Pololu Robotics & Electronics
Menu
My account Comments or questions? About Pololu Contact Ordering information Distributors

Pololu Forum

PCB assembly tools and equipment for SMT


#1

I’m not sure my questions are appropriate to this group or forum, but that’s okay… just ignore this post if it’s out of line. I’ve been designing a “vision system” which is ultimately is supposed to be part of an insanely advanced “robotics” application, but is also perfect for many other applications. In essence, its just a moderately high-resolution digital video/still camera with controller that spews the images out to an attached PC via standard gigabit ethernet via standard RJ45 jack/cable. Well, its a bit stranger than that actually, because the image-sensor PCB (which I call an “eye”) is extremely simple, but up to 4 of these can be connected to the controller PCB (which I call the “quad”) so the controller can capture and forward 4 image streams to the PCB over the one gigabit ethernet connection.

Anyway, none of the above is relevant to my questions — except maybe to entice someone to answer my real questions, which follow.

I designed and prototyped dozens of digital devices and products [more than] several years ago, so all the components were standard DIP packages with 0.100" centers (plus non-SMT resistors, capacitors, etc). After several years of science, engineering and software (mostly), I return to design more hardware devices and find… MUCH has changed.

Most important, SMT (surface mount technology) is totally mainstream. In fact, 90% of the circuits I want to design with are only SMT. So all those years of prototype experience is now ancient… and obsolete.

Anyway, I finished the schematics and PCB-layouts for these two PCBs. and emailed the gerber files off to get prototypes PCBs made. Now I need to buy whatever tools, supplies and equipment is necessary to build these prototypes. To abuse myself further, I chose lead-free components and assembly, since that too seems the modern way.

What I’d like to know is, “what assembly/soldering techniques should I adopt, and what tools, supplies and equipment must I buy to build these prototypes?”. Specify brands and product numbers too, if feasible. I’m not starting a PCB assembly biz, and don’t have $50,000 to spend, but these do need to be reliable and of commercial quality… because hopefully they will become commercial products once all is finished.

Now I’ll try to describe the types of components on the PCBs, since that information is probably necessary to make informed suggestions.

PCB #1: the “eye” : 2.80" square, 0.0625" thick, 6-layers
PCB #2: the “quad” : 5.80" square, 0.0625" thick, 8-layers

Both PCBs have components on both sides. Except for 40-pin headers to connect to ribbon cables, and a very few other oddball components, ALL the components are SMT.

The PCBs have 4-mil trace/space widths, 8-mil holes, and all "normal"
characteristics (soldermasks and silkscreens on both sides, and 0.125"
holes in all 4 corners of both PCBs). All the internal layers are GND and PWR (various voltages), except one more signal layer on the quad.

The PCBs contain several BGA components (with 48, 56, 96, 256 balls), several QFN components (with 8 to 48 pads), several QFP components (with 100 to 128 pins), and several SOIC, SO8 and similar components. Most resistors are 0402, and most capacitors are 0603 (but a couple dozen are 0402 and 0201).

The smallest pad-to-pad (or ball-to-ball) “pitch” on the BGA and QFN packages is 0.500mm or about 20-mils (0.020"). I didn’t really grasp how small and close those pads and balls were until I received samples of some of these components! The first little box I opened (max8717) was in a 28-pad QFN package. At first, I thought the box was empty. Then I noticed these tiny little square flecks of black plastic in the corner of the box, and took one out to see what it was. Then I noticed the other side had a bunch of tiny silver pads around the outside! Yikes! That’s what happens when you work on PCB layout software at typical magnifications of 3200% to 10000% — you have no freaking idea how tiny are those pads, balls and pitches! AARG.

But that’s the way it is today, and I need to install and solder the components on these PCBs… and end up with solid, reliable PCBs. I’ve watched a number of videos on the internet, but almost all of them seem vastly too crude for me (soldering on a skillet or $25 convection oven) or vastly too expensive ($50,000 pick and place machines and so forth).

There MUST be an intermediate way! Please say yes! :astonished: And describe.

I did run across a few possible clues and ideas (at least I think so).

