-
As I dig through the feeders I am realizing how important it will be to splice the tapes since the leaders are needed to thread a feeder.
Does it need both sides of the tape as leader ?
I looked at http://www.tapesplice.com/tapesplicerandtoolkit.htm but the starter kit is $600. Seems like it works well, but I was wondering if there is a less expensive way to get started. I have lots of high value parts in short strips that I have been using in my manual tape holders.
Anyone have any SMT tape splicing tips?
Remember Digikey/Mouser do a reeling service to add leaders to cut tape, subject to a MOQ
I'd think you could probably do short tapes from a fixed feeder like a tray.
-
Digi Key does the splicing as you mentioned for about $7 per reel. I would not take long to justify the splicer I guess. My current set of PCB's have a total of about $175 in components. That value keeps me from buying in full reels at this time. About 20 components would be subject to the "Digi-Reel" fee for a month of production - so about $140 in reel fees. That is not too bad actually since it arrives ready to rock.
I have about 100-150 cut strips already delivered that would be nice to use. I can buy or build a cut tape feeder as well, but that would probably be just as much effort as splicing and loading into a regular feeder. I seems like a great thing to have, but hoping to find a 'beginner' priced tool for now.
It needs the carrier tape and the cover tape to thread - seems like about 6" or so is about right.
-
That's one downside of using "proper" machines - they assume you use everything by the reel...
Though I've seen some short-tape feeders on Mydata - not sure if these are standard or a special variant
-
Yes, "proper" machine can mean "proper" prices. This Quad has a vibratory feeder for parts in a tube. The tops are interchangeable for different sizes. Of course mine was delivered with PLCC version and I need SOIC-8. I got prices of $995 from two sources. It is a flat piece of aluminum with 8 shallow slots milled. EXTREMELY basic manual machining job for $995?
All the parts are expensive. The reason this is still a good idea is because I have been able to do the work. If a technician came out to get thing up and running, it would not likely be worth it. Or, at least I could not have of afforded it.
I will feel better when PCB's are being made. For now, it still seems like a great way to 'graduate' from hand built prototypes and small batches to the next level. I can tip-toe into higher sophistication and higher volume. -
How short are your strips? You can program the machine to do picks in an X/Y (or just X or just Y) grid which is how the machine handles waffle trays. And since it has vision correction, all you really need to worry about is getting a clean pick on the part.
If it were me, I'd take a scrap of aluminum and just machine several slots in it, mount that on the machine and pick from it. Or if you don't have that option, you could just use a ruler and line the cut strips up on a piece of flat aluminum and just use double sided tape to stick them down.
Moving forward, I like the Digikey reeling service and I use it for high value parts that I dont want to buy full reels of. There are also always local companies that will do reeling, even of loose components. -
Oh and as far as splicing... I would never pay $600 for a tool that didn't do the job almost completely automated. Reasons are..
1) I rarely have tape feeding problems, but it's always with spliced reels. And I've had it from numerous sources, so it's not just one supplier - it's just that spliced tapes don't feed as reliably as intact tapes.
2) Splicing reels myself is super easy. A rubber cutting mat, sharp scissors, some kapton tape and about a minute of your time produces just as good results as "pro splicing". You can get Kapton tape at Uline
http://www.uline.com/Product/Detail/S-14532/High-Temperature-Tapes/Kapton-Tape-1-Mil-1-8-x-36-yds
and just a 2" strip of Kapton on the top and bottom side of the cut strip is fine. I haven't noticed any difference in reliability between Digikey/Mouser spliced reels and the ones I do myself.
You can buy blank tape with no components (but with the plastic top cover), but most reels have blank tape at the end of the reel also. And all reels have some at the start... the Quad 4C feeders need maybe 6 inches of blank reel, and most components will have a fair bit more than that.. just trim it off and there is your lifelong supply of blank-but-covered tape. -
Just found this video on the PPM'd IVC
http://lowpowerlab.com/blog/2014/05/13/ppm-visit-and-quad-ivc4000c-pick-and-place-demo/ -
Just found this video on the PPM'd IVC
It has the Windows 7 logo so it must be cool
-
Oh and as far as splicing... I would never pay $600 for a tool that didn't do the job almost completely automated.
