Neoden 4 pick and place

[Ichan]

Maybe your Neoden 4 was crated differently then ours but you only need to remove the screws from the bottom edges and the entire top of the crate simply lifts off and is even reusable. At that point the machine simply lists off its shipping mounts. See my pics in the first page of this thread.

Yeah, we realize that after the top had been opened 

Here some video taken yesterday, when we took pictures of its internal and setting up the hardware.





-ichan

[Ichan]

Sorry can't play with the machine today, have to do something else more important related to the project. Only able to take some pictures of the head before i leave the workshop.









The camera can not be seen, it is inside the H shaped block on the front - i do not want to open it for now, afraid to loose the offset setting.

-ichan

[Ichan]

Received a package from Neoden yesterday. We installed the new "blow" pump.
...

@TheSteve, which one is the "blow pump"?

-ichan

[H.O]

Thank you for the photos Ichan!
That looks like a nice compact design of the head though I'm sure some might say the cable management could have been better.
I can only spot two Z-axis motors. Are the other two just not visible or are they using just two motors for the four nozzles? Looking at the videos of the machine in operation it does appear as if each nozzle can move on its own so I'm guessing there's another pair of motors crammed in there somewhere.

They clearly have the drivers for the head mounted motors on the head itself. I wonder of they are pushing step/direction signals for each driver thru the control cable (probably not as there doesn't seem to be enough wires for that in the cable) or if there's some brains in the head as well, perhaps handling the vacum sensing pickup retry etc all on its own.

[vonnieda]

I can only spot two Z-axis motors. Are the other two just not visible or are they using just two motors for the four nozzles? Looking at the videos of the machine in operation it does appear as if each nozzle can move on its own so I'm guessing there's another pair of motors crammed in there somewhere.

There are only two Z axis motors. Each motor drives two Z axes using a cam to push the axis down and a spring return. This is the same setup they used on their previous machines (TM-2xx series) and on the N4 they've just doubled it up with two dual heads facing each other. You can see my version of the same design here:

https://raw.githubusercontent.com/openpnp/openpnp-openbuilds/develop/Images/head_closeup.jpg

Jason

[H.O]

Ah, that's how they're doing it.
I don't know if it's über clever or not so good but depending on the length of the arm pushing down on the motors and the desired Z-axis travel the rotational motion of the arm translates into a sinusoidal move of the Z-axis which makes the axis moves fast at the top and the automatically slow down as it gets closer to the PCB.

This obviously means that the resolution changes so the first 3mm of movement doesn't require as many steps as the last 3mm - the software would need to know this, of course.

Then again, if the lenght of the arm is much longer than the travel length of the Z-axis it'll be fairly linear over the full travel.

I'm redoing the head on my machine, getting rid of a ballscrew driven THK slide in favour of a belt driven design like on the SmallSMT Machines where the two heads are connected to either side of a belt driven by a single motor.

Sorry for drifting a bit OT.

[thommo]

H.O
Would be great to see some close up images of the head on your machine. Have you designed it yourself?
I know it's not a N4, but the topic right now is better understanding the design of the N4 head - and a simple photo comparison with other designs would certainly assist me, and hopefully others too.
Thnx

BTW - I assume you guys already know about this site
They sell complete PnP head assemblies with NEMA hollow shaft motors, etc
http://www.robotdigg.com/category/27/SMT,+AI+n+Laser

[rx8pilot]

The insides are a sloppy mess. No labels on the harnesses. It looks like many of the prototypes I have made over the years.

[mrpackethead]

BTW - I assume you guys already know about this site
They sell complete PnP head assemblies with NEMA hollow shaft motors, etc
http://www.robotdigg.com/category/27/SMT,+AI+n+Laser

And most if not all if those parts can also be found on aliexpress and taobao for about 25% of the price quoted on robo dig.    This was being discussed at #openpnp on free node irc the other day

http://world.taobao.com/item/524261348567.htm#detail  <--- for example.

[Towger]

The insides are a sloppy mess. No labels on the harnesses. It looks like many of the prototypes I have made over the years.

