Assembly line update

[kylehunter]

Hey all!

So this is quite the update since my last post/thread https://www.eevblog.com/forum/manufacture/assembly-line-thoughts-ddm-novastar/

That thread started with, and this whole adventure started with my company looking for a simple reflow oven to help with in-house prototyping and testing. After some great advice both here and on SMTNet, we decided against the DDM Novastar route. Instead we decided to go full on with an SMT line.

Our line consists of a DEK Horizon 265 printer, a Assembleon Opal Xii, and a Heller 1500SX oven. Most are early 2000's vintage, with the pick and place being the newest at 2008 vintage.

We received the equipment on May 23rd, May 29th it was wired and ducted in, and we had training on May 30th for 3 days.

Since that time, we have completed a production run of around 175 boards, and are about to start on a 1000 board run soon.

It definitely seems like our experience getting setup has been quite a bit different than most people have starting out. Especially after seeing all the horror stories that are in the pick and place lifestyle thread. It has of course been a massive learning curve, and still have a lot to learn, but it really has been quite smooth.

The fact that we finished the first production run within two weeks of having the equipment delivered is better than I could have ever dreamed. There have definitely been hiccups along the way, and I expect to have ongoing issues, but so far it absolutely seems like it was a great decision.

Issues we have had:
- The oven produces way more heat than I thought, requiring a second AC.
- Several Opal nozzles couldn't hold a vaccum, we are currently replacing them.
- The screen cleaner for the DEK is too high, requiring us to put a spacer on the screen. Trying to figure out a fix for that.

I genuinely would like to thank everyone on this forum so much for all the advice and suggestions. Even the people who said it wasn't worth it and a horrible idea!

I'll try and post some updates here as we go, as well as on my YouTube channel. Just posted a video going over the line: https://youtu.be/LiNGoku8Mqc

[jmelson]

It definitely seems like our experience getting setup has been quite a bit different than most people have starting out. Especially after seeing all the horror stories that are in the pick and place lifestyle thread. It has of course been a massive learning curve, and still have a lot to learn, but it really has been quite smooth.

If you buy a junk machine, or a worn-out machine, then things may  not go smoothly.  If you buy a good machine that is still in decent condition, then everything OUGHT to go much better.

I bought a well-used Philips CSM84 (so a generation older than your Opal) but it was in good condition when it arrived.
I did not get any training with it, but did get an operator's manual.  I now have 3 different vintages of this manual, and it is obvious the first one was a HIDEOUS translation from Japanese to Dutch to English.  The later ones are a LOT better.

Anyway, the machine had their feeder and placement files still in it, I rigged up a serial cable and downloaded the placement file.  With the help of the manual, the file format was quite obvious.  I wrote a program to convert my CAD package's placement format to the Philips, loaded it into the machine, and made my first board less than two weeks after the machine arrived.  I had never even SEEN a P&P machine before then.

I went through a LOT of issues related to proper solder paste stencil apertures, what pastes are good and what are awful, etc.  And, over the last 12 years, I've had a few breakdowns on the P&P machine.  Only one was serious, all the others were pretty minor stuff (cracked hoses, an intermittent conveyor sensor, etc.)

My machine doesn't have the vision option, and I really DO miss that on fine pitch parts.

Jon

[mrpackethead]

Sounds like you had someone come and install and train you?   While it is possible to eventually work out how the machines work without that. It sure is a lot easier to have some help!    Those  Assemblon / Yamaha machines are great.      And a decent printer.

Are you able to give us a rough idea of what it cost?

[kylehunter]

It definitely seems like our experience getting setup has been quite a bit different than most people have starting out. Especially after seeing all the horror stories that are in the pick and place lifestyle thread. It has of course been a massive learning curve, and still have a lot to learn, but it really has been quite smooth.

If you buy a junk machine, or a worn-out machine, then things may  not go smoothly.  If you buy a good machine that is still in decent condition, then everything OUGHT to go much better.