Well, to be fair, the tool is half that. The rest of the cost is the consumables ranging from 8mm to 56mm. Seems simple enough to deal with flat paper tapes but the deep pocketed plastic tapes for bigger components like inductors and push buttons seem harder. A website called http://www.nozzles4smt.com/ has a cut tape carrier that is adjustable to use various width tapes. Looks promising for my kind of work. They also have other consumables for Quads - nozzles, rollers, etc.Just found this video on the PPM'd IVC
http://lowpowerlab.com/blog/2014/05/13/ppm-visit-and-quad-ivc4000c-pick-and-place-demo/
I saw that video shortly before finding this machine. His operation seems similar to what I am doing. Wonder how much he paid for the package. It looks like after my machine is all dialed in and ready to go, it will be around $8k plus about 40-50 hours of labor working the various problems out. The list for everything being over $50k (maybe PPM would have given me a 'deal') seems like it is a pretty good deal so far.
The feeders have been all inspected. I made a log of them to keep track of various issues. In general, they all seem to be in working condition with the exception of the part called a Peel Roller. This is a urethane roller that pulls the cover tape off. Very simple, but for some reason, MANY of them melted like candle wax. Looks like they were not chemically stable from the beginning. I have never seen such a thing from urethane.
This is what I saw.....check out the goo.
Some were even worse....not easy to clean up the mess.
I will post more later....
-
Ok, I had to go pick up the kids.
So I now have a list of the replacement parts I need to get the feeders working. I was pleasantly surprised that the urethane rollers were really the only problems. The circuit boards, mechanics, gears, etc were all working well. Many of them had components rattling around inside. This was mainly the 8mm ones that deal with the small parts. Some had tapes tangled up inside as well. All in, it was about 10 hours to go through all the feeders, log, inspects and test. It was mind numbing, but necessary. I wanted to send them to PPM or anyone else that services feeders, but it was out of my budget. PPM charges $75 to inspect a feeder, another guy in southern California offered a refurb service for $125/ea. With the number of feeders I have it would cost $4,350 to have them all looked at - not including the replacement parts. It would say that the 10 hours I spent learning how they work and cleaning was well worth it.
I learned a lot about how they work and overall I would give the design a B+ grade. These feeders appear to have been abused pretty bad, yet they still work just fine. That is worth a good grade. They also have the mechanics exposed too much so dust and parts can fall into the PCB and gears which of course is bad. I cannot quite understand the melting roller problem. I think the previous owner may have sourced these parts from some shady place. Many of the feeders had hardware store mods to the rollers using grommets or O-rings to rig up the feeders. Some had scotch tape around the rollers that were going bad. I will be maintaining these feeders and the machine far more meticulously than that. Not because I am OCD, but because I need this machine to be reliable and a clean and well maintained machine is the beginning of having a reliable machine. That lesson was learned in my CNC machining days.
So, when the feeder parts show up this thing should be ready to roll. I am now focusing on how to program it. There are a few videos that PPM has made and they are helpful. I have the challenge of learning the software without knowing much about PnP which will no doubt make things more challenging. When I decided to get into CNC machining, I first got a CNC machine and then started looking for the power switch.
Here is a pile of melted rollers from the feeders. About 20% of them looked like this.
This is the setup to clean the feeders. Lots of coffee as seen on the left.
The feeder with the cover removed reveals the PCB and the gear train. The belt is a consumable, and sheds all over the place.
Two of the feeders were a little different and the PCB said Samsung Aerospace. Didn't see that coming. I thougt Samsung made competitive machines.
-
The pick and place machine is now up and calibrated. Technically this means that it is ready for programming. The problem is that I have no idea what I am doing.
I get the concepts - pickup part, place part, pickup part, place part. There are, as I have discovered, quite a few details that need attention. The programming manual is 390 pages and the 'introduction' to programming video is 4 hours and 20 minutes. Needless to say, I have a bit to learn.