First, it looks like $100 will buy me a “stainless steel stencil” to
help me apply consistent blobs of lead-free solderpaste on each pad. I’m not sure whether I need one stencil per PCB side, one stencil for both sides of a PCB, or whether I can buy one stencil with patterns for both sides of both PCBs on it. Any comments about this approach?

Second, it appears some low-end “reflow ovens” exist in the price range from about $500 to $2500. For the life of me, I can’t understand why a simple $50 convection oven plus thermistors and (air-mixing) fans plus temperature controller should cost more than $250, but what do I know? Is this a reasonable approach? Are these reliable? Do they support PCBs with components on both sides? Do I reflow each side separately?

Third, maybe an alternative to “reflow ovens” is “hot-air solder guns”? I’ve seen some pretty nice demonstrations on youtube of soldering small components with these devices. However, I can’t see how to solder in BGA and QFN packages with these devices… and I have BGAs and QFNs. These look great for removing and replacing misplaced components after a PCB is totally finished (by reflow oven probably), and maybe even soldering in most components with exposed leads (QFPs, SOICs, etc), but not BGAs or QFNs. Or is my imagination too limited in this case?

Fourth, I saw some advice on the internet that claimed assembling PCBs like these require a “manual (or automatic) pick-and-place machine”. After seeing a video of one of these in action on youtube, I can’t see how it helps aid placement accuracy AT ALL, though it might help keep the operator from getting sore arms quite as quickly… or maybe not. To be sure, automatic pick-and-place machines are massively “cool”, but vastly too expensive for me (looks like $12,000 to $50,000 to me).

I also saw a device called a “solder mill” at zeph.com, which could work for placing and soldering BGAs and maybe QFNs one-by-one… but leaves everything else to manual hot-air pencil soldering. This is also rather expensive… about $8000 for a complete system and kit, PLUS individual “templates and nozzles” for each BGA and QFN package at $450 per package type! Ouch!

But I’m looking for YOUR advice, especially from anyone who has gone through this (at home or work). And hey, if you have some of these gizmos in your garage (in or near Nevada), well… let me know! :slight_smile:

So, all you electronics wizards… BRING ME UP TO SPEED !!! Thanks!


#2

Hello.

I wasn’t able to read everything you wrote, but I have suggestions for how to learn more about SMT. Try visiting http://www.sparkfun.com/commerce/tutorials.php and click “Surface Mount Soldering Tutorials.” Reading all of those should give you a lot of suggestions for doing SMT on a small scale.

BGAs are hard to do. People who are assembling BGAs typically have an X-ray machine that lets them see if the chip is properly connected to the board. If your parts come in alternatives to BGAs, I would suggest using those instead.

You mentioned using metal stencils. You should also consider plastic stencils. Plastic stencils don’t have the lifespan or repeatability of metal stencils, but they are cheaper. We offer a laser-cut mylar stencil cutting service starting at $25.

I hope I addressed most of your questions; please ask if you have additional questions.

-Ryan


#3

Hello.

I think the quick answer is that you bit off way too much to chew with the double-sided assembly with BGA parts: you’ll have no way of inspecting them (assuming you don’t have friends with X-ray equipment) or fixing them, and just one bad connection out of hundreds will cause failure and endless hours of frustration looking for the problem.

I would get these boards assembled by someone else, such as Advanced Assembly. I have not used them, but they advertise low costs and small quantity.

Separately, you could work on assembling your own simpler SMT designs using some of the techniques you’ve mentioned. Once you have an idea of what’s doable, you can split future designs into “assemble it myself” and “have someone else assemble it” classes. As soon as you’re talking multi-layer boards with BGAs, you’re probably in the second category; but if you just have QFNs and other small but manageable parts, you could space them out a bit and try to avoid components on both sides to make things a bit easier for yourself.

- Jan


#4

[quote=“RyanTM”]Hello.

I wasn’t able to read everything you wrote, but I have suggestions for how to learn more about SMT. Try visiting http://www.sparkfun.com/commerce/tutorials.php and click “Surface Mount Soldering Tutorials.” Reading all of those should give you a lot of suggestions for doing SMT on a small scale.

BGAs are hard to do. People who are assembling BGAs typically have an X-ray machine that lets them see if the chip is properly connected to the board. If your parts come in alternatives to BGAs, I would suggest using those instead.

