SMT Production Line: Manual vs Automatic

[Reckless]

I am new to SMT Production Lines and was hoping to benefit from expert advice. 

I currently have but still need to install:
manual essemtec stencil printer
old juki laser pick n place x2
IR T962c reflow oven

Dreaming about a fully automatic setup where pcb goes in and completed pcba comes out.  The old jukis are running cheap because they dont have conveyors and are slow 4000cph.  The conveyor model runs $10k and is only twice the speed but 10x the price. 

Do I need to worry about manually handling a pcb array with 3000 smt parts when placing in desktop reflow oven?  Do the parts shift easily?  I have 75 QFN and 1250 0402 parts on 1 board which should take 1 hour to do.  I plan to run 25 boards a day.

If its not an issue, I would prefer to buy 10 small jukis.  Any insight appreciated.

[sokoloff]

(General engineering and life experience only)

Why not walk before you run? Particularly since the question is "should I expand from 2 to 10 manual placers?" My response question would be "why not run 100-200 boards before deciding?"

Nothing beats first-hand experience in your setup and market.

You have a bunch of equipment that's not yet installed and you're self-admitted dreaming of expanding. Concentrate instead on commissioning what you already have.

[KL27x]

Do I need to worry about manually handling a pcb array with 3000 smt parts when placing in desktop reflow oven?  Do the parts shift easily?  I have 75 QFN and 1250 0402 parts on 1 board which should take 1 hour to do.  I plan to run 25 boards a day.

The QFN and 0402 parts won't be any problem. Once they hit the paste, they're not moving anywhere. It's only much heavier and/or taller things that might be more challenging.

[mikeselectricstuff]

Manual handling is not an issue. Start with the minimum and see what the issues are before looking at upgrades.
My guess is you'll want to replace the oven first

[Reckless]

Definitely want to replace the oven already.  Not easy finding a small used non chinese machine.  Everything is generally really huge.  I want a tiny conveyor oven as our boards are small.  Preferably something set it and forget it.  Don't really need a pc on the oven.  Some of the chinese reflow ovens do look pretty. 

I did buy an EKRA E-1 SMT Vision Screen Printer Semi-Automatic machine this morning for $500.  Manual printing doesn't seem fun but we only need to do 20-40 board a day. 

I am ordering SMT components and boards and should be up and running in less than a month.  Trying to learn the Juki machine and need a few more feeders.

I was just contemplating the automatic pick and place units.  I am not buying anything more for a few months unless something is a huge bargain other than the oven.  We are walking before we run and want to see how much faster/easier things will be in a production environment.  The conveyor option may not be possible for us as we need to put 3 thru hole components on each board:  transistor, 18mm cap, smt relay or triac.  I was thinking to hand load those before putting in the oven.  Just worried that things will jingle/move on the paste.  Does anyone have experience with this?

I do know the larger machines can place relay/triac but 18mm cap is going to be hard and i dont think transistors aren't possible. 

[mikeselectricstuff]

I do have 3 thru hole components on each board:  transistor, 18mm cap, smt relay or triac.  I was thinking to hand load those before putting in the oven.  Just worried that things will jingle/move on the paste.  Does anyone have experience with this?

TH parts are often not  rated for reflow temperatures, so fitting before reflow is a bad idea. I've never seen an assembly house do this.

[Reckless]

How do they handle 3 through hole components?

I think the aluminum cap might be the only issue.  The transistor almost looks like an SMT part being fully encapsulated and relay/triac also looks designed for heat.

[coppice]

I do have 3 thru hole components on each board:  transistor, 18mm cap, smt relay or triac.  I was thinking to hand load those before putting in the oven.  Just worried that things will jingle/move on the paste.  Does anyone have experience with this?

TH parts are often not  rated for reflow temperatures, so fitting before reflow is a bad idea. I've never seen an assembly house do this.

The big issue in this context is through hole connectors. The plastic in many of these melts quite easily, as they were designed for wave soldering, where the plastic is shielded from the heat by the board.

[mikeselectricstuff]

How do they handle 3 through hole components?

hand solder, wave solder or selective solder

[ar__systems]

TH parts are often not  rated for reflow temperatures, so fitting before reflow is a bad idea. I've never seen an assembly house do this.

True but at the same time TH do exist for Pin-in-Paste installation, particularly connectors.

[Reckless]

TH parts are often not  rated for reflow temperatures, so fitting before reflow is a bad idea. I've never seen an assembly house do this.

