Neoden 4 pick and place

[mrpackethead]

How do you configure all the variable stuff that the machine will need to know.. like Vision Contrast and machine speeds etc etc.

Via a CSV file?

[TheSteve]

How do you configure all the variable stuff that the machine will need to know.. like Vision Contrast and machine speeds etc etc.

Via a CSV file?

No such thing as vision contrast. All of the settings are done via the included software, it is a windows program still.

[ttsthermaltech]

That is the one thing that seems to be missing from many of the small PNP's. Having both contrast and illumination settings for each part type is pretty much a necessity for proper vision control. Illumination color is also a nice feature, but not absolutely necessary. Every part, due to shape, package color, plating, etc will capture differently, and assuming they will all capture properly with the same universal contrast and lighting setting is rather naive.

Even a digital brightness and contrast setting for each part would help, but alas, not seen on the low dollar machines.

[mrpackethead]

That is the one thing that seems to be missing from many of the small PNP's. Having both contrast and illumination settings for each part type is pretty much a necessity for proper vision control. Illumination color is also a nice feature, but not absolutely necessary. Every part, due to shape, package color, plating, etc will capture differently, and assuming they will all capture properly with the same universal contrast and lighting setting is rather naive.
Even a digital brightness and contrast setting for each part would help, but alas, not seen on the low dollar machines.

This very much impressed me last night when i went through the demonstration of the SmallSMT system. It seems to have excellent recognititon of its parts, and much of that is due to the fact that you can tune the vision for every part, or fiducial etc.  a Gold plated fiducial will behave differently to a HASL coated one, etc etc.

And it was pretty easy to setup. 

[Smallsmt]

However, I bought a machine from Sunny (HotHotSMT) which sold me a machine made by Yushengtech. Both Sunny / Yushengtech would not stand behind the product that they sold me. Take a look at the Yushengtech website (www.yushengtech.com) and you will find the exact machine that SmallSMT is offering.

I think some days why are you doing this business?
You're trying to improve a situation in order to protect the customers and you receive people trying to pee on your legs.
We have brought a machine in the state that it is ready for use and in addition improved which simplify the use of all points.
We offer what many Chinese manufacturers do not.
A reliable support, feedback on customer requirements and further development of the existing product for the benefit of all customers.

My reputation is clean I do my business for more than 30 years and most of my first customers are still working with me. So I don’t need to tell lies or hide something .

We are the general sales agent for all Yusheng machines sold outside of China so Yushengtech sells in China and we do no support or give warranty for machines bought in China!

That's the last post about this topic!
I am sorry but I need to clarify!

Best regards
Michael

[thommo]

TheSteve

You made this comment below some days ago:
The parts tend to walk to the left on the X axis as we go up the Y axis. This is occurs even with the use of fiducials. The panel we have been testing with can be seen in one of the pics I posted previously. It is 3 boards in a 1x3 matrix. We were using the fiducials at the edges of the panel. The last thing we tried was using fiducial markings on each board instead of the panel. This seems to make the placement plenty usable but does leave doubt in my mind over larger boards as these really aren't that big.

Am wondering what you did to fix the 'wandering' issue that you've described.
Is it just that you work board-by-board now, or was it something to do with setting the 'initial vertical [Y axis] alignment' of the board?

Alternatively, did you detect any of the motor drivers missing some 'steps'?

[mrpackethead]

Alternatively, did you detect any of the motor drivers missing some 'steps'?

Poetnially hard to detect..

[Smallsmt]

@thommo,

Did this problem change if you change the speed?
If not it is a calibratiton problem and your machine need to be adjusted for travel way on each side.
Please put a steel ruler on your pcb and check the travel distance.

[mikeselectricstuff]

That is the one thing that seems to be missing from many of the small PNP's. Having both contrast and illumination settings for each part type is pretty much a necessity for proper vision control. Illumination color is also a nice feature, but not absolutely necessary. Every part, due to shape, package color, plating, etc will capture differently, and assuming they will all capture properly with the same universal contrast and lighting setting is rather naive.

Even a digital brightness and contrast setting for each part would help, but alas, not seen on the low dollar machines.

Seriously? That's just ridiculous. Nothing at all to set any vision thresholds or anything....?
My machine may have particularly finnicky vision, but I have to set individual thresholds for pretty much every part, and some, especially inductors with dark bodies can be really critical. At some point I intend to replace the feeble red illuminator with a really bright white, and stop the cameras down to make them less sensitive to ambient.