I bought a well-used Philips CSM84 (so a generation older than your Opal) but it was in good condition when it arrived.
I did not get any training with it, but did get an operator's manual.  I now have 3 different vintages of this manual, and it is obvious the first one was a HIDEOUS translation from Japanese to Dutch to English.  The later ones are a LOT better.

Anyway, the machine had their feeder and placement files still in it, I rigged up a serial cable and downloaded the placement file.  With the help of the manual, the file format was quite obvious.  I wrote a program to convert my CAD package's placement format to the Philips, loaded it into the machine, and made my first board less than two weeks after the machine arrived.  I had never even SEEN a P&P machine before then.

I went through a LOT of issues related to proper solder paste stencil apertures, what pastes are good and what are awful, etc.  And, over the last 12 years, I've had a few breakdowns on the P&P machine.  Only one was serious, all the others were pretty minor stuff (cracked hoses, an intermittent conveyor sensor, etc.)

My machine doesn't have the vision option, and I really DO miss that on fine pitch parts.

Jon

Yeah that's what I gathered in my research as well. It seems like virtually all of the horror stories that people spoke about were with old machines that hadn't been maintained, or bought online (eBay) without testing or servicing.

I'm planning on doing the same thing going from CAD to P&P. Does your machine use VIOS/Vios txt files as well? If so, could you send me the file format specifications that you have? I know most things in the Vios txt file, but not all. Especially linking the parts to the database parts.

Sounds like you had someone come and install and train you?   While it is possible to eventually work out how the machines work without that. It sure is a lot easier to have some help!    Those  Assemblon / Yamaha machines are great.      And a decent printer.

Are you able to give us a rough idea of what it cost?

Yeah we did. Luckily we live near a company called PFI of Florida: https://pfipcb.com/ (thanks to @reckless for the referral). They fully serviced/refurbished the machines, and did our training.

So for the printer, pick and place, oven, two straight conveyors, one turning conveyor, fume filtration system, delivery, setup/calibration, and 4 days of training it was around $50k.

In my research it seems like It was a touch high as compared to the machines themselves. But considering everything else that was included, it was more than fair in my book. Especially since none of use had ever touched an SMT line before, it would have taken months to get up to speed without them. Another great thing about them is that they offer support and maintenance on the equipment going forward. Since they are only an hour or so away, it is incredibly economical for them to handle all that.

What's crazy wrt service costs, is I reached out to Heller about some grease questions, and they wanted to quote me for a PM service. For one day of service it was quoted at like $4k including travel.

[mrpackethead]

I'm planning on doing the same thing going from CAD to P&P. Does your machine use VIOS/Vios txt files as well? If so, could you send me the file format specifications that you have? I know most things in the Vios txt file, but not all. Especially linking the parts to the database parts.

<quote>

Vios is a nightmare of a file format! Its really pedantic. I've written a bunch of python to generate it, but its still not perfect, and i still use the machines them selves to do the last step of the process.

in my generated vios files, i only go as far as 'naming' the part, i dont' try to set all the specifications for the individual parts.        As i've added new parts to the machine, i've added the to the machines component database on the machine..    After i load the generated Vios onto the machines,  I use the parts lookup function to let the machine do all the changes.

[jmelson]

I'm planning on doing the same thing going from CAD to P&P. Does your machine use VIOS/Vios txt files as well?

I think the VIOS file is for vision setup on the newer machines.  Mine has "UFOS", which has 3 parts.
There is a machine or "MCH' file for machine setup (basic machine configuration, offsets between cameras and nozzles, feeder and mech. alignment position, etc.)  There is a "FDR" file which defines the feeder type, size, and also trays.
And, then there's the mount file for each board, which has X Y R coords, feeder, and which nozzle to use.
If you have the vision hardware, there's another file.

If so, could you send me the file format specifications that you have? I know most things in the Vios txt file, but not all. Especially linking the parts to the database parts.

