Recumbent position power loss

This is another article in the vein of the 1 minute hill climb, looking at (but not necessarily answering!) issues around recumbent and upright power production.

Background

I haven’t ridden many recumbent miles since the 200km Erit Lass last September, but I have kept up my five mile commute on my upright hack bike that whole period. I consider myself to be quite badly out of recumbent-specific conditioning (despite coming in 21st in my category on the Tour o’ the Borders).

tour of the borders recumbent
The new Optima High Baron cleans up the field on the Tour o’ the Borders…

As I recently started getting serious with the Optima High Baron, I thought it might be interesting to benchmark my power on the two platforms now, and look at whether it converges over time. In my opinion it’s also an interesting window onto the experience of novice recumbent riders who try to transition from their upright bikes expecting a gain in performance.

I’m going to display the data from two PowerTap stationary trainer intervals. The sessions were uncontrolled except that I tried to shoot for my highest average power in each case over a 12.5 minute interval (interestingly, in both cases I ended up picking things up at the end – bad pacing?).

The intervals were on different days, first thing in the morning before work. For what it’s worth the recumbent went first…

Interval Upright Recumbent Power loss
1 minute 355W 296W 19.93%
5 minutes 290W 239W 21.33%
10 minutes 275W 220W 25.00%

A 25% drop in power means that on a climb the upright would pull away one mile for every four miles the recumbent rider travels – it’s a pretty big gulf!

Context

[edited to add this section, as I felt the article originally was lacking in context]

Consider this: received wisdom often tells us that recumbents do not climb as well as conventional bikes because they are heavier, and/or because they are less efficient due to some combination of frame flex, or drivetrain friction due to the idler systems involved. However, I believe this is demonstrably false.

In the first case, we can categorically evaluate the impact of extra weight using simple science, with or without a power meter (see my article on bike weight and performance).

However, with a power meter it’s possible to precisely define both limiting factors of weight and power output and, as in this case, I believe climbing performance will always turn out to be limited by reduced power output and not by increased weight.

However, by measuring power at the wheel and not at the cranks we do leave open the possibility that the drivetrain is consuming 50W or more in increased friction, however unlikely that may sound. My answer to this (and ultimately to all questions of inefficiency in the recumbent bike proper) is that if the same work is being done by the human body, but it’s just being lost somewhere en-route to the road, we should see a very similar physiological impact to the activity.

This is manifestly not the case, as in the two intervals discussed below: the recumbent one is not only far fewer watts, it had manifestly lower cardiovascular demands (even though it was as hard as I could push the pedals), relative to the upright session.

I’ll follow this up as promised with something that includes HR, although that’s of only limited use in evaluating demands on the body, as we’ll see…

[back to the original article:]

Torque vs cadence

torquepower

Comment

Don’t read too much into a sample size of one, however, counter-intuitively the power gap increases with duration. You would expect proportionally greater failure to generate power over short periods if the upright position simply allowed a greater mass of muscle to be recruited (albiet inefficiently), as many hypothesise.

As I’m sure the difference between recumbent and upright riding position tends towards equality in the very longest events, this initially suggests a reverse-U shape. Clearly more investigation is required… I personally think this is showing both a fundamental difficulty in producing recumbent power but more importantly a significant and specific lack of conditioning in a couple of key muscles that is inhibiting overall performance.

While it’s tempting to blame inefficiencies in the physical recumbent drivetrain, I don’t think this is a significant issue because my cardiovascular reaction to these two intervals was very different – the upright session left me feeling nauseous and faint where the recumbent one left me a bit sore but after a short break, I felt able to match it again.

Exactly not what you’d expect if the power output was actually the same in both positions, but the recumbent frame was losing it to friction ‘upstream’ of the PowerTap.

I’ll make an effort to do proper 20 minute max efforts on both platforms and keep an article update with how the figures change and (hopefully) converge as I start to assume better form.

