New powermeter rumours
Moderator: robbosmans
-
- Posts: 622
- Joined: Fri Feb 10, 2012 4:27 pm
Anyone with no engineeiring background who's saying it's 'easy' to make a powermeter should just shut up...
Engineering is a compromise. making something that will fulfill the purpose ( e.g. making a power-measurement device) is rarely difficult. making something that fulfills the purpose within various constraints is the problem.
anyone with basic knowledge of electronics and coding could make a powermeter. but it wouldn't be very accurate, nor precise. first you'd have to accomodate all the different crankarms and chainring standards, not to mention Bottom bracket standards, so you'd have to test, test, test and test again. not to mention you'd have to make sure it was consistent, and this can vary with each component batch the manufacturer makes- so you'd also need to test, test test to make sure calibration is correcnt.
next you'd need to filter out various noise- much has been said about drivetrain loss (for powertap) radial acceleration, et al, so I won't mention them here. it's even worse for pedal based systems, which explains the problems the vector and the look systems have.
next you'd have to figure out your interfacing with youe head unit. not so much a problem for wired units, but wireless units do also need to figure out a way to reduce noise in transmission- I've heard ANT+ headunits can have as much as 10% inaccuracy due to the transmission protoca=ol. ( I'm niot sure. I'm an ME, not an EE.)
now you'd have to make sure it is survives in a hostile environment- years without mantainence, next to no calibration, waterproofed. this also means more testing. see where that got ergomo.
now we'll have to find ways to mass manufacture, which means sourcing for suppliers and factories in the far east, probably. which also means ensuring that they have good QC, negotiations on contracts, and build to your specifications. high quality electronic packages meshed to a high quality bicycle product is not something the far east is particularly known for. japan and korea, maybe. not so much china and taiwan. whcih drives costs up. good quality strain gauges that will last you a long time and give reliable readings are also diffucult to find.
after that you'd actually need to make sure it coul SELL and be competitive in the market- i.e. it has to be light, pretty, and cheap.
the prices will come down as more and more people get interersted in developing a PM, but a powermeter is probably much more difficult to design and sell than say, wheels and hubs.
sad to say but I don't think the stagesone meter will come to carbon cranks. the composite nature of carbon means that the strain is highly dependent on the layup, which may differ slightly from individual crank to individual crank. makes predictable measurement and manufacture that much ahrder.
Engineering is a compromise. making something that will fulfill the purpose ( e.g. making a power-measurement device) is rarely difficult. making something that fulfills the purpose within various constraints is the problem.
anyone with basic knowledge of electronics and coding could make a powermeter. but it wouldn't be very accurate, nor precise. first you'd have to accomodate all the different crankarms and chainring standards, not to mention Bottom bracket standards, so you'd have to test, test, test and test again. not to mention you'd have to make sure it was consistent, and this can vary with each component batch the manufacturer makes- so you'd also need to test, test test to make sure calibration is correcnt.
next you'd need to filter out various noise- much has been said about drivetrain loss (for powertap) radial acceleration, et al, so I won't mention them here. it's even worse for pedal based systems, which explains the problems the vector and the look systems have.
next you'd have to figure out your interfacing with youe head unit. not so much a problem for wired units, but wireless units do also need to figure out a way to reduce noise in transmission- I've heard ANT+ headunits can have as much as 10% inaccuracy due to the transmission protoca=ol. ( I'm niot sure. I'm an ME, not an EE.)
now you'd have to make sure it is survives in a hostile environment- years without mantainence, next to no calibration, waterproofed. this also means more testing. see where that got ergomo.
now we'll have to find ways to mass manufacture, which means sourcing for suppliers and factories in the far east, probably. which also means ensuring that they have good QC, negotiations on contracts, and build to your specifications. high quality electronic packages meshed to a high quality bicycle product is not something the far east is particularly known for. japan and korea, maybe. not so much china and taiwan. whcih drives costs up. good quality strain gauges that will last you a long time and give reliable readings are also diffucult to find.
after that you'd actually need to make sure it coul SELL and be competitive in the market- i.e. it has to be light, pretty, and cheap.
the prices will come down as more and more people get interersted in developing a PM, but a powermeter is probably much more difficult to design and sell than say, wheels and hubs.
sad to say but I don't think the stagesone meter will come to carbon cranks. the composite nature of carbon means that the strain is highly dependent on the layup, which may differ slightly from individual crank to individual crank. makes predictable measurement and manufacture that much ahrder.
The big issue with ANT+ is that it requires data every second, on the second. If crank-based systems are measuring rate of pedal rotation with a magnet, they get a signal only once per pedal rotation, which is not going to be exactly every second. So you need to make assumptions how to line up torque (which is measured whenever you want) with cadence (where a signal is generated only once per pedal rotation) which yields potentially large errors. Ideally these systems would send power numbers once per rotation, and the head unit would display them as received. SRM, Quarq, and Rotor all have this issue.
