Desk CNC 3020 (or similar) accuracy and practicality for SMD PCB prototyping?

[sigxcpu]

For hobby use Eagle->pcb-gcode->Universal G-Code sender->GRBL->chinese controller (3040 in my case) is a very good path. I've done tests at 0.5mm pitch and they came pretty good.

[mikeselectricstuff]

Looking for Chron Mill on YT brings up some serious CNC porn! I particularly like how trapdoors open, stuff comes out, does something then dispappears back behind the door.
 

[G0HZU]

G0HZU, I should maybe of been a bit clearer and mentioned modern machines.

Ok thanks for the extra info...

Mine only has crude Z capability via a solenoid and some manual controls.

It's also worth checking the tool change vids, as the unique feature of this mill is the super fast toolchange with the spindle still turning.

The worst thing on mine is the manual tool changing. I think the T-Tech engineers all went on vacation when finishing the design on mine and someone from WW2 designed the tool changing system on my machine. It is a basic manual toolchuck with a little hex lockscrew. Totally horrible and it is the worst thing about the whole machine. I look with envy at any machine with decent auto tool changing.

T-Tech kept advising me to run it at >50"/min for some tools. I just checked the manual and the max milling speed is between 60"/min and 80"/min. However, I rarely run it at >24"/min. So I guess it is just above 'slow' by modern standards. It sounds wonderful and smooth and tight when running at 50"/min but my ignorance of its capabilities means I apply sympathy to the mechanics and run it much slower.

sigxcpu's image at 0.5mm looks pretty good. What amazes me about my machine is that I can mill part of a board, switch it off and carry on again the next day after re initialising it and it will still mill in perfect alignment on the second day. When i look at the relatively crude mechanicals of it all I still can't fathom how it is able to do this so precisely...

[mc]

Mike, CNC is very much a case of, how big is your cheque book?
If you want cutting edge, you need to check out the Japanese 5 axis machines.
Here's a couple of Daishin Seiki's videos.
This one shows how intricate they can do, but the machining speed is pretty slow due to it being titanium-

This helmet video shows faster machining as they're starting with an aluminium billet, although the speed has to slow down for the smaller tools -




I've actually got a 3040 machine, and I was pleasantly surprised by how accurate the basic machine is. The biggest two issues I found with it, are the thrust bearings used to support and control the ballscrew endfloat are utter rubbish, and the motor to ball screw couplers are a pretty rubbish design (they're prone to failure, aswell as being a bit springy under load).
I replaced the thrust bearings, which are just standard radial ball bearings with good quality ones, and it made a major improvement to the machine. I've done nothing about replacing the couplers, as I'm only using mine as a CMM/probing machine, so load is minimal, plus I fitted glass scales to make it closed loop so any play in the couplers/ballscrew doesn't affect the probe trigger point.

[mikeselectricstuff]

I've actually got a 3040 machine, and I was pleasantly surprised by how accurate the basic machine is. The biggest two issues I found with it, are the thrust bearings used to support and control the ballscrew endfloat are utter rubbish, and the motor to ball screw couplers are a pretty rubbish design (they're prone to failure, aswell as being a bit springy under load).
I replaced the thrust bearings, which are just standard radial ball bearings with good quality ones, and it made a major improvement to the machine.

Improvement in what way? An issue I have is vibration/tool judder when cutting acrylic. Are these mods likely to improve this?

[mc]

Improvement in what way? An issue I have is vibration/tool judder when cutting acrylic. Are these mods likely to improve this?

I noticed the thrust bearing issue the most due to very noticeable roughness when jogging. I initially thought it was perhaps a bit swarf in the ballnut, however when I stripped things down, the ballscrews were smooth to turn. It was when I reassembled things and tried adjusting the ballscrew endfloat, the roughness reappeared.
When I stripped the shields of the bearings, they only had 6 balls and lots of play.

Roughness may contribute to vibration/tool judder, however it could just be an issue with general machine rigidity. I've never personally cut plastic on a mill or router, but I do know it can be a major pain to get a good cut. You need to ensure you're taking a reasonable cut and not rubbing the work, and also you need to make sure you have no heat build up. There are cutters optimised for plastics, which generally have far sharper edges, and are available in specialist materials to improve cutter life.

[Fred27]

I have no problem cutting acrylic as long as it's cast and not extruded. Extruded melts too easily (and it stinks when lasered too). My mill is a much smaller Proxxon MF70, although I'm thinking about a 3040 for a larger cutting area.

[johansen]

Improvement in what way? An issue I have is vibration/tool judder when cutting acrylic. Are these mods likely to improve this?

I noticed the thrust bearing issue the most due to very noticeable roughness when jogging. I initially thought it was perhaps a bit swarf in the ballnut, however when I stripped things down, the ballscrews were smooth to turn. It was when I reassembled things and tried adjusting the ballscrew endfloat, the roughness reappeared.
When I stripped the shields of the bearings, they only had 6 balls and lots of play.

I thought 608 bearings had 7 balls, are yours even smaller?