[elmood]

Hey folks, I received the Neoden 4 yesterday along with a reflow oven and stencil printer. Everything is pretty clearly labelled. I also found that the stand for the machine was bent, probably during shipping.... but the holes were tapped and everything went together fine.

The machine booted up and I activated everything using the test menu and all things seem to function. I haven't started placing components yet. One rather minor thing that could be improved would be an option for a monitor arm and a keyboard tray to attach to the machine. I'm planning to hack something together myself just to clean up the work area. My initial thoughts are that the machine is really solid. It does move the X/Y motors at a rather jerky and alarming speed. Another thing I noticed is that there isn't an easy way to access the internal electronics. I didn't look underneath (the thing is pretty heavy and takes 2 people to put it on the stand) but the entire bottom frame appears to be a single piece. No side access covers or anything.

My up-looking camera appeared to be installed on a bit of an angle. (which I fixed) It also had some fingerprints or gunk on the front glass. The downlooking camera appears slightly rotated, which I haven't tried to fix yet. When clicking on the screen to recentre the image, the points move to the centre pretty much perfectly so perhaps it's all fine.

Neoden sent me the latest software within about an hour of my request and I was able to update the machine. The update was about 850KB. The fact that it runs Windows XP is of no consequence to me... I figure either it's a licensed version, or could be considered abandonware by this point. In any case it boots up fast and goes straight into the app. There is no network connection so I'm not worried about viruses. I haven't investigated the file formats yet and it's not clear where the master machine settings get saved... I want to make sure I can back everything up somehow. Just for fun I tried running the updated software on a Windows XP sandboxed VM and failed to load because an OpenCV DLL was not found. 

Here is a link to some pictures I've taken so far. I'll update it as I discover more: https://capybara.org/~andrew/pnp/

[sam512bb]

Quite simply it was the size. We can easily move the Neoden around and get it through standard sized doors.

Good day TheSteve,

Understood... indeed, these machines are heavy (450kg) and somewhat large ...  As I mentioned before, they are tanks.

Cheers,

Sam

[TheSteve]

TheSteve

You made this comment below some days ago:
The parts tend to walk to the left on the X axis as we go up the Y axis. This is occurs even with the use of fiducials. The panel we have been testing with can be seen in one of the pics I posted previously. It is 3 boards in a 1x3 matrix. We were using the fiducials at the edges of the panel. The last thing we tried was using fiducial markings on each board instead of the panel. This seems to make the placement plenty usable but does leave doubt in my mind over larger boards as these really aren't that big.

Am wondering what you did to fix the 'wandering' issue that you've described.
Is it just that you work board-by-board now, or was it something to do with setting the 'initial vertical [Y axis] alignment' of the board?

Alternatively, did you detect any of the motor drivers missing some 'steps'?

I don't believe we have seen any lost steps. A nozzle alignment seems to be what corrected the problem, we also made a small tweak to the theta adjustment on the upward facing camera.

[sedelman]

I have an opportunity to purchase a brand new Juki KP620 (Zevatech PM575) for $13K US. I went to see the machine and it is in pristine condition. The original owner purchased the machine new in 2002. It appears to have a DOS based control software and uses a 1.4MB floppy to read the SMT files. I found an FDD to USB converter for about $100 that allows me to use a USB flash drive instead of a floppy. My question is, has anyone fed this machine from Eagle? I had a look at the freeware software called PCBSynergy that appears to be able to convert from Eagle to the binary format that the Juki is expecting. For a dual camera system that can place up to 4000 cph (single head) with 102 x 8mm feeder capability and automatic nozzle changer (8 nozzles), it seems it makes more sense to me than a Neoden4. Anyone have a perspective that I may be missing?

[rx8pilot]

I have an opportunity to purchase a brand new Juki KP620 (Zevatech PM575) for $13K US. I went to see the machine and it is in pristine condition. The original owner purchased the machine new in 2002. It appears to have a DOS based control software and uses a 1.4MB floppy to read the SMT files. I found an FDD to USB converter for about $100 that allows me to use a USB flash drive instead of a floppy. My question is, has anyone fed this machine from Eagle? I had a look at the freeware software called PCBSynergy that appears to be able to convert from Eagle to the binary format that the Juki is expecting. For a dual camera system that can place up to 4000 cph (single head) with 102 x 8mm feeder capability and automatic nozzle changer (8 nozzles), it seems it makes more sense to me than a Neoden4. Anyone have a perspective that I may be missing?

