Measuring instantaneous power on bicycles

One can arrive at a reasonable estimate by knowing the gradient and the angular velocity of the cranks, and making a small allowance for wind resistance.

For example, many years ago a crank rate of 90rpm on a hill of indicated gradient 1 in 12, allowing for the weight of the machine and the cyclist combined, plus knowledge of the wheel diameter and the gear ratio gave an order of magnitude of personal work output of 1/6hp for a short period.

Beware: as the body falls out of practice with the activity the peak power output and the ability to sustain it declines as a weak function of 'anno domini'...

In terms of kWh-consumed-per-tonne-km the bicycle is streets ahead (sorry!) of all other methods of transportation, though an electric train is not far behind.

You may think me mad , but I'm considering this project; I get charged £1 per cubic meter of potable water. By collecting rainfall and pumping it to the loft , I could use this to flush the w.c. Each w.c. flush costs about £.01 .My notion is to attatch a simple pump to the back wheel of a bike and pump water up to loft storage (say 4 or 5 m high).Not sure it's worth the hassle yet , but it would be fun. You didn't explain your motive , but a hydraulic power lift (which is essentially what i'm considering ) would give a good indicator of peddle power even if you're only returning water to a sump.

(Say) 100kg of water per person per day x (say) 7 metres lift x 9.81ms^-2 x (say) 2 persons per home = 13700 Joules per day.

Could be done by a pump rated at 0.16W continuously: well within the capabilities of mammalian muscle whether it uses a bicycle or not.

Thanks for the figures. I'm just waiting on some decent weather , and I may get the spanners out.Who knows ,I could start a mini green revolution.

"Each w.c. flush costs about £.01."

Is this the same as "spending a penny"?

Spot on! Hope my project is not Just a 'flush in the pan'.

The pump idea was interesting, but I wouldn't be able to get instantaneous power measurement, which was the goal of this project. What I want to do is make a digital or analog readout, completely integrated into the bike.

I'd wonder why you'd want to measure output from the cyclist as opposed to measuring possibilities for human power. In other words, measuring "a biker" could mean Lance Armstrong or it could mean me (um...not that there's much difference, mind you...), and you'd have only a number in space.

But as I recall, the aerodynamicist Dr. Valkenburg (spelling?) did a lot of work measuring output/pound as applied to airplane design before they built a human powered flight across the English Channel.

That's a great exercise in every meaning of the phrase!

Maybe you could calculate the number of miles in the average commute, the number of comparative calories, and come up with an engineered diet plan that'd make it onto Oprah...

This kind of device would be for monitoring your output so you can learn to pace yourself better. I wanted to do this because exercise machines that I've used seem to be very inaccurate. It told me that I burned 1480 calories in 1 hour 40 minutes, but I was only generating on average 150 watts. Of course, you burn more or less calories depending on how efficient you body is, but Lance burns about 15-17 calories per minute, and according to that machine I was doing the same. I'm in decent shape, but there's no way I'm anywhere near burning that many calories per minute, and the number of watts seems low. If I was working that hard on a real bike, I would have been going 18 mph on a mountain bike (on pavement), so I know that both of those numbers were just way off.

With a watt meter I could then come pretty close to calculating the number of calories that I actually burned. With an accurate number, could compare myself to someone else and keep track of progress.

How about some kind of load cell fitted to the peddles (or shoes ) + an optical sensor mounted on the frame to moniter revs around the crankshaft. I don't think a load cell increases the effort (??) , and you would have to deal with fluctuating effort throughout each peddle (eg is there only effort during 1/3 rd of a peddle on the down-stroke). Combining measurements with peddle radius and a bit of electronics could feed a handlebar mounted meter.Best of luck , Kris.

That's exactly what I was thinking. I was thinking of integrating a load cell into the crankshaft somehow, but the problem is you would need to measure the torque on both pedals. Then you have another problem besides that it's rotating, and wires will need to connect to it somehow... you have two load cells one would have compression on it, and the other would have tension on it from the other foot pulling up on the toe clip. I had another idea of somehow putting the load cell into the 3rd sprocket on the front, because that's where I do most of my riding anyways. There is some room there, I'm sure it's possible. Maybe there is enough room inside the three front sprockets that it could be integrated inside that assembly. And no I don't think a load cell would increase the effort. It measures the stress (or is it the strain...) on a particular piece inside it, and the movements are very very small.

