. It would be good to have a different data set to try the calcs out on and see how the results compare. If you got stupid results before then there is definitely something wrong with the tool you used on that occasion. Either its sensors are very inaccurate or its calculations are wrong. There's nothing wrong with the physics.

By the way, do you know how much loss there is in a typical transaxle? I still can't find any data on that - it seems to be hard to come by for some reason. There will be a hundred factors which influence it (load, oil viscosity, gear profile, temperature etc.) but there must be a rule of thumb somewhere. Any ideas?
_________________
1981 931 series2 Euro spec

ptheskil · Joined: 03 Aug 2004 Posts: 128 Location: Essex, UK

Update: According to Wikipedia (please forgive me!!) Apple uses the LIS302DL accelerometer from ST in the iphone. Here's a link to the datasheet: http://www.st.com/internet/com/TECHNICAL_RESOURCES/TECHNICAL_LITERATURE/DATASHEET/CD00135460.pdf

States typical offset is +/-4% (I measured 2.5%) and all 3 axes are +/-2.2% accuracy (at approx 3 sigma based on me eyeballing the histograms). It's a pretty good sensor.

I'm happy with that.
_________________
1981 931 series2 Euro spec

Khal · Posted: Tue May 10, 2011 6:34 am Post subject:

I have to agree with morghen (how'bout that, eh?).

Please, at least tell me you weren't holding your hypePhone in your hand as you made the runs?

Either way, I cannot see how this could give you a anything other than a rough guesstimate of horsepower/torque (I mean, for example, what happens to the data when you hit a decent bump?). As an aside, I work with survey-grade GPS and, trust me, the hypePhone has FAR from survey-grade GPS. I wouldn't trust it's accuracy for any kind of high-precision measurement.

Still, better than nothing, I s'pose. And good fun.

And, of course, if it's giving you numbers more than 10% higher than the factory quoted figures, that gives you good braggin' material down the pub! So, by all means enjoy it!

Just don't be disappointed if you get it properly tested and find out it's closer to 150hp than 200hp...

ptheskil · Joined: 03 Aug 2004 Posts: 128 Location: Essex, UK

Looks like I'm on my own here.

Khal · Posted: Tue May 10, 2011 7:43 am Post subject:

morghen · Joined: 21 Jan 2005 Posts: 9095 Location: Romania

I think that you know exactly what you have achieved with this and dont need me to argue with you what it is

I understand the joy of working something like this out...i get that too when i'm on interesting projects.

I'll get the name of the app i used..just for fun..to see how that does on your setup.

just to compare setups, the software that i used requires that the phone is fixed in the car with one axis aligned with the car's direction of travel...of course any degree of error..induced other errors

we used a sticky pad and went for it.

apart from the points i made above..i think we had the following problems:
1. car starts kinda' sideways when i do a hot start from standing...so that kinda' upsets the trick.
2. because of OEM(read Old Embarrassing Material) suspension and quite some torque...the car lifts the nose and puts down the tail quite a lot on the first three gear changes.(and wobbles back and forth just after)
_________________
Supercharger and EFI kits
https://www.the924.com

flosho · Posted: Tue May 10, 2011 9:20 am Post subject:

I didn't read this shitstorm of a thread, but my uneducated opinion would be to replicate the iphone app test about a dozen times and then average them and then dyno the car 3-5 pulls and average that.

Would that not be THE most accurate method of testing the app?
_________________
[This Space For Rent]

Khal · Posted: Tue May 10, 2011 9:39 am Post subject:

fiat22turbo · Joined: 18 Jan 2006 Posts: 4040 Location: Portland, OR

I view this like I do a chassis dyno. It produces a number that you can use to base your changes on. It isn't a number that will mean much to anyone else, but as long as the number is repeatable, then don't worry about it too much.
_________________
Stefan
1979 924 Carrera GTS (clone-ish)
1988 944 Turbo S (Silver Rose)