I started layout construction using Old Pullman turnout kits - the only ready to run turnouts available at the time did not conform to NMRA standards (this was before Atlas started making turnouts again). I had also planned to hand lay a number of turnouts from scratch in the yard (this was on the Elmira Industrial plan which has a number of curved turnouts in the yard). I had some trouble with the old pullman kits, got bogged down scratch building turnouts, and then Atlas came out with their new line of 2 rail turnouts and track and I decided to use those for everything. I've recently realized there are some real benefits to be gained by adding two curved turnouts, so I'm currently working on scratch building those. The following documents my first round of scratch building attempts. I'll updated it if and when I succeed with the new curved turnout attempt.

The first few turnouts went in with no trouble, but then I ran into one that I just couldn't seem to get right. I eventually realized that there were several manufacturing defects in the kit, including closure rails so misaligned with the frog rails that the only only way to fix it was to unsolder everything and reassemble it correctly. After working for most of a weekend trying to get it right, I realized that what I was doing was probably actually harder than building a turnout from scratch. (Due to my slow construction pace, it was so long since their purchase that returning the kits to the manufacturer was out of the question.)

I did some reading on turnout construction and started building a test turnout on a scrap of homosote. Some good reading on the subject of building turnouts can be found at the proto:87 web site - note the frog filing jig. Also at the Railway Engineering web site on their FAQ page - see in particular Q19 about gauge through the frog area. And Tony Koester's article on building a turnout from scratch in the December 1989 Model Railroader, or in "Trackwork and Lineside Detail for your Model Railroad" - good advice on notching the stock rails, but use the frog technique from the proto:87 article.

The frog is the most important part to get right. A mistake at the points can be solved with a little filing or bending in place (within reason), but fixing a mistake at the frog pretty much requires taking the frog apart and possibly remaking parts of it. The best way to get a good frog is to thoroughly understand what makes a good frog good and a bad frog bad.

The relationship between tire width, flangeway width, and gage is critical in the frog area - study the NMRA standard S-2, S-3, and S-4 to get a good understanding of this relationship. Failure to get these relationships right can result in a variety of problems, the most common of which is what I call "the frog pit" - when the gap between wing rails just in front of the point is wide enough so wheels fall into the pit. A common band-aid for this problem is to fill in the frog so wheels ride through on the flange - completely unprototypical, and also not very effective unless all your wheels have identical flange depths. To avoid the frog pit problem, use the minimum flangeway width for the wing rails. That, combined with making sure the actual point of frog is as close as possible to the theoretical point of frog, and minimizing the gage through the frog area, will give you a nicely working frog. The gage minimizing is necessary to make sure that the wheels can't go through at a slight angle and catch the frog point.

O scale currently presents some unique problems for frog construction. NMRA standard wheels have a 0.l72 tire width (see NMRA standard S-4). However in the interest of better appearance, the newer Atlas cars have 0.155 wheels, and NWSL makes some 0.145 wheels which they say will work with turnouts that conform to NMRA standards. For comparison, the prototypical 5.5" wheel width works out to 0.115 in O scale. To illustrate the difficulty, lets work a little math. The most treacherous part of the frog pit is the widest point - directly in front of the frog. That's a little more than twice the flangeway width (since the flangeways are at an angle to each other, the width for a wheel going through one is the width of 1 plus the square root of the width of the other times the frog number plus one (think Pythagorian theorem). In O scale the minimum flangeway is 0.064 (calculate using NMRA S-3 C - F) . So, for a #5 turnout, the frog pit is 0.129 inches wide at the theoretical point of frog. For a standard 0.172 wheel, that leaves 0.043 of the wheel tread riding on the wing rail at the point where the frog starts supporting the wheel. Or if would if the actual frog point was perfectly at the theoretical frog point. To put it differently the distance from the theoretical frog point to the point where the wing rail no longer supports the wheel is 0.215. Plenty of room for a slightly sloppy frog that will still work. But, with the Atlas 0.155 wheels, there's only 0.026 of the wheel at the theoretical point, and the distance is only 0.130. And for the NWSL 0.145 wheels it's 0.016 inches and a distance of 0.080. To put things in perspective, that's just a shade over a scale 3/4 of an inch. The prototype moves the actual point of frog back so that the "point" is 1/2 inch wide. Consistently achieving that kind of accuracy is a daunting proposition - only time will tell if I succeed or trade in my 0.145 wheels for wider ones.

The throw bar and point hinges also have some interesting tradeoffs. With code 125 rail, using one continuous rail for the closure and point makes for a very stiff turnout. Cutting partway through the rail helps a bit, but it's still almost too much for a Caboose Industries 208S ground throw more. To eliminate a possible source of shorts, I want to wire the points to the adjacent stock rail, which requires an insulating throw bar. It's easier to make an insulating throw bar if it doesn't have to handle quite so much force. So I decided to go with point hinges. Point hinges also allow you to spike to closure rail right up to the point hinge - the flexing rail approach requires more unspiked length. More rail spiked down is more rail that's always properly gauged. After some experimentation I decided to make the hinge out of a brass strip.

The hinge test came about as a result of my concern with hinge lifetime. I made up a test hinge on some scrap rail using a thin brass strip (0.010 x 0.060) soldered across the rail gap - the solder a bit back from the gap to leave more length of the hinge to flex. I then made a cam out of a bit of masonite, jammed it onto the shaft of a gear reduction motor, mounted the test point to ride on the cam, and set it running. Here's what it looked like:

Fig. 1 Point hinge testing.

The upper rail is the hinged one, the lower rail fastens the hinge to the motor. I left it running at about 80 RPM for about 21 hours over the course of a weekend, checking regularly for signs of metal fatigue. At the end of the 21 hours, the hinge looked as good as new. 80 RPM * 60 minutes per hour * 21 hours = 100800 cycles, and as my cam in rather square there were 400000 some small wiggles. I figure that should be more than enough for the rest of my lifetime.

Tune into this page at a later date for more turnout adventures.