True but at the same time TH do exist for Paste-in-Hole installation, particularly connectors.

What is paste in hole installation?  I have heard of this but don't know what it is. 

[ar__systems]

Sorry the correct name is PiP (pin in paste). Look it up, I saw white papers from some big name connector company. I used this technique to install some connectors.

[Reckless]

I have heard mixed feelings about Pin in Paste.  What is the general consensus?  Can it be used for caps and transistors? 

[ar__systems]

For caps it might be tricky, as for PiP you must have very short leads, barely poking through the other side. Also the paste needs to be applied from the bottom side (opposite to where the body of the cap is, so likely you will need a second run of stencil and reflow).  And if you cut cap's leads insertion might get a bit difficult. Another issue is that the cap will not be very stable until it is soldered, so handling will be difficult. And of course you need to find a cap rated for PiP.

Bottom line I would not try it for caps. Just use several proper SMT caps instead. Same for transistors.

[Reckless]

Unfortunately our parts have to be thru hole on those components.  Is wave soldering the only method then for large scale production?  Can you do wave soldering after SMT Reflow Oven?  We can hand insert the parts but concerned about how to do this faster than manual soldering. 

[ar__systems]

Can you do wave soldering after SMT Reflow Oven?

of course, that is the standard practice.

[jmelson]

I am new to SMT Production Lines and was hoping to benefit from expert advice. 

I currently have but still need to install:
manual essemtec stencil printer
old juki laser pick n place x2
IR T962c reflow oven

Dreaming about a fully automatic setup where pcb goes in and completed pcba comes out.  The old jukis are running cheap because they dont have conveyors and are slow 4000cph.  The conveyor model runs $10k and is only twice the speed but 10x the price. 

Do I need to worry about manually handling a pcb array with 3000 smt parts when placing in desktop reflow oven?  Do the parts shift easily?  I have 75 QFN and 1250 0402 parts on 1 board which should take 1 hour to do.  I plan to run 25 boards a day.

No problem.  As long as you don't drop the board on the floor, it is VERY unlikely any part will shift.  Umm, wait!  ONE hour per board, and you want to do 25 per day?  Will that really work?  (Ahh, below, you say more machines.)

3000 total parts?  So, about 45 minutes/board.  However, if the machine has multiple nozzles, it may be able to pick and place several components at a time.  My Philips can do that.  It doesn't double the speed of the machine, but does speed it up a bit.  Remember, you have to hand feed all this stuff, keep up with tapes running out and feeders jamming.  Maybe it is because my machine is quite "old school" but there is a LOT of fooling around required to keep it running.  Feeders jam, parts jump out of the pockets, nozzles get gummed up with paste, the machine can't find the fiducials because there's paste on them, and on and on.  Running one machine it is not such a big deal, but if I had several machines in parallel, I'd never have them all running simultanously! 

My Philips machine does have a conveyor, and I like it, but without a totally automated line it may not save a whole lot of time over hand feeding.

If its not an issue, I would prefer to buy 10 small jukis.  Any insight appreciated.

TEN Juki machines?  How many people will you hire to run all this?  Compute the stencil print time and the oven time with the P&P time and figure out the optimum throughput.  I doubt with one stencil printer and one oven you could keep 10 P&P busy.  Unless I'm misunderstanding your 1 hour build time, I don't think you need more than, MAYBE, 2 or 3 P&P machines.  This desktop oven is a BATCH oven?  Then, it will be even slower.  Probably runs close to 10 minutes/ board.  Unless you are REAL careful, you will have MUCH bigger problems touching the board until it has cooled than worrying about the board BEFORE reflow.

Unless I am misunderstanding what you are wanting to do, it sounds like a recipe for madness.  There are machines more suited for this level of production, such as some of the Assembeon GEM machines.  They have 8 to 20 nozzles, and go over to the feeder rail and pick up a full "rack" of parts, then go to the board and drop all of them in a couple seconds.  What I'm saying is that ONE machine that can plant a bunch of passives quickly may be a MUCH better way to go than a whole room full of slow machines that you have to babysit all day long.

Jon

[jmelson]

  The conveyor option may not be possible for us as we need to put 3 thru hole components on each board:  transistor, 18mm cap, smt relay or triac.  I was thinking to hand load those before putting in the oven.  Just worried that things will jingle/move on the paste.  Does anyone have experience with this?