I need to find an optical encoder that is easy to work with too.

If you really want to know the torque delivered by human to the wheel you need to change the system that is used to transport the power to the wheel: the chain system is nice and light but filled with vibrances, play and dirt. (big chance that an eigen freq of the chain is one of the biggest values that you measure)

Replace it with a cardan axis system and on this you can easely measure the torque with two optical encoders. At least you need the measure the torsion and need to calculate the torque and power from it.

How about torque assistance from a small electric motor.

Just thinking outloud.... The force you refer to is directly related to your gearing. The 1/6 hp estimate for a short term max is correct. The typical 12mph flat cruise for all day travelling in Europe is more like 1/20 hp

I was thinking of doing that too... use an old 18v cordless drill with a variable speed controller. My bike's front fork was made to have room for a disc brake on the wheel, so there would be room for adding something to drive the front wheel... now wouldn't that be cool.

1/6 hp for a short term max? Now that machine that I was using said I can do 150-160 for over 1.5 hours. I think that quite a few people that are in decent shape could do 120 watts for a little while (of course I have no idea if that bike that I was using was accurate at all).

-Nick

I too have thought about this over the years.

Maybe I ll get to it some year, but here goes:

The chain has a resonant frequency in the range of Hz to tens of Hz,

depending on the tension exerted on it by pedaling (torque),

just as the tension on a guitar string programs its resonant frequency.

This resonance is always being excited by vibrations during cycling.

A small, light wieght transducer could pick up these vibes like a guitar pickup.

This signal gets amplified and digitized.

This result is then processed (FFT) by a low cost DSP

to find that resonant frequency.

Calculations are then performed to find the tension / torque etc.

You could also exract a graph of torque vs. crank angle,

are you doing your "circles"?

Power versus time / distance as well.

Very useful for training. This could be very cheap.

Hmmm . . .

That's an interesting idea, but is is necessarily resonating? Will keep that in mind. Thanks for the input. I'll talk to my friend who is a grad student in Electrical Engineering.

-Nick

I believe that it is resonating.

I m an electronics guy, and theres always excitation.

The resonance acts as a band pass filter

whos resonant frequency is variable here.

Road vibrations are largely random / white in spectrum.

The filtering would yield an observable peak in the spectrum.

Thats where the DSP comes in, to sniff out that peak.

Once you know the peak, you know everything else.

I'm not sure I understand what you're saying.

If the chain weren't moving, with lots of links and rollers adding to the mechanical vibrations, you'd have a shot at using resonance (maybe in the KHz range; I don't know) resulting from plastic deformation to calculate force from strain.

But that'd make my head spin. The math would be freakish in the real case, and it'd be easier to do anything else, I think.

Of course, if the chain weren't moving, it'd be exceptionally easy to measure force...

Youve got to think in terms of excitation and filters.

The excitation comes form all kinds of sources.

The predominant one here is road vibration which is wide band random noise.

The filter is a band pass formed by the chain fundamental resonance.

Actually the math isnt that bad.

The FFT or Wavelets or what have you are open source by now.

You wouldnt have to re invent that particular wheel.

Granted, a good bit of work would need to be done to integrate it all.

I m only an eletrical guy. I need good ME and SW people to complete it.

I think Andy Germany has had the best idea so far. It seems simple, which is almost always the best way.

The real problem I see is that I'm betting whatever you do will cost more than the commercial systems that are built into either the bottom bracket or rear freehub. Those systems are pretty accurate (one bottom bracket system claims 99% accuracy) and already interfaced to PCs through the on-bike computer.

But then, as an engineering project, perhaps the most excruciatingly, perversey Rube-Goldberg-circumnavigated and complicated way would be the best.

Maybe you should try to incorporate a mouse-run, a boiling teapot and an O-gauge flat bed car/track in the final product.

Have fun!

