In reply to Robert Durran:
> I'm not doubting the data, but that graph showing the braking force dramatically flattening off (it's not actually a decrease) for bigger loads does seem totally counterintuitive to me. The geometry/mechanism of the Megajul suggests that the braking force should go on increasing with load. I'd be interested in an explanation!
I think the geometry suggests the opposite conclusion. (Jim will correct me about this if I've got it wrong...) The pinching effect that makes the assisted-braking devices "lock" (when they do) requires that the device carabiner travel further up towards the top of the device than is usual for standard tubes, in order to literally pinch off the space the rope travels through. This is most apparent in devices like the Metolius BRD and the Mega- and micro Juls, in which a slot is cut into the body of the tube to allow the carabiner to travel further upwards.
An effect of the proximity of the device carabiner to the top of the device is that the angles the rope makes going through the device are decreased. Friction being exponential with the wrapping angle, a small decrease in angle could still, in principle, have a significant effect, decreasing the force-multiplying power of the device, viewed simply as a tube.
Jim's graphs make it evident that the assisted lockers have little in the way of force-multiplying power; they rely on the physical pinching of the rope in an opening that ins constricted under load by the motion of the device carabiner. Jim says (which is to say his data suggests) that the pinching effect does not scale with grip strength but rather is nearly constant as the load increases, and of course this more-or-less constant contribution becomes a smaller and smaller fraction of the total load as that load increases.
At high loads with thin ropes, I think it possible another effect could be at work, which is that as the rope stretches, its diameter decreases, and this could actually decrease the pinching contribution if carabiner reaches the upper limit of its travel and the rope keeps stretching.
I think our intuitions ( I should just speak for myself here so will change pronouns) have been shaped by the performance of standard tubes that act fairly consistently as force multpliers throughout their loading range. I think (incorrectly) that if I have a lot of braking power at relatively low loads, then the force-multiplying effect of the device will scale linearly from the initial high value, but that isn't what happens, as explained above, in Jim's post, and in his graphs.
Practically speaking, it seems to me that the assisted-locking devices will be fine for single-pitch belaying, where the possibility of high fall-factors is limited by the presence of the ground. The real question about assisted-braking device use comes in multipitch context, and even there we are speaking of the rare situations in which the leader falls well past the belay. These are the occasions in which the assisted-braking devices seem to perform poorly compared to, say gold-standard ATC-XP. If one is going to use these devices for such climbs, I think at the very least one ought to be gloved, and if the ropes are at the thin end of the range, I would be especially wary about the effectiveness of the MegaJul.
A number of people have mentioned the convenience of locking during rappels. Once again, extreme caution is advisable, because the locking capabilities may depend to some extent on the rope weight below the device, possibly making the device least likely to lock when the rappeller is near the end of the ropes. I'd suggest some very careful testing of the device---with the exact carabiner and thinnest ropes you intend to use---in a safe situation.
Speaking of carabiners, one of the most unfortunate features of the assisted-locking devices is the variation in their performance depending on the rope-carabiner combination. Only the Alpine Up addresses this issue by including an appropriate (should we say mandated?) carabiner with the device. For the rest, performance will depend to some extend on the carabiner type, and from what I've heard from people who have tried, this is not only a question of the carabiner cross-section but may even be influenced by the amount of curve in the carabiner basket.