The other day, I realized that my mountain bike (cheap Raleigh Canada 2000) came with a 135 mm wheel crammed into 130mm dropouts - which explains why wheel mounting was always a wrestiling match.

I read Sheldon on coldsetting, and the lumber-through-the frame, bend-one-side-then-the-other method seemed dodgy. I came up with something much simpler, which loads the sides against each other to ensure symmetry. Tried it this morning, and so far as I can tell it worked great.

I should mention that I actually am a physicist (MIT 77), and also the inventor of a sometime-world-record pogo stick (the Flybar - 8.5 foot bounce, powered by a pound of rubber), so my confidence in some pretty commonsensical premises does have an explanation other than sheer arrogance.

The method is: remove the rear wheel, and put an axle in the dropouts - with nuts on both sides, inside the dropouts. Move one of the nuts outward to load the frame. Back off and check the effect from time to time; continue until desired spreading is achieved.

(I put a couple of washers outside the nut I moved, with oil, to minimize friction.)

I didn't know how far I would have to go to exceed the elastic limit and get permanent spreading; turns out that was at about 150mm. To get a 135mm set, I had to go to 159mm. It was very easy - no large forces involved, gently winding the wrench.

The beauty of this is that Hooke's Law (which is simply that strain varies as stress, and is approximately true for most things), and the physical symmetry of the stays, ensures that both sides will flex evenly, and therefore overload and set evenly - much better than eyeballing it and checking. As for dropout alignment and so forth - bear in mind that the bike has always operated with a 135mm axle, and still does. The shape hasn't changed - just the internal stresses.

This will not work if the chainstays are asymmetric because of a dimple for chainring clearance (like many, many bikes.) Good for the frames it works for, though...

One little issue with your method. It assumes, and we all know what that stands for, that both sides will reach their elastic limits at exactly the same time and hence bend by exactly the same amount. But in the real world things conspire to prevent this idyllic set of circumstances.

First aspect would be any differences in the shaping and thickness of the stays on each side due to forming tolerances from the stay tube manufacturing. Second would be the often seen dimpling of the chain stay on the drive side that makes the two sides non symetrical from the get go even assuming the metals are of even elastic limits. And finally because the chain stays are shorter they will see more of the bending loads and reach their limits earlier than the longer seat stays so the seat stays won't be bent at all or not enough and the dropouts end up non parallel.

The Sheldon Brown method has the advantage of being able to move the bending force around so you can bend both the chain and seat stays of each side by equal amounts so that the dropout stays vertical . And by cold setting the one side by half the amount needed and being able to measure that change from the still intact opposite side to ensure accuracy of the job. You can accomplish and check your results with a simple metric ruler and completely ensure that the frame is still aligned to within the thickness of a line on the ruler. And then the opposite side can be bent in turn using the first bent side as a confirmed standard. So you get a slightly more involved but infinetly safer way to do the job that leaves no doubt about the outcome.

The problem with relying on Hooke's and other laws without checks and confirmations is that it leaves too much room for Murphy's law to come into play. And as someone that has bent, cut, carved, machined, soldered and welded lots of metal and wood in my lifetime Murphy is all too quick to enter the mix when it comes to stuff such as this. In my experience any steps that result in Murphy's exclusion are never badly chosen.

interesting. I'm basically with BCRider and zzyzx_xyzzy here for the reasons they described... in a lot of cases you won't have actual symmetry.
but on the other hand, Sheldon's method *does* seem a bit dodgy and also leaves room for its own errors.
I'd probably use either method on cheaper bikes and feel fine about it.
I've used Sheldon's method on a reasonably-light 1980's steel road frame and things worked out fine.

BC Rider: the point you make about having some precise way of measuring the set produced for each side is a good one. A stick taped to the down tube and seat tube would work; wish I'd thought of it in time.

That done, one could try the symmetrical, axle method and then use single-side loading to correct a difference if one was detected.

Symmetry isn't the only benefit of the axle method - it also loads in a very controlled way with a perfectly-aligned force that engages an intended load-bearing face.

I must say I'm bemused by the concern for precision. The spreading amounts to 1/5 of an inch - in a system where the wheel is mounted by eyeballing for straightness, then clamping with a friction device. I end up adjusting the lateral position of my brakes - because the wheel has moved enough to cause dragging, a matter of millimeters - as often as not after putting the wheel on. If that level of imprecision mattered, we'd be screwing locator bolts through axle-mounted ears into holes on the dropouts - and getting engine-type tolerances.

maybe All-thread rod will do too, if you don't have any super long axles..

Fair point about precision.
And fair point about the angle about the loading force as well (although, along the lines of your precision comment, that doesn't matter so much either)

I used a turnbuckle & washers. It worked fine. Went from 126 to 135mm. Drastic, but it's an old crap bike.

Went from 126 to 130 on another bike. Just used my hands to spring it out a bit. I agree, extreme precision isn't required.

The use of a length of 3/8"all-thread rod, two nuts and two large washers is a very old and well established method of spreading dropouts. I remember reading about this technique over 20 years ago and doing it twice on steel frames to make 126 mm frames accept 130 mm hubs with no forcing. It worked very well and, as far as I could measure, spread the stays evenly. I certainly found it easier to control than Sheldon's somewhat dramatic approach.

For myself I still like the idea of one side at a time so I can check for even bending. Having had too many run ins with Murphy over the years I tend to think in terms of "belt and suspenders" ways in order to cut off any troubles before they occur. And that explains why I'd go one side at a time.

Do we need this level of accuracy? I guess it depends. We see enough threads around here from folks complaining that their bikes won't ride hands off easily. An out of true frame would cause such troubles. So perhaps there is a need for better accuracy. Or at least a way to check the accuracy after the bending is done so any minor corrections can be done.

As for wheel placement consistency I think you may be underestimating how well a good eye can center the wheel where it passes through the chain stays just behind the BB shell. I've never had any issue with getting that good to within a mm or less. Mind you after years of shop work folks can develop fairly well trained eyes for this sort of thing so maybe it's easier for me than some others.

I've used both methods and I prefer Sheldon Brown's by far. Your method made such a mess of the job that I had to realign the frame using Sheldon's method. There's no reason to believe that the stays will allways bend equally using your method. It can also make a mess of the paint on the dropout.

It may seem dodgy, but it's not. It works, and works well.

I have used both methods and prefer the turnbuckle approach. I worry about denting or bending the seat tube, as the load is coming in right at the middle of the thin section (on butted tubes). One thing I really try to do is to clamp the stays over the bridges so that the braze joint doesn't take too much of a strain. C-clamps work fine, with appropriate padding (I use little wood wedges).

Sheldon's method has worked well for me two times. Going slowly and using a string to check alignment will yield excellent results. In one case the frame was better-aligned after the Sheldon method than it was stock.