Really, not painful

[Jet]

Why so long?? It's 250 ms by default

I've been testing with 0402's and with no adhesive or paste on the board. 0402's are light and need a long release time if there's nothing to stick it
to the board. The reason is some of my parts I don't want contaminated with adhesive. I find the adhesive from the postit note type and double tape adhesive
make the parts stick in the tape pockets when put back in their holes, which messes up pickup.

I don't plan on using 2.5s in production, just testing.

I am using different parts and observed this effect on different parts (but mostly on heavy parts as i remember, but i had not tried 50 ms with 0402).

0402's are very "tacky", I'll be changing to a vacuum and puff mechanism and also adding a vacuum sensor to detect parts on the nozzle. Also, it's hard to gauge
accuracy (or even care about it too much), until placing 0402's. It does show up with 0603's and even 0805's, but generally not bad enough to have it not
pull in on reflow. Then any inaccuracy is even more problematic on small pitch long parts or larger IC's... small rotation errors can result in a large errors over a long span.

Additionally problems with tapes are the components are in pockets that are larger than the component, for example I have some 0402 caps that fit their pockets quite
well, and some 0402 resistors that don't (maybe 10% space around), so the caps are better for testing alignment.

Then nozzle dance, resistors for example, typically aren't perfectly flat for example and are coated on top, so they often move a bit on the nozzle on pickup.

But the good news, is that Bottom Vision solves all that and I'm so pleased that I have accurate 0402 placement and fine pitch long connectors with it. However I needed
to come up with a totally different calibration method to the suggested ones to get this kind of precision, they're just not accurate enough for consistent 0402 placement.

I wouldn't normally be placing 0402's, but I've had to design a miniaturized high density board for a medical application that has a size constraint, which
is too problematic to place by hand and I have to make a number of them.

Frankly I'm amazed how accurate the Liteplacer can be.

[wormball]

What software gives more accurate result without vision? I heard that the openpnp is very inaccurate, unlike the liteplacer software.

Additionally problems with tapes are the components are in pockets that are larger than the component

Yes, especially sot23-5 and sot23-6 (but not sot23-3 or sc70-6). Liteplacer is almost unable to place mentioned parts properly.

But the good news, is that Bottom Vision solves all that and I'm so pleased that I have accurate 0402 placement and fine pitch long connectors with it.

But it takes quite a time. I think it is possible to tune the liteplacer that it will place 0402s properly without vision. I actually did that but since then i played with liteplacer's settings and had not tried to place 0402 again.

[Jet]

What software gives more accurate result without vision? I heard that the openpnp is very inaccurate, unlike the liteplacer software

I would say the opposite. I've used Liteplacer software for a while, without being able to place 0402's accurately.

WIthout Bottom Vision, OpenPnp and Liteplacer have equal accuracy (afterall it's the same machine), but with Bottom Vision in OpenPnp it blows
the Liteplacer software out of the water in terms of accuracy (check out both my pictures in this thread with and without bottom vision).

I'm not knocking the Liteplacer software though, it doesn't have Bottom Vision and Juha's working on it, when it does it should have equal accuracy to OpenPnp.

Also, the OpenCV Pipeline in OpenPnp is more stable under varying lightning conditions as it has more capability. I started with OpenPnp just over a week ago and I'm about
to begin making boards tomorrow.

I'm in the last stages tonight, 0402's, SOT23-3's, 30pin 0.4mm pitch mezzamines and FPC connectors are all placing perfectly. Currently working
on auto part recognition of the remaining parts tonight.... which is pretty cool, I can put multiple diode arrays, fpc, mezzamine, the one 0805 I couldn't reduce in size, and
put them al in a bundle together and any orientation, and it will auto identify them and place them. So basically I just have a bunch of tapes for the 0402 (easier to keep them in their tape),
and a bunch of loose parts in an area.

And the best part, to train for a particular part with the loose part recognition, I just take an image of the part on the table, the same vision pipeline works for everything in the area, very cool.

Yes, especially sot23-5 and sot23-6 (but not sot23-3 or sc70-6). Liteplacer is almost unable to place mentioned parts properly.

So for me, I'm just dropping em on the table rather than pick from the tape it's easier. Depends how many you're using. If picking from tape, you'd need bottom vision.,
SOT23-3 are always very loose in the tape.

But it takes quite a time. I think it is possible to tune the liteplacer that it will place 0402s properly without vision. I actually did that but since then i played with liteplacer's settings and had not tried to place 0402 again.

Yes, bottom vision is an extra step, you have 1 extra move and 2 Z moves and any small adjustment moves (typically just rotation), the pipeline only takes a fraction of a second.

And it depends what you consider acceptable. 0402's will get pulled in under surface tension, however I found every few picks without bottom vision they would be out by too much. 0402's can
also be problematic when not correctly on the pads, more than 0603's. I prefer to have mine centered in the pads, rather than having only part of the component on the pad and relying on the reflow
to catch it and not tombstone or reflow it off center.
Then when you get into fine pitch connectors, BGA's and other fine large IC's, it becomes unworkable without bottom vision IMHO.

[Jet]

After many month’s and a hair pulling.... finally in production with OpenPNP, upvision and an accurate machine.

