Upgraded mechanics

[mrandt]

In case you have not seen it yet, Protovoltaics considers their machine finished and posted a new video on YouTube / Hackaday:

https://youtu.be/Ny99KzsySHE

There are many things I really like about the machine:
- Dual camera vision eliminates the need to move camera to needle location and back - ingenious
- Magnetic (?) automatic nozzle changer seems to work nicely -> https://youtu.be/Ny99KzsySHE?t=1m10s
- UI looks very functional -> https://youtu.be/Ny99KzsySHE?t=21s
- Automatic feeders look sturdy and fully functional, similar to commercial ones -> https://youtu.be/Ny99KzsySHE?t=2m23s
- IC placement works using bottom camera for alignment -> https://youtu.be/Ny99KzsySHE?t=3m16s
- Panelization is included in software

Other interesting things in the video:
- Vacuum pump seems to be modified aquarium pump (Tetra or similar) -> https://youtu.be/Ny99KzsySHE?t=4m37s
- CNC control board is SmoothieBoard
- No idea what the software is written in and which computer vision libs it might use - but it runs on Linux in the video

Protovoltaics claim they will sell the machine fully assembled in a basic configuration (assume few feeders only?) for under $6k and are about to start a Kickstarter campaign.

I think one of their staff is member of this forum - so congratulations to your achievement; great job!!! I wish you much success when marketing this thing. Especially the feeders set you apart from all other P&P kits in that price range.

What I do not like though is the fact that everything seems closed source - at least I could neither find any CAD drawings of the parts nor source code of the software published anywhere on the web.

Looking at the machine, I think they benefitted a lot from other peoples willingness to share - XYZ frame seems to be somewhere between ShapeOko and Inventables X-carve design, nozzle changer is much like Jason von Nieda's OpenPNP hardware, the converted aquarium vacuum pump I first saw on Felix Rusu's blog, controller is an open source one, feeders are certainly inspired by endless forum threads - list could go on. This fact is even acknowledged in the video (bullet 10): https://youtu.be/Ny99KzsySHE?t=5m12s

While I understand and appreciate the commercial interest, I think it would be fair to give back to the maker community and publish at least some helpful aspects of the implementation for those who want to make their own machine. These people would probably not buy a ready made one anyways... Just my two cents.

As a potential customer, not having the software sources would be a deal killer for me and I am pretty sure I am not the only one. What if I needed a certain feature but cannot implement it? What if a bug prevents me from completing a job and Protovoltaics cannot fix it right away? What if Protovoltaics goes out of business or simply shifts priorities? So Protovoltaics will not see me back their campaign - but I still wish them success as I think the more players in the semi-pro P&P field, the better

Anyways, I think just seeing a complete machine in action gives us plenty of inspiration and there are certainly some aspects here that LitePlacer can learn from - that is why I am sharing my thoughts here.

[protovoltaics]

In case you have not seen it yet, Protovoltaics considers their machine finished and posted a new video on YouTube / Hackaday:

https://youtu.be/Ny99KzsySHE

We do not consider the machine finished, but we are at a point were we want to start selling it as a product (I believe at our price point, and the fact our software is Closed Source, we are not a competitor of Liteplacer, but an alternative). We are still working to make it better everyday, and that will remain our main goal. Any software updates we make will be available for download on our website, and if we ever make a hardware upgrade those upgrades will be available through our webpage as well.

There are many things I really like about the machine:
- Dual camera vision eliminates the need to move camera to needle location and back - ingenious
- Magnetic (?) automatic nozzle changer seems to work nicely -> https://youtu.be/Ny99KzsySHE?t=1m10s
- UI looks very functional -> https://youtu.be/Ny99KzsySHE?t=21s
- Automatic feeders look sturdy and fully functional, similar to commercial ones -> https://youtu.be/Ny99KzsySHE?t=2m23s
- IC placement works using bottom camera for alignment -> https://youtu.be/Ny99KzsySHE?t=3m16s
- Panelization is included in software

Glad you like so many things about our machine
Yes, the nozzles are magnetic.