Hey Alzie,

One of my grad student friends in Electrical Engineering mentioned the same type of thing you were talking about! Hey said that in industry (especially aeronautical) they send electrical impulses down parts of the airplane to check for really high stresses in components.

He talked to one of his professors that told him how to use filters and stuff to get the data that you need. I don't quite understand how it works, but he seems to know. Apparently there are a lot of things that change with the tension in a chain. There would be some kind of frequency response from the impulse (from my understanding, I don't know), and then the resistance changes.

So, I apologize for doubting your input so quickly, without being very well informed. I guess no one else who added their thoughts heard of it, but you are right, that should work. I was also told that the road noise from vibrations shouldn't be too much of a problem either

-Nick

It is quite easy: stress has influence on the resistivity of a metal (and on other conducting stuff)

When you stress that high that you get fissures you will see that the resistivy is going up faster as the conductive path narrowes locally.

A second method is reflection measurement: induce a pulse on a spot and wait for what is comming back (it comes quite fast)

Then you need to compare it whit what you saw last time you did it: the extra pulses could be a fissure. Stresses hava also influence on what is comming back (level and timing)

It is quite impressive to see what you can do with this technique, but to have it on a bike to measure instantanous power while driving your kids to school is a bit over the edge. Mostly it is one or two persons in the world, working for the constructor, that do these measurements as a part of the release of new structures (to prove that after a series of tests no fissures are present)

For a bike my suggestion is to keep it simple and try to do something with timing differences.

My understanding was that this technique was best suited to fixed or solid components, not something as flexible as a chain, as every link and roller will have its own (slightly) different characteristics.

You're right: a chain can't be monitored this way.

It is for the frame: you could measure the stress level on specific places in the frame.

Yeah, both of you are right. Plus each link has its own bearings, and the wax lubricant inside of them might effect the resistivity too much to make it work.

Maybe for the frame of a road bike this method might work, but the impulses on the frame from even tiny bumps will be much much higher than the pedaling will ever do. On a mountain bike, I think that it would be out of the question that it wouldn't work, because going over bumps and everything really fast imparts incredible forces on every part of the high performance frame.

Nick

I'm still enamored of the frame flex idea for several reasons:

- The natural harmonics of the frame tubes, as exited by impacts, would be far outside the frequencies we want to monitor (in other words the individual frame tubes ring at a much higher frequency, and the frame as a whole rings at perhaps 100 hz -- also much higher.

- The frequency were are looking for is relatively constant, and changes in frequency also occur smoothly. Further the magnitude is, for relatively long periods of time, fairly constant (as compared to road shocks, which are very random in frequency and vary dramatically in amplitude).

- Although I like the simplicity of using only a couple strain gauges, one could also add a crank pulse generator, so that the desired frequency could be determined and analyzed more easily, without having to fully filter out all the extraneous stuff to discern it.

- The really dramatic road shocks are apt to be fairly straight up and down through the frame, whereas the front down tube will twist with pedal cadence (from chain pull) so if the strain gauges are mounted to be most sensitive to twist, they will be less sensitive to up-down shock.

So, just as you can pick out a tone from among all the noise sources in a factory, I'd think you might be able to pick out the pedaling frequency and magnitude from frame data. Maybe... or maybe not.

The frame flex idea sounds a lot more promising after reading what you said! I think that we will end up doing it both ways, the other being as Andy Germany said with the strain gauge, like the image that I put in #27

The frame flex idea would be hard to make available to a bunch of different bikes, because every bike would be different, even of the same models would vary slightly, but once you get it right for that particular bike, it would be the best option. That's why I liked the thing that fits on the chain, because it would only be minor calibrations to change from bike to bike because of different sprocket, and crank arm sizes.

Nick

You are right on the difference between shocks and the moment induced by the crank.

But I fear that with the frame stress you can only measure the total stress level. (Von Mises)

For later processing of the signal it is indeed better to have a signal from the crank.

Except maybe that pedaling puts more of a sideways torque on the frame while bumps are up and down.

