Building new PnP machine from scratch

[Eagle Media]

Hello,

I'm new to Open PNP but have got experience with building complex machines.
In this particular case I have to make nice balance between money and value.

Target: PnP Machine with at least 4 nozzles and mutliple cameras.
I plan to make all electronics by myself and use as many standardised parts as i can.

I have read pretty much on this board, but I still have got some questions :)

Controller with at least 6+4 Stepper outputs (X, Y, N0, N1, N2, N3): Decided to go with a smoothieboard or similar.
Q: Anybody using (or build) the JuicyBoard from https://plugg.ee/ ?

As for CV I can go with
a) Multiple cameras like: https://www.youtube.com/watch?v=67qTpT4VUt0
b) or Fly-Over like: https://www.youtube.com/watch?v=wRx1ZB9lNok
Obviuosly for faster systems they are using highspeed cameras with at least 50, 60 or 100 fps.
Q: Can OpenPNP handle the amount of data coming from multiple cameras?
I have got a solution with 1080P25 with no lag at all and crisp clear macro images.
Q: Can I feed HDMI Video to OpenPNP?
Any ideas for Video I/O Hardware it doesn't make lag?

Q: Which CV method would experienced OpenPNP user prefer?
Multiple Cameras or FlyOver?

Nozzle? I don't have got any Idea about the nozzle quality right now.
As it's posible I'll stuck with single nozzle Set-Up Im asking experienced OpenPNP eser:
Q: Which nozzle is preffered for single nozzle Setups-easy nozzle (mechanical) changing function.

Q: How much of You are using hollow shaft Nema 8 for the nozzles?

Q: Which moving technique is preferred for the nozzles?
Single Servo for each nozzle or the Swing Servo for two nozzles?

Stepper Driver: I have build several driver allready, incl one with the high current DRV8711.
Later on I'll try to switch to ODrive.
Q: Anybody using (or build) ODrive in his machine?

In another thread I read something makes me worry:

Trampas Stern wrote:

Currently as I OpenPnP does not have the X and Y move commands independent, if you sent command to move X axis,

then sent command to move Y the controller would not move the Y axis until the X move was complete.

So it has to send the X and Y together so both axis are moving at same time, for fastest moves.

With independent controller this limitation is removed.

Q: Is this still the case?

Intelligent tape/roll feeder: I'm thinking about some kind of Auto-Setup function, like:
Insert the feeder and the feeder tells the machine which part is inside, as: 01080603P100N stands for:
Feeder Position1, Tape Width: 8mm: Type: Capacitor, Value: 100nF

The idea behind this intelligent feeder: Sometimes it's more efficient to reorganise the tape position for better placement speed.
With this kind of feeder You simply swap the the tape/roll to another position-which can be reaced more faster.

This position is on the last place on my ToDo List but I consider this kind of setup seriuosly as
I plan to use some AVR for the feeder which can easily handle such kind of additional information.
Q: Whom should I ask for the integration of this Interface (Software side-Open PNP)?

Q: Is there a group of people currently build a machine I can join or
anybody interested to build one with me?

I'm located in central Europe.

Regards,
Rick

[Thomas Langås]

Q: Is there a group of people currently build a machine I can join or
anybody interested to build one with me?

I'm located in central Europe.

I'm in the process of designing and sourcing parts to build a machine.  The idea is:

* At least able to place 0201 (maybe even 01005)

* About 1000 mm/sec speed

* About 560x480 ++ build area

* About 100-120 feeders / strips

* Two sides of feeders (opposite of eachother)

* ballscrews + rails for movement

[Jason von Nieda]

I'll answer what I can, and perhaps others will join in:

 

As for CV I can go with
a) Multiple cameras like: https://www.youtube.com/watch?v=67qTpT4VUt0
b) or Fly-Over like: https://www.youtube.com/watch?v=wRx1ZB9lNok
Obviuosly for faster systems they are using highspeed cameras with at least 50, 60 or 100 fps.
Q: Can OpenPNP handle the amount of data coming from multiple cameras?
I have got a solution with 1080P25 with no lag at all and crisp clear macro images.

In general, high FPS is not valuable in PnP. I run my machine at 5 FPS. What is valuable is the ability to quickly capture a single clear image at a precise time. This is how commercial machines do it. OpenPnP does not currently support triggering like this. In general, OpenPnP's vision is pretty slow compared to commercial offerings.

 

Q: Can I feed HDMI Video to OpenPNP?
Any ideas for Video I/O Hardware it doesn't make lag?

As long as the video source appears as a camera in your operating system, it will generally work. OpenPnP uses OpenCV for video capture. 

 

Q: Which CV method would experienced OpenPNP user prefer?
Multiple Cameras or FlyOver?

Most people use two cameras: one looking down, one looking up.

Q: How much of You are using hollow shaft Nema 8 for the nozzles?

I'd most people are using hollow shaft NEMA 8s.

 

Q: Which moving technique is preferred for the nozzles?
Single Servo for each nozzle or the Swing Servo for two nozzles?

I think you are referring to the Z axis here? I'd say there is about a 50/50 split between using rocker style and rack and pinion style.