If you want to throw money at it, VXB has 708A bearings and they are the same size as the 608 bearings which are in most small milling machines.
at 13$ each though its a bit insane, for that price i would rather build a new bearing block and use 10mm or 15mm bore DG bearings (which max out at about 15 degree contact angle)

it is pretty easy to brinell 608 bearings when preloading them but often times its contaminates in the bearings, not a failed race that is causing the problems.
yeah its a lot of work but you can pry off the seal, wash the bearing out and regrease it and put the seal back on it.
but at 79 cents each its not worth the effort.

--you do have preloaded bearings, right?

[mc]

--you do have preloaded bearings, right?

Of course. This is one of china's finest engineered machines 

The bearings were complete rubbish. 6 balls with noticeble play/roughness turning them by hand. I thought I maybe still had them, and was going to take some pictures, but I must of thrown them out.
I'm not sure how many balls are in a good quality bearing, but it's certainly more than 6.

I did look at various options, but fitting some good quality standard 608's was the quickest option. If I had been aiming to eliminate play, then I would of modified things and used some proper thrust bearings. Either needle roller or AC with a custom housing, as the ballscrews aren't long enough to fit a proper bearing block.

[johansen]

The 608 bearings i bought for 79 cents each have 7x 5/32" ball bearings in them (yep, its 3.969 mm)

the 708A bearings have 10 or 11 balls in them, and combined with the 30 degree contact angle means you can double the load, that said vxb lists the same load capacity as the 608 bearing which doesn't make any sense.

you should be able to stretch the leadscrew by preloading it from both ends rather than the usual double bearings on the stepper end.
alternatively you can use the bearings in the stepper to support the ballscrew 

[mc]

708A's are not commonly availble in the UK.

Preloading the full length of the ball screw is bad practice, especially on a machine where the material it's mounted to has a different expansion rate. One end should be fixed, using a suitable thrustbearing setup, and if long enough/turning fast enough the other end should be supported by a floating bearing.

My 3040 has two 608s setup to provide the fixed end at the opposite end from the stepper, with a solitary bearing (IIRC it's bigger than a 608) to provide the floating support at the stepper motor end. The stepper is then connected via a flexible coupling.


From what I've seen of my 3040, the engineering principles are actually pretty good, it's just the actual components used leave a lot to be desired.

[johansen]

the bearings aren't strong enough, so you're right in that regard.
but if the preloading is by a stiff wave washer, you'll be alright, and the frame can flex more than the expansion coefficient.

[oakkar7]

Machine uses ball screws on all 3 axis (I wouldn't bother using a machine with t-screws on any axis if TSSOP, 0.625mm or 0.5mm pin pitch, is ,....

The board looks almost exactly like the simulation:

Orangecat,
If so, the machine is an original 3040 without any modification. How about controller? Is this original one?

[orangecat]

The machine in the video:

uses an unmodified machine (i've just cable tied the air blast and vacuum hoses on). I modified the control electronics to enable me to have coolant pump and air blast control via g-code since the controller i bought didn't have connections on the back for these. The parallel port of the control electronics was connected to a beaglebone black running LinuxCNC (headless).

[RogerClark]

I bought a CNC 3020 to do some other milling, mainly wood and plastic, but thought I'd try to give it a try for milling PCB, however the accuracy seems too low even to drill accurate holes for 40 pin DIL sockets for modules or microcontrollers e.g for an Arduino Nano socket.

It appears that older machines suffered from missing steps, because of the lack of decoupling on some lines to the stepper drivers, and also under driving the optos which resulted in rather curved squarewave pulses.

However all the mods for the two issues that people have found seem to relate to some older control board, and my machine has a JP-382A control board which is quite different from the previous model.



I see some people have posted some tests on small boards with small IC's and fairly fat tracks, and I expect mine would be fine with that sort of thing, but the accuracy over larger distances seems too low if I use drill holes of 1mm.

It could also be that I need to attach a scope to my driver board and look at the same signal lines which the old board had issues with, but the issue may also be mechanical e.g. slack or hysteresis in the lead screw systems etc.

So... I was wondering what other users have experienced with their machines.

If the precision is too low even to drill holes, then that's fine, I can still use the machine for other stuff, but it would be good to know whether with some minor changes, whether I could make the thing usable for PCB drilling at least.

[mc]

I'd spend sometime going over the basic mechanics, before worrying too much about the electronics.

The basic frames, are generally OK, but are let down by poor thrust bearing setups. I've got a ballscrew version, and the ballscrews are actually pretty reasonable (I wouldn't say they'd last long under moderate loads, but they're certainly a major improvement over the leadscrew option), but the thrust bearings that were fitted to mine were total rubbish. I just opted for a named brand version of the fitted single row bearings, but swapping to a similar sized angular contact bearing would increase durability/accuracy more.
However, my machine was built solely for digitising/probing parts, so the original controller was removed, with a new controller and linear scales fitted to provide high accuracy.

[RogerClark]

Thanks

I'll focus on the mechanics