Joined Feb 2009
3K Posts | 2+
siebenbürgen süße heimat!
so is ten kate willing to spend more on engines than bmw? they could easily find the money to throw 30 engines at their campaign if thats what it takes to get that extra 1000rpm
The engines get rebuilt b/c they tune the top end as tight as they can get it to increase volumetric efficiency. Over time the valvetrain accuracy begins to wane, and the likelihood of engine failure rises and the power output declines slightly. Thus, companies like Aprilia try to get gear cams. Gears flex less than chains, which makes the valvetrain more accurate. Aprilia increase lift and duration. Volumetric efficiency increases. The engine makes 3-5hp more throughout the entire rev range. Competitors cry foul.
The top end clearance and accuracy determine reliability. Moto2 engines go three races b/c the top end is detuned with conservative lift, duration, and compression. An aggressively tuned Honda WSS barely makes it through a single race weekend. The rev ceiling is roughly the same for both bikes. If revs isn't the primary determinant of reliability, neither is average piston velocity.
really surprised that apparently the cbr1000 revs higher than the larger bore zx10r.a twin reving to 12k ? awesome
great work geo, thanks for posting
I'm starting to buy that maybe there is something preventing the revs from going through the roof, i can not get my head around the fact that bmw don't have at least an additional 500rpm over their competition
I don't know why I didn't think of this before. F1 teams have been using acoustics to measure rev ceilings since the 1980s. Rumor has it that the FOM used acoustics to show rev counters for onscreen graphics when the teams refused to provide the data.
Geo's data isn't entirely conclusive b/c the Rea number is very strange, but the shift points of the BMW and Aprilia do appear to be quite low. The BMW can rev above 15,000rpm, and the Aprilia isn't far behind. Laverty and Haslam shift at roughly the same points as Sykes. The only reason I wonder if the revs are a bit higher than actual is b/c the Suzuki would be dead to rights at those rpm levels. Yes, Zook are bad, but they aren't getting lapped by satellite BMWs.
Thanks for the data, Geo. If you feel inspired to do further analysis on 2012 onboard footage, I'd be grateful, but no rush.
Joined May 2008
3K Posts | 79+
Southern Aridzona - Hell's doorstep.
Yea, I've got the bug for this...
The Rea sample was one of the questionable ones. There was a brief 1/2 seconds of silence between the babbling fools, the signal level was low, and it might represent an overrun or backshift situation.
The Rea sample was one of the questionable ones. There was a brief 1/2 seconds of silence between the babbling fools, the signal level was low, and it might represent an overrun or backshift situation.
Joined Jul 2007
5K Posts | 1K+
unda cheese
I would love to, but my personal computer bit the dust last year. All of my sound editing and video editing software is gone. If you know of any non-intrusive freeware for Mac, I'd be happy to take a look.
Edit: Found a freeware program, but it doesn't count the samples so my measurements probably aren't as accurate. However, using the highest accuracy settings I could manage, I calculated 14,755. We are in the same general wheelhouse.
Joined May 2008
3K Posts | 79+
Southern Aridzona - Hell's doorstep.
http://sourceforge.n...audacity/2.0.0/
Load file, give it a listen.
Highlight a short section of the waveform, play it. Repeat until you reach a good section of max-revs. You'll only need 1/4 second of clean sound.
With the short section of interest highlighted, Analyze > Plot Spectrum
Axis - Log Freq
Size - 4096 or higher (something bigger than default.)
You'll see a nifty little plot.
13483:Audacity_Spectrum.jpg]
As you move your cursor, the software will jump to the nearest peak and display that frequency. Very handy. Depending on the firing order, you'll see a number of harmonic peaks. Pick the one that makes most sense. In the case of the Aprilia, the first peak, in the ~100~125Hz range is the easiest to use. Just multiply that value by 120 (other bikes, with differing firing orders, harmonics, and number of cylinders will differ) and you should have RPM.
OR
Zoom in on the waveform and count the seconds between complete 1-rev cycles. The reciprocal gives HZ, and go from there.
Enjoy
Attachments
Joined May 2008
3K Posts | 79+
Southern Aridzona - Hell's doorstep.
Cool. Glad to know I'm not going insane.
Given what Krop said about the WSBK rev limits, it seems Aprilia's CRT effort is perhaps something special.
(Assuming the source has not been time-compressed, etc.)
Now, go check out the Rossi onboard video on Youtube at the MotoGP channel.
Aren't those measured RPM too low?
16,7xx for the Duc and 16,2xx for Honda?? What am I missing?
16,7xx for the Duc and 16,2xx for Honda?? What am I missing?
Joined Jul 2007
5K Posts | 1K+
unda cheese
Cool. Glad to know I'm not going insane.
Given what Krop said about the WSBK rev limits, it seems Aprilia's CRT effort is perhaps something special.
(Assuming the source has not been time-compressed, etc.)
Now, go check out the Rossi onboard video on Youtube at the MotoGP channel.
