Advice on setting up a low-mid volume SMT line for <$130k?

[SVFeingold]

Hi everyone,

After much googling and research I feel like I've hit the point of diminishing returns on information and I'd like to hear from people that have experience with this kind of equipment.

I'm potentially acquiring a contract that will allow me to set up a small SMT line for a client. The idea is to offload the work of sourcing/assembly from the client onto me to provide a more or less turnkey service for the client. They are on board with this idea and like the concept of a quick turnaround PCBA service that (for the near future) would be primarily dedicated to them. Initial quantity is ~2.5k boards by November with another 2.5k by end of year.

We are both aware of the risk involved but excited about the opportunity. This first batch will cost them ~$50 per board, including PCBs, BOM, and PCBA. The follow on batches I think can be done for <$30. They're very happy with these prices. I for one am thrilled about the opportunity to set up my own proper lab. My background is as a mechanical engineer and a EE hopeful - I've mostly focused on electronics and board design the last several years and I even do this at my day job, though nowhere near the level of some of you here. I live in Atlanta and with Georgia Tech and a budding maker community nearby I think I could offer a good service to small business and hobbyists locally. Enough rambling:

The budget I'd like to stay under is $130k. This can be pushed a little bit if necessary but is about as high as I'd be willing to go. At $150k I'd be working for "free" and basically being paid in equipment, but I am willing to do that.

I've looked at a lot of equipment, and this is what I've settled on so far as a preliminary setup:

• Philips Assembleon Topaz Xii (2005) - $38,800
• MPM Accuflex Stencil Printer - $24,800
• Heller 1706EXL Reflow Oven - $16,000
• Mirtec MV-7L AOI Inline Inspection System - $25,000
• Additional Assembleon Feeders - up to $4,000
• HEPA Air Filtration System - $1,500
• PCB Conveyors - up to $4,500
• Ultrasonic Cleaner for Stencils - $2,000
• Sil-Air Air Compressor - $4,000
• Esimated Freight Cost - $10,000

Total: $130,600

My priorities are:

1) Reliability
2) Ease of use
3) Availability of documentation, parts, spares, etc...
4) Flexibility
5) Speed

I rather took a liking to the Assembleon machines. By all appearances they are workhorses and appear to have a good interface and a plethora of feeders and parts on the used market. I have experience with these types of motion systems and would feel comfortable doing most maintenance. I looked at the Neoden4 (and read through all 49 pages of the thread here) but I'm apprehensive about it and would prefer a "proper" machine. I think it will be a false economy. I also looked at Manncorp but new machines with feeders are just too expensive. There's a good deal on eBay for pretty new machine with a bunch of feeders but it's more than twice what the Assembleon is.

I really don't want to cheap out on a stencil printer. The MPM Accuflex seems like a common solution that works well. Same for the reflow oven.

I will need conveyors to string these machines together and a 90deg turn conveyor to make it to the reflow oven (small room). The oven will either need external ducting or an air filtration system with no outside exhaust. Very eager to hear thoughts on this topic. There's a cheap lot of 5 conveyors including 2 90deg rotators for $1500 on eBay. Pulled from production and untested, but I'm definitely willing to roll those dice and feel confident I could fix most potential issues.

The AOI system I'm not sure about. 99% board yields without an AOI system would be acceptable for the near future. Would that money be better spent elsewhere or is it a necessity even for low volume? Is that even a good AOI system?

Another thing that gives me pause is the 11mm maximum height on the Topaz Xii. There are a couple of Assembleon Emeralds that can handle taller (20mm) and more oddly shaped parts. Would that be a better choice? This initial project won't hit that height restriction but a few of my others will (RJ45 connectors, caps, etc...). Lower throughput but that's OK for now. If I do get 2 machines I'd prefer to stay within one "ecosystem."

$10k seems to me like a reasonable freight cost if I shop around, all machines are in the US with one only a state away. The space I'm looking at to house these is a steal, old industrial building, concrete floors, 3 phase.

The air compressor is pricey but it's practically silent. As I will be in the same room with it all day this is a necessity. I'm having a hard time finding air requirements for these machines so I may be able to downsize to a cheaper compressor.

