Philip I'm sure you could do as follows on your mobile phone no less than I can!
So to take an Austerity 2-10-0 going round (or not) a 3 foot curve, ignoring its pony truck's problems: (I can't find the figures for a 9F just now)
The wheels are each 5'3" apart, (21mm), wheelbase is 21' (84mm). Like the 9F the centre pair are flangeless, the 2nd and 4th have thinned flanges. Just for this example I'm seeing how it works with all the sideplay on the second and fourth axles, outer axles no sideplay:
84x84=7056.
3 feet is 914.4mm. Multiply by 8 = 7315.2. Divide 7056 by 7315.2 =
0.96mm would be the sideplay required each side for a flanged centre driving wheel. 75% of that is
0.96 multiply by .75 =
0.72mm each side.
Frames would need to be 1.5mm narrower than distance between wheels inside bosses (leaving a full 0.06 margin of error!....so that's when the fun starts, getting that in actuality - but at least we would know what we were aiming for).
**************************************************************************************************************************************************************************
We could reduce the requirement by taking into account the slop in P4 and any gauge widening, which I think is what Bill is advocating. (Worth mentioning the 9F had a minimum radius of 400 feet - 5 foot 3 inches for us...!)
If we have a 17.67BB the slop each side is 17.67 + two flanges (0.8) = 18.47. Take that away from 18.83 = 0.36 total slop,
0.18 each side.
A spreadsheet also from my son shows that the P4 triangular gauge in theory gives 0.13mm gauge widening at 3 foot. (As per prototype we would have 0.25mm
). Can't quite work out in my head whether that would apply each side or whether we'd have to halve it, but I think it would apply in full.
0.18 + 0.13 = 0.31mm.
So take that from 0.72, our sideplay requirement is down to 0.41, or frames 0.82 narrow - i.e. making them 1mm narrow would give nearly 20% margin of error.
That's taken me around half an hour, but probably, if I were making the thing which I'm not, saved hours of experimentation, and given me a certainty of what should or should not be possible.
I wrote all this down here so that it can be corrected if it's wrong