From the overview of the materials, it seems that a lot of it is targeted to assembly houses that are constantly changing the whole machine. In my case, I will get a setup dialed in and rarely change anything. The way it is designed is very flexible allowing anything to be located where ever I want it. While I appreciate that flexibility, it comes at a price. When things move, they have to be calibrated and verified. It is not difficult, but it takes enough time that I will try to avoid changes if I can.
The software seems to be stable so far, but it reminds me of the learning curve experience with Eagle CAD. It is clunky and not intuitive. It does not really go with typical Windows conventions. For example, there are buttons with fly out menus. At first glance, the menus all look like buttons and there are buttons that are just buttons. You have to just know which buttons are really a menu. Also, there is not really a logical flow that I can tell yet. Bear in mind, I am a PnP newbie as well so maybe it will be more apparent when I am more versed in the task. I have yet to find any kind of a work flow to help guide a new board. The videos and the printed documentation do not seem to cover the task as a flow - they just go from menu to menu describing what functions are available.
I love the software flexibility that I see so far. There are many defaults that the user can define, but almost everything can be overridden. This is very similar to what I need in CNC machining software. The defaults get me in the ballpark very quickly and while still allowing me to finely tune every detail if needed. For example, you can control the vacuum pressure on each and every part as well as the delay time to allow the vacuum to build. Handy for heavy inductors, but not needed of 0402 passives. you can easily control speed as well to avoid having a heavy or wobbly part from slipping after it has been through the vision system which would result in a bad place.
Not sure how other commercial machines deal with alignment, but this one has two methods. The 'Quad Align' is a system that is attached to the placement head and deals with alignment in while in motion. It is essentially, a line scan camera on one side and a light source on the other side. It looks at the shadow of the part while spinning around. The Quad Align system can detect X/Y/T and Z offsets relatively well. It is fast and does its whole routine while traveling from the feeder to the PCB. For fine pitch parts and BGA's the you can choose the upward facing camera to get really precise placements. This is also where the flexibility of the software shines. You do not have to choose one or the other - you have have an attempt with Quad Align, and if it fails the system can do a camera alignment. You can also define what to do if it fails alignment - try again, try x times then stop, or just stop.
This level of detail is why I wanted this machine over something like a TM240a or one of the other low-cost options. This machine can deal with just about any board and any part with ease. As I get past the weirdness of the software design, I feel like it will be quick and easy. I did write a script to get my data out of EAGLE in a way that perfectly matches what the machine needs - including nozzles and package types. This will slip into the machine defaults nicely and get programs up and running fast (I hope)
I have no reference for the DOS versions so I cannot say if any features in my Windows upgraded version are new or any better.
All the feeder parts arrive tomorrow, so I should have 100% of my feeders ready to go. -
Glad it's (nearly) working.
You will be jumping for joy when you first see the machine working, and will instantly realize why a "real" PnP is such a difficult thing to build.. .it's all the little things that make it hard.
On your feeders, I bet someone used something like toluene or MEK to clean them... I've never seen urethane just melt like that, but if it's not cross-linked, it will just absorb the chemical and over time, turn to crap. Did you end up buying rollers @ McMcaster? What specific parts did you buy, if I may ask?
As far as programming, a couple of tips... when you program your pickups, you want to get to where the nozzle is just about touching your part. The easy way to tell is to get it close and turn on and off the vacuum. As long as you see the part jump up a little, the nozzle isn't touching. Jog the Z down a few thou until the part doesn't move when turning on/off the vacuum... then jog down maybe 5 or 10 thou more for a good pickup location.
With placements, take a blank PCB and apply thin double sided tape. Jog the part down slowly until you just see the Z spring/bellows start to compress, and set that as your placement Z level.
I've never had to adjust the pickup pressure - but you do want to be sure you are using the right nozzle. For larger/heavier parts, use the nozzles with the rubber grommets that seal against the part (SOIC's, inductors, etc). And make sure you have the offset between the nozzle position and camera position set dead-nuts accurately. This is critical when you are using the camera and crosshairs to set your X/Y location for pickups or placements. Take the time to get it perfect.
