I have good news and bad news. First the bad news:
Quote:
Originally Posted by bwilson4web
Ah, I see what you're doing.
When I get home tonight, I'll find the original data . . .
|
This data comes from March 2007 but it also corresponded to when my computer was running out of disk space. It looks like I probably deleted that data when in the spring and summer I ran a series of gasoline tests.
The good news is I have two more graphs from the same set of data from high-speed hill climb and a descent of the same hill in "B".
HILL CLIMB AT 80 MPH
Now there are two points where the traction battery load briefly goes to zero. These are the interesting points when all power is just the engine with power split between the mechanical and electrical paths. To improve accuracy, I've used PowerPoint to expand and read out the values:
Reading out the values:
- Seconds 1034.5, 12250 / (-1 * -12500) ~=98%
- Seconds 1037.3, 12750 / (-1 * -13000) ~= 98%
Notice we can see a distinct rise in transaxle temperature from the mechanical and resistive losses of MG1 and MG2. So we're looking at:
- 2% * 12,500 W ~= 250 watts to heat up the transaxle
This is just the electrical losses of power flowing between MG1 and MG2. Losses in the inverter take a coolant loop that passes through the transaxle case. This loss does not include the mechanical path losses.
DESCENDING HILL IN "B"
Here is the descent:
The problem here is the engine is being driven by the energy captured from MG2 and we do not have a way to measure how much was sent to MG1 for the counter torque needed to drive the engine. However, we know it is fairly small since MG1 is about 300W and the transaxle temperature did not increase. It is also dealing with only 10,000 W of power, 1/5th the amount on the hill climb.
CONCLUSIONS
The only thing is 98% seems too good to be true but it is only the electrical path, the part that is independent of the mechanical path. The subsequent mechanical losses through the power-split-device, chain drive, reduction gear and differential are not measured. The earlier Knoxville reports included these mechanical losses. But there is additional room for error due in part to the limits of the Graham miniscanner.
The Graham miniscanner uses a fixed rate to read the data with a little over one second per cycle. This means calculated values were not sampled at exactly the same time. To address this problem I used a linear approximation between points proportional to the data element time-stamps. Better than nothing, it is not exact. Also, we are assuming the metrics returned from the Prius ECUs are close enough without going through a calibration cycle.
Bob Wilson