BJM

iTrader: (1)

I was not sure what forum to put this in but here goes.

A few years ago I bought one of the original G-Techs. Although it seemed to work reasonably well, I didn't like the black box aspects. I hated not knowing what fudges it applied to calculate power. For instance, it has to guess what your car's aero drag is. It also seems to guess at what rotating inertia the engine has. How does it ignore the big surge during a shift with an automatic.

So I broke into it and tapped off of the accelerometer and used a cheap data logger to record the output directly. After plotting it in Excel I can do whatever I want with the data. I can do some things the Gtech can't do, like start a run at speed, record aero drag of the car.

The following curves are from my car, a stock 2001 SS Camaro M6 convertible. The only modification is a lid. With me in the car and 3/4 tank of gas it weighed 3806 lbs. Since then I have added sound deadener and a subwoofer so I am currently assuming 3872 lbs. Stock rolling wheel diameter is 25.1", measured by rolling the car several tire rotations and measuring the distance travelled. All the data shown is derived from the single accelerometer, no other data was used.

Figure 1. One thing that is interesting is the huge noise level in the output. This is the same data the Gtech is using to calculate is precanned routines. The newer Gtech is doing some serious curve smoothing to produce the smooth lines I have seen in their advertisements.

Figure 2. Power derived from the curve above. This is a pull in 3rd gear starting from 60 kmh (37 mph).

Figure 3. Torque from the power curve above.

Figure 4. Different run, ease out in 1st, floor it, lots of wheel spin through 1st, shift to 2nd, and then to 3rd. Notches in the curve are my fairly slow shifts, about .7 seconds each.

Figure 5. Power from each gear overlayed. Rotational inertia from spinning up the engine and transmission and wheels, absorbs torque, so 1st gear is lowest. 2nd gear is higher, 3rd gear is highest, in these curves I have subtracted the aero loads. At 62 mph, I measured ~18 hp of drag power.

BJM

iTrader: (1)

Figure 6. Torque derived from the previous figure. 1st gear is wavey due to wheel spin, lost traction about 3 times. Measuring torque accurately requires no wheel slip, the accelerometer can only measure net acceleration.

Figure 7. One interesting thing is the debate about where the optimal shift points are supposed to be. So I worked out overall thrust on the car. Although I have not done so, I could easily record each gear separately and where the curves cross, you have the optimal shift point, no more debate. This method works well because all factors are accounted for. Even if you calculate power incorrectly, measuring vehicle thrust is simply the accelerometer measurement times vehicle mass so it has no approximations it. Again, 1st gear looks crazy due to wheel spin. Since the second gear ratio is so much taller, the thrust on the car is well below what the tires can handle and the car does not slip. Shift points in this case appear as speeds but I could replot it as thrust versus RPM to calculate shift points.

Figure 8. RPM is derived from my velocity calculation which is inherently much smoother than the accel data so the RPM curve is very smooth looking. The calculations also have no understanding of slipping the clutch it looks like the engine idles at 0 RPM at 0 mph. You can also see the progressive ratios of the T56 where the RPM drop is less and less after each gear.

jimmyblue

iTrader: (11)

I had thought about doing this same thing (or buying a
couple of AD accelerometer ICs and cobbling straight up).
Do you have good g/V constants figured out for the
G-Tech's circuitry? I have one too and would like to hook
up raw (or perhaps low pass filtered, averaged) voltage
to my enhanced I/O cable to log acceleration w/ the
data stream, for tuning feedback purposes.

BJM

iTrader: (1)

I actually went as far as designing a full accelerometer and op-amp circuit when I noticed those cheap ebay deals on the old Gtechs. I actually have 2 of them in the car with the second one pointing straight down. This way I can resolve the full acceleration vector in the vertical/horizontal plane. But so far the vertical axis doesn't really help in any way. For the new model Gtech it means it can self level but that is all it can really do.

g/V is pretty easy, point it straight up and then straight down and you have +/- 1 g defined. My data logger dumps out g's now. One thing I found is the Gtech is not thermally compensated very well. Hold it in your hand for 5 minutes and it drifts off a fair bit. I taped it onto the window and try not to touch it at all.

As far as the EIO cable, I originally thought it was the way to go until I enquired about it. It only logs as fast as the frame rate from the car (up to 40 Hz) which is too slow. The cheap data logger I bought does 120 Hz per channel which is about the same as the Gtech's own internal rate (the old model anyway). Go here for the data logger.

