Author Topic: Neoden 4 pick and place (Read 597387 times)

rwb · « **Reply #775 on:** March 21, 2016, 04:51:54 am »

After looking at what is for sale on those 2 website it's pretty clear how revolutionary the NeoDen 4 is and can be if we can all figure out exactly how to get the machine up and running without needing a week to set it up.

The machines being offered for sale are crazy old large dinosaurs. All the equipment and parts look like they come from the 60's

I can see how people say they are build like tanks but it looks like you would need to rebuild the electronics to get them up and running on modern software.

These smaller PNP machines are a step in the right direction for sure. Now they just need to work on the software quality.

mrpackethead · « **Reply #776 on:** March 21, 2016, 04:53:35 am »

Quote from: rwb on March 21, 2016, 04:51:54 am

After looking at what is for sale on those 2 website it's pretty clear how revolutionary the NeoDen 4 is and can be if we can all figure out exactly how to get the machine up and running without needing a week to set it up.

The machines being offered for sale are crazy old large dinosaurs. All the equipment and parts look like they come from the 60's I can see how people say they are build like tanks but it looks like you would need to rebuild the electronics to get them up and running on modern software.

These smaller PNP machines are a step in the right direction for sure. Now they just need to work on the software quality.

there might be a few bargins from time to time, but this isnt an option really for the market that is emerging..

thommo · « **Reply #777 on:** March 21, 2016, 05:28:24 am »

Hi Ichan,

I'm curious to know how you determine the measurements you've made.

You said in a previous post that the machine/board measurements are out by 0.2mm per 100.0mm. Where are you measuring these points from and by what means?

On a different note, I'm certain you're buoyed by MTN's success. I'm sure it'll get there in the end. That said, it is still very odd that you are the only one who seems to see the 'step' increments not relating to real-world measurements. Could it be the SW version, or just a mistake

Quote from: Ichan on March 21, 2016, 03:36:30 am

Quote from: TankSparks on March 20, 2016, 09:37:16 pm
I pointed out your problem, reread my post #738,

TankSpark, i already test it as on reply #742 - tried to set the X to be the same for left-botom and left-top also same Y for left-top and right-top, the problem is still.

I am pretty optimistic today, seems using fiducials in panelized mode and scaling my data to fix the dimensional difference will make my day, hopefully .

I am now setting it up to use all the components.

Quote from: MTNELECTRONICS on March 21, 2016, 02:48:26 am
A bit of an update: I have now produced dozens of panels (thousands of boards) with the machine, and I can say that it has been working great for me now that I understand the "ins and outs" of the machine. I have been placing quite a few 0603 and even some 0402 and the placement accuracy is excellent---good enough that I don't have to worry about it as long as it is set up correctly. It has taken some experimentation and experience to get the feed, pick, and place settings to where things are reliable enough to let the machine run with hardly any intervention. You really have to spend some time and tune in each part to get it to the point of excellent reliability; some parts work well with the default settings, but many require some significant adjustment.

I have been using 3 fiducials on the panels ("single" panel fiducial in the software), and I don't get any placement offset, but maybe I am setting things up in a different order than you are. I always use "manual" mode, even though I import the coordinates (offset in excel) and then mark the fiducials manually in the machine the first time without moving the board. I have not had as good of results using fully imported coordinates of the panel + fiducials. It seems like for some reason the angle/offset of the board does not get updated correctly if the board is not perfectly straight. Backwards, I know. I have also done some panelized boards where each panel has its own fiducials, and they are also accurate.

I have had one peel box burn up and quit working. I replaced it with a spare and went on my way, but I haven't contacted them yet about it.

Begging to see it in action please...

-ichan

ServoKit · « **Reply #778 on:** March 21, 2016, 07:45:50 am »

Out of curiosity: Are the PCBs fixed or clamped down while on the conveyor? Or are they only hold by friction and can move more or less freely (longitudinal and/or transversal)?

