VisionBot Pick'n'place

[l0wside]

EDIT: Alexandru has let me know that further improvements (including double-sided boards and improved speed) are to come. Updated my comments accordingly.

After Alexandru added the data of the VisionBot to my overview (https://drive.google.com/open?id=1s9YmD-L69UXG5paMiz0pwO06J9nptdcopGrSa_s9t_E), I am taking the liberty to comment on the machine.

At 3000$, he is the undisputed cost leader. The cheapest Chinese machine (TVM802A) is 3400$ plus shipping plus duties (1.7%), so it will end up somewhere just south of 4000$.

What stands out is the automatic nozzle changer and the huge maximum component size (30x30mm is a lot). Alexandru has confirmed that the machine is capable of reliably placing 0.5mm pitch (some customer places a 144pin TQFP with 0.5mm pitch).

So, right now, only two points remain:
- 1000 cph is more in the league of the Liteplacer than the Chinese machines. Alexandru has assured me that he is working on this issue.
- The PCBs are currently on just a flat table, making PCBs with components on SS impossible to use. He is working on this point as well.

The feeders look extremely similar to Volker Besmens´ approach. They seem to only pull the cover tape; any misalignment during feeding is compensated for by vision. 24 feeders is more at the lower end of the field, which is ok looking at the price tag.

The software seems to be really complex, hopefully it is equally powerful.

Max

[alexandru]

This weekend I will get back in the production line and we will make a new video showing VisionBot placing with the new Juki nozzles 0402 and TQFP with 0.5 pitch. One of our customers Rockna is placing TQG 144 pins Spartan at everyboard with our Bottom Vision, pitch 0.5. Our main focus was not the speed, but because I see that all you guys are asking for this (although I don't believe any of you is placing 10,000 components/h with n4) we will increase the speed a lot in the next models in two weeks. We limited the speed from the firmware because the steppers and the belt can move also very fast. The technology is just from any 3D Printers and very similar to any other belt machine. Also guys because I am not an electrical engineer how do you really calculate the CPH? (doesn't it matter how the BOM is made of, which feeders are you using)? In my opinion the CPH is totally a really bad approximation and most other PnP have a very bad approximation of their CPH to the real CPH.  One can change the pcb holder to assembly double sided boards. I don't believe that the current speed can turn you down to look similar to a kit.

These are some pictures

[ar__systems]

-- 1000 cph is more in the league of the Liteplacer than the Chinese machines. Alexandru has assured me that he is working on this issue.
-- The feeders look extremely similar to Volker Besmens´ approach. They seem to only pull the cover tape; any misalignment during feeding is compensated for by vision.

I doubt much higher speed is realistic unless frame is considerably beefed up. The machine will fall apart from vibration if things start moving faster. The frame seems really flimsy. At least with Chinese you get a really solid mechanics.

Good idea for camera assisted feeder. But pulling on just the cover is not - it creates jerky movement and causes parts jump in the pockets - that can actually be seen in the video.

[alexandru]

I doubt much higher speed is realistic unless frame is considerably beefed up. The machine will fall apart from vibration if things start moving faster. The frame seems really flimsy. At least with Chinese you get a really solid mechanics.

Good idea for camera assisted feeder. But pulling on just the cover is not - it creates jerky movement and causes parts jump in the pockets - that can actually be seen in the video.

We will see in the next weeks. The servos which are advancing the feeders at a certain speed don't make the components to shake and to jump from their pockets. The feeders' servo speed is variable and can be set from the software.  Dave filmed PnP 200 components from feeders and none of them jumped from their pockets

[ar__systems]

We will see in the next weeks. The servos which are advancing the feeders at a certain speed don't make the components to shake and to jump from their pockets. The feeders' servo speed is variable and can be set from the software.  Dave filmed PnP 200 components from feeders and none of them jumped from their pockets

Ok then, what happens in the video at 1:39? I think I see and LED turned on its side.

[ar__systems]

At the end of the video the presenter announces "all nicely placed" but per my standard LEDs are placed not quite nicely. Most of them are significantly shifted up.

[alexandru]

Ok then, what happens in the video at 1:39? I think I see and LED turned on its side.

Due to the not-perfect Vision Settings for that LED feeder which was done by him, the VisionBot did not detect one of the 192 component from the pocket, so the machine didn't pick it actually.

Cheers

[alexandru]

As I promised in the previous days, I did a simple PCBA test on a sheet of paper placing 0402, 0603, TQFP and SOIC. The 0402, 0603 was with Top Vision only and we can also run with Bottom Vision which I do another video test next week.  The quality of the PCBA batch on sheets of papers is really bad because of the bad fiducial detection and the paper sometimes moves and rotates. I didn't have a lot of time to do multiple tests and now I am doing on site support to one of our national customer who is 500 km far from us.

https://youtu.be/s4jcPyHZ2c8

Also I didn't have the exactly components from the VB dummy PCB's footprints on the stock here.

LE: the sheet of paper was not perfectly flat. Maybe we will film in a few hours the VisionBot assembling a complex board at one of our customer today.