There are three other things: 1) PCB feeding and unloading. Although I'd be OK with doing it by hand for now that will get old fast.  2) Double sided assembly. Is there something major I overlooked that's a necessity for double-sided assembly? Without a glue dispenser it will be limited to placing only low-mass components on one side but that's fine. The reflow oven however has a mesh belt...how would I put double sided boards on that? 3) Through-hole components. For low volume work I'd be fine placing and soldering them by hand. With the right equipment I could probably get through a hundred boards in an hours or two. Even so, would either of these machines be capable of placing through hole components using a pin-in-paste method?

So that's it! Do you see any major bottlenecks I didn't consider? Any input on the quality of this equipment...perhaps I'd be better served with something else?

This is a lot of questions but I would really, really appreciate input from folks who know these things in and out. And it goes without saying that once I get this set up I will be absolutely delighted to offer super cheap PCBA for EEVBlog members while I learn and get comfortable with the machines. My goal is to be able to take a simple job (i.e. Arduino board) and be able to get the line programmed within 4 hours.

Again thank you all so much for any advice!

[ahbushnell]

Did you do a business plan?  Cash flow?

Sounds like fun.

[SVFeingold]

I have a full-time job. My motivation to do this is primarily because I want to, the money is secondary. In other words I am not intending this to be my primary source of income, although if it turns out that way I won't complain. Cash flow wise as long as I can cover consumables and maintenance I can live with a low profit on the first runs while I gather empirical data about it. In other words if I miscalculated and need to spend $500 out of pocket to cover solder paste or something that's OK.

[mikeselectricstuff]

I hear mixed opinions on AOI from the assembly contractors I've used over the years. Some use it all the time, some only on bigger runs due to setup time, some never.
If your basic process is good, then the need for AOI may be questionable in terms of value for money. How many things will it catch that won't be caught in final test, and how many of those are things that could have been fixed by improving the preceding processes?
Every large purchase needs to be considered on a cost/benefit basis, and until you have the whole thing set up, you don't really know how much value AOI will add. It may be better to spend the money on improving the process than finding it slightly earlier it when it fails.
In your position I'd hold off on AOI until you're up & running, and analyse your failure rates, as without the figures you can't determine the value of AOI.
You may  also want to look at alternatives - high pixel count consumer cameras are cheap.  I wonder if a very simple system that takes a high-res pic and then does a "flicker comparator" type thing with a reference image would give you, say 50% of the value of proper AOI for a tenth the price.

I'd probably agree that most of the current generation  of Chinese machines leave a lot to be desired, certainly for a minimum intervention setup, in an area where labour for babysitting is expensive.
The Neoden 4 in particular, though it gets a lot right, like 4 heads and a conveyor ( albeit with its own issues), it's a bit of a car-crash when it comes to feeders and software. 
I'd certainly be looking more  towards machines that use Yamaha feeders, as feeders are THE problem for doing pick & place reliably. 
 However there's no escaping the fact that you can buy several Chinese machines for the price of one decent used machine from the "Old Guard". I have no doubt that for some use cases, one or more Chinese machines could offer much better overall value and flexibility. Few places could afford to have a "spare" Mydata just in case, but if the machine only costs $5-10K, having multiple machines means that any single failure will never stop production, just maybe slow it down a bit.

While it's true that stencil print quality is probably the most important thing to get right for maximising yield, $24k seems  rather a lot for something that can be done semi-manually pretty quickly - at least one fairly large assembler I visited a few years ago manually printed everything. They rotated the job round multiple staff, so they all got good at it and nobody got bored doing nothing else all day.
Also, with the availablility of cheap stencils from China, especially if you regularly have new designs you probably want a printer that can use unframed, less-than-huge stencils.

[SVFeingold]

Great info! Makes total sense about the AOI. Problem is I really don't know what's involved in programming one. I've watched a bunch of programming walkthroughs from various manufacturers. If it takes 2 days to setup it's not worth it for 100 boards. For the 5k order that's funding this there will almost certainly be test fixtures that could probably cover it. In any case it's a consumer product, a few warranty returns are baked into their business model.