What strikes me is how hard it is to get at the gubbins. I saw the photo of the wiring etc on my phone and initially though they were inside the base pedestal and just a matter of opening a side panel/door, until I watched the video.  I would not want to trouble shoot it, take all the feeders etc off, unbolt from the base and then a two man job to take the unit off and turn it over on its back.

[thommo]

Guys, I seriously don't think manufacturers like NeoDen, Charmhigh, etc are designing their product on the basis of how much tinkering you can do with them.

It is my best guess, and sincere hope, that they are just designed to work.

As this is the N4 forum, I believe it is true to say that almost the entire thread history does not point to build quality, or issues about the HW failing to function.

It appears that the 'almost whole' amount of issues relate to software, and maybe some firmware, and then issues of communication.

I, for one, must admit that if i were to purchase a N4, the last thing I would be expecting to do is pull it apart and tinker. I just want it to load my components, and replace the machine I have now (me).

[TheSteve]

You need can leave all of the feeders on the machine to access the bottom. Unbolting from the stand is only 4 bolts, so it is pretty quick. It does take two people to lay the machine back. It isn't something I'd want to do too often.

[TheSteve]

Received a package from Neoden yesterday. We installed the new "blow" pump.
...

@TheSteve, which one is the "blow pump"?

-ichan

I have attached a copy of your pic with the blow pump circled. In my pic it was removed as we had to replace it. They ship 220 VAC models on all machines so those running the machine on 120 VAC have a lower blow pressure. The pump is a standard aquarium style diaphragm pump.

[rx8pilot]

So, in time, when the software is dialed a bit a users start doing 8hr days - we will see if it holds up.

Sent from my horrible mobile....

[thommo]

Hi Cathy,

Please find below the 'list' Elmood referred to in response to your post last week.

As he pointed out, only the first 2 items are mentioned in YOUR list - it seems none of the others are noted anywhere.

Given Neoden's forecast of approx just 1 item/task per month, it becomes a serious concern that it may take a very long time to address these issues. Note that this list is also NOT comprehensive but, without reasonable feedback and response times, its value is open to question.

Cathy, can you please provide us with the confidence we're all seeking; that Neoden is serious about delivering a solid, bug-free product and responds in a timely manner to its Users' concerns?

Thanks - Peter

1.   Vibration feeders can't be started in a job
If you forget to turn on the vibration motor in the manual screen the job will run and then fail to load tube parts. It should be a job option: "use vibration feeder"

2.   Pick test turns pump on and can't turn it off
It should either put the part back in the feeder or throw it away in the trash box properly. The "Pick test" button should be a toggle button instead of a latching type.

3.   Impossible to force a ‘single part only’ for machine vision and placement
If there are multiple parts in the chip list that use different nozzles the machine will load each nozzle before imaging / placing. In the case of large parts it might be desirable to only pick up one part and then place it to minimize the chance of it falling off or imaging incorrectly. This is not possible unless the chiplist is organized carefully to prevent multiple nozzles from being used sequentially. There should be a "block" command to insert into the chiplist to place all parts currently on the head and then start over with empty nozzles.

4.   Imaging does not work correctly using Large Component vision setting
There is no documentation as to how big a "Large Component" is. (and Neoden didn't tell me either) It takes multiple images and stitches them together somehow, but if there are other nozzles in use (see above) it gets the incorrect centre / rotation for the part.

5.   Parts outside placement area are not detected until part-way through a job
If you start a job and a chip must be placed outside the mechanical limits of the machine, this is not flagged until the part is placed. It should be simple to scan the chiplist at start of a job and warn of parts that might be outside the area. This is mostly a job setup problem where the board was not advanced into the machine enough before the PCB position was set.

6.   Recommended nozzle sizes from Neoden seem wrong
For QFP and larger ICs the larger nozzles have trouble getting enough suction. Better results were found with the next smaller nozzle size than what is recommended. The small (XN07) nozzles work well for 0805, 0603, SOT23, etc. tiny parts. There is an even smaller nozzle option (XN03) that is not included in the standard package.

