Neoden 4 pick and place

[luiHS]

 

Yesterday I tried some new custom nozzles that I ordered, very short, to put 10mm high capacitors, and it has worked very well for me. It requires a bit of study, design the PCB properly and order the positioning of the components, all to avoid collisions between high components, even other nozzles with high components already positioned on the PCB, because the Z axis only moves 10mm.

I thought I would have problems and it would not be possible, but no, it worked for me, I'm very happy. I already have several PCB that now with these very short nozzles, I can mount 100% of the components with the Neoden4.

I just need to try to add a 0.5kg weight to the 24mm feeder peeler to see if I solve the problem of peeling the tape, and all my problems will be solved. It is hard to start this machine, that everything works well and can put all or almost all the components, but it seems that in the end the problems are solved.

The great design error of this machine, is that the Z axis only moves 10mm, if it moved 15mm, the machine would be perfect and could put all kinds of components without problems. Also some feeders are very badly designed, the peeler of the 24mm feeders has very little strength, and the 16mm feeder has a hole for the outlet of the tape too narrow that prevents using some components.

Some small bugs in the program should also be solved, and one that is quite serious, although it has a solution, is the misalignment of fine pitch chips (LQFP100, LQFP144, etc...), when the nozzle has to rotate them to position them on the PCB.

[charliex]

where'd you order the nozzles?

[luiHS]

where'd you order the nozzles?

Neoden made them for me, also the custom IC trays.

[charliex]

ahh ok, i'm going to order some off aliexpress and mod them

[girts]

ahh ok, i'm going to order some off aliexpress and mod them

You have ideas how to proceed with ceramics?

[luiHS]

 
I solved the problem with the XL4015 component on a reel, in a 24mm feeder, I put a 0.5kg weight on the tape and it peels perfectly.

I still have problems placing a micro SD card socket, placed in a 24mm feeder, with the component in a reel. This same component put on the IC Tray, the machine picks and places it perfectly.

As seen in the video, when it is on a reel, the nozzle loses the component. I suspect again of a bad design of the 24mm feeder, the hole through which the nozzle picks the component is too narrow. This SD card socket has 4 small metal tabs, 2 on each side, I think they rub against the sides of the hole and that's why the nozzle loses the component. The only solution I see is to make that feeder part, customized, with which the hole was 2 mm wider, I think the component would pass without problems and the problem would be solved. I have asked Neoden to make that modification and send me the customized piece to replace it, I do not think it costs much, they just have to file the sides to make the hole wider.

[jmelson]

Yup, if the component scrapes against the side of the pocket, it doesn't take much of a touch to break the vacuum and have it drop the part.  I've had to file the cover blade on a few of my Philips feeders so that some of the largest parts don't touch parts of the feeder on the way out of the pocket.

Jon

[luiHS]

 
It has taken me a while to convince Neoden, but finally they are going to make me some custom pieces for the 24mm feeders, with the hole 4mm wider, so I can use the micro SD sockets that I have. These are not for free, but are not expensive.

[charliex]

ahh ok, i'm going to order some off aliexpress and mod them

You have ideas how to proceed with ceramics?

grinder attachment

[Styno]

Yup, if the component scrapes against the side of the pocket, it doesn't take much of a touch to break the vacuum and have it drop the part.  I've had to file the cover blade on a few of my Philips feeders so that some of the largest parts don't touch parts of the feeder on the way out of the pocket.

Jon

When using the biggest nozzle on my machine I really have to pull the get the component off the nozzle, and does so with an audible *plop*.

Is this the difference between a light vacuum pump (Neoden) and 6 bar vacuum venturi or difference in nozzle design? Perhaps the head does not stay down long enough to build a good vacuum before lifting the component, can this be confgured in the Neoden software (with the 20 year old Dima software you can)?

[luiHS]

Perhaps the head does not stay down long enough to build a good vacuum before lifting the component, can this be confgured in the Neoden software (with the 20 year old Dima software you can)?

Neoden4 allows you to configure two delays, one after Picks the component and another after Placing the component. For sensitive components such as fine pitch chips, I set both delays to 300, and for the rest of the components to 100. This is fine for positioning components accurately, without affecting the vibrations of the motors.

