Bringing PCB assembly in-house with Dima Optimat?

[Styno]

Since I found this forum most helpful, I registered to ask some questions as well as try to contribute something in return. So here’s the story which may sound familiar to many of you:
I work for a small software company which core business is providing networking access and monitoring/control of remote industrial installation, from few tens to many tens of thousands I/O. Despite being primarily a software/system integration company, we always deploy some local data processing using a combination of off the shelf low power ARM platforms (nowadays called fog computing in marketing terms) and custom interfacing pcb’s. Assembly of these has always been outsourced. But with the rise of (sorry, another buzzword) IoT the volume and diversity of in-house designed electronics has sharply gone up. Plenty of meetings with management later and it is decided we in-house the pcb-assembly.

There are many reasons, money far from being the #1:
Contract assembly can be a bitch and take a lot of time and effort documenting. Especially with some of the weird glue-logic constructions we come up with  Also, the turn-around time is often long (8 weeks+) because we do small batches: 200 pcs is considered a large order for us but not for them. Being able to produce a single/few panel when needed without too much hassle is a huge benefit. For over 20 years we’ve used an Essemtec manual pick and place together with an Essemtec IR batch reflow oven for prototyping and small production runs. But volume is starting to go up sharply and working on the manual PnP all day, days on end is no fun.

So, based on the experience/advise from members of this forum we decided to dip our toes in and shop around for an second hand PnP instead of a new cheap Chinese, stencil printer and convection reflow oven. A local broker advised:
- In-line half-automatic stencil printer. Mainly for repeatabilty and the option of using a less skilled operator. And because half-automatic is not that more expensive then guided-manual.

- In-line Dima Optimat SMPM-2000 PnP machine. This machine from 2002 has top and bottom vision and room for ~105 8/12mm feeders. It should be able to place 0402 and 0.5 pitch components. Dima’s are said to be good for high mix, low volume and therefore easy to setup (e.g. Teach-in function). It’s mechanics are simple and robust. The vision system and other components are widely used and still available. The UI is DOS based (under WinXP) and easy to use, e.g. Local language (Dutch) is selectable. The feeders are mechanical and dirt cheap (€30 on Ebay) so I plan to get as much feeders as possible and have the components for the most designs permanently on the machine. A review of 6 pcb designs learns that ~70% of the components uses 8mm tape and that about 60 slots are needed, so there is still a little room left for more designs.
According to the broker this particular machine has seen few running hours.

- In-line 5-zone wire-mesh conveyor full convection oven (4 heating, 1 cooling? Not sure about that).

I expect the quote to come in at about EUR 25k.

We are aware that the PnP is dead slow with a gross rate of 2000 cph however approximately 1 or 2 panels per hour is much faster than manual work anyway. This is enough for us for the moment. I reckon that, when we we are lucky enough to need more capacity or improved handling, the experience with this ‘cheap’ setup will have paid for itself and tought us a lot. Plenty of space in the workshop, so size is not a problem.

Any of you have experience/tips for this particular PnP? What are it’s weak points besides being slow, what should I ask the broker or check before buying? When the deal goes through I hope to be able to share the experience as well.

Edit:
Dima Optimat SMPM-2000 datasheet.pdf

[SMTech]

I looked very closely at Dima (New) before going Essemtec, and they seemed a good match for what you are describing, however I  was looking at the HP-100 & MP-200 machines. At that time I did notice several older Dima models like the Optimat come on the market and they were always very cheap and that was 7 years ago. The feeders were quite a nice compact design that seemed to be used on their newer platforms.
 25k all in for a line isn't terrible and Dima being a local company to you probably means you have better access to help and parts than people elsewhere, when I was looking my biggest concerns in the UK were the very low installed base in the UK and the fact Dima themselves were very dispensing obsessed - it was all they promoted at events etc. Behold today Dima is gone, slurped up into Nordson, who have kept.. dispensing.

