Neoden 4 pick and place

[rwb]

I found it for $35,000 without feeders.

http://www.pcbunlimited.com/pick-and-place.php

[rx8pilot]

To match the feeders I have on my Quad, it would cost about $91k. At that level, it needs to make a lot of PCBs flawlessly and with an easy setup routine.

I think the target audience of the N4 are those looking for the 'occasional' run, prototypes, modest volumes, etc. A full-up commercial solution eats up too much money for the small operation, but they are worth it if you are fully engulfed in production every day.

If it were possible/available, I would get an N4. Scrap the software completely and pay an additional $10k for really nice/easy/reliable software. At that point it would be a $25k solution that could cover anything my small business would need.

[rwb]

I think if a software solution could be created to fix the current issues for 10K then it probably would not be too hard to get the current and future owners to chip in to get it created.

[TheSteve]

Board 1 of 2 is done and 2 of 2 is in progress right now.

It didn't go too badly overall I guess. There is video of much of it, but it isn't mine to post.

Some things to note:

-The machine can't count and often gets a number +1 or -1. Our panel is a 10x10 matrix, it skipped the first board, no idea why. When we wanted it to skip board 60 it changed the number to 59. This isn't the first time we have seen this type of problem.

-The feeders require so much tuning with some parts it is nuts. We have SOIC-8 IC's in a reel and it likes to spit the parts everywhere. You have to get the feed box and peel box settings perfect, and that is a setting we haven't been able to find yet. We have had the same problem with our FT232RL's.

-We had a thicker 8mm part that wouldn't feed properly. Turns out there is a plastic spacer in the feeder that can be removed so thicker parts work better.

-When stopping/starting it sometimes forgets where it last picked a part from a tray and can go back to the beginning.

-If you have 0805 sized parts that fit quite well in the tape pocket they can probably be placed without the camera, if the pocket is quite large then use the camera.

We don't plan to build more of these boards so we threw together a quick mount to hold it. We are also using a tray made of cut tape to use up our stock of crystals.

[rx8pilot]

I think if a software solution could be created to fix the current issues for 10K then it probably would not be too hard to get the current and future owners to chip in to get it created.

$10k would only be viable if a number of people signed up for it. 20-25 users would be enough for a small group to be interested. 40+ users could be a genuine business.

It needs to start with a developer - not a coder. Get the concepts and logic worked out before the coding whiz starts hammering into stone.

[sam512bb]

I think if a software solution could be created to fix the current issues for 10K then it probably would not be too hard to get the current and future owners to chip in to get it created.

The problem with this is that the independently created software would only be applicable to a particular version and/or manufacturing run of the N4... What happens if Neoden changes the hardware design a bit (i.e. slight hardware I/O change, etc> as they sell more units?  Sadly, the custom software will be somewhat at the mercy of the hardware.

Cheers,

Sam

[vonnieda]

Full disclosure: I am the author of OpenPnP.

If ya'll are serious about retrofitting a Neoden 4 with new software, please check out OpenPnP. OpenPnP is a working, Open Source pick and place platform that can be used for nearly any machine.

I don't know of anyone who has done a Neoden 4 retro fit with it yet, but others have done TM-240 (older Neoden machine) retros. We also recently have a Zevatech retrofit and there are numerous examples of people DIYing their own machines with OpenPnP.

Now, OpenPnP doesn't have all the features that the N4 software currently has - the biggest two being bottom vision and conveyer support. Bottom vision is my current focus and will be the next major feature release. Conveyer support is on the list but pretty far down, since few people (in the area OpenPnP usually serves) need it.

Anyway, I thought I'd mention it because there's really no need to develop a new solution from scratch. OpenPnP has a completely modular model for machines and it can be adapted to run nearly anything. I personally think the N4 is a great candidate for a retro fit. It's an incredible hardware platform but it really needs better software to make it shine.

So, if anyone would like to talk about doing a retrofit of this machine with OpenPnP, please come check us out! We have an active mailing list and IRC channel which you can find at http://openpnp.org/. I will personally devote time and resources to helping make it happen.

[mikeselectricstuff]

Conveyer support is on the list but pretty far down, since few people (in the area OpenPnP usually serves) need it.

True, but the ability to reload a long panel manually ( with fids on each section) would be a useful facility, and probably not hard to do.
 

[vonnieda]

Conveyer support is on the list but pretty far down, since few people (in the area OpenPnP usually serves) need it.

True, but the ability to reload a long panel manually ( with fids on each section) would be a useful facility, and probably not hard to do.

Not quite sure what you are getting at. In OpenPnP you can certainly load a panel in and (re)start the job. Fids are checked on each board. We just don't have automatic conveyer support yet. It's definitely something that will get added eventually, but it's low on the priority list (for me) since it's not a very commonly requested feature.

