OdinGuru
09-06-2003, 12:29 PM
In my efforts to understand what has been happening with the RX-8 power "loss", I've undertaken an analysis comparing the original Mazda specs, Mazda "published" crank specs, and the real dyno data gathered by other members of the forum.
I began with the original crank torque curve found in some of the original Mazda promotional materials. I took data points from this Nm curve and converted them into HP and ft-lbs. The resulting peak numbers from my sampling this graph (which includes some level of inaccuracy) were 247.1 HP at 8500 RPM and 162.2 ft-lbs at 6000 RPM. Since these numbers are very close to the original published specs for the RX-8 (247 HP @ 8500 RPM & 159 ft-lbs @ 6000 RPM), I decided that this torque curve was indeed correct. The resulting HP from this curve is plotted on the graph below.
Next I plotted the HP results of the best dyno run I've found published on this site so far (also using my graphical data point sampling technique). We clearly see the charateristic sawtooth pattern discontinouties at each of the port opening RPMs that have been seen in every dyno so far. My sampled data has a peak of 189 RWHP.
Now the interesting part comes when I wished to estimate the amount of drivetrain losses. Much debate has gone back and forth on this site about the drivetrain losses being too high (>17%) in comparison to other RWD cars. My understanding of the behaviour of the losses accross different RPMs is limited at best, so someone stop me if I'm completely off base, but I modeled the drivetrain loss as a constant percentage reduction of the HP/Torque at any RPM. Then I ignored the dyno data >6250 RPM and matched the crank data with a constant drivetrain loss to the <6250 RPM dyno results. It turns out that a 14% drivetrain loss matches the two plots almost perfectly <6250 RPM (see graph). Also 14% is a much more typical number when compared to other RWD cars.
So if we take the drivetrain loss of the RX-8 to be 14%, where does that leave us in >6250 RPM range? Well, way down. At 14% loss and from the crank curve, the peak should be about 212 RWHP. Except the dyno curve shows a RWHP peak of only 189. This is down about 11% (equivilent to adding more than 300 lbs to the car). The Mazda stated reduction of the peak HP to 238 represents a 3.6% reduction. A plot of an additional fixed 3.6% reduction accross the RPM shows that this revised number seems resonably accurate for <7250 RPM, but is still an overestamate for the peak HP at 8500 RPM which should be about 205 RWHP. A plot with a constant 11% reduction instead of the Mazda stated 3.6% does indeed match the peak RWHP, but does not accuratly track the rest of the dyno HP plot, and thus stating an 11% loss is also mostly innacurate.
It is clear that the overall shape of the Rensis output curve has been altered from what was in the original crank curve. This should be obvious due to the large discontinouities in the dyno results, and it seems clear to me that the source of these discontinouities is exactly the primary cause in the reduced top end power (as the lower plot seems to agree with the published specs). Mazda has already stated that its reasons for reducing the published peak HP number were due to changes needed to meet emissions in the US. It seems to me that the only way to accomplish this "last minute" would have been to change ecu mappings (different cats etc. would have been infeasable in a short time frame and would have obviously changed the HP spec so they probably would have retested). The ECU change could have been a case of left hand not knowing what the right hand was doing (no visable design change), which would explain the screw up if no one notified the testing department (and if it was last minute).
Also I can't think of any possibly mechanical change that could cause the type of discontinouties we see in the resulting dynos. Other people finger the discontinouties as port opening and closing artifacts, and I think that is only partialy true. If the HP/torque is truely that much lower with the switched over intake track, then the swichover points are WRONG, and should be placed at higher RPMs (where the curves meet without discontinouty). Yet again the points are controled by the ECU.
While normaly an ECU modification would not gain large amounts of HP in a NA engine, that may not be the case here if a de-tuned setup has specificaly been put in place. If this is truely the case, an optimised setup may gain as much as 23 RWHP at the peak, but may not provide any gain <6250 RPM. Looking forward to appling my $500 debit to this kind of mod.
Observations and reactions are most welcome.
Caveat:
If someone else has an understanding of the behaviour of drivetrain loss at different RPMs, please chime in. If there is a good reason that the drivetrain loss is a higher fraction for higher RPMs (significantly non-linear in an increasing manor), then this will have implications for the above analysis. This analysis is based on my very basic understanding of kinetic friction from physics class/textbook. It says that the fictional force between two moving objects does not depend on the relative velocity of those two objects (does depends on coefficient of kinetic friction, normal force, and surface area). Thus a constant fraction applied to the torque at all RPMs should be accurate, also since HP is directly related to torque, the same fraction can be applied directly to the HP.