Upright stationary trainer interval – 12.5 minute best effort

Edit: I’ve added a screen grab from Golden Cheetah to the one from Ascent (I had to crop the latter quite aggressively to get it to fit):

recumbent-power-comparison-new2

gc_df

Recumbent stationary trainer interval – 12.5 minute best effort

Edit: I’ve added a screen grab from Golden Cheetah to the one from Ascent (I had to crop the latter quite aggressively to get it to fit):

recumbent-power-comparison-new1
gc_bent

Food for thought! I’d love to hear any thoughts in the comment section below:

See also discussion on this post in BROL.

28 Comments

  1. Robert

    What are the green and brown lines? Can you show how your cadence and crank torque compare?

  2. Dave

    Hi Robert,

    I’ve added screen grabs from Golden Cheetah which are a bit more useful! Also a combined torque/cadence scatter plot.

    thanks,

  3. Robert

    Thanks for adding the plots. Is it possible that you have the upright and recumbent plots switched? It looks like the plot for the upright has a long steady effort at 200 watts, while for the recumbent it has a long steady effort at 250. The torque-cadence plots look quite different. Were you altering the resistance by shifting gears or changing the resistance setting on the trainer?

  4. Dave

    Thanks, as you say the plots were uploaded the wrong way around… oops!

    The resistance is controlled by changing gears. Although the same wheel and tyre are being used, I think the clamping force is relatively poorly controlled (it’s just a big handle you flip to press the roller into the wheel, like a sort of QR). I’m not sure that is significant however, as it’s ‘downstream’ of the PowerTap.

  5. Dave

    Robert, I don’t suppose you have any thoughts on better ways to look at the ‘input’ side of the power equation, other than using an HR strap?

    I’m planning to redo this for a longer interval, perhaps 20 mins, with HR (the aim of these sessions wasn’t to compare the platforms at all, I only realised it would make an interesting article afterwards).

    I’m pretty sure it will show lower HR for the lower power recumbent session, but I’m not actually sure how meaningful that is. For instance, say it’s easier for my heart to pump blood to my legs in a horizontal position (higher veneous return) – a lower HR might be possible at equal power.

    On the other hand, it could be that the leg muscles have evolved to assist in circulation through muscle flex in the vertical position and lying back would actually force the heart to beat harder. I don’t really know how to control for this (short of oxygen tests in a lab?)

  6. Robert

    I hesitate to suggest this but I think the best way to get an idea of the “true” difference in power production is to do a ramp or MAP test. I hesitate because you’ll need to do it twice, once on each platform, and there’s a small “training” effect so you can get slightly better after you’ve done it once or twice and know what to expect. Anyway, the MAP test is a pretty standard way to evaluate aerobic power (you’re probably not that interested in anaerobic neuromuscular power anyway) — it’s what the BCF uses. Here’s a quick write-up.
    http://alex-cycle.blogspot.com/2006/12/map-testing-where-failure-is-success.html

    I wouldn’t do this on consecutive days. It can be a tad unpleasant. You’ll need a good fan for convective cooling. It helps to have an assistant providing sufficient motivation (i.e., screaming at you) for the last step of the ramp, and to have a bucket ready should you throw up. If you look around the interwebs you can find calculators that will estimate your VO2Max from the MAP test results.

    I find it interesting that your cadence-crank torque pattern looks so different. I can see where you shifted gears on the recumbent trainer ride. That plot tells me that you achieved higher ending power on the DF using a combination of higher cadence and higher crank torque. I sort of expected that for the final “sprint” but it appears to apply for the longer period at 200 vs. 250 watts, too.

    In the MAP test, you can shift gears however and whenever you want in order to maintain power for each step. I’d use a 25 watt per minute ramp — it makes the suffering end sooner.

  7. Robert

    On re-reading my last post I realize I may have been a bit cryptic about the cadence-torque comparison. The resistance of a stationary trainer, once it reaches operating temperature, is mostly determined by the roller’s speed so, generally, a speed of X km/h will correspond to a power demand of Y watts, and vice versa. But there are different combinations of gearing that you can choose to produce X km/h, and I’m interested in whether in the controlled environment of a stationary trainer you freely choose to use the same cadence and gearing to produce 30 km/h on the recumbent and on the DF. Although you didn’t do a direct comparison at equivalent speeds in your test, an eyeball comparison makes it sort of look like you *didn’t* choose to use the same cadences and gear ratios.