 

Trampas Stern wrote:

Currently as I OpenPnP does not have the X and Y move commands independent, if you sent command to move X axis,

then sent command to move Y the controller would not move the Y axis until the X move was complete.

So it has to send the X and Y together so both axis are moving at same time, for fastest moves.

With independent controller this limitation is removed.

Q: Is this still the case?

This doesn't really make sense in the context of OpenPnP. OpenPnP never sends partial movements to the controller. Every movement contains a full 4 axis position. 

 

I plan to use some AVR for the feeder which can easily handle such kind of additional information.
Q: Whom should I ask for the integration of this Interface (Software side-Open PNP)?

See https://github.com/openpnp/openpnp/wiki/Setup-and-Calibration and in particular, https://github.com/openpnp/openpnp/wiki/Setup-and-Calibration%3A-Feeders.

 

Q: Is there a group of people currently build a machine I can join or
anybody interested to build one with me?

That is this group. Lots of people currently building machines!

Good luck, and have fun!

Jason

[Eagle Media]

Hello Thomas,

As I see your requirements you are going to build a serious industry quality system.
As this is my first setup I have to start with more simply machine.
Is this your first pnp machine? I guess not?

Which nozzle/head/camera system are you plan to use?

Regards,
Rick

[Thomas Langås]

This is my first PnP machine, but another member of my team has been building 3D printers from scratch before (using his custom design) so there's some mechanical experience.  Also, I've asked around a bit to get info from others who have built one or more PnP machine(s).

--
You received this message because you are subscribed to the Google Groups "OpenPnP" group.
To unsubscribe from this group and stop receiving emails from it, send an email to openpnp+unsubscribe@googlegroups.com.
To post to this group, send email to ope...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/openpnp/8b94f5df-b1a5-406f-a7de-fb702a9eba25%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Trampas Stern]

I had in the past built several machines, CNC, 3D printer, etc.  I will say the PnP has been the hardest to build. 

The precision and accuracy required on the PnP machine is much more demanding that others. For a simple example, consider homing. On a 3D printer if you home the machine and it is 0.1mm off in X axis then it is fine, just your print will be slightly off on the bed.  On a PnP machine if your homing is off then you have setup all your feeders again. I did not listen to people on the list and tried using mechanical switches for homing on PnP, and I learned they are not accurate enough.

On my PnP at the moment I have 32 feeders, 16 slot feeders (Yamaha CL feeders) and 16 cut tape strip feeders, so if homing is off I have to spend around 15-20 minutes setting up the feeder location correctly. So yea I will be implementing optical switches and maybe camera based homing. 

Currently I on my machine I have mixture of plastic brackets and parts (milled HDPE and SLA 3D printed). I have found that the flex in both is more than machine can tolerate, and now I am milling aluminum brackets. I highly recommend not using plastic (3D printed or otherwise) for parts that are required for mechanical stability. I know that if you designed the part correct you can use plastic, but my experience is that it is not worth it.  I have the same feeling about wood, wood is great for a base board, but it should not be used for parts requiring tight mechanical tolerances. 

When designing your machine start with the feeders, that is determine what you will use for feeders and design around those first. For example the CL feeders are huge, cheap and rock solid, but due to size and such they have to be designed into machine, it is not easy to add as an after thought. 

Setting up the X and Y mechanics is the "easy" part of the PnP.  The head design and camera system is where you will spend your time. The tolerance on the head from squareness to nozzle run out will consume your time.  Every part of the head (camera, nozzles, etc) has to have tight mechanical tolerances. The only thing that does not have tight tolerance is the Z height, the Juki nozzles have springs to help with this. 

Do not expect this to be a quick or cheap build. I started thinking sub $1000, right now I have $1500 in feeders alone and at least another $1500 in other parts for which I marginally have a working machine.  To get something close to the production level of the TVM920 you will pay as much as a TVM920 plus 6-8 man months of time. Like me it might cost more because of bad decisions along the way.

OK now that I have hopefully scared you.  I am willing to bet that everyone with a working PnP has gone through several iterations on the mechanics. I personally I have found it better to fail faster. That is no amount of people writing stuff in a group feed will help you more than trying to populate your first board. There is a lot to learn between here and there, so I recommend you build a machine as fast as possible and try to build your first board. That is this first machine is a throw away, it is like buying a book to learn a subject, once you know the subject the book is useless. You buy the book for the knowledge, build/buy your first machine for the knowledge of how to build the next one. 

To save you the most time and money find someone local to you that has built a machine and go visit them. This is the best way to jump start the education. 

 

 

To unsubscribe from this group and stop receiving emails from it, send an email to openpnp+u...@googlegroups.com.

[Eagle Media]

Hello,

Thank You for your honest opinion. You don't scared me :)
I know the first machine will be not as cheap I would like but
I want to bring down the costs of an entry level pnp machine to less than 800 USD.

I plan to build all electronics and mechanics in the next days for testing the software.
I have to see the capabilities of the software to be able to make final decision about this project.