You mentioned using metal stencils. You should also consider plastic stencils. Plastic stencils don’t have the lifespan or repeatability of metal stencils, but they are cheaper. We offer a laser-cut mylar stencil cutting service starting at $25.

I hope I addressed most of your questions; please ask if you have additional questions.

-Ryan[/quote]Thanks for the suggestions. Yes, I did read the sparkfun pages, and they were fairly helpful. I was especially concerned at their comments about buying a $2300 reflow oven only to find it did not produce quality results. $2300 for a toaster oven that doesn’t even do its job? That’s terrible! Sadly, after searching and reading the internet for the past week (about 12 hours per day), I found a great many horror stories about reflow ovens.

Frankly, I think anyone who is purely motivated by profit could make a freaking fortune by making a convenient, high quality desktop reflow oven at about the $2500 price point, because the current manufacturers are utter clowns and morons.

If you guys at pololu have a quality reflow oven, would you rent me some time on it? Alternatively, if you do not have such a device, would you provide space for me to locate the one I am likely to buy this week. This is a $4500 desktop unit made by a company called “APS Novastar” (I think that’s right), the model number is GF-B-HT, it weighs about 100 pounds, and requires a 220V power-plug. But from everything I have read, this is the only unit that has consistent praise from current owners. If you provide a home for this beast (about 37" x 28" x 15" high), maybe you guys would find it useful for your own work too. My other alternative is to buy the complete system from www.zeph.com for about $9000, which looks like well designed equipment but more for “rework” than making original PCBs. If I go that way, perhaps you’d find that useful to have avalable. You and I seem to be on opposite sides of the spectrum: I tend to make big, huge, elaborate PCBs (and projects), while you guys make tiny but very modular building blocks. Together, we have the spectrum of insanity covered pretty well, I’d say! Hahahahaha.

Yes, I understand BGAs are difficult, and my decision to adopt lead-free components and solder-paste makes everything even more difficult. But that’s the nature of the game when you intend to do quality, professional commercial products, right? BTW, between the PCBs, components and other equipment and tools I need to finish these prototypes, I expect to pay about $30,000 to $40,000 - so I’m trying to be as cheapskate as possible consistent with high quality, but I’m not being “absurd” about my expectations… hopefully. I’m also buying a 5-gigasample-per-second 4-trace oscilloscope, so you understand this is a serious project to me.

I’ve been designing and developing products full-time since before I got out of high-school, so I understand what it takes, and I invest the required time, effort, savings and resources to “make it happen”. This has been my entire life work for more than a couple decades already. So, while I am massively behind the curve on SMT technology, otherwise I pretty much know what it takes… and “been there, done that”.

I was going to order stencils from you, but for the number of square-inches of area it appears the stainless steel stencils are no more expensive. So now I’m not sure which is wiser.

I really should drop by and visit you guys, and see whether we can help each other in some ways or other. Is your place openn to “just walk in”? If not, would you like to meet sometime and “compare notes”? Max (AKA “bootstrap”) @ 702-270-6167.


#5

[quote=“jan”]Hello.

I think the quick answer is that you bit off way too much to chew with the double-sided assembly with BGA parts: you’ll have no way of inspecting them (assuming you don’t have friends with X-ray equipment) or fixing them, and just one bad connection out of hundreds will cause failure and endless hours of frustration looking for the problem.

I would get these boards assembled by someone else, such as Advanced Assembly. I have not used them, but they advertise low costs and small quantity.

Separately, you could work on assembling your own simpler SMT designs using some of the techniques you’ve mentioned. Once you have an idea of what’s doable, you can split future designs into “assemble it myself” and “have someone else assemble it” classes. As soon as you’re talking multi-layer boards with BGAs, you’re probably in the second category; but if you just have QFNs and other small but manageable parts, you could space them out a bit and try to avoid components on both sides to make things a bit easier for yourself.

  • Jan[/quote]
    Thanks for the comments and suggestions. I’ll check out aapcb.com and see how that looks. The other places I’ve quoted from want much more to build my prototypes than I’d spend to buy all the top-line, premium-quality equipment I need to build them myself (perhaps not including the xray system, which I have no idea the cost of).

