Hovering over "Buy"... care to sanity check my thinking?!

[brumster]

Very nice :)

I have an offer from a forum member to go and take a look-see at his machines, and a general chat, so I'm going to take him up on his offer. Think I will feel much better once I've seen a DIY-constructed device, then make a decision.

Breaking out onto a separate topic a little.... on the electronics side of things... the stepper drivers and the motion controllers. Just want to check my understanding; is the following right?

Motor Drives

These are really nothing more than stepper driver ICs, of which obviously there are a plethora out there with various ratings in terms of voltage, current, stepping capability and safety features. Bolted on to that are the relevant amount of heatsinking, protection/filtering but ultimately they take a low-voltage signal from a microcontroller or parallel port - with appropriate optical isolation - and move the motors as requested. One tick = one unit of movement (in the appropriate direction, of course).

There are differences in terms of matching them to the motor (number of wires for unipolar, bipolar) and you can get analogue and digital ones, digital ones being better all-round but presumably more expensive. There are trade-offs with how you drive the motor in terms of microstepping, torque, heat generation and so forth. So you need to know your motors in conjunction with your drivers, and also what you're going to use them for (what's most important - torque or speed).

You can get ones with direct parallel port drive (so no motion control; driven straight from PC) but then the honus is on the PC as the controller, hence all the requirements in terms of it being able to keep up. Each pin on the parallel port controls a function, which limits you somewhat to the number of axes/functions you can make use of.

Other alternatively you can get drives controlled from a separate motion controller in between the drivers and the PC.

Motion Controllers
These do the interfacing between your PC (via parallel port, USB or ethernet port) and the motor drives, but instead take a proprietary protocol and convert that to the native signals for each axis drive (they don't take G-code). The benefit of these is they can layer on all sorts of fancy stuff like inertia concepts, but the big gain here is really offloading the requirements from the PC to keep up with the frequency of signals needed to instruct the drivers directly. The motion controller interprets the movement commands (move X from 0 to 100) to how many steps that needs on which axis, at what speed, and then takes on the job of reliably hitting the driver with the requisite clock signal and appropriate-timed control signal.

You potentially gain some speed/reliability on your machine and generally it's the way forward unless you just want to have a play at slow speeds direct via parallel.

When you guys refer to BOBs (Breakout Boards) you're generally talking about motion controllers, although you can get BOBs that combine both of the above.

It strikes me that keeping them separate gives more flexibility as the drivers need to match the motors; if you get a combo BOB you'd need to be sure that the drivers side of it matched your requirements and they weren't going to change.

Point, laugh, poke the shit out of the above where I've got it miles wrong :)

[Clive S]

Motor drives say like the AM882 (Typical) are digital and have stall detection etc. are used after the BOB (and can be connected direct to the PP BUT not recommended) also they are used after the motion controller.

Motion controllers are generally fed from either USB or Ethernet port and they produces the pulses much better than the computer as they generate their own. Some motion controllers also have the BOB built in.


The Gcode is controlled from the machine controller ie Mach3 or Linuxcnc etc and fed to the PP or the motion controller

[brumster]

Does Mach3/LinuxCNC send actual gcode to a motion controller though? I thought not? It interprets the gcode on the PC, then some plugin relates that to wire signals that are specific to your motion controller (via a plugin on Mach3 for example) and then the motion controller converts that to the relevant pulses/etc...?

[Clive S]

Does Mach3/LinuxCNC send actual gcode to a motion controller though? I thought not? It interprets the gcode on the PC, then some plugin relates that to wire signals that are specific to your motion controller (via a plugin on Mach3 for example) and then the motion controller converts that to the relevant pulses/etc...?

