Yeah those videos make it look so easy...

but it ain't like that.
cool logo video how they straighten out the fringes...and that is how it does happen over many tests..and figuring stages.
read this link...this is how a machine parabolizes
you want to control it by a computer...does pretty much the same thing...
http://www.astrosurf...achines_eng.htm
here's log of making a mirror
http://www.astrosurf...to_12,5_Eng.pdf
lots of position changing and not staying on any zone for much time...ie technique.
Thanks I would love to learn these techniques someday..... 
The older polishing and grinding machines I have seen at the beginning of my career often mimicked the motions of human polishers and grinders, so it had some large flat tool like a traditional pitch lap that was suspended over the optic, moving back and forth over its whole surface, with many interruptions and measurement steps in between based on a trial-and-error strategy, so the human operator's ability seemed to be very important as you say (I use mostly flat or spherical optics, only a few aspherical or elliptical). You could see by eye how the grinding or polishing was going on, I liked to visit those places.
But these newer machines I saw were similar to the tradition of metal machining (instead of controlling the shape of slopes, it is thinking in terms of removing Z microns from the surface at position X-Y-R-theta-phi)...... so had a very large circular grinding tool that was rotating rapidly, and suspended vertically with respect to the optic, contacting a small area at a time. It moved in a completely different way than shown in your web page, it did not make large strokes across the whole optic, instead it concentrated on smaller areas at a time. Only a small number of tool passes were needed as I remember...... like 3 or 4 (coarse fast - tool change - fine slow), even I could do 2-3 micron and a shiny surface! I was so happy.
There was not a lot of possibilities to adjust the machine during machining - I cannot see the workpiece because there is a protection door and the room inside is flooded with a shower of coolant..... and even if I press my eye against the window, sometimes the tool is far away from the window so I cannot see anything. If I get worried and open the door the safety trips the machine mid-program, and even can damage the workpiece and tool if I do it at the wrong timing, so everything was done automatically ("door open" was the last command of the program). Door could be bypassed and kept open, but I was always told then the temperature gradient could make the machining precision worse, it is a safety hazard, and the coolant could mess up the floor. During machining, I was looking mostly at the computer display, which sequence was now active. I could interrupt or modify the program in real time, but I avoided that because often I would make a mistake in my haste. Maybe I am wrong; you are right, I do not know how they do the last polishing passes to get sub micron figures. But you see the advantage of this method is that many kinds of aspheric shapes could be created and tested, not only paraboloids or hyperboloids. Even arches or half-cylindrical, half-spherical shapes could be created with similar high precision.
I could simulate how the optic would be machined using CAM software. Measurement was like, "wait, select measurement tool, goto coordinate position, measure". I know it is more complicated than that....... but I could easily measure 1000 height points if needed and spit out a data file, and control the next pass automatically to compensate..... though I did not do that. I was always nervous that I might collide the stylus or laser device to something (the window is so small!)..... and well it was implied I should not play with the machine anymore. The machine I used did not have a laser interferometer to measure the optic, so I don't know how that is put into the loop. Once the program and prescription were fully set up, however, many identical parts could be ground. Of course my programming was terrible so many failures and damaged workpieces.....
To grind a very clear surface, I remember the selection of the diamond tool and its angle and adjustment were extremely important, if wrong the surface would lose smoothness, and we let the supplier adjust this and changed it frequently. This tooling was horribly expensive as I recall.
I was told in a factory, maybe 1 person could be looking at 3 or 4 machines as they simultaneously ground optics. This part is speculation, but the single company we are discussing about now: mass-produces Schmidt correction plates and Cassegrain mirrors (5, 8, 9.25, 11 inches) and Newtonian mirrors (3, 4, 5, 6, 8, 10, 12, 14, 16, 18, 20 inches) and Ritchey-Chretien mirrors, as well as all diameters of achros and apos (80, 100, 120, 150 doublet and triplet) using ED glasses, fabrication of lenses for eyepieces, also the AR and HR coatings of all optics, design, fabrication, and assembly of the telescopes and computerized GOTO precision mounts and eyepieces; all this without any production delays at extremely low cost, most of the important tasks within 1 factory according to reports. Is it natural to assume armies of trained opticians are doing this? Specialists of mirror fabrication are often not the specialists of lens or prism fabrication, then there would be too much redundancy and one would need to fire or hire people based on the production needs. It would seem to make more sense that there are a few highly trained specialists who initially design the CNC routine and set up the machine and prescription, then the operators do not need to know the tiny specifics. When a production run of a certain optic is needed, the program and maybe part of the tooling is exchanged........ I guess? Otherwise such a huge number and variety of products could normally not be maintained, at such an incredibly low price point and relatively high quality.
Edited by X3782, 15 August 2018 - 07:46 PM.