Sorry for the delay. Work, sleep, all those kind of things getting in the way.
Decided to check, or should I say re-check the squareness of everything. Noticed that the z wasn't quite square to the gantry so I'm in the process of taking it all apart and realigning it all and triple checking for any binding. Should have it all back together and hopefully aligned in the next couple of hours. Fingers crossed.

 

@Gray, you were spot on sir. The problem is in the gantry construction, i.e. Ever so slightly twisted towards the centre. There is a length of 2040 running horizontally sandwiched between two bits of 3mm al plate for the top rail of the gantry which is slightly lower on one side, causing the z to deviate from square with the bed. Rectifying the problem now.

 

Finally got everything as square as I possibly could with the limited tools at my disposal. Went for another test cut to find that the cuts were only marginally better than before.
I guess at the end of the day, each step towards trying to eliminate as many mechanical problems helps. After much head scratching, umming and arring, I finally decided to turn on the backlash compensation in mach3. Things are looking a lot better already. Even with only around .3 of a mm on each axis.
Now it's time to start dialing it in!
I know it's an easy fix but it will suffice until I get the second lot of belt glued into the v-slot, which by all accounts seems to clean up a lot of the backlash problems some people have been having.

 

Today is a good day! Installed the dual belt and what do you know, backlash all but gone. The machine is finally cutting like it's supposed to.

 

Hi guys, back again. Well, I have been cutting and making stuff, but not quite there yet as far as accuracy goes.. Still getting slight deviations of up to a millimeter i.e.
Depending on the start point, the circle start and end points do not meet perfectly. Still trying to work it out. Most of my trouble I think resides in what I'm calling the Y Axis(the gantry/carriage).
If push hard enough on the gantry plates (x direction) I can get it to move a bit under a millimeter.
One thing that was interesting in the many tests I have been doing to figure this out is as follows:
If I type G00 Y10 for example the carriage moves 10mm as expected, but it seems to take a good second or two for the last 1/10s of a millimeter to hit 10mm exactly. I have been measuring this with a dial gauge. Anyone have any ideas what could be causing this? I have repeated this test at various points around the machine to get similar results almost every time. There doesn't seem to be any real binding to speak of.
So far I have managed to work out/measure that the backlash for x is around 0.52mm and 0.675 for the y. This was tested by moving the axis in one direction a fixed mount, then reversing it by the same amount and then calculating the difference.
Now when using these figures in the anti-backlash compensation in Mach3 it actually makes things worse if I do a circle.
As far as tool deflection goes, I am going to raise the bed 50mm, thus reducing the travel of z axis to around 50mm vs 100mm which I think was a tad over zealous as far as ridgidity. This will put the spindle mount more inline with the gantry vs being well below.
Just out of interest, what sort of backlash should I expect using GT2(3mm pitch) belt?
The doubling up of the belts helped a lot, but there is still a little ways to go till this beast is dialled in.
I'm almost contemplating going to T5 1" width belt. But then I think to myself "The OX" uses GT2/3 belt and nobody seems to be having any major issues. There has to be more to it.
Hope this all makes some sort of sense.
Don't hesitate to ask if you are not sure what I mean.
I'll be happy to go into this more if it means I get a little closer to figuring this out.

 

Thanks for the thought FRIDGE.
Yeah, rack would be awesome. I think it would definitely help a lot!
It would mean a complete redesign though. Not completely out of the question though. I have been looking at that 3:1 reduction setup on cncrouterparts wondering how I'd fit it into the current design. Think this would help a little too.
For now, I'm going to start by raising the bed to reduce the z travel. This should definitely increase the overall ridgidity. Then maybe try and go with slightly less than the current 20 tooth if I can find them. 33.33333 steps per millimeter is just not quite enough purchase. probably won't gain a whole bunch of torque but it might be worth a go?

 

Hi amos, been going through a similar sort of thing. There are so many variables at play that could be causing the problem. Just out of interest what's the height of the gantry above the work surface?
If you grab the tip of the bit (while the spindle is off! ), how much movement/ play do you get if you push and pull on it with a bit of force? What I'm getting at is basically the amount of z deflection could be an issue. It has been for me at least. So much so I have now raised the bed on my machine so that the router mount is pretty much in line with the gantry for most of the work I am doing. (My machine is not an OX, just to clarify.) Not really doing any sort of 2.5/3d cutting at the mo, so 50mm of z travel total is more than enough for what I am doing.
Calibration is definitely a tricky one for sure. As I said earlier, there are so many variables. As my machine is not an OX. I have had to go through all the possibilities by trial and error and have found calibration is kind of the last step, once you have nailed down all/most of the mechanical issues. Which for me has been quite a journey. Getting there though.
One thing that has helped in my build tremendously is the addition of the second belt. This is where you use double sided tape to stick one belt facing teeth up inside the v slot and then have the normal belt on top of that. Bear in mind I am using GT3 belt in this application as the thickness of the belt/double sided tape is key here, in order to make sure that both belts are in full contact while under a v wheel. Recently snapped a belt (GT3) actually and didn't have any more to replace it with so I managed to do the double belt thing reusing some old GT2 belt. I didn't remove the GT3 that was in the v slot, just laid the GT2 straight on top without sticking down. By tensioning both belts, which was a little tricky, it seems to be working a treat.
Anyway, enough rambling from me. If you solve your problem, I'd be interested to hear how you did it!
Happy cutting.

 

Amos, have a look at the shapeoko wiki. Some really helpful info.