The idea with the stencil printer is to reduce the amount of babysitting. I am for sure not opposed to a manual printer as well for smaller jobs. Is there anything you'd consider a "must have" for an automated stencil printer or are they all more-or-less the same? The staff in this case will be just me. If it comes to it I can find help but if an automated printer will reduce my stenciling labor time from 5 hours a day to 30 minutes a day, that's a win. Would you agree?

When you say Yamaha feeders, would that require a Yamaha machine or are certain feeders interchangeable between brands? This is one area I really don't understand yet. Would you consider the Assembleon a poor choice for this reason? I.e. feeder quality/support/capability vs. a different brand machine?

What are the "go-to" chinese machines with readily available parts/feeders? Juki, MyData, etc...?

Another thing I lack the experience to know is which brand offers the best user interface/software. This is a big one for me, I already use enough buggy software and the smoother this goes the better.

Love your channel btw!

[SVFeingold]

One other thing I considered if I dropped the AOI would be picking up a second Assembleon Emerald, that way I'm covered for just about any part and have a little bit of redundancy if one machine fails. Is that a good deal for $60-70k for both machines? I assume feeders are interchangeable between the two and the programming process is very similar.

[Kjelt]

I am no expert on this subject but want to share this: three years ago I spent a week testing my companies  pcbs in an assembly factory. I talked to the guys and manager and remember what their main concern was: service.
In their contract there was a fixed delivery clause, each day late there was a huge penalty. In that week I was there the flying probe testmachine failed. They had a 24/7 service contract on all their machines since they could not afford to loose the production time. The service guy came in two hours later, a replacement part had to be ordered from Japan, it was express courier flown in and the total delay was less than one day. Cost was huge though.
So how are going to cope with a defective machine, can you fix it and get spare parts within a few days, have a backup plan in place?
Another thing I remember was that the ESD pre cautions were ultra high, special clothing, shoes etc. Another thing to take care of is the floor. Good luck on the enterprise.

[SVFeingold]

Interesting. I've heard really good things about the Heller ovens. Is there a Chinese oven you'd recommend?

The HEPA filter is something that's so cheap that I much prefer to have one to keep things clean. I've been meaning to get one for my home lab for forever. Atlanta can be a dusty city.

That's great to know about the stencil cleaner, I have to clean the stencils I do manually quite often but of course that's a much lower quality process.

On the automation I agree it probably wouldn't take long to load and unload boards, and of course I'm willing to do this for a while. The real issue is that if the line is running for 8 hours, and I have to walk up and do something every 1-5 minutes, that means I can't really do anything else even if the time required to do it is only seconds. The more hands off the better, if I only need to come change a feeder, load boards, etc... once an hour that's a world of difference is terms of time management and operational mindset. I'll be in the room doing other work, designing boards, programming, etc.. and having to get up every few minutes is a major distraction.

Mike and blueskull, on the stencil topic how do you feel about the stencil frames that let you swap out stencils and tension them with only one frame? That would certainly cut down on the cost and bulk of stencils. For small stuff I can get a decent manual printer that could use smaller stencils.

Kjelt: Couldn't agree more, I think service is super important. For the foreseeable future my advice to anyone who has very strict timelines would be "use someone else!" I'm certainly aware of the possibility and so I wouldn't feel comfortable guaranteeing extremely tight timelines. This is however where I'd like advice - who are the usual suspects for service? Are there well-regarded aftermarket sources for parts? Do all manufacturers require you to "buy in" or register just to get them on the phone for a used machine? I.e. Manncorp charges $5k just to "register" a used machine.

I am still developing a plan for how best to implement ESD protection. I'm thinking floor treatment, grounding straps, and perhaps ionized air blowers if they are actually effective.

[SVFeingold]

Here are a few alternative machines:

Samsung CP30V - $6,000

Samsung SM320 - $35,000

Juki KE-2030 w/ 95 feeders - $18,500

Juki KE-750E w/ 75 feeders - $10,500

Are these machines a better deal? I'm having a hard time finding specs like maximum part height. How do they compare in terms of used/aftermarket parts, support, and software?

blueskull: It won't run 8 hours/day every day for certain, but I'd like that ability or at least an easy upgrade path to that. I.e. I'd much rather spend the money/time on shipping and rigging heavy equipment initially rather than have to deal with moving stuff around to accommodate a slightly higher level of automation if I want to upgrade .