Well, you have to be careful not to TOUCH any of the SMT parts when placing the through-hole parts.  That takes some CARE, but if the through hole parts are on the edge of the board, it is pretty easy.  Otherwise you might need tweezers and maybe a block to rest your wrist on when dropping it in.
I have a few parts on one of my boards that I put on after P&P, they are of a size/shape that is not compatible with my P&P machine.  No big deal for just a couple parts.

But, there is no issue with "jiggle", I pick up the boards after P&P, visually check alignment on the fine res parts, place the big parts and place on the oven rack.  As long as I don't bump the parts with my fingers, no problem.

You will have to arrange for the right amount of paste to be put on the pads for the through-hole parts.

Jon

[jmelson]

I do have 3 thru hole components on each board:  transistor, 18mm cap, smt relay or triac.  I was thinking to hand load those before putting in the oven.  Just worried that things will jingle/move on the paste.  Does anyone have experience with this?

TH parts are often not  rated for reflow temperatures, so fitting before reflow is a bad idea. I've never seen an assembly house do this.

Yup, I've (automatically) placed several thousand small aluminum electrolytics and then done 247 C reflow with lead-free solder.  I have had exactly ONE cap blow it's top.
For connectors, you have to make sure the plastic parts can handle the heat.

The data sheets ought to specify what parts can and can't take this.

Jon

[eeviking]

Pin in paste works great for connectors. We do a lot of Phoenix connectors and even 25pin d-sub as PIP.
But you will need step stencils to get more paste in the holes.

Don't know how big caps you need but you can get fairly big smt caps.

If its only a few leaded components a good alternative to wave solder is a dedicated selective solder machine.

For wave solder you will need to glue bottom side smt components or use pallets to cover them.

[jmelson]

Unfortunately our parts have to be thru hole on those components.  Is wave soldering the only method then for large scale production?  Can you do wave soldering after SMT Reflow Oven?  We can hand insert the parts but concerned about how to do this faster than manual soldering.

There is also dip soldering and selective solder.  (Selective solder machines are going for CRAZY high prices right now.)  This is where a pump pumps up a fountain of solder and the board is moved in XY patterns above it to solder through-hole parts.

If you have no parts on the back of the board, and the TH parts go on the TOP side, you can do dip solder.  This is where you have a large solder pot and you dip the board onto the molten solder.  I have a roughly 3" square board with ** 40 ** through hole parts!  I have a 5 x 7" Chinese solder pot.  I mask two connector positions that have to be hand-soldered to the bottom of the board later, flux the board and place all the parts by hand (UGH!)  Then, I preheat the board on a 100 C hotplate for 2 minutes, pick up the board with a custom gripper and dip it into the solder pot for 10 seconds.

It took me a little while to get the technique down, but it now works very reliably.

If these TH parts go on the bottom, then selective solder is probably the only way to do it.

Jon

[jmelson]

Unfortunately our parts have to be thru hole on those components.  Is wave soldering the only method then for large scale production?  Can you do wave soldering after SMT Reflow Oven?  We can hand insert the parts but concerned about how to do this faster than manual soldering.

For THREE components per board with only a few pins each, just hand soldering them is NOT a big deal.  Not in comparison to 45 minute/board P&P time.

Jon

[Reckless]

Unfortunately our parts have to be thru hole on those components.  Is wave soldering the only method then for large scale production?  Can you do wave soldering after SMT Reflow Oven?  We can hand insert the parts but concerned about how to do this faster than manual soldering.

There is also dip soldering and selective solder.  (Selective solder machines are going for CRAZY high prices right now.)  This is where a pump pumps up a fountain of solder and the board is moved in XY patterns above it to solder through-hole parts.

If you have no parts on the back of the board, and the TH parts go on the TOP side, you can do dip solder.  This is where you have a large solder pot and you dip the board onto the molten solder.  I have a roughly 3" square board with ** 40 ** through hole parts!  I have a 5 x 7" Chinese solder pot.  I mask two connector positions that have to be hand-soldered to the bottom of the board later, flux the board and place all the parts by hand (UGH!)  Then, I preheat the board on a 100 C hotplate for 2 minutes, pick up the board with a custom gripper and dip it into the solder pot for 10 seconds.

It took me a little while to get the technique down, but it now works very reliably.

If these TH parts go on the bottom, then selective solder is probably the only way to do it.

Jon

Since it's 3 parts x 80 boards =120 components, the DIP method sounds like it will take might work.  Everything is on the same side.  We are not populating the back of the pcb. 