No I was over simplifying it the Mach3 converts the gcode to pulses which then goes to the PP or motion controllor

[brumster]

Understood, thanks

Contemplating whether to make my own drivers and motion controller, but sounds like the latter is out! Too much coding effort anyway, I see there's an open source one out there and the code alone took 4 years of development... so no, I won't be going that route :)

[Clive S]

Understood, thanks

Contemplating whether to make my own drivers and motion controller, but sounds like the latter is out! Too much coding effort anyway, I see there's an open source one out there and the code alone took 4 years of development... so no, I won't be going that route :)

Good drivers are essential as they also take account of resonance, current control etc If you buy decent drive they can of course be used again if you decide on a bigger machine

[brumster]

Most I've seen seem to be configurable via jumpers or DIP switches so, presumably, as long as they can match the voltage and current ranges of the motors, presumably I can overspec on them now and re-use... makes sense. Ta

[JAZZCNC]

These are really nothing more than stepper driver ICs, of which obviously there are a plethora out there with various ratings in terms of voltage, current, stepping capability and safety features. Bolted on to that are the relevant amount of heatsinking, protection/filtering but ultimately they take a low-voltage signal from a microcontroller or parallel port - with appropriate optical isolation - and move the motors as requested. One tick = one unit of movement (in the appropriate direction, of course).

In simplistic terms yes but in reality then theres much more to good drives than this. How it handles resonance is very important and what makes the difference between poor and good working motors. Resonance cripples Steppers and the drive needs to be able to deal with this.
Making your own is an option but not one I'd recommend and I've seen many people take this route only to end up buying drives after much frustration.

You can get ones with direct parallel port drive (so no motion control; driven straight from PC) but then the honus is on the PC as the controller, hence all the requirements in terms of it being able to keep up. Each pin on the parallel port controls a function, which limits you somewhat to the number of axes/functions you can make use of.

Other alternatively you can get drives controlled from a separate motion controller in between the drivers and the PC.

Not exactly correct. You can get drives with built in Pulse engine so they don't need a PC Parallel port or External Motion controller. IE Stand alone
People often get the wrong idea about how Mach3 or Control software works. The control software doesn't create pulses it plots a trajectory data based from G-code and then hands this to Pulse engine which then does the Motion control side of creating pulses.
In standard trim Mach3 uses a Driver to do this function (paralllel port Driver) which then crunches some numbers and outputs the required pulses thru the parallel port to the drives directly (Usualy Via distribution board called a BOB).
If using an External Motion controller then it hands the trajectory data to The Plug-in for that Motion controller. The controller then does all the number crunching and send the pulses to the drives. The difference being External motion controllers can crunch numbers much faster and output much cleaner pulse stream which makes BIG difference to performance and reliablty.

When you guys refer to BOBs (Breakout Boards) you're generally talking about motion controllers, although you can get BOBs that combine both of the above.

It strikes me that keeping them separate gives more flexibility as the drivers need to match the motors; if you get a combo BOB you'd need to be sure that the drivers side of it matched your requirements and they weren't going to change.

No BOB is just a Distribution board for the I/O signals. Often Optoisolated for protection. Commonly you'll have Motion control board and connect to this 1 or more BOB's for Distribution of the signals.
Better Motion controllers Like those from Cslabs provide direct connection and remove the BOB from the picture.

BOB's, or should say Cheap BOB's are a common cause of many troubles with CNC machines and often under estimated to there importance.
They take all the Signals then distribute them to where needed so if the components used are slow or low spec then any signals going out will be degraded or slowed down.
Often you'll see people use External motion controllers with MHZ frequency capabilty and connect to BOB with 100Khz bottle neck components.
Like wise you'll see drives with Max frequency of 200Khz and people using Ex controller with much higher Frequency capabilty connected to 100khz strangling BOB so all that power is totaly wasted.

So what I'm saying is BOB is very important and while cheap works at the slower end it's pointless getting a External Motion controller for the speed and quality pulses and connecting to Cheap nasty BOB's.

A Good machine needs to have a balanced system with no weak link.

[brumster]

Thanks everyone; really good input. Got a good understanding now :) going to run some numbers tomorrow and come up with a "high level plan".... :D

[brumster]

Right then!

Dismissing the idea of a large, steel-constructed machine with serious aluminium milling capability. (1) I don't have the space to do it justice, (2) I strongly suspect a converted Bridgeport would be a far more sensible option.

So I've downgraded to a desktop-sized machine, and I'll focus on doing PCBs with that. When I move house and have more space, I'll contemplate the bigger build. So I'll use the desktop as a learning exercise. Which leaves 3 flavours to go for :
1) The aforementioned "bugger it and just get a Chinese 6040" for £900.
2) Build something completely on par with the 6040 myself, like-for-like.
3) Build an overspec 6040-like machine, mainly to learn on construction but also invested in the parts so they can be re-used on a bigger machine in the future.

