Case in point, the skilled machinist knows how to code backlash compensation. There are several G code routines to accomplish that depending on your controller. And what knee mills rock? In standard operation, the spindle is always dead center of the knee. How much error does that generate from "rocking", six zeros past the decimal maybe? I get confused over the insistence that good work isn't possible with reasonably good equipment and attention to detail in the process. I'm not intending to be snippy, I just don't know how an ABEC 9 bearings and $2000 a pop ballscrews are going to make the difference between perfection and scrap in a one-man home shop, dig?
I have never advocated for high end components here for the hobbyist???? I didn't bring 1 micron scales into the discussion!. Just the opposite. I brought up the cost of the components to show the folly in it for the "typical hobbyist". Several threads, ones that you yourself have been involved in I have advocated for slow speed but high torque closed loop steppers over servos as the cost/benefit ratio is not worth the higher rapid speeds for a hobbyist and I realize, that decent rolled screws and acorn or such control can result in mediocre end of CNC's but quite an acceptable accuracy for many hobbyist needs. (.002 to .003" reliably).
In my opinion its a large leap in cost to make any manual machine into my idea of ok precision (.001"+/- reliably) Just the cost of a low end controller that can handle true basic feedback such as an Oak is triple or more than say an Acorn.
In post 7 of this thread I suggested in the last line that if better precision is desired, then an older bed mill than needs a control retrofit would be a better option. A position that you and Jim liked in post 21 ??????? Seems if you have a difference of opinion than Jim here you get attacked.
As far as rocking, I have not seen a knee mill yet that doesn't rock as you move X axis from one end to the other unless the knee is locked, most of the time several thousands of an inch. Maybe most projects it doesn't matter but many times if perpendicularity of drilled/bored holes is required it makes a difference. One example is I machined some combining gear box housings half's and If I didn't lock the knee when boring one bearing bore to the next, the shafts inserted into the bearings wouldn't be parallel to each other and would have troubles when being assembled to the other half of the gearbox housing.
In post 15 I mentioned point to point compensation for drilling holes G60. Enlighten me for backlash compensation G code for circular interpolation?
There is no substitute for a tight machine. I'll concede that you are not going to convert a <$10K knee mill in the average home shop to the accuracy of a multi-million $$ DMG-Mori by adding linear scales. Will my machine interpolate perfect circles? No. I have about 0.004'' backlash in the Y axis and it will hold 0.001'' on roundness. It's just a matter of the system being able to compensate for the backlash as best it can. Linear move deviations are normally better than I am able to measure. My knee mill is actually just a little better than my Haas.
Really the most important parameter is repeatability. If the machine will repeat, then you can compensate for the accuracy. If you need perfectly round holes, then use the boring head or a reamer.
First, you were the one that brought up uber accuracy in post #4 talking about 1 micron scales. What good is 1 micron(0.0000393701") doing when you have .004" backlash or 101 times the resolution of a 1 micron scale? Even 10 micron scale(10 times better resolution than the back lash) , or even 5 is a much more cost effective solution. Many lower end commercial machines(still in the 10's of thousands of dollars only have 2 micron scales, and the cost as you go up in accuracy is exponential. This is the only reason I started down the path talking about higher end machine components. The rule of thumb in metrology is that you measuring tools need to be 10 times the accuracy of what tolerances your trying to measure, why do you need 100?.
Second, who wouldn't use a boring head or reamer when ever possible, even in the 23 million dollar machines I worked on most of the hole tools were such, Its the only way to hit 13 micron+/- tolerances on the parts I helped make. But boring heads/bars are difficult to use on under cut grooves, under cut bores, tapered bores and large threaded holes, but accurate circular interpolation makes it easy and the tooling cheap for one off/low count jobs, like many hobbyist have?
Also boring heads/are not always possible to use, and that is what's liberating about CNC. For example, one day I was tasked with making a replacement oval hydraulic piston for an autoblok steady rest. I don't know how I would of done that on a mill of any kind with back lash and have good cylindricity on the ends. Let alone where the circular path goes straight if it had back lash would result in a tool mark which would allow fluid to pass. I don't know how the seal groove would be done either without CNC? Done on a VF3. I am baffled; because while I don't consider haas to be high end machines by any stretch of the imagination; all of the ones in the tool room I have used were able to at least achieve .0005" tolerances without much issue if paying attention to the tools.
Also CNC allows good accurate profiling that previously was only possible with tracing attachments an pantograph milling machines. No back lash allows lead in/lead out without tool marks on perimeter outside profiles as well. Back lash limits climb milling and its superior surface finish to conventional milling.