Personally however I would go with a used machine where there is more support and maybe even a use for your feeders when you decide the machine needs to be quicker, something like a MyData or Europlacer (who make machines for YOU) or one of the Yamaha made machines.

Seeing as I run one I can also recommend the newer Essemtecs for the work you describe, there's a Paraquda listed on adoptsmt right now....

Printer, semi-automatic can mean so many things, personally I would either go inline fully automatic (built like tanks lots of them out there, hopefully with extras like SPI), or offline and choose something like the Reprint Miniko. None of this mylar film rubbish.

A 4 zone oven will probably do you fine for most things, more zones gives you more control if you are targeting a very specific profile and helps get heat into big awkward parts, more length gives you more throughput. If you know now you use big heavy parts or big heavy/ceramic boards, maybe you might want something beefier. Also worth considering are small Vapor phase ovens, these take up very little space & have excellent heat transfer., the downsides are they can emphasise poor stencil/footprint with a higher risk of tombstoning, and they are not Electrolytics best friend.

As a subcontractor we run jobs in exactly the sizes you describe, the only time we would normally quote a lead time in the 8 week region would involve metalwork or sourcing issues. Bigger places do bigger jobs that require more planning, some have lots of knowledgeable people others have just one or two and a room full of monkeys, that can mean the wait is actually for the overworked non-monkey to set your job up.

[mrpackethead]

I'm about 3 years down the road from the place you are today.  THe best thing i ever did was stop relying on CM, and brought it in house.   I've gone Yamaha ( now have 2 Yv-100iis) and will be adding a third slightly newer machine..    I've also had chinese desktops and i'd just say avoid them, it will just result in pain.      I'm using Vapour Phase and its great.   I dont' have too much problem with electrolytics,  ( we dont' use lots though ).. Tombstones are a matter of good solder paste application..  dont' underestimate solder paste application.. that is where 90%+ of your issues will start.

[Styno]

I looked very closely at Dima (New) before going Essemtec, and they seemed a good match for what you are describing, however I  was looking at the HP-100 & MP-200 machines. At that time I did notice several older Dima models like the Optimat come on the market and they were always very cheap and that was 7 years ago.

First: thanks for your thoughts! The local broker also has an MP-200 machine available and initially I had my eye on it as it's more capable on all accounts but he did not recommend it for us because it's more complex and needs more maintenance. E.g. he said the servos need periodic calibration that can only be done by a trained technician.

The feeders were quite a nice compact design that seemed to be used on their newer platforms.

As I understand, the newer machines use intelligent feeders where the Optimat uses head activated pure mechanical ones. The intelligent feeders are an order of magnitude more expensive and, since we will try to keep our components on the machine at all times, the intelligence does not add much value for us.
 

25k all in for a line isn't terrible and Dima being a local company to you probably means you have better access to help and parts than people elsewhere, when I was looking my biggest concerns in the UK were the very low installed base in the UK and the fact Dima themselves were very dispensing obsessed - it was all they promoted at events etc. Behold today Dima is gone, slurped up into Nordson, who have kept.. dispensing.

Personally however I would go with a used machine where there is more support and maybe even a use for your feeders when you decide the machine needs to be quicker, something like a MyData or Europlacer (who make machines for YOU) or one of the Yamaha made machines.

I'm not entirely sure but i have the feeling there is still expertise around for the Dima machines, but this may be specifically a Dutch situation. I'll ask about this next week.
60x 8mm feeders is about $1800 @ Ebay prices. So, not being able to reuse them on a possible future machine is not a big loss. The vibration and tray feeder are a bit more expensive though at ~$600 each...

Seeing as I run one I can also recommend the newer Essemtecs for the work you describe, there's a Paraquda listed on adoptsmt right now....

Yeah, I drool over such a machine but, alas, I reckon this one will come in over twice our current investment for the PnP alone? Especially with a bunch of feeders included. Pretty sure that won't pass upstairs at this stage. Sure, when assembly is your core business the situation would be quite different.

