[matthew798]

Well let's consider the facts:

It used to work a little while ago, so barring some freakish motor damage or you doing some modification your not telling us about, it has to be something that you've done to the machine since.

The only things I can think that you might be doing to the machine are:

Changing bits
Tightening rod nuts
Changing feedrates

The most likely culprit is the bit change. Are you using the proper collet? It may feel like it fits, but you need a collet made for the specific size bit you are using. I ordered a pack of 20 different collets on ebay for 10$. Like I mentioned before, are you tightening the nut enough? Also, you can check manually for any play in the machine. Try to move one axis at a time and feel if its loose, even a tiny bit.

If, like you mentioned, the axis are moving properly, then it HAS to be the bit. Let me know how it goes!

[dmha]

Well let's consider the facts:

It used to work a little while ago, so barring some freakish motor damage or you doing some modification your not telling us about, it has to be something that you've done to the machine since.

The only things I can think that you might be doing to the machine are:

Changing bits
Tightening rod nuts
Changing feedrates

The most likely culprit is the bit change. Are you using the proper collet? It may feel like it fits, but you need a collet made for the specific size bit you are using. I ordered a pack of 20 different collets on ebay for 10$. Like I mentioned before, are you tightening the nut enough? Also, you can check manually for any play in the machine. Try to move one axis at a time and feel if its loose, even a tiny bit.

If, like you mentioned, the axis are moving properly, then it HAS to be the bit. Let me know how it goes!

You are right about mofication. I had been pissed of with this router since I bought it. This Info maybe much usefull to many CNC3020 or similar Routers:
First I found that the motors cable are bad, since I measure wire by wire from controller to Motor the Connection with Multimeter (conductivity with the buzzer)
I was so suprized that a wire of X-Axis at a specific turn inside the plastic chain stop making contact. So I replace all the motors wires.
Then I realize I m loozing steps in X-Axis again, so I changed the capacitors of the Board. It improved much!
But again sometmes steps were missing. I suspect a bad design of the controller and this has not selection of division of steps (i.e if you want to make it half step, 1/4, 1/8 or 1/16.
It was fixed on PCB to 1/16.
I thought why the heck do I need 0.00125mm step on PCBs???
Isn't it a big unneccesary risk of losing steps?
So I decided to design my own controller thought 0.01mm is enough (half step in my case)
Now with my controller Works good and I have selection of microsteps resolution.
As I said before I made for many modifications like tight again the screws, and play with different feedrates, accelaration and microsteps resolution.
Also I changed the mounting of PCB from double sided tape to pins holding.
There was also a hit of the spindle on the table with speed
After all these modifications I noticed the problem we are discussing.
I suspect if it is something wrong with pcb-gcode or Autoleveler and I run a Mach3 wizzard for cutting Circles.
This circles was VERY GOOD!

After this I realize that the machine is OK, so there maybe a matter of Software settings
I changed for half step to 1/16 microsteps , 1/8 and finally again to 1/2 step.
Change to G-code the parameter "max isolation", it was "0", i changed to 0.2mm.
Feedrate to 1000mm/min and acceleration to 100.
After all theses midifications just the last hour the pcb engraved fine!
I design just triangles and circle on the pcb for testing at width 0.1 0.2 0.3 0.4 0.5mm and the pcb seems almost perfect now !!!!! I ll post the Picture later.

Sorry But I still dont know why before 2 days was "some tracks wider and some thiner" and now they are very good!!! I neved milled 0.1mm track!
So because I dont know what was the reason, I am afraid if this happen again in the future

P.S I also suspected if the PCB is moving while milling because the pins holding the Board have 1.13 mm diameter and the dril I used is 1.2mm, so I thought maybe it is moving. I mounted again with double side tape.
Same result...
Maybe the Wood under the pcb is moving? Tighted the screws more... just for the history...

I am using the machine's collet I never changed it. It is ER11 - 3.175 and ALL my bits are 3.175 so they fit in this collet. Because I also suspected the collet I order a NEW one. Now it seems it wasn't the reason since I think it plays a role only for the run out. Please pm me about this Super price 10$ (the link) for the set of collets. It would be usefull to have such a set for future Works.

[dmha]

This is the today work.
Tracks are with order 0.1 0.2 0.3 0.4 and 0.5mm wide
Dimension of the pcb is 30 x 30 mm.
Feedrate = 1000mm/min , Accel = 100,

PCB-gcode settings : Isolation : Default: 0, max: 0.5mm, Step size:0.25mm. Etching tool size 0.55mm.
Milling Feedrate = 250mm/min, Z down = -0.25mm.

Controller steps are set to half step (1 step = 0.01mm).

Bit nominal 0.1mm / 45Deg (chinese)but it is actually engraves with this settings around 0.55mm.
Probed with Autoleveler.

It is a very satisfied result with such a router.
I guess it would be better with a spindle setup with less run out.