Just a thought, to keep the electronics on the frame and not on a rotating part and also to allow each gear to be checked, how about a small deflector roller that puts a slight bow in the run of chain between the top of the crank and the rear wheel. If this roller was supported by a fairly strong spring and the spring was spanned with a strain gauge (or whatever they call them today!), you could measure the deflection of the spring, this being as a direct result of the tension being put on the chain via the pedals! BOTH PEDALS AT ONE TIME OF COURSE!

You may need to add an inch or two of chain to make it viable!!!

If the chain went over one roller, then under the measuring roller, then over another fixed roller, this may improve measurement. All rollers would need to be liberally dosed with good ball races and be completely free running......

The reason why I specified rollers is that the gear change means that the position of the chain changes in both horizontal & left/right.....but also chain cogs, free running that can slip left and right might also work....

The unit needs to be near the rear of the frame, just before the gears, so that the crank gear change is not unduly affected.....

Once calibrated for each gear, which could be readily done with weights attached to the crank in various positions, I feel that this has a chance to do what you want to do!

The idea of measuring crank deflection, on both sides and on moving parts does not appeal personally!!

That is a great idea! I think that is the most viable option so far. I think that would be very possible with a very stiff spring and a potentiometer to do just that. I don't know that much about electronics, but with the knowledge that I have it sounds like that with a linear analog output from both the optical encoder and the potentiometer, you could use op-amps to do the multiply/divide functions to get some kind of power meter. Print a different scale (multiple scales calibrated to each gear) to put behind a cheap analog volt meter for the output.

Now I know that the displacement of the spring would not be linearly proportional to the tension on the spring, and I think it is a problem because of that. Maybe as the tension goes up, the extra sprocket could push a rack and pinion type thing that has an elliptical gear attached to the pinion, that turns the potentiometer, that is allowed to move in such a way to accommodate the change in radius of the oddly shaped gear. (so as it gets pushed up more, the radius goes up and turns the pot more). You would probably have to do some calculations, or test it out with the spring to see what kind of gear you would need. That might be a way to make the non linear vertical displacement of the chain to become linear, but there might be an easier way to deal with it electronically, I don't know. I'm not sure how I would deal with a non linear signal.

Now if you used a load cell that doesn't cause any displacement, there wouldn't be a problem, but the cheapest ones I saw in a global spec search were hundreds of dollars!

After typing all that I think I realized what you were saying... have two gears below, and one above. The one above could push the chain between the two gears below, and when the chain tightens, it could deflect the top one, which would be spring loaded. That force or movement could be monitored with a load cell or potentiometer. Now if it wasn't allowed movement with a load cell with that setup, would the force be linear to the tension? This is a good idea, because it could be made so that the side to side movement in gear changes wouldn't be a problem, because it could be made to move with the chain! I think this has some promise.

Thanks,

Nick

I have made a motor assisted bicycle where I control the motor torque based on the chain tension. The chain tension being sensed by a simple sprocket on a spring loaded arm engaging the tension side of the chain. The arm movement was transfered to a wire wound pot using a standard cycle brake cable. The potentiometer was used as a potential divider and the variable voltage used to decide the pulse width of a standard PWM motor control. The result is something akin to a power steering. The harder you pedal; more the assist. If you do not care to pedal it can also be used as an electric bike using a contol on the handlebar to pull the potmeter. But of course then the range is only about 8 km. I confess that the chain tension is not really measured accurately in this set up. But I did notice that the chain tension fluctuates over a wide range during a single rotation of the crank due to the varying component of leg effort contributing to the torque with varying crank angle. In fact I had to add an air damper to the tension sensing arm to smooth out the input to the motor control. It is this information that I thought might help you in your work.

You need what is called a "strain gauge", which will give an electronic output proportional to the deflection on the spring, no matter how minute - even amounts that the nacked eye cannot see - it measures the "strain" in the spring!

The spring could be a flat piece of stiff but springy metal and the good point about springs are that they are linear!!! They bend twice as much when double the weight is applied!

A further refinement would be to have a position sensor on the crank, so that you measure say every 5 degrees (or smaller), thus the timing would come from the crank itself, so no matter how fast or slow you rode! 72 Steps in 360 degrees. You just need a further sensor to note when the crank is at position 0° (TDC in a petrol engine so to say!).