Printer, semi-automatic can mean so many things, personally I would either go inline fully automatic (built like tanks lots of them out there, hopefully with extras like SPI), or offline and choose something like the Reprint Miniko. None of this mylar film rubbish.

We use stainless steel stencils. The printer in question is appropriately named    SMTech AVP300. Not sure what 'SPI' means but it has fiducial recognition.

A 4 zone oven will probably do you fine for most things, more zones gives you more control if you are targeting a very specific profile and helps get heat into big awkward parts more length gives you more throughput. If you know now you use big heavy parts or big heavy/ceramic boards, maybe you might want something beefier.

We build primarily small compact low power stuff, the biggest items being thin RF modules (2 by 3 cm max) and Ethernet jacks. I hope it's a significant step up from our IR batch oven which heats panels not entirely even and cannot follow lead-free reflow curves quickly enough. I guess anything has higher throughput than the PnP. I'm even considering stacking the assembled boards and than bake them at the end of the day, avoiding keeping the oven on but idling most of the day. But not sure if that is worth the effort vs kWh's saved.

Also worth considering are small Vapor phase ovens, these take up very little space & have excellent heat transfer., the downsides are they can emphasise poor stencil/footprint with a higher risk of tombstoning, and they are not Electrolytics best friend.

The broker had a few IBL batch VP ovens on hand but did not recommend them, primarily because we apparently do not really need that and the significant price hike over an inline convection oven.

As a subcontractor we run jobs in exactly the sizes you describe, the only time we would normally quote a lead time in the 8 week region would involve metalwork or sourcing issues. Bigger places do bigger jobs that require more planning, some have lots of knowledgeable people others have just one or two and a room full of monkeys, that can mean the wait is actually for the overworked non-monkey to set your job up.

Thanks for that insight. Yes, our current contractor is quite large and recommended to us by another company that does much larger volume simply because that is their core-business. I'll reconsider other contractors as well.

Thanks again for your input. I hope my reaction makes sense as apart from manual assembly this area is quite new to me.

[SMTech]

From what I understand the AVP300 to be, seems like a good choice to me, and I think they still supported/refurbished by Reprint, who possibly made it (there is a link of some kind).

Mylar is technique used by many semi-automatic printers, basically you put the PCB in its fixture, put a mylar screen over it and print. Typically there is an area outside the print area, so you slide the fixuture out, now you finely tune your alignment of your mylar print onto the PCB either by eye or with the assistance of cameras. Once your happy, remove the mylar and print away. A PLC controller handles the squeegee speeds pressures etc.

SPI is solder paste inspection, or AOI for your printing, some automatic printer have a camera fitted that can do this at some level.

[Styno]

We won't run this line full stop or even every day, certainly not the first year. So I was wondering:
- do you have to clean such an (semi) automatic printer after each use? I guess the simple answer is yes.
- so how do you clean the machine? Lots of alcohol and paper towel? There are two wiper blades closely spaced, it seems a lot of dirty work to get that clean...

Currently, after manual printing we wipe the stencil clean on a flat table using paper towel with alcohol or petrol, then lift the stencil to clean the residue off the table and repeating this process until it's judged to be 'clean enough'.

Any tips?

[mrpackethead]

I wash our stainless steel stencils with a PCB board wash solution, which does a really nice job of it.     ( the solvents in teh paste is not Alcohol soluable  ( henkel GC10 )..  i give them a first scrub down, with that and hot water, and then they get put in a ultrasonic bath for 10 minutes, to get the last little bits out.    They then get air dryed..  I'd do that after the days effort is done,     Your stencil printer may have a clean stroke as well.

If you are making lots of the same product, its well worth getting two stencils, ( or even three ).  If you damage one ( it does happen ), and you can have one being cleaned while one is on the machine.   These days with stencils being only $40-50 is not overaly expensive

[Gary.M]

Sodium percarbonate in hottish water will clean your stencils super clean, no ultrasonic needed. Very cheap too.