  8. Dave

    Yes, higher cadence on the recumbent didn’t feel as comfortable and/or stable. I was happier dropping it down a bit.

    However, that might be an artefact of the bike not being as well balanced on the trainer and it will be interesting to look at that again with a longer set of intervals. (It would be interesting if it was a genuine effect of the riding position).

  9. john scarlett

    interesting stuff – but which recumbent?
    seems there are many more variations within recumbent design than with uprights
    would you get different results with a different recumbent- more upright seat vs more laid back ,relative height of bb to seat, front or rear drive etc..?
    not to mention frame materials,flexing and how many somersaults the chain has to do before it reaches the drive wheel.
    my unscientific observation is that the two more upright seated recumbents i know –a P 38 and the streetmachine are much better at climbing than the more laid back fuego.
    in fact with the uss on the street machine and its great stability, I can inelegantly raise my butt out of the seat to get extra thrust from some of my body weight on to the pedals -but not fit enough to sustain that for long. I am trying to develop the muscles and stamina to pull on the pedals as much as i push – something it is easier to do on a recumbent than an upright and could help to close the power gap?
    Actually with your impressive power output ,I think you should be permantly attached to the national grid!

  10. Dave

    Hi John,

    This was on the High Baron. I agree that body angle especially is likely to play a big part in this.

  11. Alan Russell

    Hi Dave,
    Thank you very much for your truly fascinating and though-provoking article. It confirms what I have experienced in my three years of riding an ICE Sprint 2FS recumbent tricycle; my legs seem to be the limiter on the recumbent, but on my upright Orange P7 bike the limit seems to be the rest of me. And while I’m much quicker uphill on the upright, I’m greatly faster on the flat and more than twice as fast downhill, in the same conditions on the recumbent.
    Thanks again,
    Alan.

  12. Dave

    No problem Alan, glad you found it interesting.

    I’m looking forward (sort of!) to doing some longer sessions with an HRM to build on this…

  13. Nick F

    Another interesting article on a subject close to my own heart. Not having the cash to invest in a powertap myself, I’ve been waiting for someone else to do this kind of experiment for me! Well done Dave.

    Some off-the-cuff comments.

    Firstly, I am personally also convinced that recumbent performance is limited by physiological factors due to the rider’s position, rather than any mechanical inefficiences in the machine. An easy way to test the possible criticism that your lower power output on the recumbent may be due to losses somewhere in the drivetrain (or the frame) is to use a powermeter at both the driven hub and the cranks; that way Pin and Pout can be measured and compared for both recumbent and upright machines. You either need to have even more cash for this :-), or to beg/borrow a pedal or crank-based powermeter from someone. I’m really amazed that nobody with access to both types of meter hasn’t already done this.

    I also rely on HR readings as a gauge to how much effort I’m putting in, but I’ve read that this doesn’t necessarily correlate well with power output. (Can anyone else comment on this?)

    Secondly, as I have commented before, I have personally found HUGE performance variability across hill climb trials from one day to the next, both on upright and recumbent bikes. This variance is large enough to confound any genuine performance difference between the two machines. Either I am a particularly inconsistent performer, or many data samples are needed to extract the true “signal” from the noise. In terms of sample size I see little point in combining data from different riders, however, since now you are merely adding additional variance from the difference in physiological capabilities of different riders. What is needed is many more samples of comparative test data from yourself; sorry Dave, you’re just going to have to do more work!! Additional test data from other riders (i.e. comparing their own upright/recumbent powers) would then add generality to any trend that you find for yourself, but combining (i.e mixing) the data sets would introduce statistical flaws since the experimental conditions would not be identical. I’d be happy to volunteer, if someone would lend me a powertap!

