OpenPNP - Discussion Thread

[mrpackethead]

OpenPNP is a reall viable option as well it seems.  Jasons recently got The vision stuff sorted out which pretty much places it in the same league as any of the chinese machiens, except that the code is all open.

[l0wside]

I´d love to include OpenPNP in my overview, but frankly, I cannot figure out how.

OpenPNP lists three machines (http://openpnp.org/hardware/):
- OpenBuilds. The concept looks to me like Volker´s PP4 approach: open source, but nothing which is commercially available.
- Teton: did not find any documentation except for a couple blog entries at hackaday.io
- Zeva 460: somewhat historical (found a for-sale-as-used ad from back in 1998)

So, I´ll skip the latter two.
I´d like to include the OpenBuilds version, but I did not find any performance data or cost indication. Diving into the Github files to do a shadow calculation is more than I am willing to do. If someone comes up with the data, I´ll be happy to include it.
What I´d love to see is LitePlacer with OpenPNP, but that´s just my personal opinion.

Max

[mrpackethead]

Its going to be differnet fthan the other stuff.

The very nature of OpenPNP is going to mean that there is a lot of different builds..  People are converting $300 shape-okos! through to designing reletivelky high end dual gantry machiens.

[vonnieda]

I´d love to include OpenPNP in my overview, but frankly, I cannot figure out how.

OpenPNP lists three machines (http://openpnp.org/hardware/):
- OpenBuilds. The concept looks to me like Volker´s PP4 approach: open source, but nothing which is commercially available.
- Teton: did not find any documentation except for a couple blog entries at hackaday.io
- Zeva 460: somewhat historical (found a for-sale-as-used ad from back in 1998)

So, I´ll skip the latter two.
I´d like to include the OpenBuilds version, but I did not find any performance data or cost indication. Diving into the Github files to do a shadow calculation is more than I am willing to do. If someone comes up with the data, I´ll be happy to include it.
What I´d love to see is LitePlacer with OpenPNP, but that´s just my personal opinion.

Max

Hi Max,

I'm Jason, the author of OpenPnP.

The BOM for the OpenBuilds machine is here: https://github.com/openpnp/openpnp-openbuilds/tree/1.0#bill-of-materials. It's about $917 all told.

This is my machine design and I primarily designed it as a development machine. I wouldn't suggest the build for someone who just needs a pick and place up and running right now, but if you like building and hacking I think it works well. I can't give you any performance data on it yet as I've been working feverishly to get bottom vision going. My goal is to get that finished (just about done, now) and then once I can accurately place my test board I will put some performance data out there. Right now I can do 0805s and 0603s pretty well "blind", but ICs tend to be sloppy. The bottom vision is fixing that.

The Teton technology machine is by a fellow named Anthony Webb who is active on the OpenPnP mailing list and a few others. I'm not sure if he visits this forum, but I will forward this thread to him. The BOM for that machine is listed on the left side of it's homepage. It's https://docs.google.com/spreadsheets/d/1ijdlkk7pOZnqEVCU9lDlx2OBI44QKNkokIkmXVnPd3g/edit?usp=sharing.

That machine is similar to mine but uses more off the shelf and generally better components. I will likely be building one quite like it soon to replace my machine.

The Zeva link is a retrofit, basically showing how you can retrofit an existing machine to use the OpenPnP software. This is a particularly important concept in the OpenPnP ecosystem. The software is designed to run *any* machine, not just mine.

Happy to answer any questions you have, and if this is a topic you are interested in please consider joining our mailing list: http://groups.google.com/group/openpnp

Jason

[Martin Hodge]

I have a NeoDen TM220A that I would like to see about retrofitting. But how does OpenPNP talk to the outside world? Is there a standard API or preferred interface hardware?

[thommo]

Hi guys,

Just putting it out there - has anyone had any experience working with and creating a controller [or just controlling] a manifold of solenoids to drive the likes of Yamaha's pneumatic feeders [as used in Charmhigh's latest PnP machine offering]?

http://www.charmhigh-tech.com/sell-new_chmt530p_pneumatic_feeder_smd_pick_and_place_machine_vision_system_mark2-3728907.html

Although more expensive than a 'drag advance system', they are still much less expensive than comparable electric feeders [about a quarter of the cost - ~$USD70 for an 8mm feeder].
Seems if we can get reliable 'sections' of the machine quarantined, focus on driving other parts seems far more doable. For those wanting something a bit more robust than what the current Chinese 'drag advance' methods offer, this may provide a workable solution

The work that Jason from OpenPnP has done to date [now including recent machine vision additions], begins to make the prospect of a reliable, custom machine a lot more feasible.
Then include other machine vision work that we also see on this forum like Optical Inspection, and things start looking very bright indeed.

[vonnieda]

I have a NeoDen TM220A that I would like to see about retrofitting. But how does OpenPNP talk to the outside world? Is there a standard API or preferred interface hardware?