As for “biting off more than I can chew”… thats what I’ve been doing for decades, and I have no regrets about that. Fact is, I’ve never started a serious project that I didn’t finish, and I’ve sold or licensed most of them to companies, and most became commercial available products. Maybe this is the one to force me to “cry uncle” and “eat my words”… but I tend to doubt it (though I always expect a few obstacles to hurdle along the way). I do hope (and suspect) that PCB-assembly houses DO NOT xray every BGA and QFN on every PCB that comes off the line. I can’t believe the statistics are THAT terrible… IF… the reflow-oven system and process-profiles are of sufficiently high quality. Do you know the answer to this?

So, these PCBs are my “test case”… as well as being the project I intend to complete. Understand, this is not only a hobby to me, this is “what I do” (develop products). I developed lots of hardware products from computer systems to robotics over the years, but focused on software the past 15 years (while the world switched to SMT). So now I must “catch up” all in one giant leap (in to the fire, and over the cliff… perhaps).

We shall see. Until then, I’ll appreciate all the advice-from-experience that I can get! So thanks for that, and more is always welcome.


#6

I am skeptical about small batch reflow ovens, but I guess they’re supposed to work for small quantities. I don’t know if it makes much sense to “rent time” on our oven since presumably you’d need the stencil printer and space and time to populate everything. It wouldn’t make sense for us to do it on regular work days, but we’d still have to have someone here running the machine(s) for you. Given that I’m not too optimistic about the results, I’m hesitant to go that route.

We host the LVBots robot club here every other Thursday evening, so you could come by then if you want to discuss your project in person.

- Jan


#7

[quote=“jan”]I am skeptical about small batch reflow ovens, but I guess they’re supposed to work for small quantities. I don’t know if it makes much sense to “rent time” on our oven since presumably you’d need the stencil printer and space and time to populate everything. It wouldn’t make sense for us to do it on regular work days, but we’d still have to have someone here running the machine(s) for you. Given that I’m not too optimistic about the results, I’m hesitant to go that route.

We host the LVBots robot club here every other Thursday evening, so you could come by then if you want to discuss your project in person.

  • Jan[/quote]After spending 12 hours a day for the past week “researching” SMT PCB assembly techniques, equipment, devices and tools, I have a few “guesses” about reflow ovens, which perhaps agree with your comments. To simplify, my main conclusions are: any batch reflow oven with IR heating elements is asking for serious trouble (shadow-problems), and any batch reflow oven with less than 2KW of heating power is also asking for trouble (uneven heating). The vast majority of batch reflow ovens suffer from both those faults, though a few only suffer from one of them.

The $4500 unit I mentioned from APS is almost the only unit that has neither of those drawbacks… and it has a great big triple-pane window on the top surface so you can actually watch what’s happening to the solder joints as it happens. I believe in first-hand observation, and also this creates the ability to hone the heating profile for each PCB based upon those observations. Still, I had a long talk with someone at APS today, and even he admits that single-zone ovens like that just can’t be as precise as multi-zone conveyor belt ovens. Their smallest multizone oven is $8500… more money than I’d like to spend on that one item alone. If I had a unit like that, I’d be tempted to manufacture the products I’m developing, rather than having them assembled in some cheapo asian location. But as much as I like having total quality control over everything, I’d rather not get into any kind of production biz.

From your comments, I’m guessing you guys have a multi-zone conveyor oven. If so, that’s impressive, and good going! Nice to know you have enough biz to justify such a beast. Sounds like you guys are a few steps up from sparkfun. I’m also guessing you guys have an automatic pick-and-place machine, in which case you guys have a very cool setup over there! I’ll try to stop by some thursday night and find out. I’m not sure anyone will be much interested in my projects, though - at least it is much larger and bulkier (and expensive) than the components I see on the various robot examples on the LVBOT website. But then again, ya never know.

I totally understand you not wanting to deal with me (or you) assembling my PCBs at your place. No problem. If I do decide to buy all this junk to make my own prototypes, they still might come in handy once in a while for the more crazy robot builders in town. For sure they’ll be sitting idle in the garage 99.9% of the time. Or maybe I misunderstand the nature of those BOTs… maybe they’re all made from kits (nothing custom), and thus all robot-builders (except you guys and me) have no need for custom PCBs. Just call me robo-dummie.