From these sensors and knowledge of the gear in use, accurate speed could also be displayed!! Two Hall effect sensors on the frame and some very small magnets on the crank itself, as near to the gear teeth as possible would suffice for full positional info....

I believe the strain gauges are often quite reasonably priced (depending on the type) and I am told, relatively painless to integrate electronically! I believe prices start in the single $ area!

Talk to someone with more exact knowledge of them, or start a new Blog about strain gauges....and interfacing of them!

A further good point is that if the unit was built with some thought, it could be fairly easily placed on a different bike in a few minutes.....( a bit longer for the crank input sensor if used)...in case you use an old bike for design work and when finished, to put it on your "good" bike for example!

Good luck.

"You just need a further sensor to note when the crank is at position 0° (TDC in a petrol engine so to say!)."

Early auto systems used twin sensors - one measuring starter ring teeth timing, the other to detect a slug at TDC. Later models use a half-size tooth at TDC, so only one sensor is needed.

Ok, the strain gauge is a good idea. It looks like those are fairly cheap. I was way over complicating it with using a potentiometer to measure movement in the chain. Before, I was thinking that w/o a load cell, when there would have to be measurable y-axis movement in the chain, the tension to y-axis relationship would not be linear, but with little or no movement, it obviously wouldn't be a problem.

I think the three gear idea that moves with the chain is the best bet. Make it so that it's allowed to move side to side and up/down

Sprockets with a lot fewer teeth of course, that was just the easiest image I could find with google to manipulate. I still think though, if the movement of the top sprocket is considerable, it won't be linearly proportional to the tension... Do you agree, or am I thinking of this wrong? Maybe the strain gauge will be accurate enough with very little movement.

Maybe that could just kind of float on the chain, and be tethered to the frame with a cable.

If you read what I wrote, that is basically what I said!!

My further thoughts this morning are that with a suitable sensor, one of the cranks sets of teeth could be used as timing for the circuit. It does not even matter which one you use as they are physically linked to each other! It does not matter which one actually has the chain on it either!! A double bonus.

The larger crank with more teeth will deliver more impulses per revolution, so the resolution will be finer! But if that is not needed, then use the smaller one as it is usually nearer the frame. A suitable position for the sensor would be just behind the crank where no chain gets in the way! The type of sensor required is one that has a magnet built into it, you can check with a piece of steel. These induce into moving steel parts (not stainless though!) enough magnetism that the coil surrounding the magnet will pick up impulses from each tooth as it moves past. Don't forget the TDC sensor, to tell the electronics when a revolution starts!! I would recommend still, a Hall effect for this one!!

With a little bit of ingenuity and a suitable small cylindrical magnet, you can wind your own coil around it and make your own sensor - very, very thin wire and a lot of turns. Make the "nose" of the magnet as small as the end of a tooth on the crank. Place them "nose to nose" for best pickup. The output sine wave amplitude will vary with speed, so make sure that low speed is good, attenuate over a diode network for higher speeds if needed! Gap will need to be around a millimeter or so.....good bearings required on the crank!!

Best of luck.

Maybe you could take this idea a step further. As the torque on the pedals increases, everything will flex a little. That will alter the signal picked up. The change could be translated into a torque figure. So one sensor could pick up torque and RPM.

Andy Germany, that is an interesting idea. I've heard of similar things measuring angular velocity using the same principle on small-tooth gears. Now I think I understand how that works. Now the output with a setup as you described would be an AC sine wave. If it was purely an AC output, I suppose it could be rectified and turned into a DC output. The way I'm thinking of this, is that we would use a minimum amount of circuitry as possible, maybe only needing one opamp to do the entire operation. Really, all the operation has to be is analog torque voltage times analog rev voltage, with some constants in there, which would be put in the calculation with resistors. The arbitrary outputs of the rev-o-meter and the torque meter would have to be calibrated anyways.

Maybe you had a better plan, taking into account each revolution, torque and rpms?

-Nick

As a cyclist myself, I am yet to find a set of front gears with less than 1mm of runout, regardless of how good the bearings are (I always use sealed units). Most are made of pressed steel, which would excite the sensors, but be susceptible to twisting. The lightest are cast alu alloy - so would not work with magnetic sensors.