Rather than trying to get a whole board done first, take a blank PCB and use the double sided tape and just try to write a program to place one part - then a few parts, then a few different parts. Then figure out the repeats... component repeats and PCB repeats so you can do groups of components and panels of PCB's.
Once you have that down, you pretty much have it all. I don't bother failing over from Quad-Align to the upwards camera... Quad Align is 99.9% accurate and is used for all passives and SOIC's. If you get into bigger stuff like QFN's or TQFP's, I'd use the upwards camera 100% of the time - it's slow compared to Quad-Align but accuracy is more important for higher value parts. -
With placements, take a blank PCB and apply thin double sided tape. Jog the part down slowly until you just see the Z spring/bellows start to compress, and set that as your placement Z level.
How accurate is the baseplate of the machine where the pcb is mounted in the Z axis, compared to the head.
Of differently stated, do you have to do this calibration for all parts or just some at the edges and in the middle of the pcb? -
Don't know about the quad, but once the PCB height has been measured, it should be able to do this automatically from the component height data in the part library ( which is also used to determine the focus height for the fixed camera).With placements, take a blank PCB and apply thin double sided tape. Jog the part down slowly until you just see the Z spring/bellows start to compress, and set that as your placement Z level.
How accurate is the baseplate of the machine where the pcb is mounted in the Z axis, compared to the head.
Of differently stated, do you have to do this calibration for all parts or just some at the edges and in the middle of the pcb?
It would be a major ball-ache to have to do this for every part.
-
With placements, take a blank PCB and apply thin double sided tape. Jog the part down slowly until you just see the Z spring/bellows start to compress, and set that as your placement Z level.
How accurate is the baseplate of the machine where the pcb is mounted in the Z axis, compared to the head.
Of differently stated, do you have to do this calibration for all parts or just some at the edges and in the middle of the pcb?
I haven't measured it with a dial indicator, but one of the critical adjustments of the machine is coplanarity of the X/Y axes in relation to the base plate. From seeing how my machine reacts, it can't be out more than a few thousandths of an inch, if that.
PCB's have a fair amount of spring in them, especially larger boards or panels towards the center. The Z-axis of the Quad has a spring loaded bellows system (basically just two rods that slide inside one another with a spring between them). You want to program your parts such that during placement, the Z spring will slightly compress - this ensures that the part is always being pushed down into the paste regardless of variations in PCB thickness or coplanarity offset errors.
As far as setting the parts, you set pickups and placements. A pickup is essentially 4 datapoints - part number, X, Y and Z. A placement is 3 datapoints - X Y and Z. So you can set the placement height for every part, and in general when programming a board you need to set different Z values for something thin like an 0803 vs something thick like an SOD-123. However you normally just copy/paste the Z value from prior placements - I don't think you would bother to change it unless there was specific reason.
EDIT: Something to keep in mind is that with the standard conveyor system, the PCB is on rails and is pushed upwards to lock it in place - meaning the top surface is always in the same plane, whether the PCB is 1.6mm, 0.8mm or whatever.
With a fixture system where the PCB is placed down on a flat location, variations in Z height would need to be accomodated. However, there are "mod codes" (aka - settings) for global X/Y/Z offsets, per-pickup X/Y/Z offsets, per placement X/Y/Z offsets, per program X/Y/Z offsets - and you can change the program itself. -
You will be jumping for joy when you first see the machine working.....
I will go straight through the roof with excitement! This has been a LONG time coming. For me, its not just getting the PnP, it is the beginning of having a complete electronics design business. I am managing to make a living even while hand assembling all the PCB's. Moving into assembly automation will give me a LOT more time to focus on designs and continuing education.On your feeders, I bet someone used something like toluene or MEK to clean them... I've never seen urethane just melt like that, but if it's not cross-linked, it will just absorb the chemical and over time, turn to crap. Did you end up buying rollers @ McMcaster? What specific parts did you buy, if I may ask?