This seems like an all new thread. Should not take away from the focus on the Neoden4.

[mikeselectricstuff]

I have an opportunity to purchase a brand new Juki KP620 (Zevatech PM575) for $13K US. I went to see the machine and it is in pristine condition. The original owner purchased the machine new in 2002. It appears to have a DOS based control software and uses a 1.4MB floppy to read the SMT files. I found an FDD to USB converter for about $100 that allows me to use a USB flash drive instead of a floppy. My question is, has anyone fed this machine from Eagle? I had a look at the freeware software called PCBSynergy that appears to be able to convert from Eagle to the binary format that the Juki is expecting. For a dual camera system that can place up to 4000 cph (single head) with 102 x 8mm feeder capability and automatic nozzle changer (8 nozzles), it seems it makes more sense to me than a Neoden4. Anyone have a perspective that I may be missing?

If you have the space that looks like a super bargain ( assuming the price includes feeders)

[mrpackethead]

I have an opportunity to purchase a brand new Juki KP620 (Zevatech PM575) for $13K US. I went to see the machine and it is in pristine condition. The original owner purchased the machine new in 2002. It appears to have a DOS based control software and uses a 1.4MB floppy to read the SMT files. I found an FDD to USB converter for about $100 that allows me to use a USB flash drive instead of a floppy. My question is, has anyone fed this machine from Eagle? I had a look at the freeware software called PCBSynergy that appears to be able to convert from Eagle to the binary format that the Juki is expecting. For a dual camera system that can place up to 4000 cph (single head) with 102 x 8mm feeder capability and automatic nozzle changer (8 nozzles), it seems it makes more sense to me than a Neoden4. Anyone have a perspective that I may be missing?

If you have the space that looks like a super bargain ( assuming the price includes feeders)

I'd just be cautious about parts avaialblity for a machine that is getting on 15 years old.  Juki are pretty good, but theres stuff that is now hens teeth to get,  If the machine is running, its probably worth buying for parts alone.

[sedelman]

@rxpilot it is not my intention to take away from this thread. I'm deciding between a Neoden4 and a Charmhigh CMHT48VA and have been presented with the opportunity to buy the Juki. I have personally verified the opportunity to be genuine, so I am asking people with more knowledge and experience to judge the opportunity. Can the Juki be made to work with Eagle output without too much grief, has anyone done it?

[rx8pilot]

I have never used a Juki, but I am a (reluctant) Eagle user with a Quad P&P. Being a DOS based machine, I would hazard a guess that the Juki uses a plain ASCII text file for program data. If that is true it is fairly easy to create your own ULP to create a program or import data. I wrote my own ULP after studying the program data and sample import data. Eventually, I ended up with a single ULP that can essentially create a complete program.

This is also probably true for Neoden4 and most others that import data. ULP writing requires some effort to learn, but its not too bad.

[sedelman]

@rx8pilot the Juki uses a proprietary binary file format with file extension P5A. The PCBSynergy software is a converter that can output up to 11 different file formats used by various PnP machines.

[mrpackethead]

@rxpilot it is not my intention to take away from this thread. I'm deciding between a Neoden4 and a Charmhigh CMHT48VA and have been presented with the opportunity to buy the Juki. I have personally verified the opportunity to be genuine, so I am asking people with more knowledge and experience to judge the opportunity. Can the Juki be made to work with Eagle output without too much grief, has anyone done it?

I'd considered the Neoden4 and had a looka t the Charmhigh system as well, then I found the SmallSMT systems and they gave me a demo last week. I'm used to usnign Juki and Yamaha systems for production, and to be honest I wished the software running on those was as good as what the SmallSMT system offers.  I'm looking at a small machine for our office, and i'd suggest you take a look at this as well.

[Jefferson]

Pay attention to the size of the components 0201, 0402 as your P&PM can provide it. I'm in doubt for precision accuracy, nozzles, etc. Neoden 4 win old P&PM.

[elmood]

Well I worked with the machine a bit today and so far the software is a bit different than I expected. I wrote out a chip list for the machine as a CSV file, but the way the Neoden software configures the job seems strange to me, and so far I couldn't get it to work. If anyone has ideas on what I might be doing wrong please let me know... this is my first PNP machine.

