I sat down last weekend and read this entire thread. Like others here, I'm very curious to get some hard data for the P85D so that we can see exactly what it's really doing.
The question, of course, is how to get that data. 65 pages so far, and we still have nothing but the dashboard meter and maybe some of the REST API data. We don't have a single well-done full dyno run. I appreciate those here that are trying to find a suitable dyno unit that has the required features (can measure independent front/rear wheel drive) and the proper size (can handle the P85D's torque). However, this seems unlikely so far.
I'd like to propose an alternative, that while not as accurate, will give us enough data to analyze the two primary questions in this thread:
1. What is the combined peak output power of the P85D's motors? How does it compare to the 85D? Is it near Tesla's claimed power levels?
2. Does the P85D decrease it's power output given different states of charge (the original question at the beginning of the thread) ?
I believe we can gather enough data to reach reasonable educated conclusions to these questions using a good dyno app for the phone. I know this was briefly discussed in this thread and deemed fairly inaccurate, but I believe this method may be more accurate that it was given credit for. I tested my 85D with PerfExpert, a dyno app I purchased, and it appears to me that the results are accurate enough to give meaningful input to the questions above.
The results of my dyno test on an 85D are here:
https://network.perfexpert-app.com/results/IFML9GbQCz I posted this in another thread, but I think it's actually more meaningful here.
As you can see from the dyno graph, this has several expected characteristics of the Model S power and torque curves:
1. Torque is maximum and constant from 0 to about 37 MPH.
2. Power ramps up linearly from the start, then levels off to a near-constant value at about 37 MPH, where torque begins to decrease.
3. The absolute value of the power level at peak (361 kW) has a very high correlation with an independent power measurement from the REST API (which gave 370 kW).
4. While not shown in the dyno run, I also did a 0-60 timed run with the app, and it gave me 4.44 seconds for the 85D, right on the spec.
Given that the app appears to be generating torque/power curves that correlate very well with expectations and other independent measurements, I feel it would suffice for giving us the data for the P85D that could shed light on the questions above.
The app does not use GPS on the phone at all, it only uses the accelerometers. PerfExpert's web site claims that this is because the GPS in phones is not accurate enough to get these measurements, and can only be read from between 1 and 5 times per second (1-5 Hz) depending on the phone, which is not enough. They claim that the accelerometers in phones can be read from 10-200 Hz, depending on the phone.
The PerfExpert run above for my 85D was taken on a Samsung Note 4, which is a quite new phone with good hardware.
To minimize the potential absolute error and run-to-run error, and to minimize bias in individual phones, we should do the following if we want to properly collect this data for analysis using this app:
1. A highly accurate curb weight of the P85D needs to be found using a scale. The curb weight listed in the Model S manual does not give different curb weights caused by differences between battery pack size (60 vs 85), motor size or quantity (85 vs 85D vs P85D), or options with additional weight (air suspension, dual chargers, ultra high fi sound, power liftgate). Each P85D that is tested needs this accurate curb weight.
2. Proper calibration procedure for the accelerometers must be followed specifically as outlined in the PerfExpert manual.
3. A variety of different phones should be used to compare results across different accelerometer hardware.
4. A consistent set of parameters needs to be entered by each tester into PerfExpert (except curb weight).
5. Multiple tests with different states of charge should be tested to be able to answer the 2nd question.
6. Other dyno apps should be tested as well to see if they are consistent, and evaluate which one might be the most accurate.
7. It is known that the dyno apps have some inaccuracies that cannot be compensated for. The accelerometers in the phones are less accurate than a chassis dynamometer. Some phones may not have enough processing power to accurately collect the data on the test runs. The dyno app is actually back-computing all figures making some assumptions, including drive train losses, that may not be accurate, especially for the Tesla. Nevertheless, the comparison of results between different runs and different cars will enable us to deduce reasonable answers to the questions even if the absolute values of the parameters are off.
Please note: I an not affiliated with PerfExpert in any way, other than I bought it to test my 85D. I won't post the link here, just search the Google Play store (Android) or Apple app store (iPhone) and you'll find it. Read their FAQ on their web site, these guys really seem to know what they're doing and appear to have taken great pains to make the app as accurate as possible.
A poster above had it quite correct -- 65 pages and we're still discussing things in a complete vacuum of data. At this point, ANY reasonably accurate data is better than this, even if it's not 100% accurate.