Which Juki nozzles are you using?
Do You have got some Links for a serious Shop they don't selling garbage nozzles?
I don't understand the build of thoose adapters for nozzles: https://www.robotdigg.com/product/517/Nozzle+w/+Adapter+to+5mm+Hollow+Shaft+Stepper
Can I easily remove thoose nozzles from this adapters (I want to use nozzle changing system)?
What is the purpose of thoose screws?

Is there a build in option in OpenPNP for CV homing?
Do You know the optical resolution for optical end stops?

Regards,
Rick

[Bernd Walter]

On Sunday, June 18, 2017 at 7:05:17 PM UTC+2, Eagle Media wrote:

Hello,

Thank You for your honest opinion. You don't scared me :)
I know the first machine will be not as cheap I would like but
I want to bring down the costs of an entry level pnp machine to less than 800 USD.

Good luck - even if you know what pieces to use.

I plan to build all electronics and mechanics in the next days for testing the software.
I have to see the capabilities of the software to be able to make final decision about this project.

That's a good idea.
I currently have the gcode electronic with open steppers on my bench, not wired to the machine yet.
Just a couple of hours ago I've got a raspberry camera running with OpenPNP and I'm quite happy that the latency seems to be ok.
 

Which Juki nozzles are you using?
Do You have got some Links for a serious Shop they don't selling garbage nozzles?
I don't understand the build of thoose adapters for nozzles: https://www.robotdigg.com/product/517/Nozzle+w/+Adapter+to+5mm+Hollow+Shaft+Stepper

I've bough a set of 9 via Aliexpress, but they actually cam from robotdig.
The quality seems to be ok, but I have no experiences yet.

Can I easily remove thoose nozzles from this adapters (I want to use nozzle changing system)?
What is the purpose of thoose screws?

The screws are the ugly part...
You have 3 of them to center the nozzle.
It's a good amount of work to change it.
I've seen a video from one of the Chinese machine vendors on how to setup the nozzles.
After you've set the screws you are supposed to add a rubber ring to seal the screwholes from airleaks.
I've bought 4 picker head because of that, but the robotdig head only has 16-17mm Z movement, which is quite tricky.
A two picker head is available with bigger Z motors and I assume those have more Z movement.

Is there a build in option in OpenPNP for CV homing?

Yes
You home normally and then finetune via fiducial.

[Bernd Walter]

On Sunday, June 18, 2017 at 1:51:46 PM UTC+2, Trampas Stern wrote:

I had in the past built several machines, CNC, 3D printer, etc.  I will say the PnP has been the hardest to build. 

The precision and accuracy required on the PnP machine is much more demanding that others. For a simple example, consider homing. On a 3D printer if you home the machine and it is 0.1mm off in X axis then it is fine, just your print will be slightly off on the bed.  On a PnP machine if your homing is off then you have setup all your feeders again. I did not listen to people on the list and tried using mechanical switches for homing on PnP, and I learned they are not accurate enough.

What else would you use for that?

When it comes to Z leveling on 3D printers mechanical switches weren't that bad.
But you can always adjust your coordinate system with G92 instead of reconfiguring every single location.

On my PnP at the moment I have 32 feeders, 16 slot feeders (Yamaha CL feeders) and 16 cut tape strip feeders, so if homing is off I have to spend around 15-20 minutes setting up the feeder location correctly. So yea I will be implementing optical switches and maybe camera based homing. 

Currently I on my machine I have mixture of plastic brackets and parts (milled HDPE and SLA 3D printed). I have found that the flex in both is more than machine can tolerate, and now I am milling aluminum brackets. I highly recommend not using plastic (3D printed or otherwise) for parts that are required for mechanical stability. I know that if you designed the part correct you can use plastic, but my experience is that it is not worth it.  I have the same feeling about wood, wood is great for a base board, but it should not be used for parts requiring tight mechanical tolerances. 

When designing your machine start with the feeders, that is determine what you will use for feeders and design around those first. For example the CL feeders are huge, cheap and rock solid, but due to size and such they have to be designed into machine, it is not easy to add as an after thought. 

 Exactly one of my problems.

I've build a solid frame and the Yamaha feeders won't fit anywhere without redesigning a major part of the machine.

[Eagle Media]

Am Sonntag, 18. Juni 2017 19:45:03 UTC+2 schrieb Bernd Walter:

On Sunday, June 18, 2017 at 7:05:17 PM UTC+2, Eagle Media wrote:

Hello,

Thank You for your honest opinion. You don't scared me :)
I know the first machine will be not as cheap I would like but
I want to bring down the costs of an entry level pnp machine to less than 800 USD.

Good luck - even if you know what pieces to use.

Yes, I know it will be a challenge.
 

I plan to build all electronics and mechanics in the next days for testing the software.
I have to see the capabilities of the software to be able to make final decision about this project.

That's a good idea.
I currently have the gcode electronic with open steppers on my bench, not wired to the machine yet.
Just a couple of hours ago I've got a raspberry camera running with OpenPNP and I'm quite happy that the latency seems to be ok.

I tested the raspy cam as well. The resolution, response and frame rate is great.
Much better than the cheap chinese microscope cams.

How do You get the video signal to the PC?
 