    Third point; I agree with Robert’s comment that you should consider a more “objective” method of determining max power delivery. When using the classic “30min time trial” method to determine my LT threshold, I too found it difficult to judge the appropriate effort and found myself putting a spurt in at the end. 12mins or so is a long time to sustain a genuine max power level. I would suggest using a much shorter time window (after a consistent warm up period) so that you are genuinely going “flat out” during the measurement window. Even here, knowing when you are going flat out is very tricky; that seems to be why it’s often considered to do these kinds of performance tests on the road with other, stronger riders to give you the competitive incitement to really push to your limits. Watching a video of professional cyclists on a climb might act as a substitute if you test on a home trainer.

    All good stuff. Keep it up!

  14. Dave

    Great comment Nick, thanks!

    The problem with short efforts is that they are even less representative of how people actually ride than my 12 minutes was (I freely admit that I chose 12 minutes at random, by the way!)

    The difficulty as you say is determining a sub-maximal limiting factor that is equivalent between bikes, so they can be compared. Heart rate is, apparently, indicative only – although I intend to do some equalised-HR trials they won’t be the last word for this reason.

    There is a local lab that will do gas analysis for submaximal efforts but it would be very expensive to have both positions tested this way, especially as a series would be needed to track my acclimatisation if this was to have any real value.

    I’d love to borrow a second power meter at the front end, they are not easy to come by unfortunately.

    I’ll keep chipping away at this from different angles though. Please keep dropping by and let me know any ideas that may come to you!

  15. Dave

    On the gas analysis front: I also suspect it would show something reasonably obvious: the muscles themselves are roughly equivalent in efficiency (so X% oxygen take up corresponds to Y watts in either position) it’s just that maximum oxygen uptake is capped (glass ceiling) in the recumbent position by the inability to fully recruit the muscles.

    I really don’t know how you could isolate and study that. It would need electrodes and be beyond my reach, that’s for sure…

  16. Robert

    NickF: in the DF world, lots of people have put a power meter on both “ends” of the drivetrain to measure mechanical losses, but either no one in the recumbent world has done this or else they just aren’t willing to share the information. (There’s actually a way to measure drivetrain losses with just a single crank-based power meter, without a hub-based power meter at the other end, but it’s a lot of trouble.)

    Instantaneous correlation between HR and power is poor but if you smooth over an appropriate period of time and do a couple of other tricks you can improve the fit — however, you still have to deal with heart rate drift, plus environmental conditions. Basically, what you really want is not heart rate but rather cardiac output — but HRM’s can’t measure that since stroke volume isn’t constant. That makes heart rate a poor proxy for cardiac output and, thus, a poor control for power comparisons.

    The results on comparisons of DF and recumbent power are mixed:
    http://www.ncbi.nlm.nih.gov/pubmed/2289491
    http://www.ncbi.nlm.nih.gov/pubmed/12423185
    http://www.ncbi.nlm.nih.gov/pubmed/16222538
    http://www.ncbi.nlm.nih.gov/pubmed/20386919
    http://www.ncbi.nlm.nih.gov/pubmed/23135372

  17. Nick F

    Hi Robert. I’d be very interested to hear more on “the couple of other tricks” to improve the correlation of HR with power output over a reasonable time period. Are we talking of the order of 30mins here, as in a LT threshold TT test?

    I agree that much academic research is ambiguous on the DF/recumbent power issue. But partly this is because there is no standard for “the recumbent position”. Many articles even compare uprights with relaxed seated positions – as in the P38 or GoldRush – which they classify as “recumbent”. Most modern, sporty, SWB, European bents have much more reclined riding positions than these.

    Even the paradigms they use for their comparisons are very varied. TBH I feel most of these papers are churned out by the need for academic researchers to publish in quantity to justify their research grants, and don’t really focus on the issues that recumbent riders and builders need to understand to bridge the performance gap with DFs. Is it the body angle per se, or the height of the heart in comparison to the feet, or the more open body angle eliciting different muscles? Is it the muscles not getting enough fuel quickly enough, or not getting rid of the lactic acid quickly enough? Such academic papers generally don’t address these questions in a form that is genuinely useful to pragmatic progress.

