Here are excerpts from a paper that John Forester wrote on bicycle brake performance in 1973, comparing disc brakes with coaster brakes. Bike Tech published a version of this.

Power Absorption Tests

Because of the distinct possibility of regulation [This was when the Consumer Product Safety Commission was proposing oven tests for rim brakes but no oven tests for coaster brakes] based upon inaccurate or non-existent data, I decided to test the power absorption performance of rim brakes and coaster brakes under actual use conditions. I selected a test hill on one of the local cycling routes, providing by topographic map 2040 feet of descent in 4.0 miles (9.6% average grade) which was descended in about 9.67 minutes (211 FPM), only a few seconds difference between runs.
Equipment

A normal racing/touring bicycle was equipped with two rim brakes and one coaster brake. Fitting the coaster brake required removing the derailleur system and modifying the coaster brake sprocket to accept the narrower chain. 27 x 1-1/4 wired-on tires were used to eliminate the limitation imposed by the rim cement used for tubular tires. The rim brakes were Weinman side pulls about 16 years old. The coaster brake was a used New Departure (which is not made under other names) which had been disassembled, inspected, lubricated, reassembled and tested for normal operation before the tests. The preliminary operation test consisted of a quarter mile of level riding, alternating sprints and braking. Brake operation was normal in every respect, but even with the low level of power absorption the shell became painfully hot to the touch, estimated 180F or 120F rise.

Instrumentation

For temperature measurement small chips of Tempilstic temperature indicating crayon were attached to the metal surface by a single layer of tissue paper and one layer of transparent adhesive tape. When the metal reached the melting temperature of the crayon, the chip melted into the tissue paper, making a stain visible through the clear tape. Confirmation of the highest temperature reached during the series of runs was made by later peeling off the tape and tissue paper to determine whether the wax had melted without staining the tissue. No cases of invisibly melted wax were found. Rims were instrumented in intervals of 25F from 150F to 275F. The coaster brake shell was instrumented with 275F, 300F, 350F, 400F and 500F indicators.

For brake performance during the run only subjective measurements were made. The test cyclist is highly skilled and experienced, and the close times for the successive runs indicate that he can estimate and control his speed very consistently. Since some portions of the test run are three times as steep as others, and some road turns are very tight, the brakes had to be used in the normal manner -- a continuous variation in the amount of deceleration required. This quickly disclosed to the test cyclist variations in brake performance of the coaster brake, and presumably would have done so for the rim brakes had irregularities occurred.

A stopwatch was mounted on the bicycle handlebar to measure the time of each descent.

Test Procedure

Three runs down the course were made. At the bottom of the course the temperature indicators were read and recorded, the running time recorded, and other observations made. The bicycle was returned by car to the top. The runs were made in the following sequence: 1) Both rim brakes. 2) Rear rim brake. 3) Coaster brake.

After all runs the temperature indicators were disassembled and checked for invisible melting. The rim brakes were examined for wear and operation. The coaster brake was disassembled, examined, and several of its parts subjected to hardness testing and compared to similar new parts.

Test Data Summary

1) Two rim brakes produced a normal downhill run for the course, as it had been run in normal cycling trips. Braking force available during was always adequate, no perceptible change in performance during the run. Front rim temperature peaked between 175F and 200F (rise 105F to 130F), rear rim did not reach 150F (rise less than 80F). The disproportion between front and rear rims is the expected result of two factors. First, the rear brake has longer, more curved cables, thus achieving less brakeblock force for given lever force. Second, experienced cyclists use the front brake harder than the rear, to achieve desired deceleration without skidding the rear wheel.

Although there was some rubber transfer from blocks to rims, as is usual, there was no perceptible change in brakeblock/rim clearance or level position.

2) Using the rear rim brake alone produced a run equivalent in the technical sense to the two brake run, but the cyclist depended upon the presence of the front brake to provide deceleration in any emergency situation which might arise. The front brake was used twice at blind corners, for less than a second each time. The rear rim temperature registered the same as the front rim in the two brake run: more than 175F but less than 200F (rise 105F to 130F).

