LitePlacer.com forums

For those who didn't notice, Robotdigg will sell you the feeder block that you see on all those Chinese pick-and-place machines... only $150.
I got one, and I can say that (a) it is gorgeous, of absolutely stunningly high quality, way better than expected for $150 and (b) it works magnificently as a drag-feeder.

I don't get commissions or any other kind of incentive from RobotDigg or any other vendor. But I do find out about cool parts like this when other people post about them (i.e. Karl's source for Juki nozzle holders) so I try to return the favor when I can.

I'm currently working on converting it to a Vbesmens-style pull-the-cover-tape using $2 steppers (yes they do exist: google "28byj-48")

I saw that block and was quite interested, but before dropping $150 on it I was keen to see more specs, e.g. what dimensions, number and type of lanes. Hard to tell from the small picture.

I am also looking at a Volker Besmen type feeder. I decided that the most critical part was the final delivery part, I attempted to 3d print but the accuracy is not there. It really needs to be machined, either alu or POM etc. The rest of the mechanism is fairly tolerant, apart from sprocket hole detection. I don't have facilities for milling, so an off the shelf part would be great.

The 28byj-48 steppers are cheap, but quite slow although they do have good torque. They are also unipolar, so to use them with common bipolar drivers they need the "cut the red wire" mod. I was planning to avoid the cost of a stepper driver and use a DC motor instead.

In theory, a stepper would avoid the need for a sprocket hole sensor, as it can feed exactly 4mm. But, I think that the alignment would drift over time as the feed is unlikely to be exactly 4mm, so it would need periodic alignment via some method. Perhaps by camera?

bobc wrote:I saw that block and was quite interested, but before dropping $150 on it I was keen to see more specs, e.g. what dimensions, number and type of lanes. Hard to tell from the small picture.

Yeah, I agree. I took a gamble here and it turned out really well. The small lanes are standard 8mm (the size used for essentially all resistors+capacitors). There are four 12mm lanes, which is adequate. Unfortunately there are only two 16mm slots (the size used for SOICs with more than 8 pins) which is not enough for me. There's one 24mm slot (FPGAs, etc) but I never use it; anything in a tape that wide is expensive enough that I've probably only got a few of them and splicing on a leader is more hassle then just doing a cut-tape feed.

It has M6 tapped holes in the bottom for very secure mounting.

bobc wrote: I am also looking at a Volker Besmen type feeder. I decided that the most critical part was the final delivery part, I attempted to 3d print but the accuracy is not there. It really needs to be machined, either alu or POM etc. The rest of the mechanism is fairly tolerant, apart from sprocket hole detection. I don't have facilities for milling, so an off the shelf part would be great.

Yeah, it's gotta be precise. Aside from the material block the rest of my feeder consists of extrusion beams and very weird-shaped PCBs. Some days I think I use PCB fabs as a cheap small-run CNC shop more than a circuit manufacturer...

bobc wrote: The 28byj-48 steppers are cheap, but quite slow

That's okay, they only need to be able to turn about a half of a revolution in the time it takes to do a full pick-align-place operation on one part. And if they're even too slow for that I don't mind making the machine wait in order to have tons of cheap feeder lanes.

I do have four very-fast commercial feeders which I use for parts that get placed many times on each board (like 1uF and 0.1uF decaps, I go through those like water).

bobc wrote: They are also unipolar, so to use them with common bipolar drivers they need the "cut the red wire" mod. I was planning to avoid the cost of a stepper driver and use a DC motor instead.

Right, the mod is very very very easy. On the other hand although unipolar wastes half the torque, it lets you run the motor with no driver -- just a quad high-current NPN bipolar, which most vendors throw in for free with the motor.

bobc wrote: In theory, a stepper would avoid the need for a sprocket hole sensor, as it can feed exactly 4mm. But, I think that the alignment would drift over time as the feed is unlikely to be exactly 4mm, so it would need periodic alignment via some method. Perhaps by camera?

That's what I do. Detecting the tape holes is very easy EXCEPT for the obnoxious clear plastic tape that Kemet uses. I cannot figure out what on earth they were thinking when they made that decision...