Other interesting things in the video:
- Vacuum pump seems to be modified aquarium pump (Tetra or similar) -> https://youtu.be/Ny99KzsySHE?t=4m37s
- CNC control board is SmoothieBoard
- No idea what the software is written in and which computer vision libs it might use - but it runs on Linux in the video

Correct again! We wrote the software for the machine using C++

Protovoltaics claim they will sell the machine fully assembled in a basic configuration (assume few feeders only?) for under $6k and are about to start a Kickstarter campaign.

I think one of their staff is member of this forum - so congratulations to your achievement; great job!!! I wish you much success when marketing this thing. Especially the feeders set you apart from all other P&P kits in that price range.

Thank you for your kind words. We are KickStaters first PnP project, so we aren't getting our hopes up too much (Much less of a market for PnP's than all those 3D printers that get millions in funding!). We have met several people at maker faires who want the machine, but don't want to buy it through KickStarter, because they have had bad experiences with not getting what the ordered before (3D printers again!). Machine will come fully to partly assembled (might have to attach head to the gantry). Free shipping (to some areas) and discounted feeders are available through the KickStarter with the purchase of a machine.

What I do not like though is the fact that everything seems closed source - at least I could neither find any CAD drawings of the parts nor source code of the software published anywhere on the web.

Looking at the machine, I think they benefitted a lot from other peoples willingness to share - XYZ frame seems to be somewhere between ShapeOko and Inventables X-carve design, nozzle changer is much like Jason von Nieda's OpenPNP hardware, the converted aquarium vacuum pump I first saw on Felix Rusu's blog, controller is an open source one, feeders are certainly inspired by endless forum threads - list could go on. This fact is even acknowledged in the video (bullet 10): https://youtu.be/Ny99KzsySHE?t=5m12s

We are using some open source stuff like the X-Carve and Smoothie boards. We are not using OpenPnP, we wrote the software. The nozzle changing was just something we knew we would need. Same with the vacuum pump just a good idea to use something like a aquarium pump .

The feeders we knew going in we HAD to have. We, ProtoVoltaics, being the first customer of this machine needed something to make our current boards, and knew cut tape wasn't going to work for us. We designed and manufactured to feeders before even starting on the PnP itself.

While I understand and appreciate the commercial interest, I think it would be fair to give back to the maker community and publish at least some helpful aspects of the implementation for those who want to make their own machine. These people would probably not buy a ready made one anyways... Just my two cents.

We do have a commercial interest, but we are a business, at the end of the day people depend on us for a pay check, and we are hoping this machine can help us keep the doors open for them
We do want to give back though, so we are going to open source our feeders so people with other machines such as the Liteplacer can use them!

As a potential customer, not having the software sources would be a deal killer for me and I am pretty sure I am not the only one. What if I needed a certain feature but cannot implement it? What if a bug prevents me from completing a job and Protovoltaics cannot fix it right away? What if Protovoltaics goes out of business or simply shifts priorities? So Protovoltaics will not see me back their campaign - but I still wish them success as I think the more players in the semi-pro P&P field, the better

Thanks again for taking the time to check our our project! We understand your position on our closed source software. Our reasoning for this is we want to have ownership of the software. It really becomes a customer support nightmare when someone changes your code, and then asks you why it's not working, or when someone basically rewrites your code, and then you have to try to learn their version before you can continue making updates. With us being the owners of the code it will make support much easier for us. Also, since it's a machine we use at our company you can be sure it will continue to be updated regularly. We have a dedicated team to make this machine as great as we know it can be. Now that we have it working we are always looking and finding ways to make it faster! We are very confident we will have it about 1000 CPH in the next two weeks.

If there are any features that are not currently present in our machine that you would need please let us know. All feed back is greatly appreciated!

Anyways, I think just seeing a complete machine in action gives us plenty of inspiration and there are certainly some aspects here that LitePlacer can learn from - that is why I am sharing my thoughts here.

The Liteplacer is a wonderful machine I have been following it religiously for almost 5 months now. I am super excited to see it at the World maker faire in New York! Hope to be able to talk to Juha while we are there maybe share some ideas!

[mawa]

I think that the >400% price difference between the Liteplacer and the protovoltaics will let other kind of customers buy such a machine.

