Hub Generators; reactive loading puts new slant on efficiency claims.

[Brucey]

some folk may remember me muttering about how efficiency tests on hub dynamos have typically been carried out using a purely resistive load (as per Stvzo testing) but that real lights, battery chargers etc are very often a different 'reactive' kind of load (in which the voltage and current are no longer in phase with one another) and this alters both the power output and the efficiency of a typical hub generator. [ I have also mentioned this long ago in relation to 'resonance' effects, whereby at a certain speed many hub generators become noticeably noisier and may produce a vibration that you can actually feel; this varies with the generator (obviously) but also the characteristics of the load.]

The good news is that most nominally 3W rated generators can be persuaded to produce a lot more than 3W. The downside is that this increased power output is rarely allied to the highest levels of efficiency. Unsurprisingly different generators react differently to different loads, so some generators show well with some lights connected and not so well with others.

Some of these tests are summarised here;

https://www.cyclingabout.com/dynamo-hub-power-drag-testing-schmidt-son-shutter-precision-shimano/

[edit; with data (and reprepared figures etc ) from this site https://translate.google.com/translate?hl=&sl=ru&tl=en&u=https%3A%2F%2Fskjegg.blogspot.com%2F2020%2F02%2Fdynamo-hubs-for-your-pleasure.html]

It is not a simple set of tests but the results show that (of the hubs tested) the SON28 was the most efficient when connected to a typical load, (and in most cases not just by a little bit either).

With some high power lights, the drag exceeds the power output of the generator by up to 15W, i.e. the lamp might be using 9W but the generator is dragging nearly 25W. You will notice this!

However if you are running a simpler, low power light (they used a 40lux Spanninga 'Nomad' model to represent this type ) then it is closer to a purely resistive load (as per Stvzo tests) and there is less difference in efficiency between hubs (in absolute terms).

For example with the B&M IQ-X as a load;



the lamp draws ~4-5W @ 15kph but incurs a 'drag overhead' of between ~6W and ~10W depending on which hub you use.

or the Spanninga as a load;



the lamp power (and light output) is less than half that of the IQ-X but the 'drag overhead' is between 2W and 5W, which is both less in absolute terms and less proportionately too.

Similar variations are seen in using chargers when paired with different hub generators.

Overall, the results are complicated but there are a few takeaways;

1) If you don't really need a super-bright light and are worried about drag, you can greatly reduce the total drag and/or the 'drag overhead' incurred.

2) there are some unexpected results; for example if you ride at ~25kph and use an IQ-X lamp there is no advantage to using a SON28 hub over the cheapest model on test, the shimano DH-3D32. Both are (at this speed and load) substantially more efficient than the other hubs on test.

So it pays to consider what your needs really are before choosing a hub and lamp combination. It is as well to note that this data was all generated with standard lamps; running two (lower power) lamps in series may be a substantially more efficient way of generating more light at speed, when you need it (in the simplest terms because generator efficiency is more closely related to the current drawn rather than the voltage), so there may be more life left yet in the old trick of running more than one headlamp at higher speeds.

All very, er, illuminating!

cheers

[hamster]

As most LED drivers present themselves as a load with leading power factor there is some effect of PF correction. It would suggest that there is scope for significant optimisation!

[Brucey]

yep, I reckon.

cheers

[Greystoke]

I've used power faction correction for years running twin halogen lamps on both bottle and hub dynamos....still do now.
I recall that there was even a secondary light from a big manufacturer that had a capacitor in it as standard.

[bgnukem]

Interesting, the results certainly accord with my experience of the SP 'hub as putting out less power at lower speeds vs. the Shimano 'hubs I've used (3N71 and 3N80), as even with 26" wheels I find I need to be riding at 6-7mph before my front light stops flickering.

The drag is also fairly high at higher speeds, I run the 'hub and lights on all the time in winter and notice the difference in summer when it is removed, at typical riding speeds of 15-20mph.

I wonder how the better quality Shimano 'hubs perform, as the basic 3D32 is a pretty heavy lump....

[hamster]

I've used power faction correction for years running twin halogen lamps on both bottle and hub dynamos....still do now.
I recall that there was even a secondary light from a big manufacturer that had a capacitor in it as standard.

I think the capacitor was for standlight function, but of course could help with power factor. It would also require rectification to avoid shorting the capacitor through the static dynamo.