Oh yeah! I did check out aapcb guys, and they are MUCH cheaper than all the other places I found previously. If I was only going to make 2 or 5 prototype PCBs, I’d almost certainly have them assemble my PCBs. Since I’m probably gonna make 25 of my 6" square PCB and 50 of my 3" square PCB… the tradeoff isn’t clear (the cost of the equipment is slightly cheaper than having them built by aapcb). But thanks much for that suggestion, because they’re obviously the most reasonable people around for that service! As an “in between” option, I might even consider having them install the BGAs and QFNs only, then solder everything else myself manually with the zephyrtronics hot-air soldering-pencil. Still thinking…


#8

I don’t know why you think SparkFun doesn’t have a big oven. They have at least one for their production, but that’s separate from showing people how they can do their own assembly with cheap kitchen appliances. They’re probably about 5x bigger than us, though as far as I know, the two P&P machines we have are similar to their two. We have a relatively small, 5-zone oven because we were limited by the power available in our building. We have an older, 4-zone oven that only needs 50A single phase, which we’d be willing to sell for around the price you’re talking about for the batch unit. I think it’s small enough to fit in a mini-van or pickup truck and run in your garage, and I think it will give you much better results than a batch oven.

- Jan


#9

[quote=“jan”]I don’t know why you think SparkFun doesn’t have a big oven. They have at least one for their production, but that’s separate from showing people how they can do their own assembly with cheap kitchen appliances. They’re probably about 5x bigger than us, though as far as I know, the two P&P machines we have are similar to their two. We have a relatively small, 5-zone oven because we were limited by the power available in our building. We have an older, 4-zone oven that only needs 50A single phase, which we’d be willing to sell for around the price you’re talking about for the batch unit. I think it’s small enough to fit in a mini-van or pickup truck and run in your garage, and I think it will give you much better results than a batch oven.

  • Jan[/quote]
    Yes, I probably drew a false conclusion about sparkfun from their very extensive “doing it on the cheap” articles.

I will try to visit you guys next Thursday night (?June 10th?) to look at the oven you mention, if you prefer to keep folks out of your place until those meetings.

Are your pick and place machines “fully automatic”? I keep seeing manual and so-called “semi-automatic” machines advertised, but I fail to understand how they improve the accuracy of component placement - unless they have optical encoders somewhere/somehow displaying the x,y,r of the tip, and some way to assure the center of the tip is exactly on the center of the part, and some way to assure the tip attaches to the part in a know rotational position.

My biggest concern now is placement accuracy. I have a few 0.50mm ~ 0.70mm pitch components (QFN and BGA), and getting them placed accurately seems my biggest obstacle (so I believe). I know they “tend to self-center” once the solder melts, but that assumes the placement is already within 1/4 of the pitch (or about 0.10mm), which is mighty tiny for manual placement AND presumes the BGA packages are accurate to 0.10mm or better - which I guess they must be, or automatic pick-and-place machines would not work).

Does anyone make a cheap pick-and-place machine? For example, do any manual units have optical encoders to let operators place packages precisely (plus ways to assure the packages are exactly centered on the vacuum tip)?

All the information you’re giving me helps a lot. Thanks for that.


#10

I think ours are considered fully automatic, though I suspect there can be varying degrees of that. For manual machines, they might have mechanisms to help stabilize your hand and allow easy rotation of the part after you pick it up (as opposed to using tweezers). “Semi automatic” is probably most open to interpretation or variation since some things are more useful to have automated than others. Our machines have vision processing to automatically detect where the PCB is and to view the components from below (looking at actual pads, not package outlines). I can imagine both of these parts being semi automated, in that you have to identify the parts that the camera sees, and still being useful for small volume assembly. In either case, getting packages “exactly centered on the vacuum tip” is unnecessary as long as the part doesn’t move relative to the tip between measurement of the part location and its placement.

There are all kinds of cheap pick an place machines, but performance will vary and different people have different ideas of what cheap means. Most of our equipment is from Manncorp, and we’ve been happy with them.