3) Using the coaster brake destroyed it. Before 700 feet of descent smoke was suspected. As 700 feet emission of smoke was obvious to the driver of the following car. At several times after 700 feet of descent the ratio of retarding force to pedal force changed abruptly, the effectiveness increasing and decreasing in a random way. At about 1700 feet of descent the brake failed to release completely and dragged for the remainder of the run. At the end of the run the hub was still smoking hot, the temperature indicators were reduced to ashes, the shell's chrome plate had turned to white powder outside the brake disc area, and the wheel required about 10 lb-ft of torque to turn it.

Dismantling and examination showed that all free grease had been melted and thrown out. A thin film of liquid grease remained only in the cold-end bearing -- all other surfaces were blackened by carbonized grease. The steel stationary brake discs were blackened, had been forced elliptical by the torque between them and their flat-sided mandrel, we distorted where the mandrel flats contacted them, and were jammed on the mandrel. The bronze rotating discs had their torque lugs partially collapsed -- those nearest the center of the hub were only one half of their original width, those to the outer end of the hub were collapsed less. The combined distortion of the two kinds of disc and the elliptical distortion of the stationary discs produced the permanent drag. The cup race of the shell brake-end bearing appeared either black from carbonized grease or blue from heat.

Robin Jones, Manager of Metallurgy for Stanford Research Institute then examined some of the parts and corresponding new or unheated parts. Bearing balls from the hot end were significantly softer than those at the cold end, and used discs (both steel and bronze) were softer than unused discs. Estimates of the temperatures required in the 10 minute exposure to produce such softening were confirmed by exposing the new parts to known high temperatures and observing what temperatures produced equivalent softening. Mr. Jones states: "I conclude from the data obtained that the temperatures reached during your downhill run were as follows: Ball bearing (hot end) over 600F; Steel brake discs, equal or over 900F; Bronze brake discs, over 800F but under 900F."

Detailed Conclusions

1. The rim brakes had reached thermal equilibrium, the coaster brake had not. A bicycle rim weighs less than a coaster brake and is made of thin material. Both (in this case) were made largely of steel. For equivalent energy input, the rim temperature should rise more. The low peak rim temperature recorded in comparison to the high coaster brake temperature show that the rim was effectively transferring its heat away. The coaster brake, however, showed much higher peak temperatures, which implies a much less effective heat dissipating ability and much longer stabilization times for any given power input. Furthermore, it showed performance variations at several times during the run, the most noticeable being the failure to release at 1700 feet. This suggests that temperatures were still increasing at 1700 feet of descent. The greater lug collapse at the center of the hub compared to that at the end suggests also that collapse was in the process of occurring as temperatures increased, but was prevented from going to completion by the end of the run.

2. The coaster brake was on the verge of catastrophic failure. The torque-transferring lugs of the bronze discs were collapsing, the process being nearly completed for the inner discs. Presumably when the disc loses its ears it will stop rotating with the hub shell and cease to operate as a braking member. This will, if the operator increases the actuating force to maintain the required braking effect, transfer the power from the inoperative disc to the remaining operating discs and transfer the failure the length of the multiple-disc stack. Assuming that the operator did not crash the bicycle, deliberately or inadvertently, the brake would cease being able to stop the bicycle and melt out the remaining brake discs, or it could jam and lock the wheel.

3. It is inadvisable to use coaster brakes for descents of more than a few hundred feet. The proper operation of coaster brakes is dependent upon proper lubrication of its working parts. Smoke was suspected before 700 feet of descent had been reached, and was obvious at that point. Frequent exposures to temperatures sufficient to make it smoke destroy the lubricating properties of the grease.

4. Any braking system that operates only on the rear wheel produces insufficient deceleration for adequate accident avoidance under common downhill conditions (and, it may be added, under common traffic conditions also). The test cyclist would not have attempted to descend the test hill with only a rear-wheel brake available, and, as was noted, he used the front-wheel brake twice on one descent to slow for corners, even though he was trying not to.

5. Rim brakes are essentially unaffected by long fast descents. Any rim brake system which produces adequate deceleration in single-stop tests will presumably be adequate on downhill descents. The temperatures achieved on the rims are insufficient to affect any of the usual brakeblock compounds.

By John Forester, forester@johnforester.com.