Choosing a Desktop Pick and Place Machine

[MR]

You do not need a high resolution with pick and place machines.
For Top Camera, I'd even see a colour camera as an advantage.

There are far more challenging optical recognition scenarios out there than for pick and place machines.

I did the optical recognition for my machine myself, and I've seen the shit Mechatronika has done.
My conclusion is:
- Optical recognition needs experience (that's it basically: time + watching the behaviour + various updates = experience), some components are just different than others.
- Or the pick and place software needs to allow some plugins which allows a programmer to add his own algorithm for certain components. even if it only fails like for 1 part out of 100 for that certain project it will be bad.

YUY2/YUYV/UYVY already represents a filtered format, it's rare that USB cameras will submit raw bayer pictures
(of course might be different when not using a controller in between which is doing that)

When looking at the YY1 videos I do not see that the format (whatever they use) matters, it's the algorithms, and for those they need more experience for some components (I'm even sure it works already as it is for many)

[glenenglish]

The bayer filter is on the imager itself , you cant fix that...you cannot undo what it does..  3 x3 or maybe 4x4 pixels are weighted into a single colour pixel.

Anyway. Good that you have sorted your mekatronica.

 What do you think / what is your opinion  about a used Jap or Korean machine compared to a 'new' chinese CL machine ?

[MR]

I could probably write a lot, we outsourced for some time and I saw how Juki / Yamaha assembly lines processed our PCBs...

Chinese lower end are better serviceable with standard parts by yourself I'd say, we have built our own machine in Taiwan (with the background knowledge of the Mechatronika MX80 since we disassembled the machine back then, and used relatively inexpensive parts too), the serviceability is in our hands now and it's a nice feeling to have everything under control.

Mechatronika is very much a standard machine just like the chinese ones. portal + Z / Theta that's it (while the chinese ones have multiple heads at least). There are so many servos out there which can be used, however Delta really shined with their international support (trying to help sorting out our Mechatronika problems which was not even their job).

The Mechatronika Machine offers far less value than chinese machinery, spare parts are a rip off (and can even be manufactured as single parts even cheaper than you buy it from them - eg nozzles, from professional CNC companies you can get 10 for the price of one from Mechatronika, the chinese Juki nozzles are also okay; I did not experience magnetic issues or a runout (the runout that applied in my case was my own mechanical calibration problem, and not related to nozzles bought from china)).

If you can go trouble free with a chinese machine for your products it's a big plus, certainly not all products are suitable for those mid / low range models and might require more rework after the pick and place process.

The Juki / Yamahas which were used for assembling our products were close to trouble free the PCBs that came out of it were 99.9% spot on and required very little rework if at all. The main issues were usually related to stencil or reflow oven -- so experience (and preparation) of the operator mattered. eg. baking parts, reflow oven settings have to match with IC specs, the reflow ovens have to operate in a calibrated manner.. it call can go out of calibration if sensors have issues.

Would I go for Juki / Yamaha / ASM for our products? Probably not, even our own machine is idle for over half a month. It has to work when we need it that's all.
Would I go for a chinese machine? With our background knowledge, our software and existing machine blueprints now, also no..
I'd rather build a second machine and chain them up.

For a beginner with less money I'd probably advise to go for a chinese machine, and maybe spend one month in deeply understanding the machine (you said you wrote your own software for it, I just don't know to which extend and what your requirement is, finally you wrote you're using a Hanwha.. so your requirement might be accordingly I guess).

There are many variables that might apply in deciding which machine to use... Some might get away with machines without vision while others badly need it.
Not having to deliver components to an assembly house, waiting for a slot, counting, assembling it taking them back and again is a big plus and time saver.

[newto]

Long story short, I have a new firmware written from scratch for YY1 and I managed to assemble a few boards with OpenPnP!

Wow, congrats.  I am sure others would be keen to try that.

Thanks. I have published my firmware:
https://github.com/mdepx/neodenyy1

It lacks documentation but I'm working on it slowly.

That's awesome! I was hoping someone would manage this (or build a custom board), because I definitely don't have the skills for that.