- Jan


#11

Hello,

It looks like you wanted confirmation of the meeting time - it will be next Thursday, June 10th at 7pm. That is really the best time to stop by: most of the Pololu engineers are usually there, and it is more relaxed than the daytime when we are all running around building things, dealing with problems, and answering phone calls. A few of our other regulars do electronics and PCB assembly and might have suggestions or other resources for you.

If you stop by a little early, we could show you our equipment before everyone else arrives, but you should probably call ahead (try the extension for tech support if you get the recorded message) to make sure that you don’t end up having to wait around in the parking lot.

-Paul


#12

[quote=“jan”]I think ours are considered fully automatic, though I suspect there can be varying degrees of that. For manual machines, they might have mechanisms to help stabilize your hand and allow easy rotation of the part after you pick it up (as opposed to using tweezers). “Semi automatic” is probably most open to interpretation or variation since some things are more useful to have automated than others. Our machines have vision processing to automatically detect where the PCB is and to view the components from below (looking at actual pads, not package outlines). I can imagine both of these parts being semi automated, in that you have to identify the parts that the camera sees, and still being useful for small volume assembly. In either case, getting packages “exactly centered on the vacuum tip” is unnecessary as long as the part doesn’t move relative to the tip between measurement of the part location and its placement.

There are all kinds of cheap pick an place machines, but performance will vary and different people have different ideas of what cheap means. Most of our equipment is from Manncorp, and we’ve been happy with them.

  • Jan[/quote]
    Yes, viewing the pads on the bottom of a part certainly is the best solution to assure they are probably placed. That seems totally “worth the wait” to pause over the camera for 1/10 second or whatever for BGAs, QFNs, QFPs, CSPs and similar. Given the way the contacts are placed on resistors and capactors (0201, 0402, 0603, 0805, etc), a mechanical centering scheme is probably more than adequate.

For a “manual” pick-and-place, my definition of “cheap” is probably something like “below $3000”. For an “automatic” pick-and-place machine, my definition of “cheap” is probably something like “below $8000”. Of course, that’s just a guess. Probably like you guys, I care primarily about quality/accuracy, and speed is almost irrelevant (unless it is absurdly slow).

I started out “certain” I’d get production made in China (or thereabouts), but obviously machines get paid even less than humans, so I’m even starting to get the “scary” thought that paying for automatic pick-and-place machine ($12,500 as a wild-guess) and a low-end 4-zone conveyor reflow oven ($8500 from APS) makes business sense. If that’s the best [or viable] “future” for these products I’m developing, I might as well buy this stuff now and not “pay twice”. My primary resistance to that is the obvious… who knows how large the volume of orders might be before they’re finished development and on the market? Plus, blowing about $25,000 on this equipment is about $15,000 more than I planned. Ouch!!!

The extreme alternative (aapcb + china-production) seems just about as viable (given what I know so far). Since I’m an scientist/engineer/inventor/designer I obviously have a psychological preference for option #1, but I need to make these decisions based upon “all things considered”.

Have you guys ever regretted buying the equipment you got… or even wondrered about your decision to bring this in-house? At your volume (just guessing), probably not. Just wondering.

I have the impression manncorp manufactures nothing, only a distributor from “far and wide”. Is that another erroneous inference on my part?


#13

[quote=“paul”]Hello,

It looks like you wanted confirmation of the meeting time - it will be next Thursday, June 10th at 7pm. That is really the best time to stop by: most of the Pololu engineers are usually there, and it is more relaxed than the daytime when we are all running around building things, dealing with problems, and answering phone calls. A few of our other regulars do electronics and PCB assembly and might have suggestions or other resources for you.

If you stop by a little early, we could show you our equipment before everyone else arrives, but you should probably call ahead (try the extension for tech support if you get the recorded message) to make sure that you don’t end up having to wait around in the parking lot.

-Paul[/quote]
Thanks much, I’ll try to remember to call and show up. As you can imagine, right now I’m running around like a chicken with his head cut off (without even moving from in front of my computer screens). Most of the time I don’t even know what day it is, so hopefully I don’t lose track and miss the meeting. I look forward to tapping into any fine-pitch SMT experience I can find.