I will never know exactly how they came to be melted globs for sure. It could have been that some random urethane house made them and the chemistry was never correct and/or the end user as you say used some cleaner that un-linked the polymers somehow? Definitely not an area of knowledge for me. As I spent considerable amounts of time cleaning up the mess, I decided to buy the 'real thing' from PPM. I have saved so much money on this project by learning and doing everything on my own, that I should just buy the right parts instead of trying to fake it.As far as programming, a couple of tips... when you program your pickups, you want to get to where the nozzle is just about touching your part. The easy way to tell is to get it close and turn on and off the vacuum. As long as you see the part jump up a little, the nozzle isn't touching. Jog the Z down a few thou until the part doesn't move when turning on/off the vacuum... then jog down maybe 5 or 10 thou more for a good pickup location.
With placements, take a blank PCB and apply thin double sided tape. Jog the part down slowly until you just see the Z spring/bellows start to compress, and set that as your placement Z level.
I've never had to adjust the pickup pressure - but you do want to be sure you are using the right nozzle. For larger/heavier parts, use the nozzles with the rubber grommets that seal against the part (SOIC's, inductors, etc). And make sure you have the offset between the nozzle position and camera position set dead-nuts accurately. This is critical when you are using the camera and crosshairs to set your X/Y location for pickups or placements. Take the time to get it perfect.
Rather than trying to get a whole board done first, take a blank PCB and use the double sided tape and just try to write a program to place one part - then a few parts, then a few different parts. Then figure out the repeats... component repeats and PCB repeats so you can do groups of components and panels of PCB's.
Once you have that down, you pretty much have it all. I don't bother failing over from Quad-Align to the upwards camera... Quad Align is 99.9% accurate and is used for all passives and SOIC's. If you get into bigger stuff like QFN's or TQFP's, I'd use the upwards camera 100% of the time - it's slow compared to Quad-Align but accuracy is more important for higher value parts.
Thank you! The printed docs and video don't even come close to offering tips like that. Taking baby steps will keep me from destroying things and save time. Resistors are nearly free.Don't know about the quad, but once the PCB height has been measured, it should be able to do this automatically from the component height data in the part library ( which is also used to determine the focus height for the fixed camera).
It would be a major ball-ache to have to do this for every part.
The Windows software has a library of part packages with height defined. The script I wrote to get the placement data out of Eagle matches the Eagle package with the PnP library, so in theory, the Z height is pre-defined for each and every part before I push a single button. With a new part, I would have to create the library entry for it. This also matches the feeder type required for that part. If all works as designed, I should be able to import data and the PPM software should be able to assign a feeder, a nozzle, and know the Z level. All I need to do is get the PCB physically in the machine. In practice, I am expecting that tweaking will required. Hopefully only a little though.
For the PCB flatness, as Corporate666 says, it will flex. The machine itself is very flat and predictable but the PCB is less so. The Zaxis is on a spring so it can deal with variation in actual Z-levels. The 'mod codes' have been replaced with 'offsets' in the Windows system - but it sounds like the same thing. You can make global changes, changes to a group, changes to a package, or changes to an individual pickup.
-
Thanks all for the answers and insights.
-
The Windows software has a library of part packages with height defined. The script I wrote to get the placement data out of Eagle matches the Eagle package with the PnP library, so in theory, the Z height is pre-defined for each and every part before I push a single button.
and of course you want to set things up so parts in your PCB libraries match the zero rotation of parts in the feeders, so things end up the right way round by default.
Something I would advise is to use different part names for different types of parts in the same package, e.g. 0805 resistors, caps, LEDs should use different package names in your PCB library, with corresponding parts in the P&P library, as things like vision parameters can differ significantly. You probably also want at least "thick" and "thin" variants for capacitors.
-
RX8,
What drove you to in-house P&P? Is the cost of doing small runs on the outside so high that this is a much more economical solution? I'm not questioning your decision, just curious. I've not had to walk this path on my own yet, but its coming. -
RX8,
What drove you to in-house P&P? Is the cost of doing small runs on the outside so high that this is a much more economical solution? I'm not questioning your decision, just curious. I've not had to walk this path on my own yet, but its coming.
Its a fair question. I have been making small batch specialty PCB's that tend to change a lot. LOTs of prototypes, and pilot runs. As the designs have become more mature and sophisticated, they have become more dense and use more fine pitch parts. Hand assembly was getting harder by the second so I looked for a few assembly houses to see what the options were.