Hi Martin,

There is a simple driver interface and several built in drivers for common controllers. There's some more information about that here: https://github.com/openpnp/openpnp/wiki/User-Manual#the-driver

I am in the process of writing a new, more universal driver that will be able to drive most basic controllers and addons. There is more information about that in this thread: https://groups.google.com/d/msg/openpnp/E8rF2faOjAg/xNJG6vo5IQAJ

That driver will be released in the next day or two and should make it easy to integrate most machines without having to write any code.

For machines that are more complex, or that the gcode driver won't handle for some reason, you can write a driver very easily. The interface is here: https://github.com/openpnp/openpnp/blob/develop/src/main/java/org/openpnp/machine/reference/ReferenceDriver.java

There's only about 6 required functions.


So, to get your machine working you would need to reverse engineer the protocol that it uses to talk to the computer and implement a driver for it, or you can replace the controller with something that is already supported. Smoothie is the controller that most people tend to use with OpenPnP but there is also built in support for Marlin, Grbl, TinyG and LinuxCNC.

Jason

[mrpackethead]

   <--- video posted to the openpnp mailing list today.

How cool is that. its a fuly 3d printed feeder..

[Martin Hodge]

There is no cover tape peel?

[thommo]

Hi Jason,

Any further updates on your Vision module?

[vonnieda]

Hi Jason,

Any further updates on your Vision module?

Hi thommo,

Algorithm development is complete and working. Now just integrating the code into the UI. Had hoped to release this past week but had to take a short trip out of town. Plan is to have it released by next weekend at the latest. Will post a video and the link to the instructions once it's ready to go. I think it's going to make things a lot better for a lot of people!

Jason

[vonnieda]

Hi folks, I've just released the first version of the bottom vision module. The release announcement with lots of additional info is here:

https://groups.google.com/forum/?utm_medium=email&utm_source=footer#!topic/openpnp/dumBEEgHg8E

Enjoy!

Jason

[glenenglish]

Jason
did some investigation into stepper microsteps on the bench .

http://glenenglish.com/2016/05/01/component-rotation
 
When I get my machine up, and Open PNP online in a week, hopefully I will be able to provide some feedback and improvements for OpenPNP.

Any idea what the error bars are on OpenPnP component rotation up vision estimation ?

Mitigation strategy for poor microstep accuracy, also
http://glenenglish.com/2016/05/02/stepper-component-rotation-accuracy-mitigation

 

[Martin Hodge]

I have a NeoDen TM220A that I would like to see about retrofitting. But how does OpenPNP talk to the outside world? Is there a standard API or preferred interface hardware?

So, to get your machine working you would need to reverse engineer the protocol that it uses to talk to the computer and implement a driver for it, or you can replace the controller with something that is already supported. Smoothie is the controller that most people tend to use with OpenPnP but there is also built in support for Marlin, Grbl, TinyG and LinuxCNC.

Jason

Attached are images of the stepper controller used inside of the machine. You can see it simply takes step & direction signals. If I'm understanding correctly, what I will need is a USB controller capable of outputting these signals along with some additional logic for air and tape-hook solenoids? Do you know if there's anything ready-made that could fit that bill and work with OpenPNP?

[glenenglish]

I'm sure you could hijack the signals going to the drivers on many common boards....
You'll just need to figure our what microstep they've configured for, and that would be a simple one iteration operation.


looking through smoothie doco
http://smoothieware.org/smoothieboard-schematic

I see through JP7 through JP15 the signals are on headers......
geez this board is nice.  you'd need to confirm what micros the TM220 board is in and tell the smoothieboard likewise.

ah nice it supports SPI ctl drivers
http://smoothieware.org/advancedmotordriver

so I can use my L6470s on it with a little work.

[H.O]

As far as I can see it only supports DRV8711 and Trinamics TMC26xx series driver chips. These are different to the L6470 in that "all" the SPI interface is for is configuration of the chips parameters - not for the actual motion which is still generated by the microcontroller on the Smoothieboard and sent to the driver chips as Step/Dir signals.

The L6470 on the other hand apparently contains a motion controller accepting high level commands over the SPI interface (cool!) but, as far as I know, the Smoothieboard does not support this mode of operation. (But please do correct me if I got it wrong!). It looks as if the L6470 does have a step-input but it seems to lack a discrete direction input so I'm not sure it's really compatible with rudimentary controller outputting step/dir signals....

Then again, the firmware is open source so perhaps your goal is to add support for "motion over SPI" to it?

[glenenglish]

Yes! there is some work for a person to do  ...  per my blog
https://glenenglish.com/2016/05/02/l6470-st-stepper-driver/
The L6470 is smooth.....

There is a smoothieboard forum  somewhere, best discussed there.
cheers.

[mrpackethead]

This will be awesome.  Jasons working on Loose part pick up! This is a game changer for low cost PNP