There's different ways to extract rate of pedal rotation, however, and they have different strengths and weaknesses. Different systems use different methods.
There's different ways to extract rate of pedal rotation, however, and they have different strengths and weaknesses. Different systems use different methods.
Visit starbike.com Online Retailer for HighEnd cycling components
Great Prices ✓ Broad Selection ✓ Worldwide Delivery ✓
www.starbike.com
Epic-o wrote:djconnel wrote:Another factor: acceleration-based cadence is challenging if you don't have a way to separate centrifugal acceleration from the bike bouncing up and down. I wonder how they make that work and how well it works, especially at lower cadences (for example, 40 rpm) on rough roads, especially since their accelerometers appear to be on an intermediate radius of the crank arm (where centrifugal acceleration is less than on the outer radius, for example at the pedal).
There is a pretty easy way to calculate this. If you place two accelerometers in the crankarm at a known distance (very little if possible) and you substract the two readings, you can calculate the radial component of acceleration. Neglecting the difference of acceleration between these two point due to the rotation of the bicycle as a whole and applying some sort of filtering you can get a pretty accurate value.
Damn. I never thought of that.
I don't think bike acceleration is an error term. What matters to power transmission is rotation in inertial coordinates, not bike coordinates, so whether the foot turns the pedal against the bike or the road turns the bike against the pedal is irrelevant: the chain is still driven forward. I'd think acoustic waves propagating down the crank arm would be a theoretical error term but too fast to be important over the time scales of interest. Another term would be if the crank deflection were changing as a function of time but this would also be a small error since the associated accelerations would be on order 1 mm/sec² : well below a 1% error budget.
-
- Posts: 1163
- Joined: Sun May 16, 2010 7:28 pm
djconnel wrote:The big issue with ANT+ is that it requires data every second, on the second. If crank-based systems are measuring rate of pedal rotation with a magnet, they get a signal only once per pedal rotation, which is not going to be exactly every second.
Is this really so big of a problem? Your pedal revolution is almost always going to be faster than 60rpm, so for every ANT+ update you'll have a new estimate of the cadence. In addition, it's not like cadence varies wildly over a span of 1 second, so even if you used the cadence value from one second ago, it's not like it's going to be off by 30rpm. Also, every other crank-based power meter has the same problem and have found a way around it, so it can't be that big of an issue. And then finally, you typically average power measurements over a span of 30 seconds anyway, so any intermediate power errors get averaged out anyway.
Yes, making a power meter is hard, but let's not pretend it's rocket science. Otherwise the dudes at Stage One wouldn't be able to glue on a 20g plastic device and have it work just as well as the other major players. I think the biggest issues are just calibration and testing as you're using strain gauges on a certain crank arm and assuming some properties of each crank arm based on a set of collected data.
djconnel wrote:Damn. I never thought of that.
I don't think bike acceleration is an error term. What matters to power transmission is rotation in inertial coordinates, not bike coordinates, so whether the foot turns the pedal against the bike or the road turns the bike against the pedal is irrelevant: the chain is still driven forward. I'd think acoustic waves propagating down the crank arm would be a theoretical error term but too fast to be important over the time scales of interest. Another term would be if the crank deflection were changing as a function of time but this would also be a small error since the associated accelerations would be on order 1 mm/sec² : well below a 1% error budget.
I think that inertial cordinates isn't the best choice. If you are put a static load in the pedals without pedaling (coasting while standing on the pedals for example) and a road irregularity accelerates the bike as a whole, the powermeter will give a non-null value of power when you aren't really generating muscular power.
And the calculation of the velocity of the pedal in inertial coordinates isn't an easy task, I would say it's nearly impossible using typical equipment (variable position of the COG, position of the ICR and bike rotation).
- prendrefeu
- Posts: 8580
- Joined: Wed Oct 31, 2007 10:32 pm
- Location: Glendale / Los Angeles, California
- Contact:
davidalone wrote:sad to say but I don't think the stagesone meter will come to carbon cranks. the composite nature of carbon means that the strain is highly dependent on the layup, which may differ slightly from individual crank to individual crank. makes predictable measurement and manufacture that much ahrder.
Yes, and along with what Epic-O had to say on this I believe you two are correct, much to my disappointment.
Regardless, that the PM is non-drive-side only, it would make for an nice training tool with easy swappability... which is not even a real word, but it is now.
Exp001 || Other projects in the works.
- HammerTime2
- Posts: 5813
- Joined: Thu May 04, 2006 4:43 pm
- Location: Wherever there's a mountain beckoning to be climbed
It's not listed in dictionary.com , http://www.merriam-webster.com/dictionary , http://www.thefreedictionary.com , http://oxforddictionaries.com/?region=us , or http://www.macmillandictionary.com , but it has been used.
http://www.day.com/specs/jcr/1.0/3.1_Swappability.html
http://blog.paperbackswap.com/help/help_item.php?id=205
http://www.candlepowerforums.com/vb/sho ... appability
and many others.
