View Full Version : Time for a purpose built motion control computer.

05-11-2010, 07:03 AM

Although fairly new to this field of CNC machines, I have many years experience with commercial CNC mills and lathes. Now retired, I missed the ability to design and machine components which can only be made by CNC. I purchased a Sieg KX3 and sourced a refurbished PC to the specification required by Mach3 and started the long journey!

May I start by commending those amongst you who have spent probably years making possible CNC machining at such a low price. The machine suppliers who liased with the manufacturers and sourced the hardware, the authors of Mach3 and last but not least those who setup and run the technical support Forums without which this voyage of discovery would have been a nightmare.

However, the experience for a newbie is like buying a new car, a full toolkit and a boiler suit so that when the car breaks down at the roadside you can adjust a wheel bearing or modify the fuel injection software so the engine will start again.

It seems to me some of the problems are down to poor quality control during assembly, yet most are due to the PC not communicating with the machine reliably. How often do we read on the Forums Z axis OK but X and Y won't move, Z axis judders, spindle speed fluctuates. Invariably the responce is Sounds like the computer is not talking to the machine correctly.

Surely the time has come to design, build and even supply purpose built motion controllers so that people like me with little knowledge or interest in electronics can get stuck in to machining. I can see much potential for a business here especially when one sees the rate at which new members are joining the Forum, World wide. Machine sales are likely to increase not decrease in the comming years and for many of us the extra cost of a purpose built controller would be worth it.

Somewhere out ther is a Hero who will pick up this batten and run with it, and the sooner the better as far as I'm concerened. They might even earn a very good living out of it and well deserved it would be.

I hope this thread is the catalyst towards reliable micro CNC machines as they can and should be.

Geoff Vaizey

P.S. Put me down for the first one off the assembly line.

05-11-2010, 07:28 AM
The trouble with this concept is the multitude of different machines and and variants of those machines that require a specific setup to be "turn key" controllers. This is where Mach3 comes into its own, as a generic controller that can be setup to suit the machine used. Mach3 has a file called mach3mill.xml, mach3turn.xml, ect, that contains the setup of the machine so you can transfer this file to another computer and your precious machine configuration can be restored.

A register of these .xml files would be a good idea with the specific machine and variant listed so that a newcomer to the cnc world could download this .xml file and get up and running.

But to most, is this defeating the object when you build your own cnc machine, the learning process, the skillset that you gain that only comes with a full on initial setup of mach3. I prefer the knowlege gained having to setup mach3, but then i wouldn't go to the trouble of building a cnc machine.

So a section in the forum with these .xml files would be nice, wink, wink.

05-11-2010, 10:00 AM

Thank you for going to the trouble of replying to my new thread.

As things stand the field is catering well for people like yourself for whom the machine and computer are as or more interesting than the machining. I also understand your comments about catering for different machine configurations with .xml files. However, had I explained my thoughts more clearly, it would be implicit I was not necessarily suggesting that Mach3 needs replacing. What I am looking for, and I would think those who are considering using such a machine for use in their business, is a computer built specifically for this application. When one purchases, allbeit at much greater cost, a commercial machine the buyer can start to machine components accurately and reliably the moment the installation engineer has departed. Currently the micro CNC market is limited to the capable enthusiast with time and patience a plenty.
I started life as a car mechanic and have over the years saved myself a small fortune in garage repair bills. However most car owners have to pay the price and have their cars repaired by a proffesional!! Surely it is possible to assemble a computer specifically for this application rather than cobbling together a time expired box of tricks which may or may not work properly. I would gladly pay a premium knowing the computer and machine are working in harmony and I can concentrate on machineing components.

Geoff Vaizey

05-11-2010, 04:17 PM
i can see where you're going with this Geoff, but there will always be a need for someone to set up the controller to suit the machine unless it's bought with the machine. As machines have different axis lengths and different spindle speeds, and more importantly different ways to drive all these things. The computer with all it's variants is a fairly standardised item and driven by one of the current controllers has a standard set of options to use. By creating a specific computer/controller setup, you will still have the problem of configuring the controller to the machine.

05-11-2010, 06:15 PM
I can see the logic, just to be able to buy a machine that has the right hardware for low latency operation and a parallel port that puts out 5v would be a good start. I'm using a twin Xeon Dell precision 670 - a very good PC three years ago but it's not great for EMC2 because the latency jitter is high and the parallel port only puts out 3.3 volts which isn't enough to run the charge pump on my breakout board.