I'm not sure this will do you much good.  My machine is a FULL generation older than yours, and has no vision.
So, I'm sure the mount file format will be different.  If you have any placement file still in the machine, it should be easy to download that from the machine and pick it apart.
It will start with a header of half a dozen lines or so that define the starting coordinates of the board, where the fiducials are and their size, and then info on multiple boards within a panel, if set up that way.
Then, it will have M0 and a bunch of numbers.  Each line is one component to mount, and there will be a string of numbers at the end that are a coded field.  It starts with the nozzle number and ends with the feeder number.
Some of the digits in the middle specify the vision template to use, I think.

Anyway, you can open the mount file on the machine itself with the "data" screen, then the "mount" screen, and it will show most of the data just as it appears in the raw file.  You can read the files from the machine via the serial port with the command :
@READ B<ABC>
where ABC is the same board name as seen in the "data" screen.  You can send a file to the machine with :
@WRITE B<ABC>

You can read or write the feeder setup with :
@READ FDR
@WRITE FDR

and you can read the machine file with:
@READ MCH

Jon

[mrpackethead]

Yes, the VIOS files are quite different.    I've pmed kyle with some details. I reverse engineered VIOS a while back.  Its doable, but it is painful. 

Attached is a VIOS file

[jmelson]

Yes, the VIOS files are quite different.

Yes, this is a bit more readable than my UFOS files, but probably a lot more flexible, too.
It looks like maybe the feeder and placement info is in the same file?

On my older machine (CSM84) there are 3 totally separate files.  The MCH file just has machine configuration and offsets, travel limits, etc.  The FDR file has info on each feeder (location, tape size, number of pushes to advance by one component, whether to use the alignment station, and then a bunch more fields if that is a waffle tray (X and Y part pitch, X and Y number of parts per row/column, what to do on a mispick (bump or return to tray) and such stuff.

The board file has the board origin, fiducial locations and size, panel info if needed, and then the list of parts to be placed.

These files must be saved and loaded separately.  The machine can store about a dozen board files, but only one MCH and one FDR file.  That's all in battery-backed static RAM.

Jon
Jon

[Reckless]

Congratulations, $50k is decent with training/support and hopefully some warranty and loads of feeders(feeders are the most expensive part of the deal many times).  Assembleons (Yamaha GEM) are good machines.  My assembleon sapphire is older uses vios and has cameras but I haven't used it in production yet (need to get 8mm feeders).

Somehow in place of getting yamaha feeders I ended up getting a Fuji CP-642 turret machine gun chip shooter which I love (ran me $2k with 400 feeders, spares, accesories). My 0402 parts are finally going down fast.  For anyone placing hundreds of thousands of 0402s daily turrets are the way to go (40k cph). Still tweaking the nozzles/feeders as they weren't run in 5+ years.  We are using dry lube to coat feeders.

I have a DEK Infinity (similar to 265).  I like DEK, I got mine for $400.  I just got 3 MPM UP2500 DOS based, which I prefer but haven't had time to setup.  Not sure how prone to failure MPM is.  The ones I got are made of steel and seem to be working after 20 years just fine.  My DEK hard drive went down and it was a headache to fix.

[mrpackethead]

I can never remember what the nubmers were called;

Opal = YV100??
Saphire = ??

[kylehunter]

Congratulations, $50k is decent with training/support and hopefully some warranty and loads of feeders(feeders are the most expensive part of the deal many times).  Assembleons (Yamaha GEM) are good machines.  My assembleon sapphire is older uses vios and has cameras but I haven't used it in production yet (need to get 8mm feeders).

Somehow in place of getting yamaha feeders I ended up getting a Fuji CP-642 turret machine gun chip shooter which I love (ran me $2k with 400 feeders, spares, accesories). My 0402 parts are finally going down fast.  For anyone placing hundreds of thousands of 0402s daily turrets are the way to go (40k cph). Still tweaking the nozzles/feeders as they weren't run in 5+ years.  We are using dry lube to coat feeders.