[SMTech]

You should only need to clean the stencil and the blades, the blade assembly normally comes out very easily for that purpose. If you've managed to get paste anywhere else I think something would have gone wrong.
I clean my stencils using a similar procedure to you, good lint free towels and stencil cleaner formulated by our paste manufacturer BLT, it is IPA based with some extra secret sauce.

 In the past we have used some generic stencil cleaner from Farnell and label remover (the stuff that smells a bit like oranges) is also surprisingly good at it. This isn't perfect, it is quite hard to flush everything out of apertures, proper stencil cleaning machines are expensive beasts (not least because commercial stencils are huge). I have other toys on my shopping list I would choose first given the budget. An improvement on this process would be a handheld ultrasonic like this http://www.gen3systems.com/gensonic-stencil-cleaner one of those is probably IRO £1500 but I keep meaning to have a go at building one using a £40 ebay transducer&driver

[mrpackethead]

http://www.chemtools.com.au/product/electrical-electronics/electronics-cleaning-chemicals/pcb-flux-remover/


^^ This is what i use. I'm not entirely sure what it is.. 20L drum cost sabout NZ$120...     

[Styno]

Thanks all for these great tips!

I've got a quote for an alternative to the convection oven: and IBL VPP-M 400 vapour phase batch reflow oven. It's a bit over €5k and about ~€1k more than the convection. But a) we don't need a fully inline oven, b) de VPP should be quite a bit less smelly and hot and more energy efficient as well as use less space, c) I'm attracted to the potentially better solder joint quality. So we'll see if that can be squeezed into the budget.

Another question:
How do you teach a new pcb on the PnP? Do you add paste and just try, manually correcting afterwards? I see that many of those Chinese promotion youtube video's use double sided sticky tape or perhaps spray glue. Since this oldie doesn't project a virtual component over the pcb image I assume I will need multiple test runs before the angles and offsets in the PnP library fit the CAD libraries.

[olkipukki]

re: stencil cleaning
I am using stencil wipes https://www.electrolube.com/products/cleaning/ssw/wipes/ 

[mikeselectricstuff]

For test placements,use low tack double-sided thin-film adhesive tape.you need the stuff that's fairly translucent so you can see the pads.
Apply with minimum pressure so it comes off easily-just lay it on.
May need to find a few types to get one that comes off cleanly.
Remember you usually won't need to cover the whole board, just a few places for one of each part. I find 25mm is a good compromise between coverage and ease of removal

[Ice-Tea]

My 0.05$: having to provide lots of documentation for a contractor is not a valid reason to go inhouse. The documentation should roll out easily out of whatever version control/managment system/CAD implemenation you have. If that is not the case, my guess is going "documentation-light" is going to bite you in the ass sooner or later.

[mrpackethead]

My 0.05$: having to provide lots of documentation for a contractor is not a valid reason to go inhouse. The documentation should roll out easily out of whatever version control/managment system/CAD implemenation you have. If that is not the case, my guess is going "documentation-light" is going to bite you in the ass sooner or later.

I say it is a good reason.   I have invested substaintial effort in our system, so that everythign is consistent end to end.   Contractors don't know your systems, and you dont' know theres.      Its not about 'not' having the info, its about understanding it.    And you never quite know when the 'understanding-light' is going to bite you in the proverbial;

[mikeselectricstuff]

My 0.05$: having to provide lots of documentation for a contractor is not a valid reason to go inhouse. The documentation should roll out easily out of whatever version control/managment system/CAD implemenation you have. If that is not the case, my guess is going "documentation-light" is going to bite you in the ass sooner or later.