I've been bouncing around CNC4YOU and eBay today, making a spreadsheet up to price these.

Option 1 we all know will do a minimal job, but I'll be replacing the electronics more than likely fairly soon. But it's a baseline, £900 for a rather average machine.
Option 2 I've worked out, CNC4YOU don't come out that well in this regard, and going eBay for the majority of it I've got a very high level ballpark price of £768... that's with the same spec ballscrews, supported rails, 1.2Nm NEMA23s, cheapo TB6560 drivers (the motor/drivers come in a 3-axis pack for £95). £175 on a spindle, 1605-C7 ballscrews for a 600x500mm working area. That price includes £130 on aluminium extrusion but there would undoubtedly be more cost on getting the relevant plates/etc for the Z-axis and so forth. So, to cut a long story short, I see no point in trying to build a 6040-like machine yourself for less money; I don't think it can be realistically done. They are great value for what they are.
Option 3, then, comes out at a £1262 plus the frame material, so if I bank on £1400-1500 I think that would cover it. This would be same size as a 6040, but like I said overspecced on the components - HGR15 linear rails, 4Nm NEMA23's, Leadshine DM556 digital drivers, the same old 2.2kw ER20 spindle and VFD, a 48v PSU, same 1605-C7 ballscrews (contemplating C5s, they're not much more), and then some guesswork on the Z-axis, cable, rails and so forth. Most of it UK sourced seems just as cheap, if not cheaper, than China. I guess the quality stuff is the same price wherever you source it from, and China has no interest flogging it!

Option 2 to me seems pointless. So it's a choice between option 1 and just get started playing with it, or option 3 and long-term investment.

I understand that £1500 buys a much larger machine but the reality is the costs are kind of irrespective of size; I know aspects of that don't make sense (do you really need 4Nm motors to mill PCBs?!) but I'm thinking re-use in the long term on the next project... and doing the construction with the components I'd use on a large build, so I can get savvy with them.

Because if I don't do that, then I really might as well just buy a 6040 and be done with it, given my size limits.

Question - by my calculations a 5mm pitch ballscrew, 200 step motor and no microstepping would give me a 0.025mm per step movement, which is just under 1 thou... and with microstepping obviously could chop that in half a few more times if need be. There would be no reason for me to gear this down would there, given worst (toughest) case the most resolution I'd need on a PCB would be half a thou... (excluding backlash and rigidity issues, I know).

Am I right in thinking that to maximise accuracy, I need to keep the thing as rigid as possible, run it at appropriate speeds, but ultimately the ballscrews will define that resolution. So the quality of drive, the ballscrews, that's all just a case of how deep your pockets are - the bit that *I* can f**k up is not making the machine square/rigid.... anything else? I know accuracy is more than just resolution, it's about repeatability as well, but go easy on me, I'm learning... :D

Basically if I go NEMA23 4Nm, Leadshine digital drivers, a capable 48V 600W power supply, C5 (let's say that for now, given the price difference to C7's) 1605 ballscrews with anti-backlash nuts... the only other bits that are going to limit the accuracy/repeatability of the machine are my fudged-up design, or bad software/CAM....?

If I can accurately mill SMD boards (typically a pad spacing of 0.25mm between pads) then I'd be happy as larry.
If that's pie in the sky then you're really falling back to through-hold DIP stuff, which really could be 0.5mm or worse, and I then wonder whether it's worth all the bother, apart from as a learning exercise and investment for future upgrades. I guess it could probably do *some* aluminium work if I was careful/slow, but I'm kind of dropping the focus on that now.

Basically, £1500 on an over-specced DIY build the same size as a 6040 (but obviously better quality).... or spend £900 on a 6040, use it for a year or two, accept it's limitations, fix it as needed, then flog it on for £400 and consider it a learning exercise... or salvage it for the spindle if nothing else :)

I'm thinking the DIY route but happy for someone to tell me I'm mad for doing it/it makes no sense :)

P.S. Sorry for the clearly indecisive warblings... I do value your feedback though :)