Breakout board.

[Woodsman]

Hi, I'm new to this, so I might have some rather confused ideas.

I've been toying with the idea of building a hobby cnc machine for ages now, and came quite close to buying one of those ebay TB6560 boards.

After reading the thread on here about repairing them, I'm glad I didn't.

My current plan is to build 4 of PMinMO's TB6560 boards, and a breakout / opto-isolator board, based on PMinMO's circuit but constructed on strip-board. Personally I think that the parallel nature of a breakout board and having the isolation plane (tracks cut) all in one straight line, lends itself well to a strip-board layout. Does anyone know of any reason(s) why this wouldn't work?

I also noticed that almost all layouts with separate drivers, take their 5V supply from one driver (to avoid connecting all the V regs together ) but ground each driver to the breakout board, I'm presuming this is ok, as everyone seems to do it? But I just have visions of earth loops forming between the boards and the central ground point. Has anyone had any problems with this? It would make things more complex to separate the secondary side of the breakout board into isolated sections, but if needs be I'd rather do it at the design stage than have to de-solder the opto-isolators.

Thanks

Peter

[James Newton]

it's very important that all your electronics agrees on what is ground. In other words, be absolutely sure that all your grounds are connected together, and solidly. You can use one regulator or multiple regulators, that doesn't matter so much. If you use one, make sure it can supply the current requirements of all the loads its driving and has a big enough heat sink to avoid overheating. If you use multiple regulators, don't connect the V+ together or else the one that is putting out the highest voltage (even by a microvolt) will be the one supplying all the current and the others won't be doing anything for you.

I personally don't see any huge point in opto-isolation: I just always use an old "throwaway" DOS or old windows machine and would never connect my nicer, newer more expensive PC's (which I can't afford to loose) to the drivers and motor. Design on the high end equipment, then sneaker-net the CAM files to the CAM machine. Old PC's are cheap or free, why risk your good iron?

In any case, building a PMinMO board is infinitely smarter than buying one of those cheap drivers that you can't repair and that isn't well designed.

What is it about the TB6560 that you like?

[Woodsman]

Thanks James, yes I couldn't see why one regulator would be a problem, someone here mentioned that because HF switching circuits were so electrically noisy, it was best to supply them separately.

Most of my recent electronic experience has been with auto-electrics, so I can't agree more about a good solid grounding point, a car is a horribly electrically noisy place (my Fluke 88 will read alternator Hz on anything connected to B+) Everything these days is switched by mosfets and hf choppers with milli-Ohm value sense resistors in the ground line (sound familiar?)
And yet the signal from a small magnetic sensor is often taped into a loom with say a 90V square pulse to a piezo fuel injector, and works too! The secret to this, I believe, is good ground connections, proper de coupling, use of twisted pairs on noisy power lines or communication lines, and sheilded for small signals.
The shield (unless it doubles as the signal ground) is only ever connected at one end, (unless were talking about audio systems where this practice can cause its own problems)

So I'm going to go with a single well heatsinked 5V reg. And a star ground point with all ground wires soldered to a copper or brass disc bolted to the centre of the chassis.

About the opto-couplers, and cheap old pc's, yes I'd agree, but I only have the one, space is an issue as well as money, and most scrap pc's I've seen, already had the HDD's destroyed, so I might as well fork out for the optocouplers than new HDD.

"what do I like about the 6560?"

First of all the price, @ $20 US for 5 allegedly genuine ones (there are lots of fakes out there) over here an L297 is twice that, before you add the cost of the power stage. also having it all on one chip with only 10 external components, keeps the PCB small and cheap too.

Secondly, it seems to fit my needs, I'm looking to build this first machine to make patterns for casting components of a better one, so I've been looking at steppers of around the 3A / 3Nm NEMA 23 size, to re use when I up-grade. I havn't seen any other 3A all on one chip solutions.