Depends on the nature of the business. For me it is very rare that I need to build a second batch of a job, so time spent documenting would be way more than time figuring things out on the rare occasion I need to re-run a job.
Chances are you'll need to sub some parts due to availabilty, maybe make other changes, so it still takes time regardless of how well documented it was originally. My pick&place has easily paid for itself from time saved dealing with subcontractors alone.

[Styno]

Again, all thanks for your valuable thoughts.

I'd like to reply to this:

My 0.05$: having to provide lots of documentation for a contractor is not a valid reason to go inhouse. The documentation should roll out easily out of whatever version control/managment system/CAD implemenation you have. If that is not the case, my guess is going "documentation-light" is going to bite you in the ass sooner or later.

In my first post I tried to explain that there are several reasons to go in-house, communication problems just being one of them. I hope there is no confusion about this.

But anyway, I'll give an example: I've developed a 2-layer interface board to fit a certain small round housing. In my CAD board drawing I only made the outline round, the two copper layers are drawn as a square around the board. The CAD renderer and every online and offline Gerber viewer I tried shows the board correctly with the copper reduced to the outline of the board with some pull-back. Using the iterate often and quick method I have several versions of the board made through three board manufacturers in China. All came back perfectly as I intended, no questions asked.
So all is well and, finally, it's time for a series production run. I send the same Gerbers with assembly instructions and pictures of the half- and final assembly to a big contractor in the Netherlands. The first response I get back is: your copper pour is square but your board is round, we can't make this. 

I don't want to spend time on that sort of communication.

[Styno]

Depends on the nature of the business. For me it is very rare that I need to build a second batch of a job, so time spent documenting would be way more than time figuring things out on the rare occasion I need to re-run a job.
Chances are you'll need to sub some parts due to availabilty, maybe make other changes, so it still takes time regardless of how well documented it was originally.

Partially, this is how our business works as well. There are some (nearly) one-off projects, however there will also be some products that get sold for many years (10+ years) in quantities of a few hundred to a few thousand a year.

My pick&place has easily paid for itself from time saved dealing with subcontractors alone.

I've got one order in the pipeline for about 200 boards that will pay for the hardware already, after that we hope to enjoy to produce many more orders on the same machines. So, the hardware purchase isn't the problem.

The unknown is: can we get producing quickly after we receive the machines (i.e. 1 to 2 weeks)? The boards aren't that demanding regarding passives size or chip pitch so that will help. The machines are serviced and delivered in working condition before delivery, so that helps too. The broker will also provide a one day training and phone support if we run into issues. Finally the broker has visited us and we talked for quite a while. He told us: the Dima is so simple, and because we already have experience in manual PCB production (i.e. check for correct orientation of diodes and basic stuff like that), and our workshop doesn't need much preparation, that he expects us to easily be able to get boards populated in that timeframe.

I don't have the experience yet to judge if he's correct but we'll see. I've been told that next week the green or red light will be given from management.

Edit:
The broker also told me he expects us to use the PnP for prototypes as well with a little bit of CAD export scripting. Also because the PnP is so simple. Again, we'll see if he's correct. I've asked him if he could provide me a contact who actively uses the same machine so I can get a users perspective.

[mrpackethead]

I would suggest 1-2 weeks is very very optimistic.

[Styno]

Yes, rightly so I guess. I'm sceptical as well but that's purely based on the experiences from this forum. From you, rx8pilot, Mikeselectricstuff and others.

While rx8pilot's story sounds like a horror movie for a commercial company that has work to get done instead of fixing broken PnP's, the machine we'll get is in good working order and has few running hours. E.g. the feeders that were on the machine looked like brand new.