I also recently discovered that the Peeler can be configured, with values for the Feed Speed, regardless of the Feeding Speed of the Feeder. My machine, installed at the factory, a 24 mm feeder, which was set at a Peeler feed speed of 48, and the feed rate of the Feeder to 16 for Crystal components. When I added a new 24 mm feeder, I only set the Feed Speed for the Feeder and the value for Peeler was by default at a low value, I think at 16, maybe that's why it did not peel the tape correctly  Although now it is corrected by adding a weight of 0.5kg, I also changed this value to 48, maybe the error was corrected by the configuration change and not by the weight, or both helped. I want to try to remove the weight, and leave only the Peeler Feeding Speed to 48, maybe that's enough.

[charliex]

yeah wrong torque setting will definitely fail to peel, its sensitive

[Kammlake]

Hello!
To quickly get to the point: I want to get rid of that Neoden4 software.

I have owned a neoden for a year or so now and it is not bad (for the money). Mechanics is ok in my opinion. However, the software is driving me nuts. Besides I want to learn more about machine control etc. and want a project to work a bit on...

I see here: https://github.com/openpnp/openpnp/wiki/Neoden-4 there is some effort done to use it with OpenPnP.

They seem however to rely on the neoden control board. I am thinking about making my own board and figuring out how the sensors, pumps, etc works instead of trying to undersand their protocol. Steppers has their stand alone drivers so no problem to use those (seems like there is some kind of feedback though...)...

Does anyone know if there is anyone that are actively working on replacing the "control system" or that has succeeded in getting rid of the software in any way?

Best regards Kammlake

[Styno]

If you really want to replace the control board, why do you want to make your own? Isn't any of the existing generic control boards suitable, e.g. TinyG, Mach3, Smoothstepper, Microstepper, Pololu, Smoothieboard etc.?

[charliex]

i pulled apart and recreated the checksum (crc16 poly 0x1021 low byte) routines, and some of the control stuff, rewrote some of the neoden UI that i found annoying (and they weren't interested in my feedback so i just changed it myself), updated the qt dlls, rewrote the serial mostly for snoop) and camera drivers ( added support for usb cameras in case i feel like changing it out), and put together windows 7 drivers for the neoden cameras so i don't have to use win xp and cam run it off my desktop. only thing i was thinking about next was updating opencv

their board goes from serial to CAN bus for some of the controls.

i dunno if its worth replacing the controller with a smoothie since its a pretty basic serial control, so usb->serial with the cameras being fed in to an app seems better to me, sure there might end up being some great unknown but it seems pretty simple overall.

i have my own pnp software that i did a long time ago, and have toyed with the idea of implementing it for the n4 but so far i've managed to get their software to do everything i've needed, mostly moving all the setup work to the eagle ulp i shared.

[Kammlake]

Hi!
Ok, sounds like there has been some work done by several people on the neoden4. Some investigation before diving in myself should pay off...

I din't really mean that I would make everything my own when I wrote that I am thinking about making a new control board. What I meant was to get open source stuff to work to drive it (and supposed some addes stuff that now is on the neoden control board would be needed). I was thinking taking the guesswork out by not relying on the neoden firmware (using open source or making a new converter and making it open source). There are probably many options to consider.

Actually I haven't really started to dig. I have the machine loaded with a job and don't want to touch it right now. When the machine is available I will have a look.

/Kammlake

[luiHS]

 
Today I received from Neoden, a custom-made piece, to replace it in the 24mm feeder. The hole in this piece is 3mm longer and it already allows me to install micro SD sockets on a reel using the 24mm feeder.

Right now, I no longer have any major problem to solve, the machine works quite well. I only need a few trays as I ordered, but I will have to wait for the Chinese to finish their New Year holidays, in the middle of February.

[tony-tid]

Hello there! I use NeoDen4 for about 2 year, and still have problem with FID marks/points. I'm using through hole for fid mark....
I will upload pictures of my settings, and the position of the nozzle head.
So...
1. I upload my CSV file genereted from Altium Designer to the NeoDen4.
2. Create two "ghost" components which I use for fid points, and write their coоrdinates to fid settings.
3. Start the mounting process -> PCB goes forward, the cam recognise the edge of the PCB and then move to FID points - but as you can see from the photos-> the head is on different position.  If I do the manual positioning - NeoDen4 do his job.

Is this because of the origin that I set in Altium Designer?
Any ideas?

EDIT:
1. I know that my current version is old ( 4.1.2 B3 )
2. I have one PCB with manual written components. ( for every component i move the head to the position and save it ) Same with FID points, and when I start mounting on it - Dont have any problems.
 
Thank you and Have a nice day!