07-11-2010, 06:50 AM
Thanks for your comments. You clearly have a far deeper knowledge of electronics than me and understand implicitly the problems, what chance have I and people like me when it comes to selecting the right computer?
Our friend i2i is concerned that my proposed computer would not be viable because there are many different types of machine. Well I would respectfully suggest that at the moment there are not only a wide variety of different machines but we are trying to control them with a wide variety of old computers which were not designed with this application in mind in the first place.
I am fortunate to have experienced how smooth, fast and accurate commercial CNC machines are, and I have no doubts Micro CNC machines can be virtually as good without spending huge amounts of money in the process. I will also bet that within our midst there are people who given a piece of paper and a pencil could formulate the specification for such a computer and the source of the necessary components before I could make them a cup of tea.
Currently we benifit from pretty good small machine tools and very good control software (Mach3) can we not complete the chain with a first class computer? One day and I hope sooner rather than later, we will look back and wonder why we tolerated second class time expired electronics for so long.
How about it guys, who's up for the challenge?

Geoff Vaizey

07-11-2010, 07:57 AM
A fully stand alone controller is an expensive proposition with a potentiality small market. Something like this is a reasonable compromise


though I've no idea how well it it works, it saves a lot by not requiring dedicated keyboard and display etc. Take a look at the other controller options on that site, the price rockets :(

John S
07-11-2010, 10:28 AM
It's not that easy and at the level we are at i.e. maximum of three pound seventeen and sixpence per week on machines, we are stuck here.

If this wasn't the case we would all be spending 25,000 on a new Haas TM1.

Classic example and it's been mentioned by i2i are the XML files, these are standard XML files that 'should' be read by any computer but they are not. We ship an XML file out for the Sieg KX machines, in that file the pulse width is set to 5, checked, double checked and treble checked.

However once a punter loads this file onto his machine, whether custom, off the shelf or secondhand, he can be faced with a value of anything from 0 to 5.
Shouldn't happen but does. Haas, Fanuc, Hurco don't have this problem as the computer is part of the machine and setup at the factory.

Low end, because it's low end, it doesn't happen. Tormach will supply a computer setup to their machine if needed but we are now getting upwards of $10,000.

The setting up is a double edged lesson, on one hand you need to know everything on the first day, steep learning curve but if you are given it as in Haas, etc, any problems and you are back to them, tied by apron strings and cheque book. Once you can manage to get thru the minefield you are set for life.
I have a big 3 tonne CNC that has been refurbed over time and nothing on it other than a major casting fail cannot be sorted at home with an RS catalogue.

John S.

07-11-2010, 03:16 PM

Thanks for you comments, you must spend half you life answering the same old questions over and over again about PC incompatability on the forums.

I do understand the desire to keep costs down so almost anyone can enjoy the benifits of CNC machining, but sales of these machines are going to continue expanding, the Trade will see to this for starters. This will necessitate you or the next volunteer spending more and more time answering the same old questions about computer problems.

All I am advocating is making available either finished or in kit form (whatever fits your budget) a computer which is designed from scatch for this application. At the very least you might get some of your life back with fewer questions to answer, and every body would benifit especially people like me with little inclination towards electronics. You never know, if the design is good it might even read pulse width as 5 every time.

I have to say I am a little disapponted that so far everyone who has replied, begrudgeingly accepts that PC's are far from ideal, but seem to imply we can't do anything about it. Well I'm sorry, I think we can do something about it and I don't believe it has to cost a fortune.

Geoff Vaizey

07-11-2010, 04:45 PM
Ok, let's run with this:

I have a little chip sitting right beside me, 80MHz clock, 1/2MB flash, 32k RAM and something like 80 I/O pins, it costs about 5.

It's a surface mount component so add 15 for a board we can work with.

We need some way of getting data to it, say 5 for USB and/or 15 for Ethernet.

A keyboard of some sort, either discrete buttons, say 50pence each, or we could possibly use a USB key pad of some description.

Display options, the possibilities are vast. Off the shelf graphical LCDs are expensive and can be tricky to interface, say 50 for a 3 inch screen - have a look around for options. Or maybe just a set of old fashioned 7 segment units, say 1 each including glue logic.

Case, say 10 to 20.

Assorted connectors 20 <shrug> always more then you think.

Power supply for logic, a wall-wart is the safe option, 10.

The above is guess work but pretty much reflect UK prices.

These are your basic hardware building blocks. We now need a full hardware design and complimentary software design, as I still know feck all about CNC I'll leave these open for discussion :)

John S
07-11-2010, 05:59 PM
Much as I would like to see different options prevail I feel we are stuck for a while with what we have.
Art when he wrote Mach3 put an immense of work into it and it very nearly burnt himself out, his words, not mine.
Unfortunately no one else seems prepared to step up to the plate and come out with the next version, especially for the same amount of money.
We have had many promises, the Gecko G100 which turned out to be a door stop, the smooth stepper which has frozen in time and lacks any support.

I know of two people working on USB controllers but one is finished and works after a fashion but the guy isn't interested in working with users, take it or leave it attitude and the other is still hot air.