7. Feeder ID settings are frustrating
It appears (not confirmed) that there is probably an I2c bus controlling the "feed box" components of the feeders. They must be uniquely assigned... extra feeders purchased come with ID 50 so they can be reassigned... but you can only plug in one at time. It would be better if the hardware port determined the ID, since the "peel box" is addressed this way. Seems like a bad hack that should have been a bit better thought out. Not a huge deal if you plan to leave your system set up in a default configuration.

8.   Button “to curr p”
The button "to curr p" (cannot be completely read on the button text) on the file edit screen is used to recalculate all of the part locations and rotation based on the PCB mark settings.  When doing this, the rotation of each part is calculated incorrectly, it seems to keep adding additional rotation each time the "to curr p" button is pressed.

Dear all,
Good day!
I'm Cathy, International Service Support Coordinator of NeoDen Tech., I'd like to thank you all for your kind attention and sincere suggestion regarding our NeoDen 4 Pick and Place machine.

We understand all the concerns of the machine users and our Team is continuously on the job, at this point, we would like to inform all our esteemed and prospective customers that , we are hereby putting a data related to our software development stages from the date of release of the first batch of machines from our factory ,Hope from the attached snapshot, one can understand that the way which we are addressing the issues.
We are committed to update each and every machine on the field with the latest update free of cost. We request our existing users of NeoDen 4 machines to be in touch with us for the new release update to get updated on your machine.

[Ichan]

I have attached a copy of your pic with the blow pump circled. In my pic it was removed as we had to replace it. They ship 220 VAC models on all machines so those running the machine on 120 VAC have a lower blow pressure. The pump is a standard aquarium style diaphragm pump.

Thanks, i missed that one. If it "blow up" on long run then i just have to go to the pet store then .

I can tolerate the rather messy wiring, seems the connections are all ok - just not tidy enough.

To this far without powering it on yet, i can say that i am happy with the machine - it is a pretty well built thing for the price.

Sunday noon in here now , i am thinking to ask a permit from the kids to go to the workshop... 


-ichan

[elmood]

I am happy to report that we built 5 boards successfully with the Neoden 4 the other day. They still need some PTH assembly so I'm waiting to see how the entire PCBs turn out before running a lot more through. We spent some weeks getting our oven worked out... don't buy the Puhui oven, it's terrible! I found an Essemtec RO06 Plus on eBay for a decent price from a local seller... this felt like overkill but I'm now convinced that it's the minimum necessary for repeatable results especially if you're doing lead-free. We now have a proper vent installed in the roof of our building... not a smell is detected during the entire process now! (as opposed to many days of horrible burning smells lingering after using the Puhui... it's still waiting for some kind of sacrifice... too bad bondfires are not permitted within city limits.

The issues we had this time which are not huge but still requires some careful tweaking:

- Mark points are hard to use... I need to try tweaking more because some of my older boards don't have proper fiducials. I'm using vias which work well except that sometimes it appears to find the correct hole and then jumps to an adjacent hole on the last correction move. (it takes about 3 pictures per hole and adjusts the head each time) One time I didn't notice the wrong point got detected and the machine started placing resistors on the rail instead of the board.  I wish there was a way to see the area that is being considered because the adjacent hole isn't even visible on the screen so I'm not sure how it decides to jump over there all of a sudden since the correct hole was already photographed twice.

- Beware of the height of taller ICs. I think I need to add another entry in the footprint list. I have a ULN2803A which is pretty tall and I think the nozzle presses down too hard and causes the chip to slide sideways sometimes. The spring-loaded nozzles are great and I never felt like it's going to break anything when the settings are wrong.

- Don't shine light inside the machine! I have a powerful LED flashlight that I was using to check the placement since it's dark inside. I accidentally pointed it at the upfacing camera while it was imaging and it put the IC down in the wrong place.

- Once a transistor got picked up by its side. There seems to be no way for the machine to detect this.