I bought the Liteplacer because my projects consist of low volume boards equipt with few but expensive parts that don't allow me to buy a 1000 pieces tape except for lets say the omnipresent 100n capacitors.
The geometry of one of my boards is non rectagular and I could not find an afordable service company for my low volumes (<100 per year)

Therfore Juhas solution is a better aproach for me as I am designing my machine to use easy to exchange panels with cut tapes and a couple of trays.
I had to redesign a great deal of the software therefore open source is IMHO a must for makers like me.

Also the use of "commonly" available mechanical and electronic components.

I wish protovoltaics all the best luck but looking at the life cycle of some kickstarters I would like my humble investment insured in some way and that is open source to the software and the hardware.

And as mrandt said I also believe that protovoltaics has taken some profit form our open exchange of ideas and maybe from the algorithms of Juhas and rezas sofware and the open source libraries we use.

Therefore it would be a nice gesture if protovoltaics would give some more detail information on the nozzle changers and their camera approach. Espeacially for the latter a discussion how they want to achieve the 0.05mm accuracy with cameras that look at 45 degrees angle. In the video at least I could not see that the whole pick up and placing cycle is really faster than Juhas "look at the hole and pick the next part" solution.

[protovoltaics]

Therefore it would be a nice gesture if protovoltaics would give some more detail information on the nozzle changers and their camera approach. Espeacially for the latter a discussion how they want to achieve the 0.05mm accuracy with cameras that look at 45 degrees angle. In the video at least I could not see that the whole pick up and placing cycle is really faster than Juhas "look at the hole and pick the next part" solution.

I assure you our code is purely our own. We only look to forums for inspiration on new features. I am not opposed to giving details on how the machine works where I am allowed. We do have two cameras mounted to the head, but we also have an upward looking camera. The upward looking camera is what allows us to place low pitch parts.

I will abstain from talking about our PnP from now on, as this is Liteplacer forum. If there is something I can answer on this forum for the benefit of Liteplacer I will try my best to answer. If you wish to discuss our PnP option please contact me elsewhere. It was only my intention to reply to mrandt's post.

Liteplacer will remain the best option for an open source PnP in the <2000$ range for quite some time. We love the machine, and want to see it continue to grow as a product. 2015 is the year of the low cost PnP.

[mrandt]

Protovoltaics, thanks for taking the time to post a detailed response.

I kindly disagree with your statement about open source software vs. support - in a pull-merge scenario, you *could* share and support the code base while still allowing users to fiddle and enhance their forks. Just use a strict "if you change anything, your on your own policy". Take famous 3D printing slicer Cura as an example: It is supported by Ultimaker but source is still available and users do contribute, but only official builds receive customer support.
But let's not discuss that in lentgh; in the end we all have to take our own decisions and run our businesses.

If you published the feeder design at some point I think that would be highly appreciated by the maker community. From a software point of view, an explanation of the dual camera processing would be most valuable, I guess. Blog about that and makers will love you even more

By the way: Full ACK to the "year of low cost PnP". All the great developments in this area are empowering individuals and small businesses and I believe this will foster innovation and creativity and might umtimately change the shape of the industry.

Thanks again and good luck + success with your PnP!

[WayOutWest]

In case you have not seen it yet, Protovoltaics considers their machine finished and posted a new video on YouTube / Hackaday:
https://youtu.be/Ny99KzsySHE

Wow that zooming text effect is spectacularly annoying. Most worrisome though:

KickStarter coming very soon

Ah, there you go, sure sign that the machine doesn't work as well as the video makes it appear to work They ought to rename the company "scamstarter".

- Dual camera vision eliminates the need to move camera to needle location and back - ingenious

No, definitely not ingeneous. If you're doing advanced vision (not just Hough!) you need to be looking straight at the board and component (before pickup). The off-axis view is not reliable for vision (but helpful for human debugging). Once I get my machine reassembled (see this) I will post a video that will blow your mind. No needle shade required for the upcam at all. OpenCV is really next-level stuff, and I have to credit the FirePick team for pre-tweaking it to perfection. I don't know about FirePick's mechanics, but the vision system they've got is amazing. I might even stop using the upcam because I can align the parts so well using the downcam before pickup (probably still need the upcam for needle-wobble compensation though).