    My first home build was a FWD LowRacer with a very high BB and fairly upright seat back, designed to give the rider a very closed body angle. In fact, I tried to replicate the same rider posture as on a DF bike, but merely rotate it backwards onto my LowRacer. I loved that bike, but not only did it give me burning aches above/inside the knees when pushing hard on climbs, it is still the slowest climber compared to my RWD Metaphysic or DF bike. OK it’s heavier by around 3kg, but this is insufficient to explain the performance differences. I would love to investigate why and where performance is lost in this bike, but not having a powertap reduces me to trying to draw conclusions from comparisons over repeated climbs at full effort of my local test hill.

  18. Rob

    My gut feel is that a good chunk of the power loss comes from not being able to open the airway very effectively, limiting the amount of oxygen you can suck in over a given period. Ideally you need to have the head tilted back – pretty similar to the ‘out of the saddle’ position, and the opposite to the recumbent ‘chin down’ position.

  19. Nick F

    Rob, I have to respectfully disagree with your gut feeling for the following reason. I personally find – and I think it is generally the case – that recumbent performance is limited by lack of power in the leg muscles, while heart rate remains well below maximum exertion levels. This suggests to me (but maybe I’m wrong) that I am NOT limited by lack of oxygen intake to the lungs. If I was in oxygen debt, I would expect to have an elevated heart rate and laboured breathing in an attempt to suck in more air.

    Note that this is not the same thing as claiming that my muscles are necessarily receiving sufficient oxygen. That I can’t say. Maybe they aren’t, and that is why their effort is limited. But what I would claim is that, if this is the case, the issue with oxygen supply is “downstream” of the lungs, and perhaps due to some physiological limitations imposed by the recumbent posture.

    My understanding is that posture-related restriction of the air supply is more influenced by constriction of the diaphragm (as in an extreme aero TT position on a DF bike) rather than constriction at the throat. But – as always – I’m open to correction.

  20. Marcus Cohn

    Can you comment on why recumbents were banned from velodromes in the mid 1920s after they began winning all the races against conventional designed bikes? This seems to fly in the face of all the analysis suggesting recumbents suffer some inherent mechanical disadvantage. Lastly, I think the world speed record for a two wheel bike was set by a recumbent – how does that figure into the data set ?

  21. Dave

    Hi Marcus,

    Chris Boardman’s upright hour record has only just fallen to Aurelien Bonneteau.

    It took over sixteen years for anyone to ride a recumbent faster than Boardman rode his upright… and that’s under conditions which are absolutely perfect for recumbent performance – indoor boards.

    We know that if you put an electric wheel on a DF and on a recumbent at the same power output, the recumbent will fly ahead. Since this doesn’t happen in real life, what else is there to look for than some issue getting that power to the road?

    The fact that the 20’s were almost a hundred years ago makes it doubly difficult to separate men, machines, and achievements. My own understanding is that the ban was more down to image and traditional manufacturer pressure than unfair performance per se.

    thanks,

  22. Saharayana

    Dave, your data are very interesting indeed. How much do you weight and how old are you ? how many miles do you ride per year? interesting enough your performance on Upright more or less my performances on my recumbent. I can sustain 270Watt on 10mn and 350 Watts over 1mn, 3000 watt over 4mn and 250 over 20mn. I ma 53 year old and weight 70Kgs.I ride 5000miles a years. You loose 20% of power on a 10mn effort when you compare Upright and Recumbent, that is about what other people report, but I think that we appropriate training you can squeeze that to about 10%, in the best case, more realistically to 15%. Look at the data that SesameCrunch put on BROL lately and he reports 29% difference between Upright and Recumbent effort on a 10mn duration effort. But he measure it in a very different way. The interesting thing is that for effort more than 90mn the max effort tend to match and it shows that recumbent are first and foremost bikes for long distances rather than hill race. The problem is that for short duration effort we involve less muscles and that the blood pressure in our legs is less on recumbent therefore limiting the intensity of our maximum aerobic effort. If you want to improve that you have to focus on interval training aiming at improving VO2 max, 4mn at 250Watts (for you) effort with 3mn recovery time in between, repeat 5 times 2 times a week. Very painful but result guaranteed. But after 2 months and half you will need a break, because you will accumulate fatigue.