Just finished 4 - 1x1 inch high density double sided boards with 75 SMD s (0402), 1 x 144 ball 0.8mm pitch BGA, WSON-6/8’s, USB-C and 0.4mm pitch connectors. Getting phenomenal placement accuracy.

Thanks to Juha and the help from people on this forum.

I would have had no chance placing this boards reliably by hand before the paste dried and my neck would have paid the price.

Liteplacer rocks!

[matthew]

@Jet, Have your OpenPnP configuration files changed since you last posted them? Any updates would be appreciated!

[Jet]

Probably not much, but I'll dig out the latest when I do another job run (likely next week)

[matthew]

So I was able to use the files that Jet posted. Seems like I had to make one or two tweaks for axis control... maybe something else. My head is still spinning.

The real problem that I've run into with OpenPnP is getting the Vision Pipeline working... I haven't. I've used the Vision Pipeline output files to monitor the output as I try to change the settings. Nothing seems to make a difference. Yes, somethings prevent images from being saved out, but nothing seems to have the kind of impact needed to really use vision. Something must be escaping me.

As others have stated there appear to be advantages in both SW packages. For now, I've headed back to LitePlacer and am cheering for Juha to have great success at implementing his list of desired features.

Some of the entries on the Job page are a little strange to get used to, but after building a 30+ component board I got the hang of it. Now onto a 160+ component board.

My major hold up at present on OpenPnP is lack of documentation. There's just enough that it seems like you'll get through it... but once in deep there are some severe holes and lack of explanation. On the Vision end of things, Liteplacer was a lot easier to get started. It has it's hiccups, but I found them easy to work through and it get you going in a pretty direct fashion, as well as holds your hand just enough to understand the necessary details even if the documentation is a little out of date in places.

A general note for anyone reading along before diving in. Tall components in cut tape can be tall enough that the fixed focus lens has trouble identifying them with the vision recognition. This will be true in both SW packages. The only question is whether the vision pipeline can adapt for it. The hardware solution is to put the TOP side of all components and the PCB at the same height... the height of the tallest component you are working with. That means lifting the homing mark a little as well. You can get away without this if the components are all relatively short, chip caps, resistors, SOICs, TQFPs, etc. Try a ceramic chip antenna on top of it's GPS chip and focus on the holes in the tape. It's close to 10% of the focal length above the homing mark. That's a general hardware issue.

[Jet]

I only use OpenPnP now. But just like LitePlacer, there's a learning curve. I found OpenPnP more capable and easier / logical to use (personally), particularly as bottom vision was a requirement for placing 0402's accurately.

Yes, it's very important the top of your PCB, your homing mark, and the top of the components are all at the same height (this would also apply to LitePlace software too).

That said, there is some amount of inaccuracy with this you can get away with if you use bottom vision too or larger components.

For example, anything from tape relies on the vision for the holes for pickup (not the component). So if you say have a 0402, it's small so the vision height is important, if it's larger, it's less important (I've found). But I am placing GPS (4mm high), LiPo Connector (6mm high), RFM95, tactile switches and USB connectors realiably, these are all wildly different height parts, I'll use the Advanced Loose Part Feeder for at lot of this and have them found automatically.

The ALP Feeder saves me a lot of time.

(I also have a custom alteration to that which does fixed rotation too, so if the focus is slightly off (some ICs have very faint small markings, that the ALP might not be able to see a pin 1 id on for example), it will be in approximately the correct orientation so focus doesn't have to be perfect.

I have a few jigs I made to ensure the top of the tape components is the exactly same as the top of the pcb, and when I calibrate/test, it's done with 0402
placements.

You need good tape hole recognition for clear/white/black and will need slightly different pipelines for each. Clear will always be an issue, so I use a Sharpie to paint the tape hole area (upto the component) black. During setup I'll align the first tape hole and the last tape hole on the strip of tape, rather than adjacent holes. I also use a pink background for vision recognition (I tried a few colors, pink worked best). This is very important, I found that the LIteplacer pipeline wasn't enough to recognize consistently, but with OpenPnP when using pink it was. Also I have pink for the nozzle background
and pink nail polish painted on the nozzles. The tip of the nozzle I use a pink sharpie (which is very important especially for 0402's with the smallest nozzle... I've sanded off the black anodization). Pink is my "green" screen.

I put up a video here that shows the pink background and the laser cut jigs I'm using. I label up the trays, and they stack with a rubber band around them to keep different projects / jobs separate.

https://twitter.com/ChasinSpin/status/1 ... 4490128384

If you want to see a 3D mockup of the board in more detail (to get an idea of what's on there and what OpenPnp can do) , scroll down / up in the timeline.

I also have a vacuum sensor. My focus is accuracy / reliability over speed. One thing I will change at some point is use feeders where I'm using tape, or something to save using double sided tape. And at some point I may try to push the speed some more. For large runs, I'd just send it out, but for small sample / prototyping runs, it's working well for my clients, saving me from neck pain/hand assembly and fills a sweet spot in job size that clients like.

[Jet]

Probably not much, but I'll dig out the latest when I do another job run (likely next week)

This is the latest (albeit a bit late)

machine.zip

(33.03 KiB) Downloaded 619 times

[wormball]

The ALP Feeder saves me a lot of time.

What is it? I can not find anything of this name both in this forum and in the internet.