- I tried setting up the rail feeder option which appears to detect the top edge of the board when bringing the PCB into the machine on the rail conveyor system. (the hardware is very smooth, btw!) But when I ran my job it never seemed to find the board. Either it would give a "PCB feed failure" or an error saying "coordinate movement range beyond the boundary" which I assume has to do with the edge not being detected correctly. In both cases the PCB appeared to be in more or less the right place inside the machine.

- The mark point (fiducial) setup seems odd to me as it needs to be done in the machine, and the coordinates appear to be absolute machine positions and not related to the PCB placement file. Wouldn't it make more sense to have these points linked to the placement file coordinates somehow? It was pretty easy to set these up and they don't have to be actual fiducials... some of my older boards don't have real fiducials but using mounting holes or other large holes seem to work fine! The camera realigns to the centre of the holes better than I could click with the mouse.

- The chiplist I loaded was relative to the board's bottom left corner, but I think the software expects absolute machine positions. There is no example in the manual PDF included with the machine on how to prepare the "chiplist". There doesn't seem to be any way to "offset" the X/Y positions to the actual loaded position of the board. I'm assuming I have to provide some kind of offset in my input data, but I'm not sure where to get this. The software knows nothing about the bottom corner of the board. It does know the mark points, the top edge (it's only Y though) and the centre position of the first part in the chip list. (you have to manually align this)

So, needless to say I'm not building boards just yet. I was hoping to try some tests with only a few components and no solder paste. But after failing to feed correctly it stops and won't continue. I've sent off email to Neoden and am hoping to learn what I'm doing wrong.

The good part is that overall I'm very happy with the build quality of the machine. There are a lot of hardware details that are clearly well thought out, and some ingenious design especially in the pickup head. I'm sure the software is actually useful because there are videos of it working on real boards. But the error messages are not very clear, and there is no help showing what to do if these errors occur.

[thommo]

Hi Elmood,

I don't have a machine 'yet', but have you reviewed all the YouTube videos that Neoden produced for this machine? Just search for Neoden 4. There are about 7 in the series and from memory, they deal with the sort of issues you are describing. I suspect that the PDF they send hasn't recently been updated.

I'd be asking for all the latest info from Neoden directly if it were me.

Good luck.

[mikeselectricstuff]

Well I worked with the machine a bit today and so far the software is a bit different than I expected. I wrote out a chip list for the machine as a CSV file, but the way the Neoden software configures the job seems strange to me, and so far I couldn't get it to work. If anyone has ideas on what I might be doing wrong please let me know... this is my first PNP machine.

Everything _Should_ be relative to the fiducials, but we are talking Chinese software here.....
If it's done right, your P&P files have x/y/theta coords of each part, and fid locations, and the only setup is to tell the machine roughly where the first fid is so teh camera can find it. It would then derive the actual position from the camera sighting the first fid, and rotation/scale offsets from second (and subsequent, if any) fids.
In my process I use a special "fid" component which is recognised by the conversion software I wrote and converted to reference points for the machine's file format. 

[elmood]

Hi Mike: Your description is what I expected... simply pointing the machine to a reference and having it offset all of the other coordinates from the file. But it seems that they are relative to the bed 0,0 point on the machine all the time. (or else I'm perhaps missing something else)

thommo: I've been studying the manual and videos carefully but so far haven't seen any example of a workflow of importing data and actually using it. I have a script that makes all the data in the Neoden format... that's no problem. But now I'm thinking that I have to do something like this:

- find the fid point or some reference component point in the machine
- write down the X/Y absolute value
- tell my script the value and have it offset all others by the specific amount
- then import my part data

But that seems kind of round-about. I'd rather be able to export the data first, then go to the machine and set up everything, and then run the job. The other problem is that if the next time we run these boards the machine is set up slightly differently, does it mean I have to re-export all the offset data from my design files again with different offsets? That would be a pain!

BTW Neoden got back to me regarding the failure to find the edge of the PCB. Apparently "coordinate movement range beyond the boundary" means that the rails thought they went too far before sensing the edge of the board. Maybe I need to put the PCB closer to the input side. And "PCB feed failure" has to do with failing to find the mark point after feeding. I'm going to try again today and see if I can make it past that part.