I have discussed this before but there is little room for pulleys of useful size if they are fixed one above the other. The main problem is the length of the arm does not allow room for two pulleys of any useful size without the radii going completely haywire. I tried shifting the arm pivot above the front changer clamp, to allow a much longer arm, but it looked much too "busy" even for my tastes. Placing the pulleys side by side on a T-shaped arm might introduce skewed forces unless the cable run was altered to suit.
The image at the top shows a rough drawing with a potential solution on the right. A shorter arm can be used without spoiling the radii of the horizontal entry and exit cable runs. The cable must not slip on its route around the pulleys so the arrangement shown on the right offers much better friction due to the much increased cable wrap. In the shallow system the cable wrap is nearer 270 degrees instead of less than 180 degrees in the arrangement on the left. Both systems demand an extra length of cable but should fall within normal, rear gear cable lengths.
The image on the left shows only 90 degrees of cable wrap per pulley. A considerable reduction in desirable cable retention friction is lost. The cable is very likely to slip unless double cable wraps are used. Double cable wrapping will use up even more useful cable length and may be fiddly to achieve due to natural cable stiffness.
Pulley size seems optimum around 20mm in diameter with an arm length of 100mm. A T-piece bolted to the bottom of the swinging arm, to carry the two pulleys, would provide effortless change of radii and thereby, pull ratio. The T-piece need only to be rotated slightly to raise or lower both pulleys relative to each other and the swinging arm's pivot. The change in pull ratio may be easily calculated from the horizontal entry and exit heights of the pulleys before clamping the T-piece pivot/fixing bolt tight again.
In the image to the right the pulleys are shown rather far apart for clarity. In practice they should almost touch for maximum, ratio setting accuracy, increased cable friction and reduced cable pull error over the ratio arm's complete arc of swing.
R1:R2 must match the change in cable pull-ratio required. Note how R1 is measured to the top of the left pulley and R2 to the bottom. The downtube cable must join the top of the left pulley before being wrapped around the second pulley on the right. The cable finally exits rearward below the chainstay to reach the rear changer, cable stop.
I finally solved the problem of finding suitable pulleys. 20mm blind grommets are made of quite hard rubber with a deep groove around the rim. The extra friction on the cable from being well wrapped around the rubber, while secure in the deep grooves, will ensure the cable does not slip. The grommets are the correct diameter to allow fine adjustment of ratio change via the radii from their rims measured from the upper arm pivot. I pressed washers inside the hollow side of each grommet to stiffen them. This allowed them to be drilled for suitable bolts in the lathe so that I clamp them firmly to the T-piece. The washer will avoid the rubber tearing off in use and ensure they remain round under gear cable tension. More washers could easily be stacked to ensure the grommets are fully supported from the inside. Or alloy bushes turned up if greater lightness is desired.
I like these blind grommets as pulleys because they are easily obtained, are made of long-lived, grippy material and require no machining. Once a suitable supporting washer is pressed into the center a normal hand drill will manage a central hole within any reasonable tolerance for concentricity. The pulleys [grommets] do not rotate but are an easy means to achieve the correct radii without causing damage to the gear cable. No cable clamping should be necessary.