P Mack

iTrader: (6)

What kind of precision error does that accelerometer have?

sevanseriesta

iTrader: (34)

WTF can you guys put this stuff in english......This is pretty cool that you guys have the knowledge to do stuff like this. I have never heard anyone doing this to a g-tech

RedHardSupra

iTrader: (1)

wow, this thing is AWESOME!

i tried to do a similar thing, just by gathering data from HPTuners. Just like you noticed though, the noise levels are quite high, and the frequency just isn't high enough.
http://redhardsupra.blogspot.com/200...readsheet.html

also, i have created a shift optimization spreadsheet, which is even better than the eyeballing method of 'where the lines cross.'
http://redhardsupra.blogspot.com/200...ft-points.html

I have the Gtech comp, do you know how to take this sucker apart in the same fashion as yours?

BJM

iTrader: (1)

Ask questions about the non-English parts, I had hoped to be clear.


One comment about the spreadsheet, if you are using dynamometer curves you will not get quite the same shift points as my curves (the ones I haven't posted ) Commercial dynamometers are not actual in-car tests, the load dynos (Mustang, etc.) load the car down to only allow the engine to rev up slowly. The large roller style (dynojet, etc) have no load control but instead use a huge inertia load (the roller) to limit the engine rev rate. Why do they do this? They want to make sure the engine is in near steady state as torque measurements are taken. As you can see from my Figure 6, the lower gears show less torque because the engine is expending energy revving itself up rather than the car. The torque is there, its just that some of it is absorbed before getting out. On an engine dyno the engine revs can be held fixed, completely eliminating the effect. The effect gets worse with the square of the overall gear ratio. Therefore in first gear the engine inertia is 7 times worse than in 4th gear. Plus the aero and tire loads are rising in the higher gears so the engine rev rate is further suppressed.

Keep in mind that many things are happening at once. The T56 efficiency is most likely the highest in 4th gear with no reduction occuring, so in lower gears the transmission efficiency is worse as is the inertia effect.

Overall my plots whould show different (lower) shift points in the lower gears due to inertia. This is completely car type dependant. Add 250 lbs to your car and the best shift points will move up in RPM slightly, more in the lower gears.

I measured the aero drag in my car but its not all that simple. For now I have done a coasting run from 100 to 60 kmh (62-37 mph) and done a parabolic curve fit. In reality, tire drag and areo drag and rear axle drag combined are not exactly parabolic its more like an exponent of 2.2 (not just 2). I am also currently extrapolating up to 100 mph when my drag measurement was done at 63 so I have to remedy that. I also did it roof up windows down, roof down would be quite different and more variable with wind direction.

For anyone wanting to do this too I will post my best curve fit for drag once I make some more measurements. The other problem with an F-Body is the low drag, it makes measurements very small, these cars have quite low drag. My old turbo Buick felt like the brakes were on in comparison.

As far as the resolution of my measurements goes, I have some more improvements to make. The Gtech jerks filed the accelerometer labelling off the chip to keep it a secret but I am pretty sure its an Analog Devices chip. My cheap data logger is 10 bits over +/-10 V but the accel is 0-5 V and in the range I am using it its about 1-4V so I am using about 7-8 bits of the resolution. The data logger has instructions on where to remove a couple of resistors to drop the input down to 0-5V which is its A/D chips actual input range. I figure I can currently resolve .005 g right now according to me holding the thing and tilting it.

I can post some pictures of the interior of the Gtechs showing where I tapped off the accelerometer. My older one uses thru-hole components (chips with pins through the board and soldered) and the newer one is surface mount (wafer like chips with tabs laying on the board and soldered).

The new one probably has 2, 2 channel accel chips in it as I am unaware of any 3 channel chips. This should be a good clue since you should find a chip mounted sideways to the board in there Then you need to find the ground pin and the pin whose voltage changes as you tilt the chip. Use a decent volt meter and you are unlikely to hurt it while poking around.

jimmyblue

iTrader: (11)

What's the value of high time resolution here? I'd think
the rate of change to acceleration ("jerk") would be
pretty low and you wouldn't lose much accuracy with
a smoothed waveform. I'd be inclined to put a 5Hz
corner on it and let the tool sample at 10Hz. I'm really
only looking for "got better, got worse" acceleration
logs, not fundamental accuracy. But sampling a noisy
waveform at 1/12 data rate would probably give you
garbage.

BJM

iTrader: (1)

Finally, someone who knows what jerk motion is

In my own circuit I planned to use a 20-25 Hz filter, so I agree with your suggestion. Since I tapped right off the accelerometer that filter just isn't there. I suspect a digital filter is used internally somewhere.

Just because this is an experiment, I wanted to see how many factors I could account for and see how close to a real dyno run I could get. I think a filter is my next step.