Status Update for December 2018

A project log for Hot Hot (Yushengtech) SMT50 OpenPnP conversion

Refactoring the SMT50 to work with OpenPnP

darryldar.ryl 12/22/2018 at 02:470 Comments

I've been MIA from HaD since early this year, so I wanted to give a status update to let people know that I have not abandoned this project.  

As mentioned in a previous log(s), the big obstacle I encountered was adapting the pickup head to work with the Juki nozzle and the snazzy clicky nozzle adapter.  I was hoping to get my 6040 CNC machine working well enough to cut metal to build something. 

The 6040 still isn't working, but the big exciting news (at least for me) is that I now have in-house machining capability! A mini-mill with 3 axis DRO, a 7" mini- lathe, a horizontal band saw for rough cutting metal stock, and all the required tooling.   

Back in the late 90s/early 00s, I had access to a full-size mill and lathe, and was able to gain some experience making very simple parts using Al, ABS, and brass. So, I'm not a complete noob, but in the ensuing years, I've been doing mostly embedded software, so definitely need some time to refamiliarize. 

Hopefully be able to make some progress on this project very soon now.  

Meanwhile, please follow and comment!

Planned upgrades: 

I want to redesign the pickup head to support two Juki nozzles. Primarily because it would double the placement throughput, but also because I don't like the existing design with the offset rotating actuator and spring loaded return. I would like to use a belt drive to get more precise control over the Z travel.  A redesign would allow the freedom to "do it right".

I also need to design and fab mounts and LED rings for the two new USB cameras.  

Since the new pickup head design will eliminate the drag feed pin (which uses the other half of the head), I would also like to rethink the feeders.  Two popular approaches for this are an independent feeder-per-component, or an indexed common feeder.  I don't like the indexed approach because it adds layers of mechanical complexity and motion control, plus control software has to coordinate the indexing motion based on the component width.  The feeder-per-component approach "hides" the complexity and can scale to support as many feeders as can be physically fitted.  So, we have a design requirement to add feeders, each incorporating an embedded micro which suggest using a multidrop serial bus to signal.  This opens up the possibility of "smart" feeders that can be hot swapped to replenish the component feed while a job is running.  With a few carefully placed optical sensors, the feeders could detect missing components, end of reel, and perhaps even estimate the number of components remaining on the spool.