I've replaced my own garage door torsion springs twice, and I'm still alive to talk about it; it's absolutely doable. I learned how to do the replacement from Richard Kinch's exceedingly thorough page on the subject, so I'm going to quote heavily from it; all credit goes to him. I'm going to provide the highlights here, but read and re-read the information there, and make your own decision as to whether you want to do the job yourself.
If you've researched this subject at all, you will no doubt have heard that you shouldn't be attempting torsion spring replacement as a do-it-yourselfer. That is generally good advice, so if you have any doubts about your abilities to do risky physical work on your own, hire the job out like everyone else. I found I was capable of doing this work with acceptable risk, because I intelligently understood the techniques, paid careful attention to methods and safety, knew how to use common tools in good condition, properly improvised the special tools I didn't have, and diligently attended to correctly performing a few moments of hazardous manipulation. I learned to do it purely on my own based mostly on bits of advice reluctantly given in Internet forums such as the Usenet newsgroup alt.home.repair. When I first wrote this page in 2002, there was no other do-it-yourself information available on the Web, and it was not until 2005 that reliable information disclosing the techniques started to appear elsewhere.
I am also an engineer, and have always done a lot of mechanical repairs around the house. I also operate my own laboratory machine shop, and do some mechanical design work there. This background helped me figure things out, but nothing that involved is critical to repairing a standard garage door.
This work is risky, but the risk is comparable to doing your own car repairs, or climbing on the roof of your house to clean your gutters. These are dangerous things that many people can do safely, but that safety depends on intelligent understanding and application of proper techniques. Professional door repair technicians, who are fully knowledgable, skilled, and experienced, report that they nevertheless are injured from time to time, despite their best efforts. Coldly evaluate your abilities and motivations, to judge whether you can manage the risks of this work for the benefit of the money and time you might save.
Determine the critical metrics of the old spring so that you can order a new one.
Critical measurements: Torsion springs come a variety of standardized sizes, so you have to carefully measure the old springs to know what to order for proper replacements. Tables of standard sizes and designs are on the Web, such as at Industrial Spring. The four critical measurements (all in inches) are: (1) the wire thickness (which I'm measuring here with a dial caliper; you can also measure the length of a number of closely stacked turns with a ruler and divide by the number of turns in the stack, measuring 10 turns this way makes the math easy), (2) the inside diameter (not outside!) of the relaxed (not wound!) coil, (3) the overall length of the relaxed (not wound!) spring coils, not including the winding cones, and (4) the right- or left-hand winding of the spring. One must glibly quote those figures to the spring supplier, otherwise one's lack of expertise will be obvious, and one will not be worthy of buying the parts.
Beware of improprer prior installations: Sometimes the existing door installation is not correct, and the old springs should not be used as a specification for replacements. For example, the old springs might have been replaced with incorrect sizes because the last repairman didn't have the right one on his truck. If your door has never worked quite right, something like this might be the cause. To correct this, you must use the weight of the door to specify the spring, either from a spring rate manual giving spring torque constants, or from the formulas below.
Obtain a replacement spring. Start by simply googling 'garage door supply'.
You certainly won't find these at Home Depot or Lowe's (although last I checked Lowe's does carry the less daunting extension spring replacements). I also have a list of some suppliers at the end of this page.
Cost was $88 for 2 pairs of springs, plus $21 shipping. (I had to order 2 pairs to meet the $50 minimum order.) They came with new cones inserted as shown at that price, so I didn't bother trying to remove and reuse the old cones to save a few dollars. The cones are quite difficult to remove from old springs and to insert in new ones, and the spring supplier will have the right tooling to do that easily. That was the best price I could find on the Web at the time, and didn't seem out of line with what parts like this might cost at at the building supply (if they only sold them). Contractors buy these much cheaper in quantities; they're just an ordinary high-carbon steel wire turned on a winding machine. I also found Web sites asking a lot more money, obviously trying to cash in on search-engine traffic from do-it-yourselfers. Others report that some local dealers sell springs at retail, but at a high price that eliminates any economy versus having them installed.
Spring pairs should be replaced together, since the mate is likely to fail soon after the first, and any possible savings in parts isn't worth the extra effort to repeat the work later.
Gather and/or prepare your tools for installation.
The two set-screws in the winding cones have a 3/8-inch square head, which fits a 3/8-inch open-end wrench or 8-point socket, or a 7/16-inch 12-point socket or 12-point closed-end wrench. I carried an extra wrench in my pocket while winding, since I didn't want to be holding a wound spring that I couldn't set because I had dropped the wrench (although one could rest the winding rod against the door in this case while picking up a dropped tool).