How would you do double-sided boards on the mesh conveyor of the Heller I linked to? Is it a simple change to a conveyor that holds PCBs by the edge or should I look at a a different oven altogether?

[mikeselectricstuff]

When you say Yamaha feeders, would that require a Yamaha machine or are certain feeders interchangeable between brands? This is one area I really don't understand yet. Would you consider the Assembleon a poor choice for this reason? I.e. feeder quality/support/capability vs. a different brand machine?

I can't comment on the big-name machines as I have no experience. It's probably as much about what you can get a good deal on used, parts availability, service info etc. as anything else.
Many Chinese mfrs have figured out that feeders are hard, and the Yamaha ones have become  a de-facto standard, and I think clones are now available - no idea how good.
There are a few threads here on various machines, but probably not enough to draw conclusions from - maybe see if you can talk to users. Software is the big issue with most. I believe there is an OpenPnp interface for at least one machine now but don't know what state it's in.
Chinese machines are designed for a market where labour is cheap. An example which I expect is common if not universal - in the case of a feed error, most machines wil just stop and cry for help, requiring attention every time it happens ( and it will always happen however good the feeders are), whearas what they should do is carry on with the next feeder, and do as much as it can, and at the end tell you what needs sorting out, so you can do it all in one go, minimising "care and feeding" time.

As for support, some Chinese mfrs seem OK-ish at fixing things that they should have got right in the first place, but you should work on the basis that there won't be any, at least not as quick as you'd like, so you want plenty of spares, and a machine that's easy to work on. (e.g. do you have to take all the feeders off to get at one small consumable).

If you are in the situation where you are making boards you have designed, there is a HUGE opportunity to streamline the whole setup process if you have reasonable access to the machine software, file formats etc. Once you get your PCB library part names matching the machine's ones, and the rotations consistent, you can write a simple utility to turn the PCB software's pick/place output straight into a machine job file, with fiducial locations etc., setup time can be no longer than it takes to pick & load the feeders.

Having in-house pick/place for your own stuff vs. offering a service to others can be vastly different in terms of time & effort and viability.

Only when you actually start using an assembly line, be it a cheap crappy tabletop or big-ass thing, will you start to understand all the little details quirks and issues that can arise, and there are lots.
If you have little direct experience, but some time, it might almost be worth spending $5-10K now on a really cheap setup to treat as an "expendable" learning tool,  which will give you a lot of insight  into what you really need and where the money would best be spent.

[mikeselectricstuff]

Mike and blueskull, on the stencil topic how do you feel about the stencil frames that let you swap out stencils and tension them with only one frame? That would certainly cut down on the cost and bulk of stencils. For small stuff I can get a decent manual printer that could use smaller stencils.

A couple of places I've used use this system,  and don't understand why anyone would bother with framed stencils, but they do. If nothing else it saves a lot of storage space.

As regards ESD, the reason big assembly places go to town is the risk ratio. They have a lot of semi-skilled staff who are there just to earn a buck, and can't necessarily be relied on to take the sort of care ( or have the sort of understanding) that a skilled operator like yourself has, so a simple "thou shalt wear anti-static jackets, shoes and wriststraps" rule is a fairly easy way to spend money to reduce risk.
The consequences  of ESD issues could be huge, so they need to play as safe as practical ( and be seen to do so), almost regardless of the probability of actual issues.
 In the same way that nowadays you often see workplaces where everyone has all manner of safety gear regardless of the actual risk, and is little more than just ass-covering. 

In your situation it's going to be much less of an issue as you (presumably) know what you're doing, so you're not going to need to spend anything like as much as you're smart enough to know how to handle boards, avoid obviously staticky materials etc. Apart from making sure the floor covering is non static-producing there's proably little else that's really necessary.
IMO much of the ESD industry is barely more than Snake Oil.

[mikeselectricstuff]

How would you do double-sided boards on the mesh conveyor of the Heller I linked to? Is it a simple change to a conveyor that holds PCBs by the edge or should I look at a a different oven altogether?