 

Which Juki nozzles are you using?
Do You have got some Links for a serious Shop they don't selling garbage nozzles?
I don't understand the build of thoose adapters for nozzles: https://www.robotdigg.com/product/517/Nozzle+w/+Adapter+to+5mm+Hollow+Shaft+Stepper

I've bough a set of 9 via Aliexpress, but they actually cam from robotdig.
The quality seems to be ok, but I have no experiences yet.

Can I easily remove thoose nozzles from this adapters (I want to use nozzle changing system)?
What is the purpose of thoose screws?

The screws are the ugly part...
You have 3 of them to center the nozzle.
It's a good amount of work to change it.
I've seen a video from one of the Chinese machine vendors on how to setup the nozzles.
After you've set the screws you are supposed to add a rubber ring to seal the screwholes from airleaks.
I've bought 4 picker head because of that, but the robotdig head only has 16-17mm Z movement, which is quite tricky.
A two picker head is available with bigger Z motors and I assume those have more Z movement.

So, basically this parts are useless if I want to auto change the nozzles?
 

[Bernd Walter]

On Sunday, June 18, 2017 at 8:29:44 PM UTC+2, Eagle Media wrote:

Am Sonntag, 18. Juni 2017 19:45:03 UTC+2 schrieb Bernd Walter:

On Sunday, June 18, 2017 at 7:05:17 PM UTC+2, Eagle Media wrote:

Hello,

Thank You for your honest opinion. You don't scared me :)
I know the first machine will be not as cheap I would like but
I want to bring down the costs of an entry level pnp machine to less than 800 USD.

Good luck - even if you know what pieces to use.

Yes, I know it will be a challenge.
 

I plan to build all electronics and mechanics in the next days for testing the software.
I have to see the capabilities of the software to be able to make final decision about this project.

That's a good idea.
I currently have the gcode electronic with open steppers on my bench, not wired to the machine yet.
Just a couple of hours ago I've got a raspberry camera running with OpenPNP and I'm quite happy that the latency seems to be ok.

I tested the raspy cam as well. The resolution, response and frame rate is great.
Much better than the cheap chinese microscope cams.

I use camera modules from Waveshare.
They use the same chip as the original Raspberry camera and are identical in software, but
come with several different lense options, of which some allow easy focus adjustment.
 

How do You get the video signal to the PC?

Basicly by the info from https://groups.google.com/forum/#!topic/openpnp/-Y1Q_e44Q20
However the sourcecode links where not available anymore.
Paul has a new Webcam driver based system, which didn't work for me because of BridJ init
errors, which is odd, since BridJ should have FreeBSD support.
I've hacked the OpenCvCamera driver to do an fg.open("http://10.1.1.195:8081/stream/video.mjpeg");
All the property setting functions are removed.
Maybe I will have to readd a few in case some of them setup camera calibration.
As a hack I will reuse the USB index field to select hardcoded strings for the two cameras.
With FreeBSD I had been forced to use a different OpenCV compilation anyway, therefor mine
already had ffmpeg support.

Which Juki nozzles are you using?

Do You have got some Links for a serious Shop they don't selling garbage nozzles?
I don't understand the build of thoose adapters for nozzles: https://www.robotdigg.com/product/517/Nozzle+w/+Adapter+to+5mm+Hollow+Shaft+Stepper

I've bough a set of 9 via Aliexpress, but they actually cam from robotdig.
The quality seems to be ok, but I have no experiences yet.

Can I easily remove thoose nozzles from this adapters (I want to use nozzle changing system)?
What is the purpose of thoose screws?

The screws are the ugly part...
You have 3 of them to center the nozzle.
It's a good amount of work to change it.
I've seen a video from one of the Chinese machine vendors on how to setup the nozzles.
After you've set the screws you are supposed to add a rubber ring to seal the screwholes from airleaks.
I've bought 4 picker head because of that, but the robotdig head only has 16-17mm Z movement, which is quite tricky.
A two picker head is available with bigger Z motors and I assume those have more Z movement.

So, basically this parts are useless if I want to auto change the nozzles?

Can't say for sure, but i personally wouldn't even try.
For autochanging I assume those to be the best choice:
http://www.betztechnik.ca/store/p32/Quick_change_Juki_nozzle_holders-_NEMA_8_5mm_OD_hollow_shaft-_STOCK.html

[PK]

Thank You for your honest opinion. You don't scared me :)
I know the first machine will be not as cheap I would like but
I want to bring down the costs of an entry level pnp machine to less than 800 USD.

I'm sure this will be a great experience for you. You'll find lots of support and advice here.   

My advice is this: 

A machine with the sort of specs you are talking about will place more than 2000 component placements per hour.

Now it obviously depends on what you are making, but our machines, when they run at about 1500cph consume between $1000 and $2000 of components per hour of operation.

As a concrete example, we are running at about 1000CPH at the moment, it looks like run will take a week and we'll place a little over 30,000 parts for an all up cost of AU$32000

Not including our time, we've spent just under $3K on the upgrade of a machine that cost (IIRC) $4500. So, in one week the machine has cost 4 times its purchase price in parts...

If you look at it this way, spending a few grand on a machine that will (in a short period of time) cost $100,000 to 'feed' doesn't seem so bad...