Hopefully I can keep my mouth shut about the intended applications for these devices. While the “first-release” products based on these devices are extremely mundane (multi-camera security systems), the others are increasingly… well, not sure what term to insert here. Probably your opinions would range from “wacko” to “absurd” to “totally freaking insane” to “delusional” to [probably something even more accurate and appropriate]. :slight_smile:


#14

Great - you will fit right in at the club. See you next week!

-Paul


#15

Oh, and I forgot to mention: if you have anything to show, such as your prototype boards, BGA parts, cameras, software or whatever, bring it over. We always like to see real stuff.

-Paul


#16

For one bottom vision option, the head has to take the part above a fixed camera and move the part around for the four sides of the part, so it takes way more than 1/10 of a second. For the other approach, a little mirror flips out so that a camera can look at the bottom of the part as it’s being taken from feeder to PCB, so there’s no time lost.

I’m very skeptical that you can get something adequate for the $25k you mentioned. “Absurdly slow” is another term in the eye of the beholder. We do care about speed; even a simple design can have several hundred parts per panel, so it’s easy for our relatively small production runs to have over 10k parts placed. It can take several hours to all day on a machine that can place 5000 parts per hour. If you have a small machine and multiple complex parts, you might need to swap around waffle trays and otherwise baby the machine through one part at a time. So, a board like what you’re describing could take a several hours each on slow or semi automatic machines. That can be great if the point is to assemble a prototype; it might not be a practical approach for manufacturing.

I don’t regret getting our machines; for our kinds of volume of a few hundred individual boards per run, with a different design running every day, I don’t think there’s a better option. Early on, our first machine sat idle a few weeks at a time sometimes, so we might have managed our money better by getting it a bit later. I got a fine pitch dispenser (e.g. for solder paste) with our first machine, which wasn’t worth it and which I wouldn’t recommend since stencils are pretty easy to get and work much better.

I don’t think Manncorp directly manufactures anything.

- Jan


#17

[quote=“jan”]For one bottom vision option, the head has to take the part above a fixed camera and move the part around for the four sides of the part, so it takes way more than 1/10 of a second. For the other approach, a little mirror flips out so that a camera can look at the bottom of the part as it’s being taken from feeder to PCB, so there’s no time lost.

I’m very skeptical that you can get something adequate for the $25k you mentioned. “Absurdly slow” is another term in the eye of the beholder. We do care about speed; even a simple design can have several hundred parts per panel, so it’s easy for our relatively small production runs to have over 10k parts placed. It can take several hours to all day on a machine that can place 5000 parts per hour. If you have a small machine and multiple complex parts, you might need to swap around waffle trays and otherwise baby the machine through one part at a time. So, a board like what you’re describing could take a several hours each on slow or semi automatic machines. That can be great if the point is to assemble a prototype; it might not be a practical approach for manufacturing.

I don’t regret getting our machines; for our kinds of volume of a few hundred individual boards per run, with a different design running every day, I don’t think there’s a better option. Early on, our first machine sat idle a few weeks at a time sometimes, so we might have managed our money better by getting it a bit later. I got a fine pitch dispenser (e.g. for solder paste) with our first machine, which wasn’t worth it and which I wouldn’t recommend since stencils are pretty easy to get and work much better.

I don’t think Manncorp directly manufactures anything.

  • Jan[/quote]
    For the “a little mirror flips out” technique, it sounds like you need one camera for each part that requires that kind of treatment… which is fine if you don’t have very many. Or maybe they have the camera on a rail so it can be moved to multiple components.

When I said “speed doesn’t matter”, I did so in the context of the automatic machines that are sold today. The slowest I’ve seen is 2000 parts per hour, which is about 15-per-hour or 100~200 per day of my larger (6" square) PCB and probably 500~1000 per day of my smaller 3" square PCB. My impression was, your PCBs are even smaller than 3" square, so — you must have a very nice volume of orders! Major congrats about that!

My product will sell for about $950 each (with 1 large PCB and 4 small PCBs), or $100 less for each small PCB removed (each small PCB is a camera). Since the biggest savings my customers will realize exists for applications that need to view through 4 cameras with one RJ45 gigabit ethernet interface, I suspect most of my sales will be the 4-camera == 5 PCB configuration.

