Schlumpf Hub Efficiency

A specific testing setup and equipment would be nice, but I think that with a power meter and some good testing protocols, it should be possible to get some decent efficiency numbers.

A recent test of mine was to ride a hill repeatedly, alternating between high and low gear, and different speeds/power. I estimated a bunch of parameters, knowing low gear should be approximately 100% efficient.

Surprisingly, I think the conclusion is decent! 4% efficiency difference between high and low gear.

Few things I think I can improve:

  1. Disc brake rubbing. I’m using brakefast, and it was near impossible to true the disc since the brakefast brake arms are so stiff. I think I will try shimming the disc rotor later.
  2. Wind conditions varied. Strava reported 7.8 kph, nearly perfect headwind direction. But, I have no idea if this was accurate. Choosing a perfectly calm day would be best.
  3. Hill is short, only about 27m. The strava segment says 30m, but my data points to it being closer to 27m. Unfortunately, this is the biggest “hill” I can get to within an hours drive. Thus, my method of using the strava segment matching may cause some inaccuracies.
  4. Other coefficients are just my best guesses (Crr, CdA, etc). Some sort of algorithm to estimate these based on my data might be useful.

Would love to hear other peoples’ experiences and insights as well.

I dipped my hub internals in ATF recently. Whether that increased or decreased or made no difference to efficiency, is part of the reason why I’m trying to make a test protocol that I can carry out.


Interesting! The results seem to similar to those efficiencies in bike gearing.
Here is one study where they reported:
The efficiency of the Shimano Alfine ranged from 90.4% to 96.6%. The efficiency of the Rohloff speed hub ranged from 95.8% to 99.5%. The efficiency of the Sturmey Archer hub ranged from 84.6% to 99.8%. The efficiency of the single speed chain drive was found to be 99.71% and the belt drive 98.0%. The efficiency of the 7-speed derailleur ranged from 97.7% to 99.4%.

I wonder which of these is closest to the Schlumpf?

You made a mistake. Lap 13 was very slow for high gear and lap 14 was very fast for low gear. Did you invert the data? If they are ok you had an high gear lap at high efficiency confusing all data table and a low gear lap quite inefficient thus making gear changing a bit non-sense

I couldn’t find the full paper for that, but looking at this one:
I think the 3 speed hubs are most likely similar, where they have high gear around 1.3

At around 200W, the efficiencies I see from the paper are around 92% to 94%, and a normal derailer system is about 94% to 95%. They authors noted that the 94% efficient geared hubs used a light oil instead of grease though. Their conclusion was geared hubs are approximately 2% less efficient.

If the schulmpf is actually 4% with light oil (ATF), that’s a bit unfortunate. But I probaby will want more data before trusting my numbers.

What I did was (H = high, L = low gear)
Lap 1-4 (H,L,H,L): normal speed x4
Lap 3-5 (H,L,H): Fast, slow, slow
Lap 6-8 (L,H,L): Fast, slow, slow
Lap 9-11 (H,L,H): Fast, slow, slow
Lap 12-14 (L,H,L): Fast, slow, slow

No mistake on that part. Lap 13 was very slow for high gear, the most bottom left blue dot on the graph. I did some super slow on both high and low gear so I can get a better spread of data across different power levels.
Yes, lap 11(high gear, high power) registers as very high efficiency, while lap 10 (low gear, low power) is very low efficiency.
I would expect low gear to be a flatter line across power levels (aka, efficiency has less dependence on power). There should still be some, such as bearing losses. I am wondering if it is due to brake rub, but no idea…
High gear, I do expect efficiency to increase at higher power, which matches the trend line.

The amount of noise is unfortunate though…
I tweaked my guesses for some variables, and this more represents what I expect. Low gear approximately 100%, with some noise.
High gear is less efficent by 5% actually.

Here is another good plot that just shows low gear is more mechanically efficient. This graph isn’t as truthful though, since it doesn’t take into account my various starting and ending speed, which I do account for in the efficiency calculations.

1 Like

I don’t trust my number very much anymore (4.5% average efficiency difference). And wondering if this protocol is doomed.

7 samples for each high and low gear.

Assuming a normal distribution, looking at just the mean and standard deviation of ones set, 80% confidence interval is +/- 0.7%.

So added together, about +/-1.5%

So my range is 3% to 6% efficiency loss with 80% confidence.
2.3% to 6.7% efficiency with 95% confidence.

To get to +/-0.5% with 95% confidence, I would need 150 laps each, which is an everesting.
80% confidence would still require 60 laps each; 3600m of climbing and descending :joy:

Not impossible, but I would need to be careful and probably enter different wind values, air density, and weight (as i sweat and eat and refill supplies) throughout the day. Though, at this point, maybe I could probably just do it in multiple rides too.

Today’s test condition was much windier unfortunately.
Nearby weather station said 10mph headwind, and the end of the ride was 9mph cross-headwind.

I tried to do more runs around 150-250W.
Again, I alternated high gear and low gear every other lap.
But I tried to do around 180W for two laps, then 250W for two laps. Though, in actuality, I was trying to maintain constant speed throughout the run, and the average power just turned out to be like that.

Deleting lap 5 from test 1, since it is a bit of an outlier:

Around 200-250W, seems like 6% to 5% losses.

No idea what exactly made high gear run 1-2% lower efficiency in today’s test…

1 Like

Was there any difference on ambient temperature? Or maybe the higher wind kept the geared hub colder and resulted in a lower efficiency.

Oh yes… it was 32°C for test 2, compared to 22°C for test 1.
Whether cooler or hotter is better for efficiency though, I can’t imagine 10°C difference to be result in a measurable efficiency difference.

One other thought is that it is dry inside, and I need more lubricant.
Only thing I can do is do more tests!

I think I will try 2 more tests, then try adding lubricant, to see if there is a difference before and after.

Also, I did a ride between the two tests where I fixed my brake rub, but I think it was back for test 2…
There might have also been spoke sounds, mixed in. 99% sure my right side spokes were going slack… Measuring 42kgf on some.

55 runs over 4 rides done!

Ignore the absolute efficiency numbers. The relative efficiency is what I’m testing.

I estimated wind speed and other parameters for each low gear lap such that efficiency is just around 100%. The wind speed used to calculate efficiency of the high gear run is the average of the adjacent low speed runs’ wind.
i.e. choose
Lap 2 low gear = 3 kph wind
Lap 4 low gear = 4 kph wind
Then, I use
Lap 3 high gear = 3.5 kph wind

Efficiency difference is around 5.4% for this range of input power. I think with 55 samples, it is “enough” for this conclusion.

There also seems to be a trend of efficiency improving at higher power.
Looking at the difference between watts lost, I get a fixed 3.1 watts loss + 4% power dependent efficiency loss.
However, it might be risky to extrapolate down to very low powers, since all my data is at greater than 150W, and just by looking at this graph, it seems way too noisy.

Deleting the “outlier” where I did 360W, I get a fixed 9.0 watts loss + 1.0% power dependent efficiency loss