I decided later to use an ordinary open-end wrench rather than this socket or a box-end wrench to turn this fastener. The extra play fits the crude square heads better, and if the cone were to accidentally let loose, an open-end wrench would be the least likely tool to stay on the head and turn into a flying hazard.
The standard winding tools are simply a pair of 18-inch lengths of mild steel rod, 1/2-inch diameter. Winding cones can have different socket sizes (such as 5/8 inch instead of 1/2 inch), so it is important to measure the socket and select a matching rod diameter. Also beware that poor-quality cones may have a sloppy fit to the winding bars, and a loose fit presents a severe hazard of slipping at the worst moment; anything more than about an inch or two of play at the handle end is too loose for safety. I bought a 3-foot length of zinc-plated 1/2-inch diameter steel rod from Home Depot for about $3, which conveniently cuts into two halves of just the right length (the store might even cut it for you if you ask). A steel supplier selling at commodity prices might charge about 50 cents or so for such a piece that weighs about 2 lbs. Drill rod would work if used in the annealed condition in which it is originally sold, but the added expense provides no benefit and the brittleness (if it had been hardened and not annealed) would worry me a bit. Rebar, threaded rod, screwdrivers, etc., are absolutely foolish as they will not fit the socket snugly. Aluminum rod is definitely too weak, and will bend under the torque that must be applied. Longer rods would make for more leverage but unwieldly swing; shorter rods make for uncontrollable swing. As we'll calculate below, the 18-inch standard tool length is an appropriate compromise. Note that you do not need 18 inches of ceiling clearance above the torsion shaft to use an 18-inch rod, since you need not swing the rods above horizontal when winding.
Remove the broken and unbroken springs.
To remove and disassemble the shaft and lift drums, the torsion on the unbroken spring must first be released. I used a ratcheting box-end wrench to loosen the set-screws while pushing the rod against the force I knew would be released when the screws let go. Later I switched to an open-end wrench for the set-screws, since some of the square screw heads were too rough to fit in the box-end wrench.
It is prudent to be prepared for torque from either orientation when loosening the set-screws. This protects you from a miscalculation of the force orientation or from an attention lapse.
If it isn't marked already, you should run a chalk line (as described below for the new springs) down the length of the old spring before unwinding, so you know how many turns it took to unwind. This will give you a starting value for the number of turns in the new springs.
By each "turn" is meant a full revolution of the winding cone. Each full turn requires four quarter-turn manipulations of inserting and switching the winding rods. My door is a very common height of 7 feet, which with the 4-inch drums requires about 7-1/4 or 7-1/2 full turns on the springs.
This is the moment of truth for the beginner, as you will be holding the full force of a fully-wound torsion spring for the first time. It is time to adopt a calm, quiet, deliberate, careful attitude of concentration. Make sure that the telephone, a bystander, or other distraction is not going to startle you or make you lose your concentration.
Loosening or tightening the set-screws is the moment of most risk, since the end-wrench is a potential missile if you slip, and your hand is close to the cone. When the wrench is removed and only the rods are in place, it would seem that the worst that could happen is that the rod is flung out and the captive spring and cone rattle around, assuming you are keeping yourself clear of the rod's radial disk of rotation, and not leaning on the rod such as to fall into the apparatus were the rod to slip out of your grasp. The torsion shaft design has the virtue of capturing the mass of the spring and cones reliably on the shaft, preventing these parts from launching themselves as projectiles, even in an accident.
The prior clamping of the set-screws tends to have pressed a dimple into the hollow shaft and to have distorted the shaft's roundness into an eccentric shape. While releasing the set-screws, I was careful to loosen them enough to let the cone swing around any such distortions. I was also careful to observe any binding of the old cones on the eccentricity or burring on the shaft. The fit of the cone on the shaft is supposed to be loose enough to avoid binding, but if it were to occur one would have to be careful not to assume the spring was unwound when in fact the cone was just stuck on the shaft. If I had a stuck cone that I could not unwind with a little extra force, then I would have called in a technician to deal with it. In the worst case, I suppose the spring must be deliberately broken with some hazard, thus releasing it for a forceful disassembly, and the shaft and some other parts replaced. But this is an unlikely situation and in this case was not necessary.
The winding technique is simply to (un)wind as far as one rod will go, where it is pressed against the top of the door, or nearly so, by the unwinding torsion. You insert the other rod in the next socket, remove the first rod, and continue. At any point you can stop and rest by leaving the active rod pressed against the door, where it will be held by the unwinding force. I would make a quarter-turn increment that way, and let go for a moment to collect my attention for the next increment, almost in a quiet, meditative alertness. While you can go from one quarter-turn and rod-swap to the next continually without letting go, working fast against the steady tension seemed to invite a kind of shakiness in my arms that was a bit unsettling. It isn't that there is much physical exertion, it is more that the tension is unrelenting, like peering over a precipice.