Dual suspension bikes also tend to give a very uneven pull on the chain, as the chain is connected between the two halves of the bike (unless the rear suspension pivots around the centre of the crank). I have tried them, but have returned to a hardtail, as I cycle mainly on roads. I think they are only suited to offroading, where the shocks experienced are much greater - and any of the suggested systems would be unlikely to cope with the harshness of the ride, or the dirt encountered on the way.

I was wondering if putting strain guages between a sealed unit bearing and its support sleeve (arrowed) would be possible, as this would protect it from the worst of the dirt.

I'll have to (attempt) taking apart the bearing, and checking out to see what is possible. I can foresee problems with electrical contacts there, with the gauge in there or on the cranks. That's why I liked the homemade strain gauge that would ride with the chain, even though it would be kinda ugly, crude and it would add a little friction to the ride.

The sine wave pickup was only for the timing - when to sample the torque detector for a new value, that remains as I described before.

It may be that the sensor will still pickup enough signal, a bit further away, it depends on mainly two factors, how strong the magnet is and how many turns are on the coil, both more magnetism and more turns will increase signal strength.

The sensor could be tried at right angles to the crank, if the left/right movement is there smaller and it may need to be only on the smaller inner crank as the flexing should have less effect there for several good reasons.....

If the worst came to the worst, then a light (infra Red please) barrier setup might be better as it is also remarkably insensitive to dirt on the sensor, just as the magnetic one is.......a small hole drilled for the second sensor for TDC might be also needed.

With regard to alumium, it may still function as alumium IS affected by magnetism, but in a completely different manner, sadly my knowledge is not enough to comment any further on that score!

A further method maybe to add a crank/timing disk, with no chain to the LH side of the bike, no torque passing through it.......

I'm going to talk to some of my EE friends, to see how they would use that signal to do the calculation electronically. I don't seem to know enough about this kind of thing to make sense of it. I understand perfectly how the sensor you described works, but don't know how to apply the result. I still don't quite understand what the sensor for TDC would be for. Like you said, maybe a timing disk on the left side of the frame, would be the best option for measuring the RPMs. I've seen disks with laser cut slits in them so fine, that the metal seems translucent in that area until you look close, and see the cuts in it. Something not so extreme as that would be pretty good I think (for using as the input to an op-amp, because the on off on off output would be so fast, it might act like an analog dc signal if a capacitor was used to smooth it out.) I know I keep turning to using op-amps to do this, as that's the only way I know might be possible! Or maybe I'm completely wrong and it's not!

I'm with Andy Germany, put yourself strain gauges on the pedal arms (whatever they're called) The crank arm connecting your foot pad to the spindle with sprockets on it. I suggest you put two (or four) strain gauges on the petal beam. One on each side, that way tension mode would be canceled, only measures the bending load. Look up strain gauges and load cells on web. Wikipedia should give you good help. Putting the load cells there would give you the (tangential) force applied to the petals, and hence the torque at the crank where you can measure RMP. This would give you many details about how the power curve changes through the cycles as well as compare right from left. Could possibly help you improve you cycling performance by encouraging you to produce the same amount of power from both legs.

I might go all out and proto up some adjustable length pedals and then have people of different heights (or inseam measurements) test ride your bike at different pedal lengths. Create a 2d chart based on rider height and pedal length to help draw a correlation to the best power output based on riders metrics. Could also categorize it from street riding to off road... Getting raw, low level data can be very useful (and over whelming) but fun to analyze to see what you can learn about it.

For info on strain gauges check out Vishay or Omega. When I was at college our ME department had strain gauges in stock for Mech Systems Lab. There's a chance much of the stuff you need is already in your department.

Best of luck.

I really like the idea of having outputs for each leg or just one on the inside somehow, but I think that would make it really complicated and prone to failure, having some of the electronics on rotating parts. The more I think about it, the more difficult I feel it would be to make reliable contacts going to the frame with dirt, bike grease and water. Even on a road bike, dust particles and bike grease would be an issue (especially the wax lubricant that I use).

I think this may end up being just a free time project and not a senior design! This should be fun...