Regards, Axel

Ichan · « **Reply #779 on:** March 21, 2016, 07:49:27 pm »

Not much time available today, only able to try single board with full components. The result is acceptable but there are some anoying pick failed issue, still need some adjustment. I make some notes of the feeder issue:

- the feeders had some play, the pocket hole is not always on the same position - need several trial to get the right pick position, using small nozzle will help.
- thick components in 8mm feeder is hard to feed, need to remove (or replace) the white plastic base (mentioned before by TheSteve).
- the feeder box and peel box have to be perfectly inline, if not the peeled cover will tend to shift and later failed to peel.
- the rear feeder can not be aligned using camera, have to use the nozzle and your eyes
- pick height default is 0.5 mm, let it be like that just change the placement height as required
- seems it is not possible to use more than one nozzle to pick parts from the same feeder (correct me)
- seems the feeder spit too many components on pick fail retry (no setting for that)

Quote from: thommo on March 21, 2016, 05:28:24 am

I'm curious to know how you determine the measurements you've made.
You said in a previous post that the machine/board measurements are out by 0.2mm per 100.0mm. Where are you measuring these points from and by what means?

I write coordinates of some points on the pcb from the CAD then put the pcb on the machine, navigate and write the corresponding points coordinates of the machine, then calculate it like below.

This way will only find the difference, not telling which one is true. I will try this again on my older machine later, should be interesting as that machine use 2um resolution linear encoder feedback (Renishaw RGH41G).

Quote from: ServoKit on March 21, 2016, 07:45:50 am

Out of curiosity: Are the PCBs fixed or clamped down while on the conveyor? Or are they only hold by friction and can move more or less freely (longitudinal and/or transversal)?

The pcb pressed by some spring through the belt on the conveyor (you can see it on "machine stup" video), held firm enough on position.

-ichan

rwb · « **Reply #780 on:** March 22, 2016, 12:06:36 am »

I saw this on Adafruits blog today and figured I would share here.

https://blog.adafruit.com/2016/03/21/updated-visionbot-pick-and-place-shipping-in-europe-manufacturingmonday-makermovement/

If only the Neoden 4 had this software with it.

https://youtu.be/ewQUH6qBHEM

I asked the guy to come join the party over here.

rx8pilot · « **Reply #781 on:** March 22, 2016, 12:35:26 am »

96 feeders (unspecifiied) for 6500. Decent start if speed is not your goal (like me)

The software looks reasonable, the hardware strikes me as rather rotten but seems to work. I will buy slow/reliable over fast/unreliable any day.

rwb · « **Reply #782 on:** March 22, 2016, 12:39:00 am »

Yea I don't need the super fast speed either.

Another video of it doing panels. The software looks to be much better.

Also you can see the machine speed up when picking parts and slow down when placing parts which is what @mikeselectricstuff likes to see

Also he says they are now using Juki nozzles and it has a built in nozzle changer.

Hopefully this guy shows up over here to talk about all this.

https://youtu.be/52fK5Ej-88I

rx8pilot · « **Reply #783 on:** March 22, 2016, 12:43:36 am »

Speed ramps are important. I wonder if they will sell the software to a hardware developer. It has the look of a toy.

rwb · « **Reply #784 on:** March 22, 2016, 01:03:26 am »

This is what it looks like to program your parts. Lots of different options, almost too many

https://youtu.be/anw01OADcbM

thommo · « **Reply #785 on:** March 22, 2016, 01:04:20 am »

Guys, we're all obviously worn out with the N4 travel path but, for mine, this machine and the N4 are galaxies apart. Sure it has more feeders, but others like SmallSMT offer this also with a MUCH BETTER HW build.

Good luck if you want to replace a reel in the middle of the stack once it's empty on this machine - it has 1 common axle.

The N4 [even though I haven't ordered one yet] now has an English speaking community who have it running successfully, including MTNElectoronics who says he's loaded thousands of boards now, and TankSparks and TheSteve too.