  23. BG

    I think Marcus has missed out that you are only measuring the power you can deliver to the wheel, not the speed of the bike. As you pointed out, LBs are faster because of aero advantages.
    While I reckon it would be interesting to investigate possible power losses in the drivetrain, that’s a separate issue to investigating the biomechanical limitations of DF vs LB. To be fair, you would need to select a period of time to train/prepare on one type of bike, do repeats; then compare with the other bike after a similar time to prepare. This may still not be a fair comparison as DF bike may well use more of the muscles we use for walking/standing than LBs do.
    Once you’ve done that, you’ll need to make the comparison for ordinaries and belly bikes to complete the set …

  24. BG

    Also, doesn’t this agree with tests that Alex Moulton did?

  25. Mike Croker

    “A 25% drop in power means that on a climb the upright would pull away one mile for every four miles the recumbent rider travels….”
    Er, dropping from 275W to 220W is a reduction of 55W, or 20% 😉
    I tried plugging these numbers into Kreuzotter’s on-line calculator (http://www.hembrow.eu/personal/kreuzotter/espeed.htm), just to see what it made of the power loss. To my surprise, with a 12% gradient a road bike will travel at 9.5km/h vs 7.3km/h for a ‘race equipped SWB’ – a 23% speed reduction. However, once the grade reduces to 5%, the ‘bent closes the gap a bit as aerodynamics becomes more significant: 19.4 vs 15.6 km/hr, just under 20% reduction.
    Of course there’s no highracer option in Kreuzotter, and the SWB’s rolling resistance is 15% higher than the road bike, but even using radial ply tyres doesn’t help that much!
    Thanks for the good work, Dave :)

  26. Steve

    Further to Mike’s reply directly above, you’ll recall that power required increases with the square of your speed.

  27. Hi. You may be interested in this new hypothesis to explain the loss of power in recumbent bikes: https://chemamancebo.wordpress.com/2017/02/28/perdida-de-potencia-en-el-pedaleo-reclinado-sobre-la-componente-gravitatoria/

  28. Interesting study. I thought I might add a couple of comments to clear up a couple of points of confusion in the comments:

    (1) Power is (roughly) proportional to velocity cubed. (Force is proportional to velocity squared, and power = force x velocity)
    (2) The effect of gravity on the legs is net zero over a complete pedal revolution, so this cannot explain the difference in recumbent vs upright power output.

    Having (fairly) recently bought a recumbent bicycle, my experience aligns with Dave’s that power output in the recumbent position is limited by leg muscles. I come from (historically) a highly competitive DF TT background and (more recently) a recreational running and DF cycling background.

    Riding on the recumbent, my vastus medialis (and to a lesser extent my vastus intermedius) were burning at 20mph and I was barely breathing hard. However, I was really struggling to perceptibly recruit my rectus femoris or my gluteal muscles.

    My suspicion is that this is related to hip angle. The hip angle on a DF will average somewhere around 90deg, whereas on the recumblent my hip angle is much more open than this (I would guess around 45deg). I think that this hinders recruitment of the glutes* which is one of the keys to going fast on a DF bike as the glutes are pretty much the biggest muscle group in the body.

    * Anecdotally, this also correlates with my experiences of long distance trail running – on really steep uphills (>40deg), I instinctively end up leaning forwards to close the hip angle, and recruit the glutes.

    It seems obvious that spreading (say) 300W of power production across glutes and quads on a DF isn’t too hard. Spreading 300W across just two of the quad muscles is going to place heavy demands upon those two muscles. Ultimately, just two of the quadriceps muscles are never going to be able to produce as much power as quads and glutes combined. However, we can probably train them to produce enough power to tax our cardiovascular system. This is probably why it takes time to develop “recumbent legs” when transitioning from DF cycling (not that I have ridden enough to get there yet!)

    There is probably an interesting study to be done into the trade off between power production and aerodynamic efficiency…

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>