Afarco

110 views
Skip to first unread message

SR

unread,
Mar 30, 2026, 3:50:34 PM (8 days ago) Mar 30
to OpenPnP
Hi, found Afarco by chance....did ever anyone use it successfully?

Chris

unread,
Mar 31, 2026, 9:03:29 AM (7 days ago) Mar 31
to OpenPnP
I bought a mini v3.2 earlier this year. So far it's a great machine once you get past the learning curve that is OpenPnP. Happy to answer any specific questions. 

SR

unread,
Mar 31, 2026, 9:48:44 AM (7 days ago) Mar 31
to OpenPnP
thanks, you are the first one i'm getting an answer from.  From all those videos afarco posted, the device looks decent. How is you experience, mis-picks, failures, technical problem, Accuracy over time goes down, frequent recalibarion needed?  

Also the Afarco team is not that communicative, which felt a bit weird almost like scam. 

I have a production running and each year we populate 6000-8000 PCB with average 50 parts. In your opition how would the Afarco mini 3.2 suite for the job?

Oscar Illerup

unread,
Mar 31, 2026, 1:29:39 PM (7 days ago) Mar 31
to ope...@googlegroups.com
He,

I have the "big" version. My experience has been up and down. At first, I had a lot of problems because there was no blow (vacuum release) on it, so I had to add it myself. After that, it ran great with 0603 components and larger. However, I needed both smaller and very large components, which it really struggled with. Therefore, I replaced the cameras with Lumen ones, which helped a lot. Now I run 0402s and modules that are over 20mm.

Another issue is that it does not have auto-squaring. This is something I really would have liked to see, as it is a rather cheap feature to add originally; however, because there are no more GPIO pins available, I cannot add it easily.

The motor controller used also does not support idle mode, which means the motors get pretty toasty. I learned this because I tuned acc and speed right after a long "off" period, and when the motors got hot, the torque went down. It lost acceleration and crashed. I have improved this by simply installing bigger motors without changing the settings, so they do not get as hot. This means I only get around 700–800 CPH (components per hour) with vision enabled.

The included PC was also too slow—maybe because of Windows, but I don't know for sure. I just extended the cables and used a PC made for my CNC (which runs Linux).

The machine is pretty rigid and feels nice with an aluminum structure and linear rails, although I suspect the rails are not the highest quality.

The feeders are very good. I ordered around 45 of them in my first order. I have been told these are V1. Around 40 of them have had no issues (only some sticky tape getting stuck in the gears occasionally). The other five had issues, but I wrote to him and he sent five new V2 feeders free of charge. In addition to those, I purchased 10 more. I have not tried the V2s yet (they just arrived this week), but they look to have been improved mechanically; they feel a lot more secure with two locating screws and are thinner.

I would probably hesitate to go for it again if he does not fix some of these issues, which is sad as it is a great machine when everything is working. I would like to see these upgrades on a future version:

  • Adjustable blow: (I just made an on/off version with a script that turns on for the period set in the blow parameter in OpenPnP).

  • Squaring of Y: (With two endstops, one for each motor).

  • Better control board: (With spare IO pins for customization and better options for external stepper drives for closed-loop upgradability).

  • Better cameras!!

  • Cost savings on the PC: Use that budget for a closed-loop stepper setup for better speed :-)

I would totally buy it as a frame kit and use his feeders, but I think the electronics are currently lacking.

To close, I would really like to mention that he is a great responder and has helped every time I reached out to him. I believe he can absolutely make this a great machine.

If you have any specific problems, you are welcome to reach out and I can see if it is something I have encountered as well.

Best Oscar Illerup



--
You received this message because you are subscribed to the Google Groups "OpenPnP" group.
To unsubscribe from this group and stop receiving emails from it, send an email to openpnp+u...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/openpnp/e5a74dfc-6085-4aaa-a98b-376ddfa41e26n%40googlegroups.com.

Chris

unread,
Mar 31, 2026, 2:04:05 PM (7 days ago) Mar 31
to OpenPnP
I wrote a full reply, and it seems to have disappeared, no idea what I did wrong, but I'm sure it was my fault. Tried to remember and retype most of what I wrote. 

My experience has been largely positive, but not perfect. I've only had the machine for around a month and a half, so I can't conclude yet about long term reliability. However the machine is well thought out and well built from what I can see. Most, but not all, of my problems have been with OpenPnP, not the machine. 

What's been good:
It seems to place parts very reliably. I've put down a few hundred 0402 so far and most are dead on. It's done fine pitch QFN packages and large ESP32 modules very well, once you get the OpenPnP vision side tuned. Repeatability seems great so far. 
Assembling it was easy, and the assembly videos for the mini3.2 are good. The included paper instructions were lacking, and I think I didn't get half of them. But the videos were good and everything is labeled well enough. 
The feeders are pretty good, other than a lack of documentation. I plan on buying more of them once I work out how many more I need. I bought just enough to get started with the machine. 