Hopefully soon, Ichan will join them.

I don't think that reviews of the N4 where normal, if not optium, conditions for producing machine loaded PnP SMT components are not followed - like having good, readable Marks/Fiducials, are true reflections of the machine or how it was intended to be used.

For mine, the rail and conveyor set this machine apart.

If we could just get one of the current, successful users to create a start-to-end video including the GUI and actual machine video, I think we'd all be converts.

Accuracy vs Speed
I don't need speed either, but the N4 is capable of it.
Accuracy is something it is also capable of according to TankSparks, TheSteve and MTNElectronics.

It's still N4 for mine.

The ONLY potential contender for me is the CharmHigh with changeable Yamaha pneumatic feeders [but only if they increased the max qty from 30 to something like double that number].

rwb · « **Reply #786 on:** March 22, 2016, 01:10:35 am »

Yea that new Charmhigh machine with the Yamaha pneumatic feeders would probably solve the feeder issues that people are talking about having with the Neoden 4 and Servokit's smaller machine. Seems like feeders are a problem with both machines eventually. That's probably why Charmhigh when the direction they did.

mrpackethead · « **Reply #787 on:** March 22, 2016, 01:20:27 am »

Quote from: thommo on March 22, 2016, 01:04:20 am

Guys, we're all obviously worn out with the N4 travel path but, for mine, this machine and the N4 are galaxies apart. Sure it has more feeders, but others like SmallSMT offer this also with a MUCH BETTER HW build.

This machine is not much more than a hobby.. There's a whole lot to be desired.. The Charm, Yunghsung and SMT machiens all are way better than this for a whole bunch less cash.

rwb · « **Reply #788 on:** March 22, 2016, 01:31:10 am »

It sure is interesting to see all these people try to solve the problem of a high quality PNP machine for under 10K.

Nobody has hit the nail on the head yet.

The NeoDen 4 comes close but it sounds like the feeder design needs updating and the software needs repaired. The hardware looks amazing though.

Based on AR_systems feedback on ServoKit's machine it sounds like its not really a reliable solution either.

I was hoping that Ichan would have his 2000 boards almost done by now. Sounds like he is almost ready to get started though.

ar__systems · « **Reply #789 on:** March 22, 2016, 02:20:35 am »

I'm not sure I like VisualBots software. The screen is cluttered, fruit salad of millions of controls. Needs cleaning up.

ServoKit · « **Reply #790 on:** March 22, 2016, 04:57:52 am »

@rwb, thommo: I don't think that the Charmhigh is an alternative right now unless you want to be a guinea pig. As I see it, the TVMs and the N4 represent the two ends of affordable vision based P&P, price wise and feature wise. Both are actively used, for both communities are beginning to form. Now all we have to do is getting the manufacturers listening to us. Obviously, both machines have their issues but at least you know what they are and can plan accordingly.

Regards, Axel

mrpackethead · « **Reply #791 on:** March 22, 2016, 05:16:49 am »

I also see the openpnp system becoming a force to be reckoned with, its come a long way in a short time frame, and you can get a slow but reliable vision based system for $1500 + effort.

I've been investigating this as an option.. there are plenty of "parts" shops in china will sell you just about everything you need. Pickup heads and feeders.. Some good plans are coming together nicely now..

thommo · « **Reply #792 on:** March 22, 2016, 05:29:19 am »

Likewise, I would love to see OpenPnP succeed, but I also have commercial requirements that can't be fulfilled with love alone.

Some weeks ago we loosely discussed options to 'pool' some funds and produce a 'deliverable' piece of equipment. Regrettably, OpenPnp wasn't able to jump in and get the work done that would be required, and no one has time to go fishing for the contractors. It seems the only way this would work however is if it was coordinated by OpenPnP.