I've been doing ICs at 20% speed and this is much more gentle and avoids the "train wreck" scenario when it doesn't think it has suction and attempts to get a new part while there is already one on the nozzle. This happened on the tube feeder and was quite hilarious! The only trouble with 20% mode is that the acceleration is so slow that the motors clearly pass through some very resonant mode which makes a disturbing rattling sound... these kinds of undamped vibrations could cause stepper errors, but so far I didn't actually see any problem with that. So I guess everything comes with a trade-off... the resonance is not apparent at higher speeds.

At full speed the machine shakes quite a lot, so much so that I'm sure the 20" LCD sitting on top is going to fall onto the floor one of these days. I think the base is the problem and could use a bit more mass / bracing. Flat sheet metal panels are generally not very stiff and I believe that the base experiences quite a lot of shear and torsion forces as the head moves around up top.

[TJ232]

About the base, just a old tip used before with high torsion/momentum equipments and wobbling stands: pour concrete inside the base, even only 40-50% up from base will make the difference. 80-100% will make it stiff like a rock.

[Smallsmt]

@elmood

Put a foil bag inside the stand and fill up with sand.

The mass will be increased and shaking avoid.

[uncle_bob]

..... 80-100% will make it stiff like a rock.

Hi

..... first make sure the stand is *exactly* where you want it to be

Second, check the floor loading ratings. A single floor factory on a concrete slab better for this sort of thing than a 10th floor office.

Third ... think about the contact area between the stand and the floor. Having a cubic meter of concrete sitting on four little legs .. errr ... If you ever need to get a fork lift under it, that big flat contact area you just went with, not quite so good.

Yes this is the voice of empirical knowledge speaking. That is 4,000 lb of concrete sitting there. No, it's not going to wiggle around any at all.

Bob

[SeanB]

Buy an old safe, and mount on the top of it. Cheap as used, and moveable with most safe movers if needed, though you can make the dollies to move it and the big pry bar that you use to get the first one under it in the first place. Gives a heavy block with a steel surface you can mount on, and the inside of the safe is a good place to store things like backup drives and such. Just open the inner cover and remove the small lock so you do not need a key to unlock it.

Won't tell you just how much the movers charge to move, even though we just moved it from one office to another, as a data safe. 300kg is not easy to move unless you have the dollies and bar, but you only need 3 people after that to get it anywhere.

[elmood]

All good suggestions but seem to all focus on "make it heavier" instead of "make it more rigid". I'm convinced that some strategically placed bracing can make an improvement without creating a "widow maker" waiting for next time we move.

Will report back.

[SeanB]

More rigid then you will find it walking across the floor on the legs. You either have to bolt it down, or add more mass so it will hold there by friction alone.

[H.O]

H.O
Would be great to see some close up images of the head on your machine. Have you designed it yourself?
I know it's not a N4, but the topic right now is better understanding the design of the N4 head - and a simple photo comparison with other designs would certainly assist me, and hopefully others too.
Thnx

I don't want to derail this thread any further so I've posted some details in another thread (my own) where you can see some photos of the current (to be replaced) "head" and what my plans are for its replacement.

EDIT: Included link to said thread....

[Ichan]

I got about two hours free time slot today when the kids and their mother go for shopping  , then i go to the workshop and play with the machine for a while. Video below is the very first powering on of the machine, i try navigating the machine (just by mouse click) and play with "manual test" section. Will need to ask Neoden for the password tomorrow as i can not access the nozzle calibration setting.

This is a boring 9 minutes video, try to follow the mouse click while watching it.



The second video is a manual programming test for one part only, it is include panel edge, panel position and angle, and mark points (use drilled holes instead of fiducials) setting. The first run attemp is failed (at 7:40), seems it can not recognize the panel if it is already in position - second attemp with the panel moved back first is success.

Next test will be the speed test, trying to get the fastest CPH this machine can do.



The workflow is somehow unusual but i got no problem so far, maybe because i already watch all the video and read many about this machine. There are some little issue like resizing the columns but again i just can tolerate it.

I feel the software is light and responsive, the image recognition is fast - good programming behind is my honest thought.

-ichan