You can see "the list" here, if you are curious: https://github.com/openpnp/openpnp/issues

Jason

[rx8pilot]

Full disclosure: I am the author of OpenPnP.

If ya'll are serious about retrofitting a Neoden 4 with new software, please check out OpenPnP. OpenPnP is a working, Open Source pick and place platform that can be used for nearly any machine.

I don't know of anyone who has done a Neoden 4 retro fit with it yet, but others have done TM-240 (older Neoden machine) retros. We also recently have a Zevatech retrofit and there are numerous examples of people DIYing their own machines with OpenPnP.

Now, OpenPnP doesn't have all the features that the N4 software currently has - the biggest two being bottom vision and conveyer support. Bottom vision is my current focus and will be the next major feature release. Conveyer support is on the list but pretty far down, since few people (in the area OpenPnP usually serves) need it.

Anyway, I thought I'd mention it because there's really no need to develop a new solution from scratch. OpenPnP has a completely modular model for machines and it can be adapted to run nearly anything. I personally think the N4 is a great candidate for a retro fit. It's an incredible hardware platform but it really needs better software to make it shine.

So, if anyone would like to talk about doing a retrofit of this machine with OpenPnP, please come check us out! We have an active mailing list and IRC channel which you can find at http://openpnp.org/. I will personally devote time and resources to helping make it happen.

as a side note.....I would love to do a retro on Quad machines. There are a lot of carcasses out there that are a really solid platform without modern controls. Swap in servos and software and it would be a sub <$10k machine with truly professional capability.

As for the N4, I could design the mechanics of the machine in my sleep and build it in the morning (slightly overstated, but not much). If a group came together and we pooled resources, my contribution would be leading the electro-mechanical development all the way to production. Then, there would be no need to worry about what Neoden will do. It would be a platform designed from the beginning to be used with Open PnP.

[mikeselectricstuff]

Conveyer support is on the list but pretty far down, since few people (in the area OpenPnP usually serves) need it.

True, but the ability to reload a long panel manually ( with fids on each section) would be a useful facility, and probably not hard to do.

Not quite sure what you are getting at.

Obviously any P&P, subject to mechanical constraints can potentially do a bigger PCB than its placement area by treating it as  two or more seperate jobs, but this would mean splitting the job into two (or more), and reloading the job files for each panel, which would be a pain.

What would be good is if you could generate a single pick/place file to cover the whole PCB, which would include additional fids for each section. The machine would use the global placement positions and fid positions  to work out which parts go in which section.
The machine would then place each section, and then prompt the user to reload the PCB for the next section. Or if there is a conveyor, send the 'move conveyor by <section length>' command.
 

[vonnieda]

Conveyer support is on the list but pretty far down, since few people (in the area OpenPnP usually serves) need it.

True, but the ability to reload a long panel manually ( with fids on each section) would be a useful facility, and probably not hard to do.

Not quite sure what you are getting at.

Obviously any P&P, subject to mechanical constraints can potentially do a bigger PCB than its placement area by treating it as  two or more seperate jobs, but this would mean splitting the job into two (or more), and reloading the job files for each panel, which would be a pain.

What would be good is if you could generate a single pick/place file to cover the whole PCB, which would include additional fids for each section. The machine would use the global placement positions and fid positions  to work out which parts go in which section.
The machine would then place each section, and then prompt the user to reload the PCB for the next section. Or if there is a conveyor, send the 'move conveyor by <section length>' command.

Ah! I see what you mean now. Yes, that is an interesting idea and a use case I have not come across before. I think we could currently facilitate that with careful job preparation but it would be entirely manual. Having that be something that software would figure out for you and prompt for movement is a really neat idea.

Jason

[rx8pilot]

Ah! I see what you mean now. Yes, that is an interesting idea and a use case I have not come across before. I think we could currently facilitate that with careful job preparation but it would be entirely manual. Having that be something that software would figure out for you and prompt for movement is a really neat idea.

Jason

The concept is certainly sound - I have done this in CNC machining applications many times. It is a great way to overcome limitations of the hardware which can grow in cost exponentially as the travel distance increases.

[mikeselectricstuff]

Conveyer support is on the list but pretty far down, since few people (in the area OpenPnP usually serves) need it.

True, but the ability to reload a long panel manually ( with fids on each section) would be a useful facility, and probably not hard to do.

Not quite sure what you are getting at.

Obviously any P&P, subject to mechanical constraints can potentially do a bigger PCB than its placement area by treating it as  two or more seperate jobs, but this would mean splitting the job into two (or more), and reloading the job files for each panel, which would be a pain.

What would be good is if you could generate a single pick/place file to cover the whole PCB, which would include additional fids for each section. The machine would use the global placement positions and fid positions  to work out which parts go in which section.
The machine would then place each section, and then prompt the user to reload the PCB for the next section. Or if there is a conveyor, send the 'move conveyor by <section length>' command.