PK

[Arie Lashansky]

Hi
I think you need to ask what people get from the open PNP platform if my tool can
Place 0402 and QFP 100 I think it is a win smaller than that I do not think is possible in practise.
I will be happy to be proven wrong.

Arie

[Eagle Media]

Hello,

The initial price of 800 USD for my machine should be with 2 nozzles and without feeders.
I'm not sure I understand Your calculation at all. Offcourse if You add feeders, the machine will costs You more.
And the time You spending seting up this machine I didnt calculated as well...

Do You want to tell me You spend over 7500 USD only for the (raw) parts for Your current machine?
Not including setup and other times you usually have to pay for a worker?

PROJECT UPDATE: I decided to drop the Raspberry as Video Source and goes with webcams instead.
The video quality is worse but there is allmost no lag. This will probably push the price of the PC it will controll the PNP machine.
I solved the mechanical contacts of the feeder to the machine.
Unfortunatelly my new solution limits the count of the feeder I can insert to 25-30 tapes on this side.

UNSOLVED: Feeder mechanics that can be manufactured cheap.

[Eagle Media]

Hello,

I'm not sure I understand what do You mean?
Currently with my budget I should be able to place components from 0603 to QFP48.
I guess placing QFP Parts will depend on the used vision system as well.
As I remember it's hard to place 0,4-0,5 fine-pitch components with cheap pnp setups.

Regards,
Rick

[PK]

The initial price of 800 USD for my machine should be with 2 nozzles and without feeders.
I'm not sure I understand Your calculation at all. Offcourse if You add feeders, the machine will costs You more.
And the time You spending seting up this machine I didnt calculated as well...

My apologies for being unclear. And a caveat before everyone turns on me!

Building a PnP machine because you want to is cool. Pushing back the limits of what a PnP machine can do is way cool. There is some evidence that things like loose part recognition and picking may have some real value for small scale production ( oh and it's really impressive)

Aside from that, what pick and place machines do is put large numbers of components on medium to large numbers of boards.  At 1000CPH the cost of the components we are placing this week is around $1000 per hour of machine run time.

So you build you $800 machine.  10 hours of use later, you have spent $8000 on components for it to place and all of the corners you had to cut (3d printed motor mounts instead of machined steel) seem a bit silly.

Do You want to tell me You spend over 7500 USD only for the (raw) parts for Your current machine?
Not including setup and other times you usually have to pay for a worker?

I estimate that we have about $10,000 - $12000 into our machine including labor purchase price, upgrade parts, The refurbishment and upgrade took something like a man month..

This is very good value for money.

 

Again, building little benchtop machines to try out new stuff is great, But building a pick and place machine you intend to use to make boards with holding low cost as a goal just seems silly , you'd be better of spending more and getting better components...  

Anyhow, that's my 2c worth,I'll shut up now....

PK

[Trampas Stern]

I agree with PK! 

Here is an analogy for you..

Years ago I worked as an automechanic to support my college habit. I purchased a cheap 4 cylinder Caravan for $800 with a blown head gasket. I fixed the head gasket and then about 5,000 mile latter the engine went. A friend had a junk yard and I got free V6 engine transmission, wiring harness,etc. I swapped the drive train into my caravan, wired up the fuel injection system etc.  It ran great, but then I wanted power windows and door locks, so I went to junk yard and added power windows and door locks.  I kept fixing the caravan, and fixing, however no matter how much I fixed the caravan it was still a POS $800 caravan. 

I learned something from the experience, which I seem to still repeat. It cost just as much to fix a piece of junk as it does something nice, the only difference is the initial purchase and what you have in the end.  After that I brought Mercedes and fixed them, at least when you are done you have something. 

 

I say I am still learning this as I did the same thing with my PnP, I tried to build something for under $800 and now have several thousands into it and it is worth $800.  Please do not repeat my mistakes, start with a good foundation so you have something in the end.

Bend Rocks has started a CAD design with a ATP5 base plate and designed for the Yamaha feeders from the start. So far this is the best start of a stable platform I have seen. 

Trampas

[Eagle Media]

Thank You for the explanation.

Well, I didn't planed to build PNP machine initially,
but a tool helping me assembling single boards more faster than with tweezers.
So, I made a semi-auto feeder and lag-free camera system.
I understand, I'm not far away from a PNP machine.

Unfortunatelly without software the greatest hardware is only a piece of metal.
This was some years ago... Some days ago I found Open PNP by chance.

I know very well, that a machine worth of 800USD will have got it's limitations.

In my case the limiting factor for this project is the software which i don't know at all right now.
This is the reason, I'll build the first prototype with 3D printed and allready existing parts.

I wonder, why did You decide to go with a DIY PNP instead tof buying a neoden 4 for example.
It's cheaper and You know it's working at least on some level.

Building pnp machine from scratch hiding to much risks to fail and as you told allready
not pre-calculated (hidden) costs which can make the DIY machine to show stopper pretty fast.
I'm asking myself if you asked yourself afterward why didnt you buyed finished machine from manufacturer?

I talked to some people they build and publish they own machines allready.
Basically non of them was ready to make updates-if needed.