What's been less good:
Two holes for mounting the X axis assembly came stripped out of the box. I reported it, and then fixed it with helicoils because I didn't want to wait for parts to ship. 
It seems to take 5-10 feed motions to get the feeders settled and tensioned right. The first few feeds can be less repeatable and cause 0402 parts to get picked up at odd angles. Once you get around 10 feeds in things seem to settle and become repeatable. This would totally be solvable if OpenPnP would extend the advanced features of the pushpull feeder to also be usable with auto feeders. 

The feeder mounting plates are designed for the 8mm feeders, so if you use a bunch of 12mm you need to skip slots, and it limits how many you can fit. I asked if he'd make a 12mm specific mounting plate so you could go from 5 up to possibly 8 of the 12mm feeders per plate (16 per machine side), and he said he can and would ship one to me when I place my next feeder order. (I was up front that I plan on buying more once I work out what I need). 

Documentation for the feeders is sparse. There are 4 settings in the feeder, but I can only find a video on how to set the RS485 address. I did get an answer about 2 other settings, but not the last one. ad is address, st is feed step size in mm, and pe is peel delay with each increment corresponding to 10ms. rt hasn't been explained to me yet and I haven't figured it out on my own. 

For some reason I cannot get the controller to talk to linux. I'm sure this is solvable, but I was told by AFARCO that they only do windows so I'm on my own. The computer I'm using is older and can't be upgraded to windows 11, so I'm stuck with 10 and keeping it off the internet. I would prefer to run linux on that machine, so someday I'll figure out what is going wrong. 

I agree with some of the comments about the controller from Oscar. I think the controller in mine is updated from the Iron machine, but there is still room for improvement. Spare GPIO would be awesome, added vacuum sensing and blow off would be great. I also had to add a ground wire to my controller and machine. The power cable that shipped with the machine was not US spec, and only had 2 pins, so I assume ground pins aren't as common where the machine is built, but I wanted one and added one. 


My issues with OpenPnP:
1. There is a bug with nozzle tip unloads that adds the nozzle tip calibration offset to unload moves, which effectively causes unloads to happen in a different location. Since AFARCO's nozzle station has tighter tolerances, this has been causing machine crashes for me at unload. I'm trying to help fix this bug with OpenPnP. Not a problem with the machine. 

2. As I said previously, There's some awesome features in the pushpull feeder like OCR/QR vision for auto assigning the loaded part, vision for pick location, etc. But you can't use it with auto feeders. I think I can get the push pull to work with auto feeders with some scripting, but I really want to see OpenPnP expand these features to other feeder types, it would make the experience so much nicer. 

3. The shear learning curve of OpenPnP has been a lot. Documentation is mixed. Some things are documented really well. Some things not at all. I tried finding a page on all the stages you can apply in the vision pipelines and there's just nothing. So you're left having to insert each stage and read the tooltips and they're often not super helpful. Documentation in particular has been my biggest sticking point with trying to learn everything. 

I wouldn't hesitate to tell someone to get one of the new mini3.2 machines if the feeder capacity and working area meet their needs. It's not perfect, but it seems largely good so far. I can't speak to long term reliability until I have more time with it, but so far I see nothing that worries me. 


Chris. 

Jan

unread,
Mar 31, 2026, 5:21:52 PM (7 days ago) Mar 31
to ope...@googlegroups.com
Hi SR!
If you take your numbers and the speed Oscar has reported, I'd say no:
8000 boards/year * 50 parts/board / 800 parts/hour = 5000
machine-hours/year is more then 52 weeks/year * 5 days/week * 8
hours/day = 2080 work-hours/year. And that assumes, that you have no
down-time and you can swap boards in no-time.

Jan
> --
> You received this message because you are subscribed to the Google
> Groups "OpenPnP" group.
> To unsubscribe from this group and stop receiving emails from it, send
> an email to openpnp+u...@googlegroups.com
> <mailto:openpnp+u...@googlegroups.com>.
> To view this discussion visit https://groups.google.com/d/msgid/openpnp/
> e5a74dfc-6085-4aaa-a98b-376ddfa41e26n%40googlegroups.com <https://
> groups.google.com/d/msgid/openpnp/e5a74dfc-6085-4aaa-
> a98b-376ddfa41e26n%40googlegroups.com?utm_medium=email&utm_source=footer>.

tonyl...@gmail.com

unread,
Mar 31, 2026, 5:40:03 PM (7 days ago) Mar 31
to OpenPnP
8000 boards/year * 50 parts/board / 800 parts/hour = 500 machine hours/year

bert shivaan

unread,
Apr 2, 2026, 10:16:05 AM (5 days ago) Apr 2
to ope...@googlegroups.com
Dang decimals

To unsubscribe from this group and stop receiving emails from it, send an email to openpnp+u...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/openpnp/86195ce3-958a-4bdc-8d80-e3e19a9230cfn%40googlegroups.com.
Reply all
Reply to author
Forward
0 new messages