Hi Rokag, the problem is that you can't mesh a straight line with a helix angle....the corners of the Acme thread will foul the corners of the roller.
The bigger the roller the more of it's flanks will interfere with the screw flank corners at top and bottom.
What will happen as you have found out in practice, is that the steel Acme screw thread cuts into the bronze straight sided roller and wears it away......this only stops happening when the bronze roller hits the bottom of the Acme thread, and gives you sideways clearance called backlash.
Having studied this phenomena by meshing a screw thread with a straight sided single grooved roller for the study purpose it became apparent that the bigger the roller the harder it is to get the two to interface on depth....about 2/3 rds max.
The only time a screw thread will mesh with another screw thread PERFECTLY is when one is right handed and the other left handed, but it does not work as a drive as one cancels out the other for forward movement, but the two threads are meshing because the opposite sides of each one are at the same helix angle......too bad they don't work.
I suggesrt that the smaller the roller diam the sharper the curvature of the contact with the Acme thread which means that there is less chance of the corners of the Acme thread cutting the straight sided roller thread.
For instance if you had an Acme thread of 50mm diam and a roller of 15mm diam you would have less line contact at the helix/groove faces at the centre line.
If the 15mm roller was meshing with the Acme thread it would have to have bearings for radial support and also thrust races for axial thrust resistance against the frame of the nut, but this would make for very small bearings and even smaller thrust races....not very practical.
So to solve the problem I suggest that the roller is made a small diam, solid with grooves, no bearings or thrust races and held against the Acme thread by two more larger rollers.
The smaller roller is on the centre line and the two larger rollers also grooved are above and below the centre line, 30 deg approx, so trapping the small roller in a triangle....the small roller can now float there both radially and axially held in place by the big rollers.
Having big rollers you can now put substantial needle bearings and thrust races to carry the load which is transmitted axially by the large rollers to the frame of the nut.
To sum up, the Acme thread will apply pressure to the small floating roller axially, and this force will be transmitted by the small roller grooves to the large roller grooves and then to the frame.
The small roller will have a better time interfacing with the Acme screw thread due to it's small curvature that makes less contact with the screw flanks than a large roller.
The larger the roller the greater will be the interference with the screw at the thread top corners.
This is a study using a single roller pack on one side, but for a practical drive it would have at least 5 roller asssemblies spaced round the Acme thread.
It would be quite bulky, but you could have a low profile model by having just two assemblies one on either side of the Acme thread and to prevent the Acme thread from deflecting up or down under load a pair of plain ungrooved rollers would be placed in the frame to contact the top and bottom of the Acme thread.
Adjustment for mesh would be by one of the large rollers having an eccentric spindle, the other fixed.....the small roller floats of course, held in space by the two large rollers.
Ian.