The good news is that there are quite a few small assembly shops a short distance away. I met with a few and showed what I have and got quotes. The prices were fair, but the financial penalty was rather severe if I wanted to run 10-20 pcb's on a quick turnaround to do some tests. It did not make much sense financially until I got to around 100pcs and could wait 2 weeks for them. In addition, I would have to kit the project and document all the details which takes a considerable amount of time. What I really wanted, was to have complete control and flexibility to to prototypes with BGA, 0201's, double sided, etc put the PCB on the bench and shake it down. Order the next revision of the PCB and put it back into the same setup on the machine and run 25 more within a few hours and on my schedule.
When faced with the cost of a 'real' machine it was hard to justify. I was expecting to spend well over $50k for a basic machine. I already have a stencil printer and batch oven so I only needed a machine with feeders. The Quads looked like a great place to start because I can pile on a TON of feeders so it can be setup for multiple PCBs at any given moment. They are also supported by PPM and a few others. When I found this machine for only $5k including the feeders, I was able to snatch it up quickly. By the time I finish, I will have put in another $2500 or so for parts plus about 60 hours of labor. Really not bad at all for the ability to assemble PCB's at any level of complexity I will ever see. With a rating of 3600CPH (maybe 3000CPH in real life), it will outrun our sales for the foreseeable future too.
I did the same thing when I built a CNC machine shop nearly a decade ago. It was awesome to have the flexibility and total scheduling control over our parts. Most CNC shops (understandably) can't deal with someone barging in that needs parts or a prototype in a few hours. Having these things in-house was also a FANTASTIC learning experience as a design engineer. I actually programmed, setup, and ran the machines. I made fixtures and tools. I studied the process in detail. When I sat down to design a new product, I knew exactly what colors I could paint with. I knew what was easy and what was hard.
This is a bit of a ramble....hope that it answers the question. I am guessing that the cost of the machine will be covered in about 3 months, the labor I have put into it will be returned in about a month. That is about the fastest ROI I could have dreamed of. The time savings will be split between new designs and sales efforts. So, I see this purchase as a critical pivot point in my business that will allow us to get over the hump and grow up. -
That's a totally legitimate answer and it sounds like you've really thought through the "payback." Thanks!
-
RX8,
What drove you to in-house P&P? Is the cost of doing small runs on the outside so high that this is a much more economical solution? I'm not questioning your decision, just curious. I've not had to walk this path on my own yet, but its coming.
Its a fair question. I have been making small batch specialty PCB's that tend to change a lot. LOTs of prototypes, and pilot runs. As the designs have become more mature and sophisticated, they have become more dense and use more fine pitch parts. Hand assembly was getting harder by the second so I looked for a few assembly houses to see what the options were.
The good news is that there are quite a few small assembly shops a short distance away. I met with a few and showed what I have and got quotes. The prices were fair, but the financial penalty was rather severe if I wanted to run 10-20 pcb's on a quick turnaround to do some tests. It did not make much sense financially until I got to around 100pcs and could wait 2 weeks for them. In addition, I would have to kit the project and document all the details which takes a considerable amount of time. What I really wanted, was to have complete control and flexibility to to prototypes with BGA, 0201's, double sided, etc put the PCB on the bench and shake it down. Order the next revision of the PCB and put it back into the same setup on the machine and run 25 more within a few hours and on my schedule.
When faced with the cost of a 'real' machine it was hard to justify. I was expecting to spend well over $50k for a basic machine. I already have a stencil printer and batch oven so I only needed a machine with feeders. The Quads looked like a great place to start because I can pile on a TON of feeders so it can be setup for multiple PCBs at any given moment. They are also supported by PPM and a few others. When I found this machine for only $5k including the feeders, I was able to snatch it up quickly. By the time I finish, I will have put in another $2500 or so for parts plus about 60 hours of labor. Really not bad at all for the ability to assemble PCB's at any level of complexity I will ever see. With a rating of 3600CPH (maybe 3000CPH in real life), it will outrun our sales for the foreseeable future too.