You can put boards in holders onto the conveyor. Once place I've seen just made triangular spacers out of PCB tooling strips taped together with Kapton and rested the boards on those - seemed to work pretty well & it's what I do in my toaster oven now.

[mikeselectricstuff]

Interesting. I've heard really good things about the Heller ovens.

I don't think I've ever seen an oven made by anyone else in an assembly place.

[ar__systems]

IMO much of the ESD industry is barely more than Snake Oil.

Most of digital electronics will tolerate pretty much any handling. But power mosfets are a lot more fragile and will break down if you are not careful at handling the boards properly.

[amspire]

IMO much of the ESD industry is barely more than Snake Oil.

Most of digital electronics will tolerate pretty much any handling. But power mosfets are a lot more fragile and will break down if you are not careful at handling the boards properly.

Back in the late 80's, I was with a company making inverters. Initially we were getting far to many failures in the field - from days after delivery to a year or more. The numbers could have been 25%. We implemented strict ESD provisions from the warehouse to assembly and the failures went down to essentially zero.

The point is without ESD provisions, you can have significant unexplained failures of you product that can fail over a long time period - probably years. Parts can be ESD damaged and yet function perfectly on the assembly line. In fact it was rare to see an ESD damaged part actually fail on the production line. Very costly, very time consuming and very bad for your reputation. With ESD provisions, there is a good chance your product will run trouble free for decades.

Mosfets are very heavily stressed devices and are more likely to eventually fail after a micro-puncture in the gate. It could be that many other devices can have similar damage but run for years before failure. The only way you can know that you probably are not sending out damaged devices is to implement strict ESD practices.

[SVFeingold]

I definitely take ESD safety seriously, and I'm looking for that sweet spot in terms of cost efficiency. ESD damage is, as you said, one of the most frustrating and hard to track down so it's worth investing a little. Naturally I don't have $50k to invest in copper floors and whatnot but I believe ESD safe floor mats + a floor treatment for the rest, ionizing air blowers, and perhaps wrist strap grounding points at each machine in case I'm touching anything would go a long way.

Does anyone know is Assembleon products are still actively supported? Looks like Philips sold off that division...? Real shame watching companies give up tech divisions in which they are strong, particularly Philips. They did it with electron microscopes back in the day. I digress...

[ar__systems]

During the second pass I simply place the board on another blank board.

[amspire]

The point is without ESD provisions, you can have significant unexplained failures of you product that can fail over a long time period - probably years.

You are right, but there are virtually no digital chips that don't have quite robust ESD protection built in.
Usually most digital chips have 2kV HBM and 200V CDM built in on virtually every pin, except for some really high speed pins (SFP/SFP+/XAUI/SDI/etc., but hopefully people using SERDES know what they are doing).

The thing is we were getting devices like the IRFP460 failing which is a 20V gate but with 4200pF capacitance. If you are talking about a chip with 1pF input capacitance, 2kV is probably very easily reached.

As an engineer, I just know the feeling of frustration when products are failing and management is asking you to check for every possible reason for failure other then ESD failure. It can be months of fruitless work. The only real way to check for ESD failure is to implement strict ESD practices and see if reliability improves.

[amspire]

The thing is we were getting devices like the IRFP460 failing which is a 20V gate but with 4200pF capacitance. If you are talking about a chip with 1pF input capacitance, 2kV is probably very easily reached.

It's nothing about voltage. It's about energy.

20V at 4200 pF = 0.8uJ
2000V at 1pF = 2uJ

So yes, the energy to blow a 1pF pin on an esd protected IC is a little more then double the mosfet, it is still a similar magnitude. The human body has a capacitance in the 100-200pF range, so basically if your body is charged to over 2000V, you have enough energy to damage many esd protected IC pins.

Simple ESD provisions eliminates this possibility.

[SVFeingold]

Hey guys, thank you for all the responses so far. I was considering the Samsung SM320 but one thing worries me - I have been googling for an hour and I can't seem to find an official website that even acknowledges the SM320 exists. Samsung doesn't seem to have an SMT equipment website - seems they've sold it? So I looked at Dynatech but apparently that website is now defunct. That brings me to Hanwha Precision Machinery which appears to be the successor. Hanwha appears to have a single service center for the US, located in California, the opposite side of the country.