Therefore, If I’m selling more than 100 of the 4-camera systems every day ($100,000+ per day revenue), I will be a very happy camper indeed. Since that calculation assumes the slowest automatic pick-and-place machine (that I’ve seen), that’s why I said “I don’t care about speed”.

Of course, if I’m getting more than 100 orders per day, I will be even happier - because I’ll get the opportunity to buy an additional $100,000+ machine to dedicate to production, while the “old one” gets dedicated to engineering prototypes.

All I can say is, I am very impressed with the volume of orders you are receiving. I don’t know anything about the market you serve, but I am very happy to hear it is that large and vigorous. The only other possibility that might be “slowing your machine down” is… you’re trying to operate on a “just in time” basis, in which case you may be spending much of your [machine] time switching over from one PCB to another. If that’s true, your volume is lower than I was just now assuming (immediately above), but you’re running your biz in a different way than I envisioned. Personally, I am strongly inclined to operate my business opposite of the “just in time” (smaller batch) manner. I’d rather have investment in products sitting on the shelf waiting for orders than the lower efficiency of the “just in time” scenario I just described. Having said that, this is “easy for me to say”, since I’ll only have 2 PCBs (at the beginning), while you have about 23 zillion. So probably you’ll now explain why my approach is totally insane for your biz, which very well might be true.

After spending all day and night yesterday looking into this issue further, I tend to agree with you about costs. Since I’m not the type of fool who is willing to “take a flier” on a used machine, and cross my fingers that “I can make it work efficiently” (not on this type of device, which is “fast developing”), it does appear you are correct about costs. My latest estimate is:

$35,000 for pick-and-place including accessories (APS or MCD or ???)
$ 9,000 for low-end-but adequate c4-zon conveyor reflow oven (APS)

So, by the time I buy a modest set of rework equipment (PCB warming bath, PCB holder, hot-air-pencil, stereo microscope, etc)… we’re talking $50K to $60K. If I was 100.000000% certain the product would be successful when released, I’d bite the bullet and “buy it”. I do have enough savings, but it violates my “rule” to never invest more than 25% of my savings in any new project.

So right now (for the next 10 minutes probably), I’m thinking I should try the “cheapest way” in the short run, to get the prototypes I need to develop the firmware and software, then go “whole hog” in January~March of next year when I start getting a reasonable stream of orders. I’m happy to buy anything now that I KNOW I will need eventually, and that includes all that “benchtop manual rework and assembly” junk I mentioned above, which is a few thousand at most (assuming I wait on the reflow oven).

What I have a very, very, very hard time convincing myself to do is opt for the AAPCB alternative. Why? Their prices are quite reasonable compared to their competition, but I’m looking at $3500 to $4500 to have all 25 + 50 of my prototype PCBs assembled. And that expenditure gets me ZERO equipment that is useful in the long term (and I’m sure I still need to buy the desktop manual rework and assembly gizmos anyway).

HOWEVER, if I was [virtually] certain I would decide to get my production “made in China”, then I wouldn’t need anything except the desktop rework equipment here, and I’d be happy to give the work to AAPCB. Have you any experience or knowledge about that option? I worry about quality, and I worry that perhaps the “cost per PCB assembled” is very attractive, but the “scrap cost” due to bad assembly is obscene. And “worst of all”, I fear they’d drop-ship untested or unreliable GARBAGE to my customers - which I absolutely will not tolerate. To keep everything “in house” is clearly the most certain way to assure quality.

So, anyway, that’s my latest thinking… and dilemma.

Keep coming with the great ideas and inside information. I appreciate it.


#18

One follow-on question, about stereo microscopes for PCB inspection. Do you have a stereo microscope for PCB inspection, or you find some kind of illuminated magnifying glass sufficient?

In either case, I worry a bit about the LED ring these gizmos have to illuminate the work. Since the light is shining directly down from around the objective lens onto the PCB and components, I wonder whether that generates glare that makes inspection more difficult. While lights off to the side probably have their own problems (namely “shadows”, unless we have 3 or 4 spread around), I’d expect such side-lighting to be superior.

Any experience or opinions on this?