I have a DEK Infinity (similar to 265).  I like DEK, I got mine for $400.  I just got 3 MPM UP2500 DOS based, which I prefer but haven't had time to setup.  Not sure how prone to failure MPM is.  The ones I got are made of steel and seem to be working after 20 years just fine.  My DEK hard drive went down and it was a headache to fix.

Thanks!

Yeah, forgot to add that with my previous post. It came with a couple hundred feeders. A lot of 8mm, with 20 or so 16s, 32s, and a few vibratory feeders as well.

[Reckless]

Sapphire is YV112
http://www.twentech.com/info/placementmachines.html

Which feeders did he include?  CL feeders are supposed to be much better than previous generation.

[jmelson]

It seems all the Gem series are 3-phase input.  What kind of initial power stage do they have?  Do they just rectify the mains and feed DC supplies, or does it have a 3-phase transformer and several rectifier stages?  I'm wondering, as I only have single-phase 240 V available here.  I'll keep running my CSM84 as long as I can, but I might eventually want to upgrade to something newer with vision.

Thanks,

Jon

[kylehunter]

Sapphire is YV112
http://www.twentech.com/info/placementmachines.html

Which feeders did he include?  CL feeders are supposed to be much better than previous generation.

Yep all CL. He threw in like 20 of the old style because he had a tough time selling them. Seems like every one wants the CL style

[mrpackethead]

You might see some 'FV' feeders as well, and some of them can be  interchanged with the CL.  However they will be very old..   The main source of drama with these PNPs is Feeders.   

I have bought clone feeders from ksunsmt out of CHina, and have found them to be really good. No problem with their product..  8mm feeder is around US$60, a bit less in quanity.  They also have lots of other spare parts, like nozzles and belts and stuff.    They have been a godsend for me, keeping my machines going.

[kylehunter]

2 month update!

So we have used the assembly line primarily with our one main client for LED boards. We have assembled around 600 of the smaller boards with around 80 LEDs each, and 20 of their really big panels, with upwards of 600 LEDs each.

The bad:
1) When people kept saying how important the printing portion of assembly was, they weren't kidding! First stencil had one LEDs apertures too big, had to have it remade. Our cleaner was adjusted too high and it hit the screen, had to recalibrate that. The rising table wasn't perfectly even, still need to fully calibrate that.
2) Our assembly line is in a small 500 sqft room, in hot and humid Florida. We really couldn't vent the fumes outside, or the negative pressure would bring inside tons of hot, humid, and dirty air. So we use a purex air filtration system, exhausting the filtered air inside, which works great. But, we can't keep the temperature under control. Wee have a 2.5 ton mini split, and now a 1.5 ton wall unit, and even with both fully maxed out, within a shift of 6 hours the temp goes from 72f-80f. Not good for paste, and this is with leaded, lead free will obviously be worse with the elevated temps. Not sure of a solution yet.
3) Feeders are annoying. Advice I wish I had known when starting, is if the pick and place misses the part on the same feeder more than a few times, and you can't figure out why, don't bother, swap out with a different feeder, and service that one later. Once we got a solid group of the CL feeders, they work flawlessly.
4) The Opal has a really really weird bug currently, in which e every other board we have to stop and restart the program, or it will misplace every part by around 5mm. It's really just a nuisance, as it only takes like 30 seconds extra each board, but you have to be vigilant, or you ruin a board.

The good:
1) Other than the one pick and place issue, that opal is an absolute tank. It just flat out works. We placed boards that are 20" long, it automatically optimized for the nozzles it can reach. We also do 30" boards by Flipping the board around, no problem at all.
2) When dialed in, the printer works well, and has the 2Di enabled. Haven't gotten around to testing it yet, but it seems nice.
3) Thanks to all the help here with the VIOS files, I have been able to start getting a system together to go from CAD - > Opal pretty fast. It's become quicker for us to use the pick and place for passives on our prototype designs (our main core business) than it is to hand assemble them. We then hand place the larger, less used components.