EMC is out there but without being derogatory it's written by geeks, for geeks, if you think windows and Mach is hard then Linux and EMC is ear of bat, tongue of newt as regards setting up.

The 'new' Mach4 is aimed at OEM's or profession retrofitters working on expensive dead machinery, anyone with less than three pound seventeen and six need not apply.

Robin Hewitt
07-11-2010, 06:24 PM
I have a little chip sitting right beside me, 80MHz clock, 1/2MB flash, 32k RAM and something like 80 I/O pins, it costs about 5.

I have a little chip, 40MHz, 64 pins, 1.88 at 1000 off, already soldered to a PCB.

On the back 4 RJ45 sockets for 3 stepper drivers and 1 phase converter, 3 switched outputs, 6 buffered inputs, 1 USB socket and one header plug to connect one of those big fat positioning knobs. On the front an RJ45 for the pendant, a spindle speed knob, 5 push buttons and one concealed RST button.

I built if for my machine so it has everything I want. Would work for anyone who built their machine around it, but people build their machine then think about how to control it. Cart before the horse perhaps, but that's exactly what I did :naughty:

07-11-2010, 06:41 PM
Dear ecat

Thank you for your previous contribution with the link to slides and ballscrews, but most of all thank you for getting the ball rolling.

I have to start by saying much of the above sailed clean over my head, but if your initial cost estimates are realistic you have confirmed my suspitions that a suitable controller need not cost a fortune. I wish I could contribute technically but I can't, however I can help with organising such a project and would be prepared to fund a prototype or two to prove the system, after which those in the team could progress however they wish.

Perhaps when you have a moment you could list the other difficult areas so that members with relevant experience may be enthused enough to join in.

Geoff Vaizey

07-11-2010, 06:55 PM

Thank you once again for the benifit of you experience.

It seems to me that there is little wrong with either the machines or Mach3, the weak link is the PC. Is it not possible to design a dedicated PC in a box that runs Mach3 and talks correctly to the machine. Ecat has made some usefull suggestions and Robin Hewitt has clearly trod this path before.

I think it is time for me to stand back for a while and see if we can put a team together.

Geoff Vaizey

07-11-2010, 06:58 PM

Thank you for your contribution, you clearly have much valuable experience in these matters.

Geoff Vaizey

07-11-2010, 07:16 PM
I suspect at the current moment, the Mach in a box (http://medw.co.uk/wiki/index.php?page=Mach-in-a-box) type of set-up, is the nearest you're going to get to a plug and play solution.

As for the Smoothstepper, Greg has been active on his forum over the past week, and it looks like things are finally progressing. His latest post (as of 14hrs ago), says he'll hopefully be releasing a test plug-in for slaved homing tomorrow provided it tests ok, and then he'll be working on finishing up the backlash compensation. So with any luck, it might be done by the time I get my new toy wired up!

07-11-2010, 09:51 PM
EMC is out there but without being derogatory it's written by geeks, for geeks, if you think windows and Mach is hard then Linux and EMC is ear of bat, tongue of newt as regards setting up.


Quite an interesting interpretation? Being an EMC2 user i would have to disagree, i actually found EMC easier to get to grips with than MACH3. What is interesting is that what you get because it is free is unconditional help and a shed load of people inputting to the code and improvements because they want to and not for a pay check. The big problem you face whether it be free or for money is that hardware dictates most of the software changes made by any developer (keeping it up to date). Fixes come before improvement and upgrades takes a lot more time to develop and deploy.

If you could have a proprietary software in a box solution, what they do is start a company like all the big boys and charge a lot more money for the fruits of your labour. Plug and play comes at a cost.

08-11-2010, 12:11 AM
Oh PCB porn, that's a beautiful looking board :) Driven via USB so you still need a PC for the user interface?

You're welcome :) but as well as point out some of the basic building blocks I was also prompting at the overall hardware costs of small scale production of a full controller, the software costs, as JohnS hinted, could be much, much higher.

If you are thinking of a one off unit or a production run of a few hundred units then dedicated hardware and software makes some sense. If the numbers are anywhere in-between then a PC platform makes good sense, especially if a complex user interface is required and doubly so if software is already available.

The problem with the PC platform is not due to all the subtle and not so subtle differences in makes and models, this can be addressed by targeting a specific board implementation. The problem is the insanely short life span of the various makes and models meaning that a chosen board will be end of life within 6 months. Thankfully, this can also be addressed, mostly, with a bit of luck and sprinkling of bat tongues and newt ears ;)

The trick is to chose a light industrial solution. VIA, Jetway and others produce such boards which often have a production life of several years. They may not be the cheapest, 50 to 150 each, but if one of these can be certified to run something like EMC, or if a version of EMC can be tailored specifically for one of these boards then a cost effective, readily available, supported and documented solution is at hand.