As this is a community I hope that feature requests they are reasonable will be made from someone.
I'm open to all suggestions and critiques as well :)

In the mean time I changed my concept by 180°

Q: Is there a "Lost Part Recognition" in Open PNP allready?
If Yes, what kind of recognition is this? Camera or CV?

Do You have got an Image or video of Your machine somewhere to show me?
I saw You are pretty active in this group :)

Regards,
Rick

[Paul Jones]

All of that is very true, but in my case it’s a matter of what comes first: the chicken or the egg? I’d love to spend 20k on a good machine, and I’ve even got the money to do so, but I’m much better off with my 2k machine till I get a feel of how successful my business will be.

I could only use china assemblers at the moment due to the cost vs volume equation, but from past experience I know I can do myself just as cheap(ish) using hand placement, and it doesn’t take weeks of back and forwards in communication.

But could I spend 6 hours a day placing 0603s on a board? Nope, I go crazy after about the 4^th day!

A normal person would just persist using a 3^rd party assembler, I used to work in industrial control + automation so building a machine is easy 😊 Every time you post an update about your retrofit you make me jealous!!

 

Paul.

 

--

You received this message because you are subscribed to the Google Groups "OpenPnP" group.
To unsubscribe from this group and stop receiving emails from it, send an email to openpnp+u...@googlegroups.com.
To post to this group, send email to ope...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/openpnp/2481f1e5-768b-4b28-b6dc-107915885a3c%40googlegroups.com.

[Bernd Walter]

Well - there is the pressure sensor support.
Not sure how the bottom camera would handle a lost part.
And some people have shown some impressive progress with loose parts feeders.

[Bernd Walter]

On Monday, June 19, 2017 at 1:39:32 PM UTC+2, Paul Jones wrote:

All of that is very true, but in my case it’s a matter of what comes first: the chicken or the egg? I’d love to spend 20k on a good machine, and I’ve even got the money to do so, but I’m much better off with my 2k machine till I get a feel of how successful my business will be.

I could only use china assemblers at the moment due to the cost vs volume equation, but from past experience I know I can do myself just as cheap(ish) using hand placement, and it doesn’t take weeks of back and forwards in communication.

But could I spend 6 hours a day placing 0603s on a board? Nope, I go crazy after about the 4^th day!

If you do this from time to time it can be very relaxing.
But you need a calm hand day, no interruptions when you are doing lots of boards, since time hardening the paste is against you.
You can stretch the working time by spraying a bit of isoprop, but not too much and only once, since it washes out the flux.
Plus even if your body allows you to do this by hand - a cramp can completely ruin your day.
Even a machine just capable of doing the chicken food (or fairy dust as others say for the passives) would be enough to be very helpful.
So the requirement for a perfect machine just from the start isn't that important.

A normal person would just persist using a 3^rd party assembler, I used to work in industrial control + automation so building a machine is easy 😊 Every time you post an update about your retrofit you make me jealous!!

I only do this for very high volume boards and even with my own machine I would do this.
There is way more than a PnP machine.
A complete assembly line with a zone oven is so much better for speed and quality.
Not to speak about a wave soldering machine for through hole components plus you buy the experience of using those machines.

[Eagle Media]

@Trampas Thank you for your comment.

Basically I plan to save money with exact planning and prefedifined manufacturing processes.
Do You have got images or videos from your setup?

Does Bend Rocks have got published something allrady or estimated the costs?
To be honest I think the Yamaha feeders are the way to expensive.
I found prices somewhere about 100-150 USD per unit.

Q: On the beginning I'll go with 5 steppers and smoothieboard.
Somebody asked for 16X stepper support in this group Unfortunatelly I cant find this thread anymore.
Somebody else answered it can be written with 60 lines of code and connect both controllers via SPI.

Q: Anybody using or build Juicyboard R1000AX from plugg.ee Labs?
What is the purpose for this I2C I/O on each Slot?

Regards,
Rick

[PK]

On Monday, June 19, 2017 at 7:37:10 PM UTC+8, Eagle Media wrote:

Thank You for the explanation.

Well, I didn't planed to build PNP machine initially,
but a tool helping me assembling single boards more faster than with tweezers.

You need an assisted manual PnP with something like a carousel feeder. It's what we used when a batch of boards was 20 or so.

They are cheap and easy to make and you can easily hit 300-500 CPH...  Now I agree that this gets old pretty quickly. But an hour a day gets you 10 boards made.....

[Eagle Media]

Thank You. The point is that I have got pretty much spare parts for a cnc machine and electronic generally
and I can get the missing parts pretty cheap-allmost for the price of the manufacturer.

Sometimes I have to deliver units in 100X and I don't want to make this manually.

Actually I asked you to show me the 10K USD Open PNP Setup :)
Is there a way to show me your open pnp machine too?

Q; How much people supporting (coding) the current build of the software (open pnp)?

[Bernd Walter]

On Sunday, June 18, 2017 at 7:45:03 PM UTC+2, Bernd Walter wrote:

On Sunday, June 18, 2017 at 7:05:17 PM UTC+2, Eagle Media wrote:

Can I easily remove thoose nozzles from this adapters (I want to use nozzle changing system)?
What is the purpose of thoose screws?