- Magnetic (?) automatic nozzle changer seems to work nicely -> https://youtu.be/Ny99KzsySHE?t=1m10s

Yes, I like that a LOT. If somebody is going to make a serious attempt at a magnetic nozzle addon for the liteplacer, I would definitely buy one. Maybe even partially fund development (but not with no-strings-attached kickstarter terms!). Maybe Juha would offer to escrow funds and release them if-and-only-if the effort is successful.

- Vacuum pump seems to be modified aquarium pump (Tetra or similar) -> https://youtu.be/Ny99KzsySHE?t=4m37s

Yes, I modded one of these, I think it's the same one in fact (dual-output Tetra). According to the $adc0 pressure sensor it is less than half as powerful as the pump that comes with the Liteplacer. But they are much quieter -- inside a box like that it would probably be quieter than a CPU fan.

What if a bug prevents me from completing a job and Protovoltaics cannot fix it right away?

Bingo. That's why I bought the liteplacer. It was the first well-designed machine I found where I knew I wouldn't be told "you have to use our software and you can't modify it or write your own". That was what kept me from buying a used Madell, some of which I did find for under $2k (in questionable condition...) or that Chinese one with no vision.

[WayOutWest]

Espeacially for the latter a discussion how they want to achieve the 0.05mm accuracy with cameras that look at 45 degrees angle.

I second that.

Even if they got it to work, they are sacrificing a huge amount of vision ability with that choice. Whatever they're able to do, they could be doing a whole lot more.

[mrandt]

I don't know about FirePick's mechanics, but the vision system they've got is amazing.

I started building a FirePickDelta "Alpha" before I switched to LitePlacer for PnP. Their mechanics are really interesting, delta robots are fascinating machines

My FPD can move its head (tool holder) quite fast and positioning accuracy would be good enough for PnP - but I never finished the PnP head and camera assembly as I started to build the LitePlacer at the same time.

For my PnP use cases, the build plate of FirePick (about 30x30 cm, leaving about 18x18 cm for PCB) is just too small. PnP could work on FPD once they develop (semi) automatic feeders to attach to the sides of the machine - but that seems more complicated than expected. They are just not there yet.

However, FPD is really a multipurpose machine that could do a lot of other things from 3D printing to electronic testing to decorating cakes... I am sure I will find a use case for mine.

OpenCV is really next-level stuff, and I have to credit the FirePick team for pre-tweaking it to perfection. [...] I might even stop using the upcam because I can align the parts so well using the downcam before pickup (probably still need the upcam for needle-wobble compensation though).

Are you using FireSight wrapper or OpenCV?

I believe OpenCV is an open source lib maintained by Itseezs, a San Francisco based company employing quite significant R&D staff in Russia:
http://opencv.org/about.html

FireSight is a wrapper around OpenCV and other tools from the Fire Pick sphere - it provides a nice way to pipeline image manipulation and detection algorithms:
https://github.com/firepick1/FireSight/wiki

Interesting question is: How can LitePlacer benefit from all these great things without rewriting the whole software?

Btw: Have you looked at OpenPnP? The group of developers around Jason aim to build a "universal" PnP software and I think it would also work well with LitePlacer hardware. They also use OpenCV for vision and have support for TinyG. Want to give it a try?
http://openpnp.org/

[JuKu]

The magnetic nozzle change is indeed a great idea. Implementing that to LitePlacer should be possible: 3d print a magnetic adapter to attach to the current Luer connector and print receptacles for the needles. If we are really clever, we design these so, that the connection is self-aligning. With optical zeroing the rotation axis, maybe the needles only need calibration once?

[protovoltaics]

Ah, there you go, sure sign that the machine doesn't work as well as the video makes it appear to work They ought to rename the company "scamstarter".

I ask you please don't call us a scam without reason. KickStarters are a great way to gather market research.

If you do not like that video there are many others on our YouTube that show the machine working and placing all the parts on the board.

Also, I'm sure you just don't understand how our vision is working, but we appreciate your feedback