[Reckless]

TEN Juki machines?  How many people will you hire to run all this?  Compute the stencil print time and the oven time with the P&P time and figure out the optimum throughput.  I doubt with one stencil printer and one oven you could keep 10 P&P busy.  Unless I'm misunderstanding your 1 hour build time, I don't think you need more than, MAYBE, 2 or 3 P&P machines.  This desktop oven is a BATCH oven?  Then, it will be even slower.  Probably runs close to 10 minutes/ board.  Unless you are REAL careful, you will have MUCH bigger problems touching the board until it has cooled than worrying about the board BEFORE reflow.

Unless I am misunderstanding what you are wanting to do, it sounds like a recipe for madness.  There are machines more suited for this level of production, such as some of the Assembeon GEM machines.  They have 8 to 20 nozzles, and go over to the feeder rail and pick up a full "rack" of parts, then go to the board and drop all of them in a couple seconds.  What I'm saying is that ONE machine that can plant a bunch of passives quickly may be a MUCH better way to go than a whole room full of slow machines that you have to babysit all day long.

Jon

Jon, you are right.  I am used to having alot of backups for old machines.  I have been finding these machines for $1-2,000 and I hear they are workhorses but they are SLOW.  I do want a fast machine instead of many slow machines BUT I look at these newer SMT machines and they seem soo complicated.  Reminds me of Mercedes, back in the 80s they were simple to understand and work on.  Now try working on a newer german car parts are outrageous, you are married to the dealer, everything is soo complex that it requires a heart surgeon to diagnose problems.  In the end you drop $60,000 repairs for a $30,000 car (been there, done that).  I like to go back in the day when things used to be simple.  I feel like you give up reliability and peace of mind for speed and unnecessary features.  That's the one thing I don't want to lose peace of mind.  I will pay extra for that.  If a machine goes down its on my head whether to call an expensive technician or buy another machine.  With a number of older machines that go down I can call repair man when I have had 4-5 go down and get them fixed all at once.  Anyway, for now this is speculation until I get my machines up and running.  I am looking for the bulletproof, never dies Land Cruiser of placement machines. 

I wish chinese machines could properly clone these old juki machines.  I was seconds away from pulling the trigger when my friends reminded me of the chinese curse.  I don't feel they will work in a production environment for more than 6 months. 

[jmelson]

Well, some of these Assembleon machines are not insanely expensive on the used market.  They are 10 years old.
Emerald is 2 nozzles, so that doesn't do you any good.
Opal is 4 nozzles.  Sapphire is at least 12 nozzles, seems to have 2 heads, can do 2 boards at once.  But, it may not be able to handle larger chips.

Jon

[Reckless]

I don't know anything about the Phillips line of products.  How do they compare to the Jukis?  What about aftermarket support for parts/service?  My biggest chip is 5mmx5mm so not a problem there.  The machines look big, remind me of the Universal machines.  I don't mind big if it makes sense.  If I can fit 2 Juki machines running 4000 cph in a place where one other machine is doing 5000 cph doesn't make sense to me unless the 5000 never goes down. 

[jmelson]

I don't know anything about the Phillips line of products.  How do they compare to the Jukis?  What about aftermarket support for parts/service?  My biggest chip is 5mmx5mm so not a problem there.  The machines look big, remind me of the Universal machines.  I don't mind big if it makes sense.  If I can fit 2 Juki machines running 4000 cph in a place where one other machine is doing 5000 cph doesn't make sense to me unless the 5000 never goes down.

My experience is VERY narrow.  The only machine I've ever run is my Philips CSM84 (Made by Yamaha, they sell the same machine as, I think, the YM84.)
The CSM84 and others of its vintage were truly built like army tanks.  The nozzles are simple steel tubes, I made my own with a lathe.  Some other machines have very fragile nozzles with complex shapes that would be harder to make.  The up-down is driven by air cylinders, although there was a Z-axis option available.  The CSM84 is 5 x 7 feet, but seems like about the size of many other machines in that class.  Some Mydata and Universal machines are really LONG!  Well, you would not want to commission a CSM84 NOW!  They are just too old.  (I've been running my used machine for slightly over 10 years, now.)  I broke a few of the air cylinder extension rods that trip the feeders when I changed some timings.  I just made replacements on my lathe.  It seems that some Chinese outfits have EVERY part you might need to keep one of these machines running, but you have a bunch of issues getting parts from there.  I have gotten several electromechanical parts and sensors from PLC Supply, and hoses and such from eBay.  I most CERTAINLY do not have a service guy come out when something breaks.  I have to fix it myself.  The most recent issue was built-up commutator dust shorting out the rotation motor.  That took a while to diagnose.