Using the second method with the plot analyzer, I selected 7.2 seconds to 7.4 seconds at size 8192, and got a harmonic peak a 123hz. According to your rpm spreadsheet, 123hz yields 14,760, which corroborates the time calculated method I used the first time (spectrum plot is much nicer), and the basic mean piston velocity calculations given the bikes stroke. Everything appears to be checking out okay, assuming we are counting the peaks properly or using the proper log function analysis.
Using size = 4096 with a smaller sample slightly closer to the first shift from 7.26-7.41, frequency analysis yields 124hz or 14,880rpm, just a few percentage points from its theoretical capabilities.
I guess everything is checking out okay. What exactly does the log function frequency analysis tell us?
Analyzed the Rossi clip using the log function frequency analysis. I found a reading of 146hz (17,520rpm) between 16.6 seconds and 16.8 seconds. Everything looks good, again, assuming the underlying math is correct and we are using the tools properly.
Joined May 2008
3K Posts | 79+
Southern Aridzona - Hell's doorstep.
Low-mid 16s are the most I've seen from the 1000cc Honda.
The few brief clips I found from the race might suggest that Rossi's Duc was dialed back somewhat. At the time when the samples were taken, he was all alone on the track with little hope of improving. He may well have decided to drop the thing into put-put mode.
Joined May 2008
3K Posts | 79+
Southern Aridzona - Hell's doorstep.
Lex, the log axis setting is just a different way of plotting the spectrum. It works better for the relatively low frequencies involved. The function as a whole looks for repeating signals and plots the relative strength of each vs. the signal's frequency. It's a faster way of calculating instantaneous RPM, but for each engine you need to verify that you're looking at the right frequency range. (By measuring peak-to-peak times on the raw waveform.)
Good news. I've been testing the Audicity software against a control, and it looks pretty good.
http://www.youtube.c...h?v=Yw2CMzqTXHw
Checked the Honda at 31.6 seconds and 37.3 seconds. Data read 151hz and 154hz respectively, which corresponds to 18,120rpm and 18,480. Quite close to the onscreen graphics.
Checked the Yamaha at 3:33.4 and got a reading of 151hz/18,120rpm, only 38rpm from the highest onscreen value I saw at 18,092 for that particular shift.
Checked the Yamaha at 4:40.3 seconds and observed 144hz/17,280. Highest observed onscreen rpm for the same period was 17,204rpm.
Checked the Yamaha at 6:25.9 and recorded 150hz/18,000rpm. Highest observed onscreen readout for the same shift was 18,048rpm.
Sampled the Yamaha at 10:10.9, but no peak was dominant (unlike other Yam samples). Sample a wide .3 second range, and observed a small peak at 151hz/18,120rpm, compared to 18,230rpm on screen.
This log frequency analysis appears to be hilariously accurate, and about 100 times easier than any wave period analysis I've ever done. These are not cherry-picked results either. I chose 6 pieces of quality footage, tested all 6 of them, and observed margin of error around 1%. I think it's safe to say that the Ducati is a 930cc or slightly less based upon the 17,500rpm, but here is an interesting twist in the plot:
Assuming mean piston velocity of 26m/s Ducati would need a stroke of 44.5mm to achieve the rev ranges we observed. What displacement is 81mm x 44.5mm? 916cc
I've seen that number somewhere before. Sounds like something Ducati would do, make the GP bike a 916 for a very superstitious rider.
http://www.youtube.c...h?v=Yw2CMzqTXHw
Checked the Honda at 31.6 seconds and 37.3 seconds. Data read 151hz and 154hz respectively, which corresponds to 18,120rpm and 18,480. Quite close to the onscreen graphics.
Checked the Yamaha at 3:33.4 and got a reading of 151hz/18,120rpm, only 38rpm from the highest onscreen value I saw at 18,092 for that particular shift.
Checked the Yamaha at 4:40.3 seconds and observed 144hz/17,280. Highest observed onscreen rpm for the same period was 17,204rpm.
Checked the Yamaha at 6:25.9 and recorded 150hz/18,000rpm. Highest observed onscreen readout for the same shift was 18,048rpm.
Sampled the Yamaha at 10:10.9, but no peak was dominant (unlike other Yam samples). Sample a wide .3 second range, and observed a small peak at 151hz/18,120rpm, compared to 18,230rpm on screen.
Assuming mean piston velocity of 26m/s Ducati would need a stroke of 44.5mm to achieve the rev ranges we observed. What displacement is 81mm x 44.5mm? 916cc
I've seen that number somewhere before. Sounds like something Ducati would do, make the GP bike a 916 for a very superstitious rider.
You,re a funny fellow. When things go over my head, I keep quiet and hope nobody notices.
You, you want to advertise the fact, and then , just when some have forgotten, you need to emphasise it again.
Joined Jun 2007
10K Posts | 1K+
Norah Head
Just out of interest, what version of Audacity are you guys using?
I have stuck too a version 1.3 Beta cos everything I got after that had a limit of about 1 minute for any processing of the waveforms. Not that this task needs 1 minute, I guess. But in the past I went ahead and processed a bucketload of stuff then burned all the waveforms to what I thought was going to be a usual full cd, instead I got a CD full of 1 minute cuts
That was when they went 1.4 so as I said I have stuck with 1.3 ( before they got all bigheaded and uppity and were happy to freely spread their good work around