Then there's the size, I have 4 nice heatsinks, each of which would be plenty adequate for 2 TB6560 drivers, but if I went with a driver ic and discrete power stage, I'd have to use all 4 just to fit all 32 FET's on, doubling the size of the case (plus I have a use for the other two heatsinks)

I havn't ordered anything yet, so if anyone has any better suggestions than the TB6560 please let me know.

Thanks

Peter

[neilw20]

TB6560 should work fine, but you need to know what you are doing and the app notes don't give any clues to the current mode traps and pitfalls.

Look at this thread and see what can happen if the layout is done by an auto rout program!!

http://www.cnczone.com/forums/859895-post23.html

Take heed of my comments and you can make them work perfectly, but one missed detail and it can all turn in to fizzle.

Rather a challenge to get it right on stripboard, but follow the ground rules and they will work. The Motorola app notes I posted will help if you are new to designing with the TB6560

I have done lots of switchmode design and current mode operated BLDC's (not like the crummy syil one), but never used the TB6560. Just a horse of a different color

[Woodsman]

Thanks Neil, I'd like to think that I'd do a better job than that autoroute program did. However, I'm not going to design the driver boards myself, as even if I were to come up with a good design, I could not make the boards for less than These HWML & PMinMO boards cost.

It is just the breakout board that I think will take quite well to being made on stripboard.

My original question was, Should the ground line(s) between the breakout board and the individual driver boards be connected? If so does this not create ground loops?

Or would it be better to design the breakout board in such a way as to isolate the secondary sides of each axis from each other? feeding 5V and ground to each isolated secondary from that axis's own driver, and eliminating any chance of ground loops.

Or am I just making this more complicated than it needs to be?

If I'm not explaining properly, let me know and I'll draw up a diagram of what I mean.

Thanks

Peter

[James Newton]

Those HWML board are interesting... I'm not quite sure I understand how he is getting all that into such a little package. If you use them, please let us know how it turns out.

[neilw20]

Treat all conductors like resistors.
When current goes through the conductor, there will be some voltage drop, however small.

Generally, if you connect the commons of all POWER circuits to a single (star) point, you ensure that the voltage drops in a conductor are not induced into another circuit.
Ground loops is an old term, but basically even if there is no 'loop' it is really some circuit picking up the voltage drop from a circuit.

It should be remembered that wires to filter capacitors often carry quite high pulse currents, maybe even a lot higher than your supply current, so a 2" wire to an electrolytic may have 2v pulses across it.
For instance, the negative from a bridge rectifier to an electrolytic on this 2" of wire can have quite a high ripple voltage across it.
Just connecting a ground at the rectifier instead of at the electrolytic can add some unexpected ground bounce, and if the loads on the circuit is of a changing nature (like a motor running sometimes, or charging another capacitor else where when a switch closes) the level of ripple or transient pulses can change.
'Oh, I get emergency stop when I start the pump' type problems can stem from these wiring habits.
If you need to connect the logic to that same common, do it at the star point.
Never daisy chain grounds, or think that point (on some cable) is ground. Each daisy chain link can voltage drop across it, which its load does not care about, except for a slight loss off supply, but if that loss is introduced into some logic or current sensing circuit, wheeee!
It may be bouncing 2v or 3v, which will severely upset any 5v logic.

Most professional machines run 24v for the low voltage logic, so that 2 or 3v won't cause a problem.
All this said, you can still induce interference in other circuits by the high current circuits, being run in the same cable tray, by inductance and capacitive effects which is another story, and the solutions are more complex than just a star point.

Saving the odd wire or two is not smart, or economic if the wire has current flowing through it, or presents a low impedance so current can be induced in it, or transients can be coupled to it.
This is why one end of a shield is left unterminated. No connection = no current, -- well almost. Oops, it has capacitance to somewhere else so current can still flow, albeit transiently. Hey the emergency stop problem just came back!

Most of these transient, fleeting, one time events can only be captured on a fast storage CRO. Not a storage CRO? You probably will never see it.