I agree it's come a long way in recent times [just look at the latest feature to include machine vision] and it certainly has potential to be a great option. But the real question is will NeoDen progress faster than even the very best efforts of an OpenPnP programme. Unfortunately, I think not. NeoDen, although we don't agree with how the prioritize things, do have a commercial manifest driving them, and they do have successful history in delivering hardware to the market.

Trust me when I say 'hang on for about 4-6 weeks' before you make any decisions in this space. I'm fairly certain it'll be worth the wait.

Quote from: mrpackethead on March 22, 2016, 05:16:49 am

I also see the openpnp system becoming a force to be reckoned with, its come a long way in a short time frame, and you can get a slow but reliable vision based system for $1500 + effort.

I've been investigating this as an option.. there are plenty of "parts" shops in china will sell you just about everything you need. Pickup heads and feeders.. Some good plans are coming together nicely now..

vonnieda · « **Reply #793 on:** March 22, 2016, 06:17:39 am »

Quote from: thommo on March 22, 2016, 05:29:19 am

Some weeks ago we loosely discussed options to 'pool' some funds and produce a 'deliverable' piece of equipment. Regrettably, OpenPnp wasn't able to jump in and get the work done that would be required, and no one has time to go fishing for the contractors. It seems the only way this would work however is if it was coordinated by OpenPnP.

I'm still game for this. I've been hoping someone would post a little bit of information about the workings of the machine so I can get a better feel for the level of effort in reverse engineering it. Some things that would help a lot:

* Screen captures from Windows device manager showing any connections the machine has to the computer.
* Manufacturer and model of the cameras, and their connection methods. Are they UVC? Do they require a special driver?
* If the machine connects via USB to Serial, a serial dump would be magnificent. There's http://www.eltima.com/products/serial-port-monitor/ which makes it pretty easy. Install that, run a job, capture the output.
* Detailed pictures of the main control board. Not the PC board - that doesn't really matter, but the board that interfaces with all the motors.
* Pictures of any other electronics involved in the thing.
* Captures of the CAN data on the feeder bus. I don't have any experience with CAN, so I don't know how complex this would be, but I suspect something like a Bus Pirate would do the job.
* Information about how the peel boxes connect to the machine and what is controlling them. I know the feeders are CAN, but the peel boxes are analog. What controls them?

If I can get some of that information I can make an informed decision on whether this is something I can get done in a reasonable amount of time and then we can move forward, but right now I am in the dark.

Further along on that topic, do I remember correctly that someone in Vancouver, BC has one of these? I'm in Seattle, so if someone in Vancouver wants to let me drive up for the weekend and poke through their machine I'd be up for that. Even better if someone in Seattle has one

Jason

ar__systems · « **Reply #794 on:** March 22, 2016, 06:29:03 am »

I hope you don't think you can decipher a comm protocol just by looking at logs of a random job, without knowing what the machine does at the time?

vonnieda · « **Reply #795 on:** March 22, 2016, 06:33:21 am »

Quote from: ar__systems on March 22, 2016, 06:29:03 am

I hope you don't think you can decipher a comm protocol just by looking at logs of a random job, without knowing what the machine does at the time?

Depends very much on the protocol. I think the likelihood is very high that it's just using Gcode with possibly some extensions for running the feeders, lighting, etc. If it's Gcode then that's enough for me to say "Yea, I can probably do this." If it's something else there's a pretty reasonable chance I can figure it out anyway. Reverse engineering weird protocols and file formats is something I've got a lot of experience with.

Jason

TankSparks · « **Reply #796 on:** March 22, 2016, 07:32:05 am »

Quote from: ar__systems on March 22, 2016, 06:29:03 am

I hope you don't think you can decipher a comm protocol just by looking at logs of a random job, without knowing what the machine does at the time?