While winding or unwinding, one must be mindful of the possibility that the spring could break during winding process itself. If that should happen while the spring is significantly torqued, hazardous forces on the winding bar will suddenly become unbalanced, and the bar will take an unexpected jump, possibly injuring your hand or anything else in its path. At the same time, the spring remnants, although captured on the torsion bar, will create a frightening racket that would give the bravest soul a start. So your winding technique should be firmly in control of the rods, and you should not be so delicately perched on a ladder such that a startle will result in a fall.
In my case, removing and replacing the relaxed springs required that I take down the assembly: torsion shaft, lift drums, and bearings. Doing that requires unbolting the center bearing plate from the wall, removing the drums from the shaft, and finally sliding the shaft back and forth out of the end bearings to remove the whole assembly off the wall.
Once the springs are relaxed and loose on the torsion shaft, the lift drums lose their tension on the lift cable, and the cable comes loose. The end of the cable is terminated by a press sleeve, which locks into a ramp on the drum. Different drum styles have a bolt or other method to fix the cable end to the drum. These steel cables are springy and won't stay in place without tension. If my pre-inspection had showed that these cables were worn or frayed, then I would have ordered proper replacements ahead of time from the spring distributor, since this is the opportunity to replace them. Standard hardware-store cable and fittings are not appropriate.
Install the new springs.
As noted above, set-screw clamping may have distorted the cross-section of the shaft and made it difficult to slide off all the hardware. With the shaft on the floor, it may be possible to restore enough roundness to proceed, using compensating clamping force to the distorted area via a machinist's vise, an arbor press, a hydraulic shop press, etc., on the shaft body. Burrs and other slight distortions on the shaft can be filed off with a hand file or touched with an abrasive wheel on an angle grinder. At some point, the condition of the shaft may just be degraded enough that it ought to be replaced.
I was careful not to assume that the previous installation correctly oriented the right- and left-hand springs on the correct sides of the center bearing plate. They could have been installed backwards by an incompetent installer, resulting in them having been wound looser (larger diameter coil) instead of tighter (smaller diameter coil) than when in their relaxed state, and if so I would have corrected them on the new installation. The proper orientation of the springs applies their reaction torque from tighter winding such that it turns the drums to lift the door. Verifying this is a rather simple exercise in mechanical visualization, but does require some care to be certain of correctness. If you were to install the springs backwards, and then start to wind them in the wrong direction, then the torsion bar will start winding the drums backwards, and not holding against the vise pliers, which should be obvious to inspection. Winding a spring backwards also tends to screw the spring off the cones; this error cannot proceed too far before the spring slips off the cones.
At this point I weighed the unlifted door to confirm and fine-tune my calculations. This is not strictly necessary, but it makes the adjustments easier to perform, if you happen to have a scale with the requisite capacity. With some helpers, we first lifted the door a few inches and rested it on blocks of wood to provide clearance underneath. Then I slid a 400-pound-capacity freight scale under the center of the door, we lifted again to remove the blocks, and lowered the door gently onto the scale. This door weighed in at 238 pounds, which is very heavy for a single-car door. Since the outside of the door carries the 3/4-inch plywood paneling to match the house, and that plywood weighs about 2 lbs/sq-ft, I estimate the door weight to be about 7 x 10 x 2 = 140 lbs of paneling with the rest 238 - 140 = 98 lbs the interior panels, hardware, and cobwebs. Knowing this total weight will help later in adjusting the torsion on the springs. After weighing, we removed the scale and blocks, leaving the door fully lowered again. Had I not had a high-capacity freight scale, I might have improvised a crude weighing device from levers and smaller weights of known mass, or a lever arm pressing a reduced proportion of the full weight onto a lower-capacity scale. Another factor to remember is that the weight of a wood door can vary with humidity.
Next, the torsion shaft is reassembled with the new springs, the drums repositioned loosely on the shaft, this whole assembly slid back into the end bearings, and the drum set-screws tightened down. I tightened the set-screws about 1/2 or 3/4 of a turn after contact with the shaft, which provides a good grip, but does not distort the shaft. The drums can be set on their old positions, if they were correctly installed, which is snug up against the end bearings to remove any longitudinal play in the torsion shaft. Now the lift cable can be reattached to the drums, and a slight temporary torque applied to the shaft to keep the cable taut while the first spring is wound. This temporary torque is conveniently applied with a pair of locking pliers clamped on the shaft, positioned such that they hold the torque by pressing lightly against the wall above the door, before you start the spring winding, The locking pliers stay on the torsion shaft until you have finished the spring winding locked down the spring cone(s) with the setscrew(s), and removed the winding bars. Then you simply remove them with the release on the wrench handle.