These dictionaries find themselves in a pickle when having to decide whether to add "new" words. It's an uphill race as words race by.
http://www.day.com/specs/jcr/1.0/3.1_Swappability.html
http://blog.paperbackswap.com/help/help_item.php?id=205
http://www.candlepowerforums.com/vb/sho ... appability
and many others.
These dictionaries find themselves in a pickle when having to decide whether to add "new" words. It's an uphill race as words race by.
Epic-o wrote:I think that inertial cordinates isn't the best choice. If you are put a static load in the pedals without pedaling (coasting while standing on the pedals for example) and a road irregularity accelerates the bike as a whole, the powermeter will give a non-null value of power when you aren't really generating muscular power.
No, because if you have accelerometers mounted at two radii of the crank arm, road irregularities will accelerate each the same: you'd subtract the two readings and divide by the fractional separation (divided by crank length) of the two accelerometers to get a centripetal acceleration.
kulivontot wrote:djconnel wrote:The big issue with ANT+ is that it requires data every second, on the second. If crank-based systems are measuring rate of pedal rotation with a magnet, they get a signal only once per pedal rotation, which is not going to be exactly every second.
Is this really so big of a problem? Your pedal revolution is almost always going to be faster than 60rpm, so for every ANT+ update you'll have a new estimate of the cadence. In addition, it's not like cadence varies wildly over a span of 1 second, so even if you used the cadence value from one second ago, it's not like it's going to be off by 30rpm. Also, every other crank-based power meter has the same problem and have found a way around it, so it can't be that big of an issue. And then finally, you typically average power measurements over a span of 30 seconds anyway, so any intermediate power errors get averaged out anyway.
Yes, making a power meter is hard, but let's not pretend it's rocket science. Otherwise the dudes at Stage One wouldn't be able to glue on a 20g plastic device and have it work just as well as the other major players. I think the biggest issues are just calibration and testing as you're using strain gauges on a certain crank arm and assuming some properties of each crank arm based on a set of collected data.
There's two issues: (1) sampling cadence only once per second, (2) the time lag of cadence relative to torque.
On (1), I looked at that here.
I concluded for low-inertia pedaling (up a steep hill, for example) the error would exceed the claimed 2% accuracy, steady state and not just transient.
Power2Max doesn't suffer from this issue since it uses accelerometers. It may suffer from other issues...
So I don't think SRM and Quarq are accurate when the pedal moves at a nonuniform rate.
DC Rainmaker mentioned that Power2Max is now offering a temperature compensation circuit on new power meters, and retro fit plus firmware update to older models - so it should hopefully now pass the Jeff Jones test.
http://www.dcrainmaker.com/2012/08/power2max-announces-updates-to-power.html
http://www.dcrainmaker.com/2012/08/power2max-announces-updates-to-power.html
djconnel wrote:Epic-o wrote:I think that inertial cordinates isn't the best choice. If you are put a static load in the pedals without pedaling (coasting while standing on the pedals for example) and a road irregularity accelerates the bike as a whole, the powermeter will give a non-null value of power when you aren't really generating muscular power.
No, because if you have accelerometers mounted at two radii of the crank arm, road irregularities will accelerate each the same: you'd subtract the two readings and divide by the fractional separation (divided by crank length) of the two accelerometers to get a centripetal acceleration.
The difference of accelerations in the radial direction is proportional to the angular velocity in inertial coordinates (angular velocity relative to the bike+angular velocity of the bike with respect to the inertial frame) of the crankarm. If you aren't pedaling and your bike is bouncing, the angular velocity in inertial coordinates of the crankarm and the power while standing isn't null. While power is the product in inertial coordinates of a moment and an angular velocity, I don't think that this contribution to total power should be taken into account. Solution? Maybe adding another sensor to give zero power when the crankarms aren't rotating with respect to the bike.
In any case, there will be an error term in the centripetal acceleration due to Coriolis effects that will become zero if the bike angular velocity in the inertial frame or the angular velocity of the crankarm with respect to the bike are null
Dan, do you know anything about the protocol SRM uses? It's not true ANT+ and the PC7 provides a sampling rate of .5 seconds. I don't know how it's different and how that changes things.
That would be a good question for Quarq as well, since their newer pms work with SRM headunits.
That would be a good question for Quarq as well, since their newer pms work with SRM headunits.
"Nothing compares to the simple pleasures of a bike ride," said John F. Kennedy, a man who had the pleasure of Marilyn Monroe.
Visit starbike.com Online Retailer for HighEnd cycling components
Great Prices ✓ Broad Selection ✓ Worldwide Delivery ✓
www.starbike.com