My solution isn't for everyone though. Having a few fast commercial feeders takes away a lot of the pressure to be fast-fast-fast. I'm happy with the compromise. I have a few fast+expensive feeders and an ocean of very cheap slow feeders so I rarely need to change reels, which I hate doing.

Seems we are thinking along similar lines. One reason I wanted a faster motor was so that a feed module could be more easily integrated to PNP software. I intend to have a push switch on the feeder, so it can be actuated manually or with a probe on the head unit, without requiring a comms link. This is a "nice to have", in practice it is probably not difficult to add a comms interface to PNP software.

I'd be interested to see how you mount the 28byj-48, I had trouble designing something that was compact for 8mm tapes.

bobc wrote:I'd be interested to see how you mount the 28byj-48, I had trouble designing something that was compact for 8mm tapes.

Here's a snapshot of the work in progress. I expect to have it all finalized in a week or so. If you want the GERBERs I can post them.

Much credit due to vbesmens; it should be pretty obvious that his automatic feeder was the main inspiration here.

The boards are designed to have an OpenBeam (i.e. 10mm/M3 mini-extrusion) running along the top and bottom edge. Each board holds six steppers, and they use right-angle box headers for daisy-chain communications.

Digitally, the boards are one long shift register of "enable" bits that cut off the power (red wire) to the deselected motors; you shift in a bit pattern that enables the stepper you want to advance, then toggle the four phase lines using a single high-current quad driver. The deselected motors won't step. This is sort of the minimal-components approach -- just one SIPO chip per panel and one PFET per motor (all other circuitry is shared across the whole feeder array). It also requires only a fixed number of GPIO pins (around 6-8 or so, depending on how you wire the NPN drivers) for as many feeder lanes as you like. Disadvantages are that it only works for unipolar wiring and the deselected motors don't actively hold their position while the selected motor is stepping -- although the gearing on the 28byj-48 is so extreme that isn't necessary (I can just barely manage to turn the motor shaft by hand).

The panel size is somewhat arbitrary; my preferred PCB fab rounds up all sizes to the nearest multiple-of-10cm enclosing rectangle so they're 10cm wide and tall enough (15cm) to accomodate as many steppers as I could cram into that width.

Oh, that is really neat, I like it

Of course, multiplex the output stage, I should have thought of that! I also realise now I could put hole detectors in series, as only one tape will be moving. One small micro could control a lot of feeders.

I guess you have some idlers to guide the tape to the spool?

bobc wrote: I also realise now I could put hole detectors in series, as only one tape will be moving. One small micro could control a lot of feeders.

Or use a PISO (Paralell In Serial Out) chip. They are dirt cheap, like $0.02/pin. I bought a huge pile of NXP 74LV165, they're great, easy-to-solder SOIC package and rated down to 1.0 volts.

bobc wrote:I guess you have some idlers to guide the tape to the spool?

I was just planning on using plain old standoffs... in other words just a peg mounted to each of the six small holes near the left end of the front side (as shown above), plus two more so the tape from the outermost motor doesn't interfere with the innermost (you can see the hole for that peg in the picture, the silkscreen lines are the intended cover-tape path).

My main concern at this point is getting the tape to do the 90-degree twist that it needs to do. You can see what I mean in this picture from vbesmens' machine (zoom in). Mine will have the same horizontal aluminum bar, but also an additional peg that the tape is pressed against before it gets to the spool. I don't know if this will work smoothly or not; we'll see soon.

Ok, having being inspired and encouraged, I have reworked my feeder module with the 28byj-48 steppers. I can also put all the electronics on that board also, simplifying the design a lot.

bobc wrote:Ok, having being inspired and encouraged, I have reworked my feeder module with the 28byj-48 steppers. I can also put all the electronics on that board also, simplifying the design a lot.

Hey neat! I like how you managed to keep the tapes straight by using non-multiple-of-90-degree rotations, I'll probably copy that on my next respin of the board.

For reference, the 8mm tape slots on the feeder block are at a 13mm pitch. So that's the amount of space you have to work with.

I also have one of these in order now. Thanks for the tip!