Thanks everyone for helping!!!

-Nick

Hi Nick, Unhook your bicycle chain and separate the link that completes the loop. Attach one end of the chain to a fixed position. Remove the rear wheel and block up the frame. String the chain over the drive sprocket and driven sprockets and attach the loose end to a spring scale, strain gauge, pneumatic piston with air pressure retracting it or other methods. When you step on the pedal that force will produce a result that can be observed and then calculated for the torque. Obviously most twelve year olds will produce a different force/torque output. What you are actually measuring is a variable, personal strength and body weight. If the bike were driven by an engine the output torque would be more finite and consistent.

You could do this test for $250 worth of junk and your time. You might go with a more challenging project for a better grade. Lead the field.

Albert Einstein said: "Everything should be made as simple as possible but no simpler."

What you are suggesting is to measure the static force possible at rest, while cycling this force can be altered considerably, depending on the cadence and the cyclists ability to straighten their leg at the required rate.

The TDC sensor is just to let the electronics know that the crank has just passed thru 360 degrees. The other sensor is to allow the electronics to have that 360° chopped up into usable segments, as I am sure that the torque will vary with the position of the pedals, so say for example you have 72 segments, you can measure the torque being produced in each one of those 72 segments, half will relate to one foot, the others to the other foot, simple!

You will need memory and a way of loading it with the values read out of the torque meter part, the position on the crank (how many teeth after TDC for example) and a way of reading the data out to a laptop for calculation/manipulation etc...

I think that it is a good project and could later be more refined and probably saleable too!!

Oh, ok. We were thinking of it differently to get the same result. I was thinking of turning the RPMs into a linear analog output, doing a whole lot of calibration by hand and using simple op-amps (and trimmer pots) to get the power output signal.

Your idea sounds a lot better, but would require quite a bit more knowledge of electronics on my part, but maybe with some cheap programmable I/O device, it would end up being easier. I've used them before, just have to remember the name. I'm thinking that this will just be another fun project, and not my senior design (The department would probably ask for more of a project), so I can cheat all I want now .

In my experience, its not the complexity of the task that matters, but how well it is carried out and documented. Many have done less well by choosing a difficult task and not completing it, so leaving too little time for a comprehensive report.

Do you want to build something that can be sold and installed on each bike?

A possible way might be to measure the strain in the chain, filter out the vibrations.

Two optical pick-ups can do it, making use of the holes in the chain. You will need a procedure to calibrate but this can be done on the fly (lowest tension is the floating 0)

The main idea is that a chain under tension will become longer, changing the timing between two pick ups that are placed over a distance on the chain.

You certainly will need some processing power.

Gwen

Talking about the phase difference between chain poinst, you could also evaluate if you could not measure the phase difference between two targets in the wheel. As the torque in the wheel increases, the outer rim rotates in respect to the centre. Of course, the stiffness of the wheel is the key here.

I'd follow this approach (optical or electromag sensors in chain or wheel). It is gonna be more reliable, no contact, and more elegant too for an engineer work.

With the wheel, you have a problem with the different materials used, resulting in different pre-stress depending on temperature changes, making it more difficult

I think I found the perfect device for doing the calculations. Parallax BASIC Stamp kits. They can have up to 32 I/O channels, plus a dedicated serial port. That one has 16 kb of EEPROM too. That one costs about $90, but there are faster much cheaper ones out there. There are lcd dot matrix displays available for really cheap, so maybe I could make it display velocity, power, and cadence. That would be cool!

Sounds like a fun project, Nick.

Here's a thought: The frame down tube just above the crank will flex back and forth with each pedal stroke. So a pair of strain gauges bonded to either side of the frame down tube could sense torque, and could also sense cadence via the rate at which the torque values cycle. No moving parts, no critical clearances, etc. I'd think analog electronics could filter out road shock, etc prior to input to the Stamp.

For measuring instantaneous power on bikes I suggest this way

For measuring the power you need to know, mechanical force, speed and time.

With one load cell under the seat of the bike you get the driver weight, as you can know the bike weight you will have the total weight of the mechanical body, by using one accelerometer, are cheap, you can get the mechanical force and finally with one optical sensor it will be easy to obtain the bike speed variation.