Checking if the lift drums need resetting: The old position of the lift drums on the shaft may have slipped or otherwise lost the the proper position, requiring a reset of the drum position on the torsion shaft. You will also reset the drums if you are replacing the lift cables, since the new cables will not exactly match the length of the old ones. Problems like uneven tension on the cables, or a tilted door, or a door that doesn't easily stay aligned with the tracks, can be due to an improper "set" of the drums on the shaft. So one shouldn't assume the old positions are correct. Setting the drums on a "fresh" part of the shaft will avoid the possibility of damaging the shaft from re-tightening in the same dimples.
There's a lot more info about adjusting the drums and lift cables, so visit Richard Kinch's page for details.
Wind the new springs.
The last step before winding is to run a piece of chalk down the length of both relaxed springs. This lets you observe the number of turns as you wind. You don't want to be busy counting turns when you should be paying attention to the winding rods.
The position of the bars in this photo was necessary to take the photo, and does not show a correct winding technique, You should not have to swing the bar up as high over the top as shown. The lower bar during winding should swing from pointing down to pointing a little past horizontal. Then you hold that bar horizontal while socketing the other bar pointing down, apply force to that (now) lower bar, then remove the (now) upper bar, wind one-quarter turn, and repeat.
You can also see I am wearing eye protection. It would be foolish to go bare and risk an eye injury.
Winding "up" starts out easy. It finishes at the proper number of turns, by which time you are pushing against the maximum torque. Count the turns of spring winding from when the springs are slack. To be sure you're winding the right direction, all you have to remember is that proper winding makes the spring smaller in diameter and longer in length as it twists "in". On the standard door (most common), this means you push the winding bars up to wind up the spring, which is an easily remembered rule. This is very apparent and should be verified during the first few easy turns. You can also think about the correct winding direction in mechanical terms, namely which way the reaction of the spring will torque the shaft and drums, which in turn will lift the cable. This should all make perfect sense before you attempt the manipulations.
By watching the chalk mark while winding, you can count the number of turns applied, and confirm the number later. My standard-size door (7 foot height) with 4-inch drums has a nominal wind of 7-1/4 or 7-1/2 turns, which leaves 1/4 or 1/2 turn at the top-of-travel to keep the lift cables under tension. After 7 turns on the first spring, I clamped down the set-screws, weighed the door again, and found a lift of about 100 pounds in reduced weight. As expected, this wasn't quite half of the full 238 pounds, nor would it leave any torsion at the top-of-travel, so I added an 8th turn. The door now weighed 122 pounds on one spring, which was ideal. After winding the other spring, the door lifted easily, with only a few pounds apparent weight. This confirmed that the spring choice was properly matched to the door design. I engaged the electric opener trolley, and adjusted the opener forces down to a safer level suitable for the new, improved balance. The door was now ready for return to service.
As with the drums, I tightened the winding cone set-screws about 1/2 or 3/4 of a turn after contact with the shaft, which provides a good grip, but does not distort the shaft.
The usually recommended rule for a door being properly balanced is that it should lift "easily" through all its travel. The door may also remain stationary if let go somewhere around the middle of the travel, but a smoothly rolling door many not show this behavior (while a sticky track will!), so easy travel is the only reliable test for proper balance. A difficult door may be due to stiff bearings or rollers in the mechanism, tracks out of alignment, etc., not necessarily the torsion spring adjustment.
Once the springs are torqued, the setscrews tightened, and the locking pliers and winding rods removed, do not play with turning the torsion bar using the winding rods. Doing so even momentarily can relieve the tension on the lift cables, which then easily slip off the drums. Replacing the cables on the drums can be difficult without repeating the entire spring unwinding-winding procedure again, and the cables can be damaged if tension is applied while they are off the drums.
Total time for me to complete this work the first time was 3.5 hours, including cutting the winding tools and the photography. I've completed subsequent repairs in less than an hour. Hundreds of people have written me to say they did it safely in a few hours after studying this essay. I'd like to hear how it goes for you.
There's much, much more detail at Richard Kinch's page on the subject. I tried to include the most important information in case that page goes down, but it's been at that URL since 2002, so I hope it's still there when future DIY'ers need it.