I did the same thing when I built a CNC machine shop nearly a decade ago. It was awesome to have the flexibility and total scheduling control over our parts. Most CNC shops (understandably) can't deal with someone barging in that needs parts or a prototype in a few hours. Having these things in-house was also a FANTASTIC learning experience as a design engineer. I actually programmed, setup, and ran the machines. I made fixtures and tools. I studied the process in detail. When I sat down to design a new product, I knew exactly what colors I could paint with. I knew what was easy and what was hard.
Totally agree with all the above - the reason I got a P&P was that it was a pain to deal with subcontractors for small jobs, and I often needed fast turnrounds (like design-start to delivery under a week).
In my case it wasn't much about cost as most of my customers could have taken that, but about the time & hassle of kitting and documentation.
There can also be issues of only finding out late that there is a problem - e.g. in on case a part got missed off the BOM, so didn't get fitted - if I'm assembling a board I designed a few days ago, I'd have spotted it straight away.
Also totally agree about the learning process - seeing the assembly process close-up is valuable input to future designs.
When I do subcontract out larger jobs, my local place now doesn't charge me any setup fee as they know it will all be ready to run.
Another issue I had was that subcontractors would only use their normal full-size metal stencils at GBP170-odd each, whereas I can use smaller frameless metal or plastic ones for small runs.
-
RX8,
What drove you to in-house P&P? Is the cost of doing small runs on the outside so high that this is a much more economical solution? I'm not questioning your decision, just curious. I've not had to walk this path on my own yet, but its coming.
I know you didn't ask me, but if I may just horn in with a real world example.
I used a Chinese based PCB assembly service in years past. The cost of assembly is always hard to quote without looking at the board because there are so many variable, but anyway, this particular PCB house has a standard base rate of $0.0125 per pad (regardless of single or double sided, or large or small parts). So 1.25 cents per SMD pad.
I have one product that just has a bunch of PLCC-4 LED's on it, some passives, a couple of chips and associated regulation. 30 individual parts in total. There are 100 pads total on the board (SOIC-8 is 8 pads, for example). So that board would cost me $1.25 to be assembled. There is also a setup fee of something like $150 or thereabouts.
I used perhaps 5,000 of these boards per year. Based on the setup fee and assembly price, it makes sense to order around 1,000 of these assembled boards at a time. That entire production run will use a half reel of some parts, 2 reels of other parts, 10 reels of LED's.
With all due respect to the Chinese, I don't trust them sourcing parts for me at all. Even when I have specified (for example) a National Semi branded 7805 regulator, I've had staggeringly high failure rates and do not believe the chips were genuine for a moment. And passives are so cheap that it is absolutely foolish to try to save $4 on a production run using Chinese sourced parts. So long story short, I would buy reels and ship them off to the fab in China. That means the $2,000 I spent on a reel of 2,000pcs SOIC-8 microcontrollers will only get half used... and it will be sitting in a factory in China for 2-3 months until my next production run.
Net/net, my run of 1,000 PCB's will cost me $4k or $5k when you add in the cost of all the parts, setup fees, assembly fees, shipping both ways, etc.
I have three Quad 4C's that I paid around $6-8k each for. In the course of a single year, the cost of the machine is more than paid for just by what I would have spent on those 5,000PCB's - which would have been over $6k, before setup fees, shipping costs etc.
The additional benefit I get is that I can buy my reel of $2,000 SOIC-8 microcontrollers and leave it on the machine - and if I need to do an emergency run of 50 PCB's, it's easy. If I need to grab 5 of them to repair PCB's that were bad, it's easy. If I have another product using the same parts, I don't need to inefficiently have thousands of them sitting on a shelf in China while I am hand assembling other boards with more of the same parts I had to buy separately.
And it's not just high value parts like MCU's... even "cheap" parts like power inductors at $0.25/ea get expensive when they come in reels of 3,000pcs and there are several equally expensive parts on a board.
So in a nutshell...
-Saves money
-More flexible to your needs
-Allows you to buy parts "just in time" vs warehousing dozens of reels of parts
-
Corp666,
No problem. Thats great reasoning. Thanks for sharing. -
How is the machine going?