If I read it right, you're running openPNP externally and passing the commands to the YY1 board over the UART in the main screen port, and the new cameras are connected to the external device? (raspberry pi?). I'm not super familiar with openPNP, but I think I have some research to do before I start messing with a machine that I didn't pay for. Do you have any links for the cameras you used?

And I'm surprised the cameras weren't just USB, I assumed they were and possibly even the small displays as well.

Next step is powered feeders (or at least some kind of lock for tapes that won't cooperate with the peelers)

[br]

Long story short, I have a new firmware written from scratch for YY1 and I managed to assemble a few boards with OpenPnP!

Wow, congrats.  I am sure others would be keen to try that.

Thanks. I have published my firmware:
https://github.com/mdepx/neodenyy1

It lacks documentation but I'm working on it slowly.

That's awesome! I was hoping someone would manage this (or build a custom board), because I definitely don't have the skills for that.

If I read it right, you're running openPNP externally and passing the commands to the YY1 board over the UART in the main screen port, and the new cameras are connected to the external device? (raspberry pi?). I'm not super familiar with openPNP, but I think I have some research to do before I start messing with a machine that I didn't pay for. Do you have any links for the cameras you used?

And I'm surprised the cameras weren't just USB, I assumed they were and possibly even the small displays as well.

Next step is powered feeders (or at least some kind of lock for tapes that won't cooperate with the peelers)

Yes, the main screen port, which is 4 pin UART (+5v, rx, tx, ground).
OpenPnP is running on a PC box, USB cameras are connected directly to the same PC.
OpenPnP talks to the motion controller (I call it main board on the firmware page) over UART.
I will post hopefully this week links to cameras and gerbers of PCBs that hold them. Also the bottom LED ring.

The first challenge for everyone interested would be to find a suitable debugger and try to read out your old firmware. I noticed there are some magnetic fields around so you need to play a bit with cables

[newto]

The first challenge for everyone interested would be to find a suitable debugger and try to read out your old firmware. I noticed there are some magnetic fields around so you need to play a bit with cables

Would removing the main board from the machine before messing with the software help? I can't imagine the cheap stepper drivers and all the other signalling and power going through the same board helps that situation. Definitely a little hesitant to risk a corrupted backup with nothing to compare it to if something goes wrong with the update process. If it was my own machine I'd be much more eager, but I don't want to try to explain to my boss why I broke our expensive hardware playing with it and need to order a new mainboard. And I swear my firmware from September 2022 has fewer bugs (or at least fewer critical bugs) than some of the newer machines people have received, so I wouldn't want to lose this version.

I'm excited to give it a try once you get more of the details up though, I know a lot of people have been interested in getting them to run more reliable software

[br]

Would removing the main board from the machine before messing with the software help? I can't imagine the cheap stepper drivers and all the other signalling and power going through the same board helps that situation. Definitely a little hesitant to risk a corrupted backup with nothing to compare it to if something goes wrong with the update process. If it was my own machine I'd be much more eager, but I don't want to try to explain to my boss why I broke our expensive hardware playing with it and need to order a new mainboard. And I swear my firmware from September 2022 has fewer bugs (or at least fewer critical bugs) than some of the newer machines people have received, so I wouldn't want to lose this version.

I'm excited to give it a try once you get more of the details up though, I know a lot of people have been interested in getting them to run more reliable software

You will need to remove it to solder down 2.54mm header for SWD, but I'd recommend then to put board back and connect a debugger permanently, this gives you a possibility to update firmware any time later. Debuggers are cheap, just leave one there forever.
Also the board has to be powered during firmware update, otherwise stm32 does not receive power.
If you want just to try and have quick access to the back of machine, you can just hold cables with fingers for the backup and update procedure.

Note that you may need like a month or so to dive into OpenPnP as things may not appear straightforward.

(not sure if a dedicated thread is needed for YY1-firmware related discussions?)

[glenenglish]

I could probably write a lot, we outsourced for some time and I saw how Juki / Yamaha assembly lines processed our PCBs...
(truncated)
Chinese lower end are better serviceable with standard parts by yourself I'd say, we have built our own machine in Taiwan (with the background knowledge of the Mechatronika MX80 since we disassembled the machine back then, and used relatively inexpensive parts too), the serviceability is is a big plus and time saver.

Nice writeup MR, very useful info for everyone.