[SMTech]

For those of you in the UK I see the Neoden 4 will be at Southern Manufacturing next week for people to gawp at and ask the sales peeps annoying questions.

[mikeselectricstuff]

For those of you in the UK I see the Neoden 4 will be at Southern Manufacturing next week for people to gawp at and ask the sales peeps annoying questions.

Who's showing it - Neoden or a distributor?

[SMTech]

AMS the distributor.

[luiHS]

Hello there! I use NeoDen4 for about 2 year, and still have problem with FID marks/points. I'm using through hole for fid mark....
I will upload pictures of my settings, and the position of the nozzle head.
So...
1. I upload my CSV file genereted from Altium Designer to the NeoDen4.
2. Create two "ghost" components which I use for fid points, and write their coоrdinates to fid settings.
3. Start the mounting process -> PCB goes forward, the cam recognise the edge of the PCB and then move to FID points - but as you can see from the photos-> the head is on different position.  If I do the manual positioning - NeoDen4 do his job.

Is this because of the origin that I set in Altium Designer?
Any ideas?

EDIT:
1. I know that my current version is old ( 4.1.2 B3 )
2. I have one PCB with manual written components. ( for every component i move the head to the position and save it ) Same with FID points, and when I start mounting on it - Dont have any problems.
 
Thank you and Have a nice day!

Why do not you place the two fiducials with standard 1mm copper round marks on your PCB? Using a THT component hole is not as accurate as using a 1mm round copper fiducial.

In your list of components, the first component is also one of the fiducials, this may not work correctly. Move, at the beginning of the list of components, a real component, for example a resistor, then align the first component again.

It is important that the two fiducials are in the design of the PCB (in your Altium project), so that their coordinates are imported correctly, with respect to the rest of the components of the board. Add them later in the Neoden software, will not recalculate the actual location of the other components.

All the coordinates, in the list of components, and the fiducials, must be real coordinates. I do not know how you import your list of components, I work with Eagle and with a ULP that calculates the real coordinates of all the components and fiducials when I input the real coordinates of the first component of the list.

[charliex]

there are a couple of things to be aware with the way the software works

1. design tool coordinate space
2. machine coordinate space

when you create the components list in the tool the XY position is relative to whatever the coordinate system is , usually lower/left 0,0 but you can put your board anywhere.
when you import into the neoden software the same XY coordinates, it is still relative to the design tools space. So it needs to be translated to the physical coordinates of the machine

the eagle script luiHS mentioned does this by taking the first component in the list and calling that the origin, then it adds on a fixed XY translation that it adds to all the components.

This translation is the difference between the origin on the design tool software, and the origin of the machine.

In the image the green rectangle is the altium work area, the green fid is where it is relative to the altium coordinate space, the blue rectangle is the physical neoden machine, and the red rectangle is the PCB you wish to place on, the red fid would be where the fidicual would be found by the camera

depending on how you generate your coordinates from alitum, and if they are translated or not depends on how you add those ghost fids, if they are in altium coordinate space, you have to add them in altium space.

so where they would be relative to the green rectangle

if you translated the coordinates as luiHS script does. Then you have to add them in the red rectangle space, where the lower left of the rectangle is 0,0 or whatever the origin was in altium

once you've added the coordinates in the right space. make sure they are all listed in the panel, and fiduciary lists in the same coordinate system, either all relative to the green box, or all relative to the red one. Then you can select the location of the first fid listed on the list , move the machines camera to the physical location of that fid press cancel and use the move everything yours says (to current po) button which is the lowest button on the right of that group. The machine should then translate everything to the right place.

That is the procedure the machine uses with the rails or when its in automatic mode, so when you do a test load with the rails use the "to current po" to test it all works before feeding it, then when you feed it should work the same.

if you create the ghost fid's relative to the red box and the rest of the components are relative to the green box or vice versa, the head will go to the totally wrong place and should be offset by the difference of the orange line.

also you should definitely update the software.

[charliex]

i've been writing up my neoden4 stuff, haven't finished yet but looks like I am going out of town unexpectedly so figured i'd post what i had so far and finish up when i get back

https://charliex2.wordpress.com/2019/01/31/neoden-4-pick-and-place/

[cgroen]

i've been writing up my neoden4 stuff, haven't finished yet but looks like I am going out of town unexpectedly so figured i'd post what i had so far and finish up when i get back

https://charliex2.wordpress.com/2019/01/31/neoden-4-pick-and-place/

Impressive work!