6 pack Universal CNC controller - thoughts?

The specs say that it will support a MOSFET capable of 20A and these MOSFET are quite big. Probably will never work on a Pololu type driver but still, 4.2A that they can deliver now is quite respectable and in line with most high end drivers.

 

Actually the TMC5160 are rated 4.2A. I'm sure you'll have to have a fan blowing on them but I have it anyway on my 3d printer.

 

I looked at the datasheet and it only claims 20A with external mosfets. Where does it say 4.2A? Also, those sockets don't look particularly special to me - I'd be surprised if they supported that much current. Pin headers start to glow at about 3A.

 

OK, well that answers it. Those Chinese products are guilty of wishful specs. Look at the TB6600 drivers that are all over amazon/ebay/ali-whatever. Supposedly 4.5A but don't even use the Toshiba TB6600 chip and the one they do use is only good for 3.0 Amp. I did a tear down on one here.

The issue is heat production and the Step Stick format simply doesn't have enough surface area to handle more than a couple of Watts of dissipation. Those little heat sinks they ship with them are simply not going to do it. Even with forced air. There is a reason why drivers like the DM542 have a heat sink bigger than an arduino mega.

 

The datasheet has a fair amount of detail. Take a look at the Breakout Board. It could be pretty easily adapted.

 

The EV boards are intended to be used as a basis for products so the docs for most of them include schematics and BoMs. A lot of them include CAD files (though never Kicad, it seems). The AO4882 mosfets are like 8A/40V - datasheet here Wish they had used some that are a little more common.

 

I'm not sure I'm getting it. The pluggable modules are a nice theory, but none of them really seem super optional for basic functionality, and are quite expensive addons, so why introduce the failure point at all? Conversely, the driver headers definitely should be modular so stepsticks are a choice- either a stepstick board or just a straight step/dir terminal board. 0.1" pin headers are awful to interface with for external drivers. It just seems like the worst of all worlds in a fairly messy package.

The "linearized" driver supply rail (at least, that's what I'm assuming from all the caps) is an interesting theory, though it'd arguably be better as a standalone external component.

I dunno. The germ of the idea is interesting, but I'd like to see it heavily terminalized, the optos and buffers built right in, and the size of the thing probably cut in at least half- too much BOM and board space on optional things, not enough on what should be mandatory things. To be this "pluggable", there's almost no reason to buy it at all, it barely accomplishes anything more than some stripboard and a handful of header strips. It could be a really beefy interface for the ESP32 board, which is really all that's required.

A term-to-term stepstick board could be interesting as a modular addon. Shame terminal plugs don't come inverted (that I'd seen) where the through-hole leg is on the plug, and the socket is screw terminal'd.

[EDIT: Decided to Google "inverted terminal plugs" on a whim anyway... They totally exist. That's the way I would go. Board-to-board stepstick breakout, or straight wiring to externals]

There are TMC5160 BOBs for $20 at the usual places- easy to add terminals, heatsinks and cases to. No need to reinvent the wheel, at least at that stage of the design process.

 

Continuing my earlier point, the problem with the step stick format is that it just doesn't go much beyond 1A. It's physics... Those old TB6600 drivers (ones based on the real through hole chip) had adequate contact area to extract multiple watts of heat to a heat sink and thus cool the chip. And a design that allows tightly clamping the package to the heat sink. The surface mounted chips used on stepsticks just don't have a big enough of channel to extract the heat to a radiator heat sink. You can push the boundaries with forced air cooling but the bottleneck is the chip package.

As to the "motherboard" approach, it's got some appeal but I agree that pin and socket adds failure risk. Or perhaps I should call it reliability risk. Plus, it isn't cost competitive versus a product that has the needed components on the base board. Perhaps the idea is that there are so many competing requirements that there would have to be a lot of different designs. One of the reasons I only make a 5 Axis BoB rather than 3, 4 and 5 Axis versions is that the incremental cost of an extra axis is minimal.

