Calibrating Steps/mm for a minimill?

Hey guys, first time poster here. I used the tutorial of Martin Barfoed to set the steps/mm for my minimill.
The steps seemed to be set right, but I just tried to mill some aluminum (just a tiny piece of strip from the hardware store)
Was going to mill a square pocket 5x5mm and 2mm deep. Well, it came out as X=4,88mm, Y= 4,52mm and Z=1,52mm. Now if I plug this into a simple step/mm calculator the value's overshoot by far (crashed my endmill the following run after adjusting the steps )
Anyone got a way to calibrate the minimill? The openbuildsCONTROL only seems to have it for gantry-style routers.

Also: I seem to have some flex somewhere, if I push down against the endmill I can visibly see it deflect :s but I can't seem to find the origin, anyone got any tips?

 

but aren't I limited by the precision of my hands this way? meaning, how close I can stick a tape or measure the difference afterwards? I'll be sure to try it out!

 

Good thing you mention it, I'm still 'learning' my feeds & speeds. I used a cheap chinese 4 flute 1,5mm endmill with a 6mm shaft and 4mm long flutes. So I'd say deflection would primarily be in the cutting part of the endmill. As for parameters:
10000 rpm (setting 1 on makita RT0700C)
47,1239m/min surface speed
700 mm/min cutting feed rate
0.0175 mm feed/tooth

roughing stepdowns of 0.4 mm and a finishing stepdown of 0.2 mm
finishing stepover of 0.15 mm with 500 mm/min feed rate.
all climb cut

I'm thinking the problem might be in the 4 flute? Chipload doesn't seem very high, but I'd think to achieve that chipload the machine just pushes too hard and might not (yet) be rigid enough for this? Also, the cut went pretty good, but that 'finishing pass' (or I suppose it was the finishing pass) seemed waaaaaaay more agressive. To illustrate: I ramped down into the pocket, cleared it out without any hassle or heavy sounds, then did one last pass at what seemed 3x as fast. (Would seem logic, as a finishing pass and to achieve the same amount of chipload when just cleaning up the walls, but it didn't really sound like that )

 

that's basicly how I had calibrated X and Y, so I'm guessing it must be tool/spindle/machine deflection that caused the cut to be undersized?

 

Alright! So to simply recapitulate; the calibrated steps/mm without any load can just be used, right? Secondly, I'll check for endmill specs. It'll be interesting to bring into perspective how it may attribute to the undersized slot Lastly, I do notice some kind of deflection on my machine when I press unto the spindle or spindlemount. now, this could be because I've made the throat a bit deeper and so it becomes spindle heavy (analogy to nose heavy on a plane). Perhaps I'd ought to stiffen that further up or reduce throat depth by a bit. That could perhaps explain the lost in depth and Y-axis (as these are the two axis mostly influenced by this "nod-type" deflection)

Edit: Would it help to post some pictures of my minimill setup? As I've changed a few things. I stiffened up the column and made it heavier by adding a second C-beam, creating a channel and filling that up with epoxy granite mixture. I've also added linear rails on Z and X (Y is still in the mail).
I'm using some 'prototype' parts printed in PLA for now (carriage plates for the linear rails, fixture plate for X and a spindle mount for Z for easy tramming)