The screws are the ugly part...
You have 3 of them to center the nozzle.
It's a good amount of work to change it.
I've seen a video from one of the Chinese machine vendors on how to setup the nozzles.

Here is the link to the video:
https://youtu.be/fO_93S8NHOY
 

After you've set the screws you are supposed to add a rubber ring to seal the screwholes from airleaks.
I've bought 4 picker head because of that, but the robotdig head only has 16-17mm Z movement, which is quite tricky.
A two picker head is available with bigger Z motors and I assume those have more Z movement.

Bigger rotation Motors, not Z...
 

[Eagle Media]

Thank you, I see you have got a solid base.
As I understand you can not change those nozzles automatically.
That I understand this correctly -> If you want to place 0402, 0805 and 2510 you have to change the nozzle manually?

Did you ever tried video aligment after turning your part?

Are you using hollow-shaft NEMA 8 for the rotation of Z?
Beside of the nozzle spring you don't use any mechanism for the hight on Z?

[PK]

Actually I asked you to show me the 10K USD Open PNP Setup :)
Is there a way to show me your open pnp machine too?

Sorry, I missed that. Here are the last two video's

https://www.youtube.com/watch?v=urPLE_9oDhM&ab_channel=PaulKelly 

https://www.youtube.com/watch?v=a_Z8eJp7LTQ&ab_channel=PaulKelly

[Eagle Media]

Thank You, it's great. Are you using servos on all places?
2 nozzles, top and bottom camera. Correct?

Are you using the downloadable version of the OpenPNP or
you made some modification? If Yes, how much % is changed/added?

Which controller are you using to drive this machine?
I'm sitting now and thinking about the controller-smoothieboard or juicyboard?

[PK]

On Tuesday, June 20, 2017 at 7:41:50 AM UTC+8, Eagle Media wrote:

Thank You, it's great. Are you using servos on all places?
2 nozzles, top and bottom camera. Correct?

 

3 nozzles, top and bottom cameras. The machine is a refurbished CSM60 

Are you using the downloadable version of the OpenPNP or
you made some modification? If Yes, how much % is changed/added?

Pretty much the downloadable version. I have built customised autofeeder code. Our feeders are advanced by the same pneumatics that lower the heads and some of them need multiple pushes before the part is picked.

 

Which controller are you using to drive this machine?

A smoothie and a slighty smart and very isolated io board we made with  

I'm sitting now and thinking about the controller-smoothieboard or juicyboard?

Seriously. Forget all that, and make a decision about what you can live with for feeders.  I know nothing about your situation, but my experience is that people trying to build <insert thing they want here> for a fraction of the $X price are usually doing so because they don't have $X.  If this is the case and you want a production class machine then, right now, there is a good chance that the cost of the feeders will swamp the cost of of the machine.  I spent US$1500 on our CSM84 and US$10000 on feeders...

If you don't have a plan, you may struggle to complete the project.

OK, so we're now up to about $7.50 of 'my 2c worth...  Just trying to help...

[Eagle Media]

I'm thinking about the controller as I have to make it.
Without controller (and driver) nothing will move at all.

With 800 USD for a PNP machine I mean the net price. Not the real price.
In real life I count someting about 2-3000 USD or maybe more.

At this time I need some (moving) hardware parts to evaluate the software.

[Cri S]

You mean building machine for selling it ?

If you want have 2-3000 USD for the PnP go with smoothie, it is the only board official supported.

If you could support it yourself, there are cheaper alternatives.

[Marek T.]

No chance to build it for $800 in my opinion. 2-3 closer to real.
Forget Juki nozzles as Chinese cheap copies of original are not worth later problems (runouts). Think out Jason's advice for Samsung CP45 nozzles, their holders seems to be more stable ad construction is more simple.

[Eagle Media]

Am Dienstag, 20. Juni 2017 07:37:41 UTC+2 schrieb Marek T.:

No chance to build it for $800 in my opinion. 2-3 closer to real.

This is exactly I want to try. Ordering parts in higher quantities from manufacturer but not from reseller.
 

Forget Juki nozzles as Chinese cheap copies of original are not worth later problems (runouts). 

Think out Jason's advice for Samsung CP45 nozzles, their holders seems to be more stable ad construction is more simple.

Yes, I noticed the quality of the Juki nozzles are very ugly and they are the way too expensive for such simply part.
I'll take a look for the nozzles from Samsung. Do You have got a link to a reliable shop for them?

[Eagle Media]

Am Dienstag, 20. Juni 2017 04:15:22 UTC+2 schrieb Cri S:

You mean building machine for selling it ?

In the first line I mean ordering parts in higher quantities from the manufacturer directly.

[Trampas Stern]

For myself I have very specific desires for PnP and having a back ground in computer vision and embedded design I wanted add vision algorithms to make my life easier for my production goals. 

I looked at the cost of various machines, but at the time figured "I could build it cheaper and better" and I did not want to invest that much money into a machine at the time.  It is like buying most tools, if you don't know what you need in a tool it is often best to buy a cheap one and learn what you need to know, then throw away cheap one and buy a good one. Alternatively you could buy the best and be done. With PnP it is the same, the last thing you want to do is spend $5k and realize it will not work for your needs.  I now know my needs and what to look for, and I am looking at used PnP machines while continuing to work on my machine.    