There's likely a difference between the time getting the first board populated vs being able to produce boards without too much operational problems and manual correction afterwards. I hope to be able to comment on that later 

[Ice-Tea]

But anyway, I'll give an example: I've developed a 2-layer interface board to fit a certain small round housing. In my CAD board drawing I only made the outline round, the two copper layers are drawn as a square around the board. The CAD renderer and every online and offline Gerber viewer I tried shows the board correctly with the copper reduced to the outline of the board with some pull-back. Using the iterate often and quick method I have several versions of the board made through three board manufacturers in China. All came back perfectly as I intended, no questions asked.
So all is well and, finally, it's time for a series production run. I send the same Gerbers with assembly instructions and pictures of the half- and final assembly to a big contractor in the Netherlands. The first response I get back is: your copper pour is square but your board is round, we can't make this. 

I don't want to spend time on that sort of communication.

I'm having trouble picturing what those Gerbers would have looked like but in general: I prefer a supplier that asks questions over one that guesses what my intentions could have been. In this case it may have been an annoyance, in other cases it could have saved your bacon.

I'll add an example: I once had an error in an outjob file, did a quick revision on holliday while people were litteraly waiting on me at the door to finish up. The 4 layer board left with top and bottom gerbers and 2 blank inner layers. They produced it like that, shorted everything out. I would have prefered someone asking me a question at that point instead of loosing a wad of cash, the assignment and face.

But I understand that your scenario may be different.

[Styno]

It has me wondering what is in the Gerbers as well but, really, I'm not going to investigate. Why should I? The point is: at least 5 different online and offline Gerber viewers show the board correctly as I intended, none had problems with my data and three different board manufactures make the board as intended. So why should I question my design or wonder whether the Gerber data is right or wrong?

But 'your Gerbers are wrong!' is the first response I get when trying to outsource assembly... They may very well be 'technically' correct, but I gave them enough information (photo's, assembly description) to work it out themselves. Not meaning to offense anyone but I already have an autistic colleague, I don't pay for a contractor to be one too!

Edit:
Re stencil cleaning: Only just found out that simply washing the stainless steel stencil with running tap water cleans it perfectly.

[mikeselectricstuff]

I would suggest 1-2 weeks is very very optimistic.

For a known-working machine with dealer support it's probably quite doable if you don't have much else to do
Just make sure you have PLENTY  of spare parts and PCBs, and make more than you need to deliver soon

[SMTech]

Well I will beg to differ on the questions front, questions are good, it means they are paying attention Just because a handful of free tools that are probably all powered by exactly the same set of open source libraries APPEAR to render your files as you seem to want them doesn't mean they aren't in someway invalid or will fail some facilities automated DFM checks. What you should do with feedback like that is work out why the way you are doing it is wrong and fix it. The only clients I we deal with who are pretty much always right, are the ones with systems and procedures in place who do things properly or who work on very high end stuff and have made the effort to understand how manufacturing works, the other guys break all sorts of design rules, have shoddy BOMs, shoddy procedures, and expect everyone around them to be psychic. I know which kind of shop you should be trying to be, amd you will certainly find out the hard way if you bring it in house and don't gear yourself that way.

However, bearing in mind we already knew how to print & reflow, our experience of receiving new machines with on-site training. Machine Deliver+commission+24hours to get to room temp = day one. Day 2 = Training using our data & products and actual product being created. As long as the person being trained on the machine has some concept already as to what is going on and your broker can provide such training I don't see why you should be able to do the same. Get advance details of exactly what data this machine likes to be fed, buy your stencils, frame adaptors if required, components for a few products, have your datasheets and dimensions to hand for all those parts, and a set of calipers for where that data is wrong or missing.

[Styno]

@SMTech, thanks for that sobering response. I had not considered the possibility of all those (cheap) manufacturers plus a bunch of opensource/free Gerber viewers using the same open source libraries. I must also accept that they (the assembly house) are way more experienced and professional than I, a physics degree that wondered by chance and interest into EE several years ago. So, yeah, it's probably my fault.

Doesn't change the fact that I hate that my lowly perfection level worked well before and now someone else determined that there is a problem with that design though.

Edit:
@Mike, I'll be looking around for thin, transparent and low-tack double sided tape. And already have several reels available for initial testing and learning. Thanks for those tips.