I've been working through the cars sorting out the issues that have been causing my freight cars to derail one-by-one.
First I asked myself do they all derail in one place, or are they derailing for another reason? I had both. Derailing at turnouts in general, and on my reverse loop in particular. The latter causing me to fret somewhat, as the whole premise of the layout is based around a very tight 14 inch radius loop that all trains have to take to enter the yard.
Another problem is that such a tight radius acts like a grade and reduces the amount of freight cars a locomotive can pull.
However, one confounding variable, as it turned out, was the fact that the drive train of my only operational locomotive (operational being defined as painted and chipped) was failing, and failed during some of the tests, which perplexed me a little until such point as I figured it out.
For me there are a bunch of obvious thing that I find cause derailments.
1. The freight car is too light, but that needs to be defined.
2. Trucks are not swiveling properly.
3. Coupler pockets are sloppy and allowing the coupler to twist when being pushed.
4. Track is out of gauge
5. Wheels back-to-back is not right, or the standard is different to the turnout check gauge.
Let me start with number one, and refer you all to the NMRA Recommended Practice RP-20.1. This recommends half an ounce per inch plus one ounce for HO freight cars etc.
So a standard 6 inch long (a scale 40 foot box car), should be 4 ounces. Weighing my box cars I find that all my Kadee PS-1 cars are 3.9 ounces (I have a very accurate set of scales). In practical terms they are spot on. My Proto 2000 C&NW 40 foot stock car on the other hand only weighs 3.5 ounces, and is therefore half an ounce too light. Also running at an ounce light is my Intermountain AT&SF Caswell gondola. However, my Intermountain NYC 1937 AAR 40 foot box car came out of the box at 4.4 ounces, and is nearly half an ounce too heavy.
Given that I've managed to get these to run round my layout my guess is that the NMRA recommendations are conservative, and that one can run lighter cars reliably, even on stupidly tight 14 inch radius curves. This fits with what I know about how the formula for the weights were arrived by; namely empirical testing and heuristic analysis.
Looking at my longer freight cars I have a C&NW 53 foot flat car (7.5 inches long) that should weigh 4.75 ounces, but only actually weighs 3.4 ounces. My Proto 2000 WP mill gondola is 8 inches in length and should weigh 5 ounces, but hefts in at 6 ounces, so an ounce over weight. Whereas my PRR War Emergency gondola, which is 8 inches long should weigh 5 ounces, but actually comes out of the box weighing in at 2.6 ounces, which is nearly half what it should be.
Looking at my shorter cars, I have a couple of Kadee 50 ton coal hoppers that weigh 3.5 ounces, which is bang on the recommended NMRA weight for their length. My two Intermountain 2 Bay AAR hoppers both have different weights out of the box. One weighs 1.35 ounces, the other 2 ounces, go figure. My Proto 2000 tank cars all weigh around 3.8 and 4 ounces out of the box, which makes them nearly half an ounce heavier than the recommended 3.5 ounces I estimate they should be.
The longest train I can run at the moment is eleven cars, its all I have that have been fettled.
So I pushed this train backwards through my reverse loop with the longest and lightest car by the locomotive, and the heaviest longest car at the other end, with a random mix of lighter than recommended, or heavier than recommended cars in-between them. Outcome, random derailments of one of the trucks on either the C&NW flat car, or the PRR War Emergency gondola, with a tendency to derail when pulled. However, both were consistently pushed around the tight curve with no problem.
In summary, weight is not the issue that people make it out to be. Yes too little is not good, and yes too much reduces what you can haul, but by and large it is not the what keeps the models on the rails.
So what have I found to be the biggest cause of unreliable running?
Trucks that don't rotate freely, with wheels that are not free rolling, and couplers that twist and foul on turnout check rails and crossing vees.
One thing I would like to mention is that the received wisdom that one truck should just rotate, while the other should be loose enough to rock as well as rotate needs to be judged carefully, as my experience shows that rocking trucks can cause freight cars to derail when pushed.
The amount of rocking needed is very small, around 0.5 millimeters in total, or for those of you who are metrically challenged, about 1/50th of an inch. I shall now get back to sorting out all my Proto 2000 tank cars, which have yet to be fettled.
NB: One final comment to the NMRA recommended practices, which is more of a question. How is it that the prototype can run an empty car at the head of a consist a mile long, and switch empty cars etc without derailing everytime? Not all things scale.
Terms and conditions apply. Errors and omissions excepted. Everything said here is based on extrapolation from a limited set of parameters. Remember everything is bigger in Texas, and eleven freight cars is hardly representative of longer train lengths.