Assembleon (Now under K&S it appears) has a service center in Alpharetta, GA, which is literally a 30 minute drive. From some threads on SMTnet it looks like they do training there, as well as support and service. If that is indeed still the case this seems like a no brainer. From my experience in the CNC world support is very important, and it would certainly ease my mind if I knew support was available locally.

Having said that, I have no idea how much I might expect to pay for training or support. If training is $5k and teaches me how to use the machine comprehensively that is not so bad, but if it's $50k... Still the only thing worse than expensive support is no support.

Regardless of other differences in the technical capabilities of the machines, would I be making a mistake choosing a machine without local support?

I'm going to call the local K&S office and see if I can schedule a visit to talk through some of these issues with them.

[SVFeingold]

Mike, it appears at a glance that Yamaha used to manufacture the GEM series Assembleon machines and so there should be Yamaha feeders that are compatible. A cursory eBay search for feeders seems to support this. Another plus in the Assembleon column it would seem...

[mikeselectricstuff]

I've always found it pretty hard to find good info on used pick/place machines. The problem it's a very niche area, and some machines have origins  from before the internet was much of a thing. It's traditionally been a high cost, high margin value, so people buying new machines would be dealt with by manufacturers' sales people.
There are some dealers in used equipment, but there still seems to be a culture of keeping information to themselves - why should they publish service info when they can sell you parts for a fortune or sign you up for an expensive service contract?
The people who use most of these, subcontract assemblers also IME seem to have a culture of not sharing what they regard as their own little tips, secrets etc. They don't seem to be the sort of people who'd do things like setting up a forum to co-operate with each other to support machines.
This is the only forum I'm aware of with discussion on the bigger machines : http://www.smtnet.com/Forums/
 
If you look at the few accounts of people who've bought old machines to get them working, they've pretty much had to figure everything out themselves.
I don't think there's an easy answer, other than for people who are going through this process to share what they've learnt to make it slightly easier for the next guy. You need to do a lot of research, but that research can be pretty hard - you probably need to actually start calling people and going to see machines in action where you can.
Meanwhile the Chinese are doing their own thing, and if they ever figure out that they need to hire a few western staff to show them what they are doing wrong ( and then actually take notice of what they say), they could have a much bigger impact on the market. The prices that most low-mid volume assembly places charge are based on amortising equipment that cost six figures, with high ongoing maintainance costs. Once you can buy a fast, reliable and easy to use machine for $5K, the economics of either doing in-house or a company setting up with a few dozen of them to undercut the old guard change radically.



 

[mikeselectricstuff]

Re. ESD, I never said it wasn't a problem, just that the precautions you need to spend money on to deal with it can be substantially different depending on the nature of the environment and skill level of the opperators.
 

[jmelson]

The Philips "gem" series all need 3-phase power.  The earlier machines like the CSM can run on single phase.  So, make sure you have the required power.  (I see you mention that farther down.)  You ought to be able to find a place where an air compressor can be placed (on the roof, behind building, etc.

$38K seems pretty high, a lot of machines are going for less.  Some of the Gem line are for super fast assembly, with 8 to 20 nozzles on them.  Do you need this level of production?

Philips Assembleon is out of this business, as far as I know, so spare parts could be a problem.  There does seem to be a pretty active grey market in spares, but I don't know the quality.

My CSM84 is limited to 6.5mm component height.  That actually is not that huge a problem, but yes, some larger electrolytic caps could be too tall.

As for software, my CSM has amazing levels of fault recovery, you can even POWER DOWN in the middle of the board, and it will pick up and finish the board when you come back up!
(I've had to do that on a couple occasions when things went really badly during a run.)  Figuring out what feeders you need and how much that will cost is important, they can really ADD UP!

Jon

[jmelson]

OK, know that there is a lot of learning to be done before everything works well.  Also, stencil design is really important to getting good reflow results.  The smaller the pad/lead and the lead pitch, the more the aperture area needs to be reduced so you don't get solder bridges.  Don't ASK how I know all about this!

Jon