2 month conclusion:

There has definitely been a sharp learning curve due to the massive amount of "stuff" that is going on, but overall, the equipment has performed great. If I knew that this is how the equipment would have operated before starting, this would have been a no brainer to do. Hopefully we can figure out the AC issue, as it really is becoming one of the biggest issues we are facing, as the paste gets very difficult to use when the temps go up.

Biggest surprise is the sheer amount of stress that this has brought in. I may have been adequately prepared for the technical side, but I was not at all prepared for the mental side. Just the fact that the entire line can come to a screeching halt if one belt, pulley, heater, etc. breaks takes a toll on you. I assume it'll get better in time, but man, I have been struggling handling it sometimes!

So, I guess that's about it. Any questions/advice/feedback is always welcome!

[mrpackethead]

1) When people kept saying how important the printing portion of assembly was, they weren't kidding! First stencil had one LEDs apertures too big, had to have it remade. Our cleaner was adjusted too high and it hit the screen, had to recalibrate that. The rising table wasn't perfectly even, still need to fully calibrate that.

THis ^^ is so important.   Do not under-estimate this.  Put lots of effort into getitng your printing process nailed.  If your PNP is running well,  bad stencils will account for 95% of your problems.  It looks so simple...  but this is were so many people come unstuck.

3) Feeders are annoying. Advice I wish I had known when starting, is if the pick and place misses the part on the same feeder more than a few times, and you can't figure out why, don't bother, swap out with a different feeder, and service that one later. Once we got a solid group of the CL feeders, they work flawlessly.

This can also be releated to the paper tape and state of the palstic leader tape.

4) The Opal has a really really weird bug currently, in which e every other board we have to stop and restart the program, or it will misplace every part by around 5mm. It's really just a nuisance, as it only takes like 30 seconds extra each board, but you have to be vigilant, or you ruin a board.

That most def sounds like a setup/config issue and not a bug.   at 5mm it almost sounds like your origin is wrong.  The location pin is 0,0, which is 5mm up and 5mm right of where the boards bottom corner will be (  are your machine running left to right, or right to left?  mine are right to left ).

3) Thanks to all the help here with the VIOS files, I have been able to start getting a system together to go from CAD - > Opal pretty fast. It's become quicker for us to use the pick and place for passives on our prototype designs (our main core business) than it is to hand assemble them. We then hand place the larger, less used components.

[kylehunter]

That most def sounds like a setup/config issue and not a bug.   at 5mm it almost sounds like your origin is wrong.  The location pin is 0,0, which is 5mm up and 5mm right of where the boards bottom corner will be (  are your machine running left to right, or right to left?  mine are right to left ).

100% Agree, but it's still impossible for me to figure out why it is doing it with such consistency. Line goes left to right. Here's a post I made on another forum about it:

"
1) A board goes in, all components are placed perfectly.

2) Next board goes in, all components are around 5mm to the +x, and 5mm to the -y.

This was 100% consistent, every other board would be perfectly placed, and the others would not be.

The "fix" that we did is, we placed the first board correctly. The next board when it was clamped, we hit the e-stop, reset and dumped the parts on the head, and conveyed the board out. The next board would be placed perfect.

Any thoughts? It's just crazy how it is so perfectly consistent. Didn't matter if the bad board had it's fiducials looked at or not, placed any components or not, once we stopped it and reset, the next board always worked.

I don't have another board/panel to test with, so I don't know if this is an issue with the machine, or just my board program, but either way, I don't see a cause. It seems like there is some crazy setting that the machine is following, because it doesn't seem like a "bug", it seems like it is doing exactly what it wants to do.
"

[mrpackethead]

does your board have a location pin hole in it?

[kylehunter]

does your board have a location pin hole in it?

Like for a tooling pin? No, it uses edge clamps.

So we did 35 panels. Every odd board was perfect, every even board was wrong. So clearly it truly is a setting or something, since it is 100% consistent.