 

I agree with your overall point, I'm not a huge stepstick fan either except where portability would be paramount, but I'd probably quibble a little on this one- LGA isn't much different to SMD in terms of physical board relationship, and you can extract tens of watts from an LGA device. It's all about whether the chip is properly designed for heat extraction and whether the board is engineered to do so (and some surface area considerations). Typically you get more heat out of the base of an IC than the top casing, so heavy isolated copper pours leading to heat pipes are the way to go (in *addition* to top cooling... LGA processors still have those!). There's limited space (maybe budget too) to do that kind of thing in a stepstick, where the larger format of an external driver allows for a reasonable spacing around the driver IC to get heat out both above and below. So you can totally use an SMD driver, but it has to be designed like an SMD project, not shoehorned into a through-hole project like the "TB6600"s.

More accurate, yeah. I'd like to see more positive connection options as plug choices, either way.

I thought about this but it doesn't seem to pass even a cursory consideration- machine tools have 3-5 axes, material manipulator robots have 3-6 axes, camera motion rigs have 3-6 axes... There's really no reason to include at least 3 axes' worth of components by default- it's cheaper and likely covers 70-95% of use cases- the BlackBox, essentially, but with more options. Axes 4-6 should be a single board option too- there are likely just as many people making 6-axis robots as are adding rotary axes to 3-axis routers.

I think that makes sense. It's designed for machine tools, just throw it all in for negligible BOM cost and save on volume.

 

Well, maybe but look at the TB6600 package - It's got mounting slots on either side that allow clamping to a heat sink. Whereas any SMD chip is is reliant on the PCB being pulled tightly against the heat sink. That places limits on how much pressure can be safely applied without bending the FR4 and a bit trickier to get the surface coplanar/flat against the heatsink. Also the TB6600 package has about 6X the surface area of the stepstick class of stepper driver chips. Unfortunately, Toshiba doesn't publish thermal resistance specs so a proper comparison is difficult to do.

Yup. At least put 3 Axes on the base board. Perhaps that upsets the symmetry of it all or makes the true cost more apparent?

 

You didn't add any IO modules, which means all of your IO is unprotected and on header pins instead of terminal blocks. Which is A) a fairly terrible idea (I bet pretty much everyone on this forum has destroyed multiple Arduinos at this point) and B) is not an apples-to-apples comparison to the BlackBox, which includes IO protection and terminals. At $20 a pop for those opto/buffer boards, and needing at least one, probably two, that's not looking quite so inexpensive any more.

If you really think it works for you at a price you like, no one here's going to disagree. Hell, I use Mesa cards and I've destroyed two of those at $100 a pop. Hardware dies or is the wrong thing sometimes, it's all part of the learning process. But I wouldn't be so quick to dismiss the institutional knowledge that's built up here either. There's a reason we bring up these concerns (even if we do sometimes find a semi-related rabbit hole to dig down).

The scalability was one of the parts I liked about it. But given that without 3 axes of motion control, basic limit/probing/spindle control/coolant ports, and an actual processor to run the code, it's not actually a solution in itself and builds unnecessary costs (BOM, pick'n'place setups, multiples of masks, etc- parts of the board manufacturing process that do not scale well at all) that get passed along to the user. So, designed properly with users in mind and not either some arbitrary sense of "freedom" or the number of items in a shopfront, 3 axis should be a baseline, not a target. Phil also designs 32-bit hobby CNC hardware, so I'm guessing he knows the exact process quite well.

Scaling from 3 to 4-6 axes is a great option. Scaling from 0 to 3 axes and from "no IO" to "basic IO" isn't really a feature at all, it's an irritation. If I only needed one axis I'd get an Arduino and a driver and that's it (that's how my camera slider worked).

That would be good, I'm guessing (because if I needed this type of solution, I'd go grblHAL on Teensy 4.1, but I don't so I don't have a horse in this race) the ESP32 could handle it. USB issues are the worst.

No one's talking about building in the drivers. In fact, I semi-regularly criticize it on the BlackBox. We're talking about stepstick driver chips vs external box driver chips and the physical parameters/limitations of each within the confines of their operating locations.