In order to process all the information you will have to use one microcontroller such as AVR from ATMEL, they are very cheap and easy to program.

For equations details you can use any basic university physics book.

Best Regards

Dear guest, what you propose is nowhere near enough, you also need to know at least the inclination of the ground and direction of the inclination and the speed and direction of the wind and apparent wind as these also affect bike movement - to name but a few. You would also need to know any mechanical differences in the bikes chain, wheels and gears due to dirt, lack of or too much lubricant etc etc etc.

It would get far too complicated, just getting sensors made/bought and then installed and calibrated!!!

You seem to forget that bike riders do not always have flat, perfect road with no potholes, same air pressure, no wind and no hills and that a new bike, a run in bike and the same bike getting old, all have mechanical differences that must be in the calculation for reasonable accuracy....!!!

And I could easily have forgotten something as well here, my list is not perfect.....but have a great day anyway.

I have recently read thru this Blog again, just to see if there are any points that could be improved upon and whether the méthod I detailed could actually be made to work. And after reading thru I feel that it is still the only method that has a possible chance, unless someone comes up with something better of course, that has not yet been described!

The method I proposed is mechanically fairly simple, it will not require and engineering degree to build, only to program effectively maybe! But that is the same with any method....

I would also be very interested if anyone makes an effort to build it, either in the way I described or of course any other way as well!!

I will be doing something with this, using your method. I probably won't get to it until this summer when I have more time, but I will definitely notify you of my progress. For now, I will just be searching for all the little parts that might come in handy for the project.

Andy Germany :

I know its probably been a while but here's a real world application of using the electric guitar pickup method on the chainstay. The sensor picks up vibrations in the chain and uses DSP to bring out torque and power. Scroll down on this page to POLAR.

http://www.biketechreview.com/archive/pm_review.htm

Nick :

Just my opinion : The project seems viable but in practical terms, a frictionless, non-contact type of measurement method is ideal since any tapping of the input energy to throw out a total output torque is not very accurate. It will not be a true indicator of performance. But if you just want to get the gist out of making a power meter go for it!!

Ron

Cozy Beehive

Best of luck NickJD, it will be nice to hear your progress and of course we will all try and help if you run into any problems, though personally my programming is a bit old fashioned and out of date!! Hardware wise I am sure I can help further if needed.

Andy Germany,

Your idea is very coincident with Polar Power System. Like Nick said, the idea is very easy to understand but application is all in electronics and computing. In fact, in the early part of this project, Polar (supposedly called CC Kinetic in the past) hired Boston Engineering, an award winning electrical and software firm to code everything down and build them a prototype.

The entire technology is now patented! See here for :

Method and apparatus for measuring power output of one powering a chain driven vehicle (patent# 6199021

Ron

Cozy Beehive

Here's another idea that (to me) simplifies the approach...

When you push on the pedal, you turn the wheel at its hub, and the force is translated through the spokes to the edge of the wheel. There is obviously a degree of deflection in the spokes - ie the hub turns, the spokes deflect, then the edge of the wheel turns.

If you had a wire in tension following the path of the spokes (joined to the hub and joined to the inside edge of the rim) then this wire would stretch (ie increase in tension) as force was applied to the hub. Measuring this tension with an in-line unit - which could be as simple as measuring the stretch in the in-line unit) would give you a result that you could transform into a power reading. This would not be dependent on which gear you were in etc - it would give (I think) a measure directly proportional to instantaneous power to the rim. Moreover it wouldn't interfere with your gearing or chain and would be all packed up neatly inside the spokes. (Hope this isn't just a reiteration of someone else's post - I skimmed through the last half after deciding to post this...)

You would need to cater for the deflections caused by road jolts, but those data would be spikes that could be filtered out.

(even if this is half baked the beauty of this forum is you get tons of expert feedback when you are wrong!)

...yup - should have read through more carefully ... -bhrescobar's post captured this idea. And unless you only want relative power (still useful) it's going to be difficult to calibrate this to give you watts - perhaps you would have to create a calibration dock that measured the power output against a known force on installation.