References:
"Best" Dyno Data -
http://www.rx8forum.com/showthread.php?s=&threadid=9939
I began with the original crank torque curve found in some of the original Mazda promotional materials. I took data points from this Nm curve and converted them into HP and ft-lbs. The resulting peak numbers from my sampling this graph (which includes some level of inaccuracy) were 247.1 HP at 8500 RPM and 162.2 ft-lbs at 6000 RPM. Since these numbers are very close to the original published specs for the RX-8 (247 HP @ 8500 RPM & 159 ft-lbs @ 6000 RPM), I decided that this torque curve was indeed correct. The resulting HP from this curve is plotted on the graph below.
Next I plotted the HP results of the best dyno run I've found published on this site so far (also using my graphical data point sampling technique). We clearly see the charateristic sawtooth pattern discontinouties at each of the port opening RPMs that have been seen in every dyno so far. My sampled data has a peak of 189 RWHP.
Now the interesting part comes when I wished to estimate the amount of drivetrain losses. Much debate has gone back and forth on this site about the drivetrain losses being too high (>17%) in comparison to other RWD cars. My understanding of the behaviour of the losses accross different RPMs is limited at best, so someone stop me if I'm completely off base, but I modeled the drivetrain loss as a constant percentage reduction of the HP/Torque at any RPM. Then I ignored the dyno data >6250 RPM and matched the crank data with a constant drivetrain loss to the <6250 RPM dyno results. It turns out that a 14% drivetrain loss matches the two plots almost perfectly <6250 RPM (see graph). Also 14% is a much more typical number when compared to other RWD cars.
So if we take the drivetrain loss of the RX-8 to be 14%, where does that leave us in >6250 RPM range? Well, way down. At 14% loss and from the crank curve, the peak should be about 212 RWHP. Except the dyno curve shows a RWHP peak of only 189. This is down about 11% (equivilent to adding more than 300 lbs to the car). The Mazda stated reduction of the peak HP to 238 represents a 3.6% reduction. A plot of an additional fixed 3.6% reduction accross the RPM shows that this revised number seems resonably accurate for <7250 RPM, but is still an overestamate for the peak HP at 8500 RPM which should be about 205 RWHP. A plot with a constant 11% reduction instead of the Mazda stated 3.6% does indeed match the peak RWHP, but does not accuratly track the rest of the dyno HP plot, and thus stating an 11% loss is also mostly innacurate.
It is clear that the overall shape of the Rensis output curve has been altered from what was in the original crank curve. This should be obvious due to the large discontinouities in the dyno results, and it seems clear to me that the source of these discontinouities is exactly the primary cause in the reduced top end power (as the lower plot seems to agree with the published specs). Mazda has already stated that its reasons for reducing the published peak HP number were due to changes needed to meet emissions in the US. It seems to me that the only way to accomplish this "last minute" would have been to change ecu mappings (different cats etc. would have been infeasable in a short time frame and would have obviously changed the HP spec so they probably would have retested). The ECU change could have been a case of left hand not knowing what the right hand was doing (no visable design change), which would explain the screw up if no one notified the testing department (and if it was last minute).
Also I can't think of any possibly mechanical change that could cause the type of discontinouties we see in the resulting dynos. Other people finger the discontinouties as port opening and closing artifacts, and I think that is only partialy true. If the HP/torque is truely that much lower with the switched over intake track, then the swichover points are WRONG, and should be placed at higher RPMs (where the curves meet without discontinouty). Yet again the points are controled by the ECU.
While normaly an ECU modification would not gain large amounts of HP in a NA engine, that may not be the case here if a de-tuned setup has specificaly been put in place. If this is truely the case, an optimised setup may gain as much as 23 RWHP at the peak, but may not provide any gain <6250 RPM. Looking forward to appling my $500 debit to this kind of mod.
Observations and reactions are most welcome.
Caveat:
If someone else has an understanding of the behaviour of drivetrain loss at different RPMs, please chime in. If there is a good reason that the drivetrain loss is a higher fraction for higher RPMs (significantly non-linear in an increasing manor), then this will have implications for the above analysis. This analysis is based on my very basic understanding of kinetic friction from physics class/textbook. It says that the fictional force between two moving objects does not depend on the relative velocity of those two objects (does depends on coefficient of kinetic friction, normal force, and surface area). Thus a constant fraction applied to the torque at all RPMs should be accurate, also since HP is directly related to torque, the same fraction can be applied directly to the HP.
References:
"Best" Dyno Data -
http://www.rx8forum.com/showthread.php?s=&threadid=9939