Pick and place machines

The three levels of machine might be

Manual: a rig that basically steadies your hand as you place parts. No guidance at all.

Semi-auto: adds a carousel or feeders, a computer, and camera to step the operator through placement.

Full-auto: hands-off assembly.

An auto line might cost $500K or so. A good semiauto p+p can cost $20K, but requires more labor per assembly. This takes roughly an hour (I think) on semi, and about 4 minutes full-auto.

It might take all day to do manually. I assume that by "manual" you mean the non-computerized thing, basically a rig to steady your hand. Both semi and full auto greatly reduce placement errors.

We have a couple of Essemtec ($20K) semi-autos and a Universal full auto, and we still use the semis for smaller runs. If you spend $20K on a semiauto, and run 5000 boards through it in its lifetime, that's only $4 per board. The cheaper semi-autos are probably OK too.

We keep a bunch of carousels loaded for the Essemtechs, in a dry cabinet, so we can assemble a few boards on short notice.

This is the one I was looking at, I assume it can do regular old SMT parts in addition to BGAs, etc?

This guy build his own, if you like spending a lot of hours

Cheers

Klaus

That looks like a special-purpose BGA placer, probably not suited to assembling a whole board. It doesn't have provision for organizing and feeding a bunch of parts.

We have a BPA placer much like that, but it's only used to place a few FPGAs or uPs on a board. The Essemtec places the majority of the parts. The Essemtec has a carousel that holds a lot of parts, and the computer spins it to present one part at a time to the operator to pick up and place. It can also feed from reels off to the side.

We've become paranoid about minimizing the number of different parts on a board, to minimize feeder setup. We'll use a quad 1K resistor pack in all sorts of resistor/divider combinations, or series two 10K resistors to make 20K, things like that.

I'm convinced you can build just about anything in order of ten multiples of 47 ohms.

I have a used Philips CSM84, (number means it can hold 84 8mm feeders). it has 3 nozzles on an X-Y gantry, so it can be set up for small passives, SOIC chips and large FPGAs and such. That is a fairly old machine, and not as precise as I'd like. But, it still stuffs 100 parts in about 2 minutes. There are lots of used P&P machines that still have lots of life remaining in them. This, of course, is a full-auto machine. I wrote a program that converts the P&P output from my CAD/CAM system to the format the machine takes. Yeah, my CSM84 is a fairly BIG machine, but I actually have it in the basement of my house!

There definitely are tabletop P&P machines from Manncorp and Madell (and probably some others). These can generally be picked up by even one person, or easily by two.

Jon

these guys are getting pretty good press for their machines (from $4k btw)

all standard disclaimers apply (YMMV etc)and I have no interests in this company

I got one of their TM240A units. It's ok, not great but does help. 0603 is the limit. It can do .65 pitch parts but no way .5 parts. The biggest problem is the placement depends on the board edges being exact which normally are not. If the board edge routing is a little +/- your part placement is also going to be off by that amount. You can adjust it but you almost have to do it for every board. For mine I'm working on a jig that will hold the board by the tooling holes instead of a board edge.

Watch your feeder count. The 240A has 27 total, 21 8mm, 4 12mm and 2 16mm (plus a "bulk" which is pretty much useless).

Like I say, it's a ok setup. I've more than paid for the unit in time saved not hand placing all the parts.

A moderate complexity board might need 100 feeders.

Doesn't the 'Digi-reel' trick buy you some relief from that?

how? digi-reels are just reels with less parts on them

-Lasse

Well, if there are 100 different parts on the BOM, you need 100 feeders. Our semi-auto machines can use tiny trays on carousels, for loose parts, and maybe a dozen or so reels. We may need to install more than one carousel, sequentially, for a given board. The full auto needs all the parts available at once.

A Digi-reel looks like just a chunk cut from a manufacturer's reel, which wouldn't help the feeder count. Mixing different parts on one reel would be interesting.

For one complex board I need double the feeders I have. I have to make 2 passes. I've gone through the BOM a dozen times trying different combos of Rs to minimize different part counts. I dont know why feeders have to be so damn expensive.

I was assuming that was the point; if a set of six boards each need 20X, 20Y, and 20Z components, you'd make a spliced-together big reel from 120X, 120Y, and 120Z parts... and presumably keep an eagle-eye on the changeover at the splice

It would be impossible to manage. Feeders often lose parts. You lose a few when you load the feeder, anyway. You'd never keep them aligned. It's also more work to put together these special reels than to change reels in the feeder.

It would be nice if you could just slap on a new full set of 20 feeders and set the others aside- without waste or additional setup.

--sp

It doesn't sound like a Digi-reel mixes parts.

I had a friend who ran an assembly house and needed more vibratory feeders, and couldn't afford them. They were about $2K each. We found some aluminum extrusions that would work, and got the rest from Good Vibrations, a local lesbian sex toy shop.