Fact is, stepsticks just generally struggle to dissipate heat above an amp or two, whether they're rated for 4A or not. That's not great for running NEMA 23s. Neither is being limited to only 24V in my book, but that's a trait it shares with the BlackBox so that's apples-to-apples for purchasing.

Many excellent hardware developers are terrible hardware designers, just like many excellent software developers are terrible software designers. Two different skill sets, usually requiring multiple people. He may be a designer but struggle to start from scratch with no inputs, altogether possible. May make a cool system, but doesn't work from a user experience point of view. Either one is perfectly normal, happens all the time.

 

LOL, same thing for a camera star tracker I built - Nano and a TMC2209 step stick (soldered in place, though). I'm always learning but one thing that has become clear to me - the cost between 3 axes and 5 or 6 is, literally, 10s of pennies of BoM cost. The cost of a 2 pin screw terminal in small volumes is $0.06 and that's one of the more expensive items. As a HW designer, I make a lot of decisions for the user. I think flexibility is good but I also want my users to have a good experience. That's why I avoid pin headers for most of the I/O. Screw terminals make the board more expensive (my Teensy BoB has 32 2-pin STs, you can do the math) but after a couple of hundred sold, I have had zero complaints about that. I'm still debating whether to offer a board that takes step sticks. I know I am losing sales because of it but am not sure I want to play in that market. I see a lot of unhappy people that have chosen them inappropriately.

 

What do you like so much about them? As I recall, there was another discussion around here not too long ago that ended up teasing out that the sensorless homing isn't all that reliable as a feature. Sounds great on paper, but in practice we found a surprising amount of people who just couldn't tune the sensing parameters to get consistent results. The 4A rating may be true at the transistor level, but we're pretty dubious about it from a physical package perspective; stepsticks are just hard to adequately cool. You could machine a dedicated enclosure for them, but at that point both cost- and size-wise you may as well have just bought some external drivers.

Mesa serial is RS422. You could probably bit-bang I2C if you tried, but it's probably not worth the hassle. Mesa cards are intended as actual industrial-grade equipment; all buffered and terminal'd, 8-32V IO (for 12/24V signal gear... I didn't realise how bad 5V signalling is for noise, even with best practices- no going back now!). The question would be how to interface with it; they're FPGAs, not MCUs, so you wouldn't be running grbl. Though I suppose you could, technically, use an ESP32 or Teensy running a 32-bit grbl, make a BOB for it to a pair of 25-pin parallel ports, and then use non-FPGA Mesa daughtercards as IO. Not sure how you'd access the serial though, you have to initialize that within the HostMot environment and set the card mode at the same time.

I guess it really depends on what you're trying to do, and what performance level you're trying to hit. It's always best to work backwards from the end result.

 

Rob, from my point of view, the attraction of the TCM stepstick drivers is that the steppers run quieter, way quieter, then the other driver. It may not make much of a difference in an environment where spindle and the dust remover make more noise then the stepper combined but it may play a role in big 3d printer and similar machines.

 

For step sticks, the TCM based products are the best. While lower noise is desirable, they are also much smoother than others. In my star tracker, I started out with a 4988 SS and at the finest step rate (1:16) they were just too "jerky", even with significant gearing down, to keep the camera steady for long timed exposures. I was about to junk the stepper approach and move to a DC motor but tried a TMC2208. The difference was simply amazing. I was able to use 1:8 microstepping for long exposures and no camera shake. This was in "compatibility" mode as a drop-in replacement. I believe it is their "Spreadcycle" feature that makes the difference. I also tried an 8825 based SS and it was similar to the 4988 - unusable.

For low power applications, TMC step sticks rock.

 

Definitely agree with this, as far as I can without direct experience. I've seen a lot of places about how smooth and quiet they are. My Ender 3 Pro is in an enclosure, or I might have considered swapping them out on that too. Would have been nice vs my a4988 Black Edition (which was a pretty decent option at the time!) on my slider too.

But yeah, anything that's cutting, you have bigger problems than stepper noise, which is why I hadn't brought it up previously. I'd also be somewhat inclined to question their ability to maintain smoothness at higher power levels, simply due to power transmission physics.