The big machines have space to hold more feeders.  You will be surprised how many feeders you need on one board.  Now that chips are getting hard to find in tubes, you have to get them on tape, and they put them SIDEWAYS in the tapes.  So, instead of the SOICs all lined up the long way in  a tube, they are the wide way in the tape, requiring a wider feeder.  At least on my CSM, anything wider than 8mm tape takes up TWO feeder slots! 

Jon

[Reckless]

Just for my knowledge, does anyone know how the older Juki 760 machines compare to the Phillips Assembleon? 

The guy who was trying to sell me Universal Instrument GSM equipment admitted that Juki's are a really good product and couldn't say anything bad about them. 

[nisma]

Do you know the limits of the machine ? as example on the mentioned juki you cannot have more then 2000 placement and not more then 100 different components and the nominal speed is from memory 3600 cph, calculate 3100cph including nozzle change and pcb change.
In this case, this probably mean, that one pcb must be assembled on two different machines and you must split the assembly and assembly time between two
machines.
Depending on solder paste used, you could stencil pcb for 2 days, in your case 50 pcb. I don't know how long you want work, in 8 hours you maybe do 8 boards,
16 with two machines.
I suggest doing a dry run of 4 boards and check including timing everything what is the effective cph you archive including nozzle changes and pcb changes.
In case you need to do 25 pcb/day and archive 16-20 boards / day this could be a problem. In this case you need a third machine.
Further probably on that numbers, when doing 8h runs, your are busy 4h keeping assembly running, this all as guess.

 

[jmelson]

Yes, until you have RUN a P&P yourself, you really have no idea of what is involved.  Maybe some other machines work better than what I have, but I kind of doubt it.  0.1uF caps give me the WORST fits!  Resistors are all fine, 100 pF and 1000 pF never a problem, but 0.1 uF from 5 different vendors ALL have the same feeding problems.  Crazy! The cover tape glue is applied with some kind of roller, and the glue application is uneven.  So, every foot or so, it goes from practically zero adhesion to so much glue that it rips the top layer off the paper tape!  It is really HARD to adjust  a feeder to handle this variation.

I run a very small scale operation, so I have reels that are several years old.  (How many places do you use a 39.2K Ohm resistor?)  So, some of the tapes go bad over time.  On some of them the cover tape gets harder to pull, or so brittle it snaps, on others the glue fails completely, and all the parts spill on the floor.  (This is mostly ICs and inductors, never resistors.  Yes, the more expensive the part, the worse it gets.)

So, the machine works great, if only I could get tapes with consistent glue.

Just wanted to mention that you have to factor all this into your manufacturing time calculations.  I can't just let my P&P sit there and run, I have to be right on top of it to keep up with this feeder stuff.

Jon

[coppice]

Yes, until you have RUN a P&P yourself, you really have no idea of what is involved.

True. I suspect some people think its like buying a printer for their PC. 

[mikeselectricstuff]

Yes, until you have RUN a P&P yourself, you really have no idea of what is involved. 

Absolutely.
P&P is not one big problem, it's lots and lots of small and subtle ones, any one of which can bite you in the arse, often in unexpected ways.

[Reckless]

I'm sure pick n place machines have alot of issues and a long learning curve.  Lucky for me I have few parts, small boards, and mostly standard components.  I do have 3 thru hole components per board but worst case I hand solder them.

I do hope to get nicer pick and place machines eventually that don't give me any real headaches even if I have to spend $20k extra for them. 

[DerekG]

So, some of the tapes go bad over time.

Yes, I know what you are going through. Try & store the tapes in a cool dark place. The optimum humidity seems to vary with the glue used. Some glues "fall apart" in humid environments (it is like the moisture in the air breaks them down) & other glues either fall apart or set rock hard under low humidity conditions.

All I can advise is that if the glue falls apart in humid environments, place the component roll firstly in an anti-static bag (do not seal this bag), then place this in a second double thickness (200 micron) plastic bag with a packet of silica gel in it. We store the silica gel in the fridge which over about 10 days sucks out most of the moisture in it. It is important to immediately wrap the silica gel in a thick plastic bag when taking it out of the fridge to prevent it absorbing moisture from the air (remember it is cold). Once it is up to room temperature, remove it from the plastic bag & place it in the 200 micron bag above.

You can keep using the silica gel indefinitely using this method.