Once I got my first prototype machine built and I tried to get OpenPnP working with it, this is when the reality hit. OpenPnP is well documented and Jason is great but it has a very difficult setup and learning curve.  No matter what I did I could not get OpenPnP working on my hardware. I found that everything has to be perfect for OpenPnP to work and figuring out what was not correct and causing problems was difficult.   I then stopped with OpenPnP have written my own software in python, GUI and all. This allowed me step by step to figure out what is off with my machine.  I have learned a lot, even simple stuff like you have to have feeders and PCB at same Z height so the fixed focus camera is focused, duh.  Again these little things when designing a machine can result in a complete tear down and rebuild. There is a lot more I have learned about setting up cameras and mechanics of the system.  There is little things to learn like don't use a white build platform with white tape strips. Other things like cut tape and strip for diodes and LEDs do not have enough surface area on bottom of tape to stick to platform with double stick tape.  Even things like with OpenPnP if the cut tape and strip is long and not perfectly straight you will most likely have problems the software can not adjust for.  Learning these things and how to work around them with in context of the software is what you learn on your first builds. 

The main thing I have learned is that mechanical tolerances are critical to stability of the machine.  There is some amount that computer vision can help correct the problems but the vision is tricky. For vision you need the right camera and lighting setup, which I don't have yet.  I have also learned that the software needs to help you. For example in my python software I have the software do as much as it can to help me setup the mechanics and inform me when something is not right, because that was my initial problem.  However at the end of the day if you have good mechanics, vision, and follow the rules OpenPnP should work.  OpenPnP really needs vision to work, you can use it without vision but generally that is not the most tested and the most used path of the software (get off the happy path and their will be trouble). 

Right now I am working on doing vision based homing in my software, here after mechanical homing move head camera over a fudical (dot) and align to that to fine tune homing. Additionally to work around slight shifts in my mechanics I am also doing algorithms to find center of parts in feeders for fine tuning pick locations.  Again all of this stuff is to deal with poor mechanical design, so I am working on fixing the mechanics at the same time.  Again to make software work, you want to stay on the happy path of the software. 

At the moment I can pick and place parts, but  I have to manually go and realign parts on PCB after the PnP is done. This is still a huge time saver over manually populating boards.  The biggest mechanical issues I have at the moment, is the plastic mechanical parts, these are going away. Second is that I need more reel feeders. The less cut tape parts you have the happier you will be.  

I am using Yamaha feeders on my machine as that used Yamaha feeders from Ebay are under $20. 

Trampas

[Marek T.]

No don't have any reliable link/contact. I've used Juki like almost everybody here and trying to fix it as have all mechanics made for this. But I'm close to decision to change it, waiting for success of CriS' update yet, which maybe will fix it by software way. If you start from beginning, consider option of Samsung nozzle parts - or just other than Chinese Juki copy.

[Eagle Media]

Thank You. This is something I did not wished to hear.

OK, I'll think about new nozzle system too.
I never liked thoose nema 8 hollowshaft setups anyway.

[Eagle Media]

My machine shoud be capable to place components from 402 to 2917 and from QFN20, QFN48 to LQFP100

 

Once I got my first prototype machine built and I tried to get OpenPnP working with it, this is when the reality hit.

This is what I'm afraid of but it's looking very promising for me.
 

The main thing I have learned is that mechanical tolerances are critical to stability of the machine.  There is some amount that computer vision can help correct the problems but the vision is tricky. For vision you need the right camera and lighting setup, which I don't have yet.  I have also learned that the software needs to help you.

I'll get some camera tomorrow and make some tests.
Which camera are you using right now?

I know about the problems with focusing the parts. Low end cameras can't focus as fast that you can use them.
Q: Is there a way to integrate "Hight Adjusting on Z for CV" for the time, the part goes over the bottom camera?

[SMdude]

"I know about the problems with focusing the parts. Low end cameras can't focus as fast that you can use them.
Q: Is there a way to integrate "Hight Adjusting on Z for CV" for the time, the part goes over the bottom camera?"

Z height is compensated for.
The base of the part will always be at the same height.
You set the part height under the parts tab.

[Eagle Media]

NOTE: These are street prices in regular online-shops.
Price in industrial quantity can be lower by 25-75%

PRICE UPDATE: ~530 EUR

Included:
2x ELP Cameras (USB2-3MP) with 12mm Lens
2x NEMA 23 Stepper (0.9°)
2x NEMA 8 Stepper (hollow shaft)
2x Linear Rails (150mm with 2 Block)
2x 16mm Rails (600mm with 4 blocks)
2x 20mm Rails (600mm with 4 blocks)
6x Pulley
2x 8mm Rails (650mm)
2x 8mm Rails
2x Cable Chain (15x30)
2kg PLA Filament *
Base Construction Block
4m 10mm Belt
10x Ball Bearing

* Current Printing Time for the Base Parts: 160h

Not Included:
Electronic and Cabling.