[Greystoke]

It was a Schmidt e6 and the secondary light wired in series used a capacitor to enable the second light to illuminate at lower speed.
I worked out the capacitor value probably 25-30 years ago for a bottle dynamo and it worked well. A few years ago I went to a hub dynamo and changed the capacitor to suit.....different number of poles so frequency changed.
Calculations are in my desk at work somewhere but I'm working from home so we're out of luck.

[Brucey]

...I wonder how the better quality Shimano 'hubs perform, as the basic 3D32 is a pretty heavy lump....

there are only four hubs in that test but one is DH-UR700-3D, which is shimano's current 'XT' ish level offering.

cheers

[pwa]

Does all this mean that when people say this or that generator / light combo only costs your legs 5 watts at whatever speed, they are understating it by a significant margin?

[Brucey]

Does all this mean that when people say this or that generator / light combo only costs your legs 5 watts at whatever speed, they are understating it by a significant margin?

if that lamp constitutes a reactive load (and most of the posher ones do, these days) then the answer is "yes".

Previous notions of drag and efficiency have mostly been based on data which used a simple resistive load to measure drag vs electrical power output. Such tests represent tungsten bulb loads reasonably well, but not much that is more modern and fancy than that.

cheers

[iandusud]

How does the 3D32 used in test compare with the 3N72 in terms of the electrickery? Is it the same generator?

[Brucey]

How does the 3D32 used in test compare with the 3N72 in terms of the electrickery? Is it the same generator?

different I think; however both have an 'aluminium coil structure' according to shimano (I'm reasonably sure DH-3N72 used to have copper windings; presumably they have revised it; it used to have an earth return but for the last few years it has had isolated outputs, so it definitely has been changed in some respects) . The shimano 'sport' series generators (such as DH-3N72) seem to have posher (smaller, lighter) magnets than '30' series generators which are accordingly heavier and bulkier as a consequence. The overall electrical performance of '30' series generators has usually been very similar to contemporaneous shimano 'sport' series generators at any given time. However there are so many different shimano models that keeping track of which are exactly the same as others is, I'm afraid, beyond me. When they change what is inside any given hub generator without changing the model number, it makes it virtually impossible!

cheers

[iandusud]

Thanks Brucey, I was expecting something like that!

[Brucey]

I found these (older) graphs (which I presume date from the time of the Edelux 1, and possibly the older model SON28)





this data suggests that the Edelux 1 isn't a 'difficult load' to drive and/or that it reacts in a similar way when driven with a DH-3N30 or a SON; the variation of power with speed seems similar in each case, and is similar to data from tests with these hubs taken using a purely resistive load.

[edit; the Edelux II produces a beam that is both much wider and brighter than the Edelux I. To my mind (in the absence of a step change of efficiency and/or reflector design) this can only be achieved by using more power and I believe that the Edelux II is likely to present the same kind of load to a generator as a Cyo premium or an IQ-X does. ]

[edit II; the above data suggests that the electrical performance (when driving that particular load) of the 3N30 generator is pretty well identical to that of the SON28. You can see this because the difference in power required to drive the two generators is the same whether there is a load or not. This means that (in this case) the difference in drag is exclusively in the bearings and seals. As tested there would have been no service loads on the bearings, only bearing preload (arising from how the hub was set up internally) and seal drag.

Now, every single new Shimano generator I have examined has had ridiculously high bearing preload from the factory. Typically with a QR type hub the LH cone can be backed off 1/4 turn and there is still no free play with the wheel in the bike with the QR tightened. This probably translates into several hundred kg of bearing preload, if you don't fettle a new hub. The seal lips have been pretty dry too. In somewhat ad-hoc tests (timing how long a spun wheel takes to run down) I found I was able to reduce the parasitic drag in a new DH-3N72 generator by at least 1W, by simply adjusting the bearings and lubricating the (double in this case) seal lips better. In this case provided the load was a 'simple' (largely non reactive) one I'd expect the actual difference between the SON28 hub and the shimano one to be so small you would struggle to measure it.]

cheers

[Brucey]

re running lamps in series and how it might affect efficiency; Andreas Oehler has addressed this here

https://fahrradzukunft.de/14/neue-nabendynamos-im-test/

albeit using a 24 ohm resistive load (rather than a reactive one). However the effect was to significantly increase efficiency. eg with a SONdelux at ~20kph, almost 4W electrical output requires only 5W mechanical input, which is a significant improvement over the power output/efficiency into a 12ohm load.

cheers