Also, on the bad boards, it checks for fiducials, and seems to find them fine

[mrpackethead]

Is the location pin, being activated when the boards are being shuttled in?    It might be that your boards are not not finding the end stop, and getting caught on it..  ( which would place it 5mm out ).   If you are using edge clamps for locating the board on the conveyor, then make sure you've got the location pin turned off in your program as well.   

I'd highly recommend using a tooling hole, and not rely on the edge clamps as its a lot more accurate, and reliable.  It does mean that your Pcbs have to be designed to include it.  For panels, i always put a 10mm tooling strip on each side.  This also helps me as i can set up the panel for stenciling a lot more accurately as well.

Does your panel have panel fids? or just board fids?

[kylehunter]

Is the location pin, being activated when the boards are being shuttled in?    It might be that your boards are not not finding the end stop, and getting caught on it..  ( which would place it 5mm out ).   If you are using edge clamps for locating the board on the conveyor, then make sure you've got the location pin turned off in your program as well.   

I'd highly recommend using a tooling hole, and not rely on the edge clamps as its a lot more accurate, and reliable.  It does mean that your Pcbs have to be designed to include it.  For panels, i always put a 10mm tooling strip on each side.  This also helps me as i can set up the panel for stenciling a lot more accurately as well.

Does your panel have panel fids? or just board fids?

If that was the case, either it would find the fids fine, and there would be no issue, or it wouldn't find them, and error out. The boards make it to the end stop even when it places wrong.

Panel and board has fids, only used panel fids, as it was plenty accurate.

Also, if anything with the tooling was goofy, it would be a random error I assume, not exactly every other?

[mrpackethead]

Also, if anything with the tooling was goofy, it would be a random error I assume, not exactly every other?

Very hard to tell without seeing it.  But i'm picking its something with your machines clamping.

[jmelson]

3) Feeders are annoying. Advice I wish I had known when starting, is if the pick and place misses the part on the same feeder more than a few times, and you can't figure out why, don't bother, swap out with a different feeder, and service that one later. Once we got a solid group of the CL feeders, they work flawlessly.

I got a homemade feeder stand from the guy that sold me the machine.  I added a bridge over the top with a crosshair right over the component location, and calibrated a bunch of my feeders to deliver the part to the same location.  This helped greatly.  Also, a little speck of dirt on the bottom of the feeder will throw it out of line.

4) The Opal has a really really weird bug currently, in which e every other board we have to stop and restart the program, or it will misplace every part by around 5mm. It's really just a nuisance, as it only takes like 30 seconds extra each board, but you have to be vigilant, or you ruin a board.

Wow, that is really annoying.  Is this on all programs, or just some of them?  I'd go over your programs carefully, looking especially at anything related to fiducial detection and panel coordinates.  it sounds like something is causing the fiducials to not be located properly.  You really want to have the fiducials in the middle of nothing on the board so it doesn't erroneously pick up some other mark as the fiducial.

I don't have vision, so my machine searches around with a single-pixel camera scheme to locate fiducials.  At least with that, you can actually see what it is doing while searching for them.  I guess on your machine it shows what it is doing on the screen, but it all happens very fast.

Jon

[jmelson]

So we did 35 panels. Every odd board was perfect, every even board was wrong. So clearly it truly is a setting or something, since it is 100% consistent.

Wait, every odd BOARD on each PANEL was off?  Are you sure your panel offsets are correct?  Maybe the board manufacturer screwed up the offsets, and every other one is different?  Oh, but that would show up with the stencil, unless the board house made that, too.

Is there any data recorded on the fiducial offsets that were taken?  As far as I know on my CSM84, there is no place to see the coordinates of the fiducials after a board is run.  That could be really helpful it diagnosing something like this.

Anyway, even on my machine, you can specify the fiducial diameter and the search area.  There is a cryptic note that specifying a larger fiducial than actually exists on the board improves accuracy.  Possibly some playing around with the fiducial parameters might improve registration of the fiducials.  there may also be brightness/contrast settings that can help.

Jon