Fortunately the Neoden has a GUI screen with all the test command buttons. Commands for feeding, suction, nozzles up/down rotate etc. So it would not be that hard to decipher the protocol.

thommo · « **Reply #797 on:** March 22, 2016, 07:55:32 am »

Hi Jason,

Great to see the activity level rise and the discussion happening again.
What I actually meant [and I understand MrPackethead meant also] was creating a machine from 'scratch'. This is no mean task, but certainly has potential if NeoDen ever went off the rails.

What is more exciting is the prospect of buying N4s, and using your OpenPnP code to drive it.

This means those of us sitting on the fence still have some options and would likely move more quickly:
put up with N4 GUI and software in the interim,
look forward to enabling custom commands via OpenPnP application when, and as, they get released, developed, or requested - or write ourselves.

Heck, it's even possible to toggle the option at 'start up' and decide who's Software you wish to run.

Let's know if I can help in any way. FWIW - is it an option to change out the controllers in the N4 if you already have OpenPnP code that works with them? I know it's an additional cost, but it may also be the quickest and, in the end, cheapest way forward. [I say this blissfully unaware of the relative cost of the controllers you have designed for in the OpenPnP project].

Quote from: vonnieda on March 22, 2016, 06:17:39 am

Some weeks ago we loosely discussed options to 'pool' some funds and produce a 'deliverable' piece of equipment. Regrettably, OpenPnp wasn't able to jump in and get the work done that would be required, and no one has time to go fishing for the contractors. It seems the only way this would work however is if it was coordinated by OpenPnP.

I'm still game for this. I've been hoping someone would post a little bit of information about the workings of the machine so I can get a better feel for the level of effort in reverse engineering it. Some things that would help a lot:

[/quote]

mrpackethead · « **Reply #798 on:** March 22, 2016, 08:07:47 am »

Quote from: thommo on March 22, 2016, 05:29:19 am

Likewise, I would love to see OpenPnP succeed, but I also have commercial requirements that can't be fulfilled with love alone.

From what i've seen, i'd trust openpNP and some custom built hardware before i trusted this N4 stuff. 40 feeders just dont' cut it for my Low Volume High Mix box. The only other thing i've seen that comes close was the Small SMT, but i'm uncomfortable about that, and how it might be supported.

I see that an open hardware / open software solution even if its a llittle less stable to start with is a much better long term solution..
everyone has to do what they think is the right thing for them, and nobody should be expecting that everyone should follow their choosen path.

Quote

I agree it's come a long way in recent times [just look at the latest feature to include machine vision] and it certainly has potential to be a great option. But the real question is will NeoDen progress faster than even the very best efforts of an OpenPnP programme. Unfortunately, I think not

I think it will.

Quote

Trust me when I say 'hang on for about 4-6 weeks' before you make any decisions in this space. I'm fairly certain it'll be worth the wait.

Anther mahine released before its ready?

Quote from: mrpackethead on March 22, 2016, 05:16:49 am

I also see the openpnp system becoming a force to be reckoned with, its come a long way in a short time frame, and you can get a slow but reliable vision based system for $1500 + effort.

I've been investigating this as an option.. there are plenty of "parts" shops in china will sell you just about everything you need. Pickup heads and feeders.. Some good plans are coming together nicely now..

[/quote]

mrpackethead · « **Reply #799 on:** March 22, 2016, 08:13:53 am »

Quote from: thommo on March 22, 2016, 07:55:32 am

Hi Jason,

Great to see the activity level rise and the discussion happening again.
What I actually meant [and I understand MrPackethead meant also] was creating a machine from 'scratch'. This is no mean task, but certainly has potential if NeoDen ever went off the rails.

Theres a couple of documented builds and boms you can follow now and get your self into a working machine for under $2000. Yeah it will be simple, ( cut strips and parts in trays.


EEVblog Main Site	EEVblog on Youtube	EEVblog on Twitter	EEVblog on Facebook	EEVblog on Odysee

Author Topic: Neoden 4 pick and place (Read 597387 times)

Share me