oh also International tolerance grades (wikipedia) are a better measure of accuracy, sorry.

be better with it all in mind i.e. vibration or excessive play in bearings, lack of workspace etc. is not something that could practically be compensated for in software, so maybe focusing on that sort of thing would be good. Also leaving bolt on points for suitable sensors etc.


Hypothetically, a milling machine that could also turn the workpeice, with essentially all the cutting power coming from the mill head though, could produce rotationally symmetric parts that are as accurate as the mill is. This is a variant of so called multiaxis milling and I have seen pages for producton cnc multi axis machines that are advertised as a replacement for cnc lathes. In between is "live bit" lathing, where some but not all of the power comes from a rotating bit, this is apparently common and desirable on cnc lathes. Since the forces are I think lower for a millor router (or grinder) bit due to the high speed and the power relationship mentioned above this might make a lot of sense. Question for machinists: what are the downsides to this approach? Is there anything you can't do with a multiaxis mill that you can do with a lathe? poor surface finish?

Sensors:
A single chip scale camera like often used on laptops might be useable as a digital microscope if the lense location was changed. The equations for focal lenght are high school physics but I forget the details (fresnel equations?) but if a lens is focussed on a 1 square mm area, and the sensor is 1000 pixels by 1000 then that can give a resoluion of 1 micron. such sensors are small so the optical paths can be too. Note a color sensor has 3 pixels, also pixel interpolation can increase effective resolution.

Such a sensor has a myriad of uses when combined with software, if it were directed at the journal bearings of the x y axis then it could recognise the scratches on the surface and determine the position of the bearing to a high precision.

3d point cloud generation using stereo imaging may be practical here (see the reprap metalica project documentation), so two could be used together maybe as a scanner. Maybe one with multiple led light sources from different directions that turned on and off? Saw hewlett packard project doing this once I think on the antikithera mechanism project. Also: as an image goes out of focus the points become circles. These circles are still detected and you could hypothetically compute the image information from their locations, I have heard I think of images being "digitally focussed" in this way. that would effectively increase depth of field

- an array of microlenses focusing each cone of the image feild on a designated pixel ? Could eliminate the focus problem Maybe such sensors are available.
-mills like the modela have sensors some of them optical check out how suitable they migh tbe
-could maybe use interferometry with stufff from old cd players for encoders and maybe a slow scanner of it is not available in product form, the gorilla tech scanner apparently works on somtehing like this don't know the details
-to compensate for bit wear etc. don't really need a scanner per se, if the measurement points are far appart on the workpeice that can still be fine, so a single distance measurement device could be okay, to use it while working would have to be have a short exposure time so to speak or the motion would blur the measurement
- this depends on the light sensitivity of the image sensors, although the image can be very brightly illuminated since unlike with a normal microscope overheating is not a problem
- top eye software might be useful with cheaper hardware says it is "open" http://www.micro-optik.com/htm/top-eye.php
-example excoder for calibration http://www.aculux.com/Alignment.htm maybe others that are cheaper, we can have the area dark if filtering out other light sources is difficult , maybe could measure the resonant frequency of a gap using an optical photodiode and laser diode, measuring frequency accurately and precisely is realtively easy, if length is 400 mm , resonanly freq is 750 MHz, if can measure frequency to withing 1 kHz precisely that corresponds to 400,000 microns/750,000 = 0.53 microns, since the light beam is straight this could potentially be used for calibration. Also highly resistant to dirt etc., could be used as encoder, this could also be done using the journal bearings as a resonator, skipping the need for any optical equipment and giving high precision, but the journal may not be straight though.

Could use a peice close to the bit for the mill, pont the 3 laser sensors at it (maybe have it a small white retroreflector like stopsigns, the geomerty of the micro prisms migh tbe an issue, changine reflection path depending on angle? still maybe okay for several micron accuracy) and measure the actual position of the head of the bit, if a diffuse or retroreflector is used, the pointing og the sensors does not have to be very acccurate, might even be better to keep it stationary for precision but then the power of the treturn signal might be an issue, it coul be made to operate ontly in the dark.

another ption might be to use a laser diode, very small light source basically,maybe a fisheye lense, flash it at a known frequency and then measure the resonant frequency of the round trip, but then inductances of the wire etc could interfere. Wan to to send it there and back along the same or known path ideally. An electrical engineer may be able to do things like this readily, the high speed counters are available to do it digitally even, "direct sequencing" radios are an example, or hybrid analog/digital approach similar to a phase locked loop? Highly accurate can oscillator components are like precision and accuracy in a tin for a couple bucks.

ideally 3 measurement devices could be placed on the mill bed or side of the lathe. They could measure the distance to a point on a calibration bit or designated point og the xyz carriage, meant to reflect the signal or whatever, to very high precision and preferrably accuracy.
Only one sensor and a rotating 2 axis turret of sorts could work, but then you need to know the angular position of the turret to very very high accuracy and precision which is hard, and the mechanical parts would have to be machined, would wear etc.

could take a hybrid approach with the 3 meausrement devices each with emitters and recievers using the same fisheye omni lense and a half silvered mirror, attached directly to the bed, then have a turret whose purpose is to shade the sensor from all reflections except the ones from the expected position of the callibration reflector, thus eliminating echos from other sources. That could so work.

Holography could be used too, should be in the dark, needs camera, laser with reasonable coherence length (diodes can in fact work sometimes) and software to keep track of the fringes in the image, use a white target attached to the cutting head, one problem might be that it would be too accurate, fringes would move too fast due to vibration for the camera to keep track of, it is also open loop, might be useful for calibration though, in that case machine would have to be designed with the system in mind, places to bolt put the lasers etc.

those digital calipers are very accurate but relatively inexpensive, they use capapcitive sensing (bit like a touchpad on a laptop), an array of electrodes presumably etched with the help of lithography (organized herarchically) detects the position of the ground electrode which is the lower jaw of the caliper

I think this guy:http://openlathe.wikidot.com/controls-discussion says he was getting 2.5 microns over 24 inches with digital readout model, whatever encoders are used for that, need to check prices