Ah! I see what you mean now. Yes, that is an interesting idea and a use case I have not come across before. I think we could currently facilitate that with careful job preparation but it would be entirely manual. Having that be something that software would figure out for you and prompt for movement is a really neat idea.

Jason

It's probably not too big a deal if the job setup needs some extra work, you could even do it with a seperate preprocessor that takes a "big board" job and generates a job for each section,
The main thing is to not have to manually load a new job file for each section, as if you're doing several panels you're bound to do it wrong at some point.
In fact, from the point of view of minimum effort to implement, if you were to (at least initially) take the approach of preprocessing or manually splitting into multiple jobs,  maybe all you need is the facility to automatically "chain" jobs together so they can be auto-loaded in the right sequence with user prompts.


 

[TheSteve]

The Neodon4 hardware is all controlled via a single serial port from the PC. Sounds like it is time to start sniffing.

[mikeselectricstuff]

The Neodon4 hardware is all controlled via a single serial port from the PC. Sounds like it is time to start sniffing.

What about the cameras?

[TheSteve]

The Neodon4 hardware is all controlled via a single serial port from the PC. Sounds like it is time to start sniffing.

What about the cameras?

Both cameras are USB.

[vonnieda]

The Neodon4 hardware is all controlled via a single serial port from the PC. Sounds like it is time to start sniffing.

I would personally love to see some traces from that And if someone has any info about the cameras (manufacturer, model, etc.) which should be available from the OS, that would be very helpful as well.

[elmood]

I believe the cameras on the N4 are just generic USB webcams. There are two EXE files ... one for each camera. And I see an OpenCV DLL hanging around there. If I were to guess I'd say that each one probably connects to a camera and then allows the main app to grab pictures / detection data when it wants. In my efforts to do my own PNP software I basically did the same thing except that the camera interface was part of the main app.

BTW, to the OpenPNP guy... do you offer commercial support at all? If there was a "ready to go" N4 version that supported the conveyor and both cameras I would be pretty much ready to give you money! I understand that a lot of open source projects are "for fun" but as I'm relying on my N4 for commercial purposes it would be nice to know that I could pay to have some reasonable turn-around on bugs and help with stuff... certain Neoden (who I already PAID for the software basically) aren't doing anything to give me confidence any of their bugs will be solved in this century. It's the same old empty promises that they took my comments to the engineers, blah blah.

Perhaps there is a bit of an income stream for your project from us frustrated N4 owners so far.


Andrew

[vonnieda]

BTW, to the OpenPNP guy... do you offer commercial support at all?

Perhaps there is a bit of an income stream for your project from us frustrated N4 owners so far.

Andrew

Hi Andrew,

Currently no. I might do something like this in the future, but it's a ways out as the software is still in an alpha or early beta state. That's on top of the fact there would need to be a development effort required to get the software working with the N4 beforehand. We'll need to figure out control protocols and machine specifics first.

Jason

[elmood]

Hey Jason,

Would a reverse engineered prototcol help or do you need actual access to a machine? I'd like to help out of possible with this.

[rwb]

Sweet! Things just got a lot more interesting with the Open PNP software developer joinging the discussion 

[mikeselectricstuff]

I wonder if it would be worth asking Neoden for info on protocols - probably not but can't hurt to ask...

BTW some pics of the internal hardware would be interesting to see...

[thommo]

Guys - you can count me in as a contributor for sure.

Just a thought, but perhaps we could deploy a simple 'commercial plug-in' approach to the OpenPnP solution. The benefit would be a 'reasonable' limit on the number of Kings [I'm thinking say 5 at $2k each to start with], the Servant had to answer to. Of course, this doesn't mean other requests couldn't be addressed from outside.

What it does mean is that each King could have access to the source code for their own custom requirements [and then could contribute it to the group if they wished].

The group could offer the NEW solution to Early Adopters  of N4 at a discounted rate - say $500-1000 per license, and everyone would benefit. This way the commercial development could continue to be refined.

Not suggesting cutting out the Open PnP either - the whole thing could be handed over to Open PnP, if agreed, once the initial costs were recovered.

We could even sell it back to Neoden if it came to that - or both.

Either way - pls count me IN as a foundation member if it gets up, and I don't even own a machine yet.

[thommo]

TheSteve

Hey, sorry your last build, and the post for it, got swallowed up in 3rd party Development discussion.

It seems that something along these lines could go a long way toward making your current [or company's] aquisitation a lot more valuable and practical though.

I've got some questions for you, but will need to come back a bit later.
Just wanted to say thanks for the post!

Board 1 of 2 is done and 2 of 2 is in progress right now.