Slashdot Mirror
Why the Raspberry Pi Won't Ship In Kit Form
An anonymous reader writes "A post at the Raspberry Pi blog shows an image containing the device's SoC and memory chip to help explain why the tiny PC won't ship in kit form. Clearly, the chips are so small, and the solder blobs required so tiny, that most people would mess up doing it by hand. Add to that the fact one chip has to sit on top of the other, and if you're a millimeter out, your chips are fried."
The post also addresses the use of closed source libraries for graphics acceleration.
BGA packages are intimidating, even to a guy who's been hand soldering other SMD packages since around/before 1990 (that being me)
Plain SMD is easy to do by hand, even the 0402 stuff.
The thing with BGA is its an alignment problem. Some entrepreneur will likely invent a magic clamp that holds the chip in perfect registration to the PCB, at which point it'll be dirt simple to solder BGAs.
I donno where the "if you're a millimeter out, your chips are fried" stuff comes from because thats /.ed. I've done analog microwave RF work where that is actually true. That is not possible on a logic level board. "oh noes, /ce has been grounded, whatever shall we do?" Well just fix the solder bridge and stop whining. Its not like you just shorted out a 20 amp 24 volt power supply thru the bias/bypass network of a microwave FET amplifier, nothings going to blow up on a digital ckt.
"Science flies us to the moon. Religion flies us into buildings." - Victor Stenger
I thought the main reason they couldn't build the Raspberry Pi in the UK was there were prohibitive costs to importing the needed components whereas the completed device was taxed differently.
Isn't this the same problem?
It would probably cost more to package the components for a kit than to assemble the thing anyway, so your kit would not only cost more, it would probably never work anyway.
Your reflow oven would need the correct temperature profile, you'd need a solder paste stencil, you'd also need fresh solder paste of the correct type - because it has an expiry date and should also be kept refrigerated.
HTPCs mess with the signal in all kinds of ways (YUV->RGB conversion is forced, even if you select YUV, it converts to RGB then converts back)
RGB to YUV is lossless in both directions.
Give me Classic Slashdot or give me death!
The chips are not available from digikey, only directly from broadcom, in large numbers.(tens of thousands)
Well unfortunately the Broadcom SoC used in this is only available to official Broadcom partners. So no, the typical person who wants to do this CAN'T acquire the parts.
I can't say I blame Liz on that score.
She's right- unless you're REALLY good, you're going to need all of the things she's talking to- and MOST of us over here won't have them.
I view the only shame being that the state engines are going to be closed source blobs. It seems that NOBODY out there in the embedded/mobile space has the forethought to open things up a bit. First one to open up an ES/VG/etc stack on their GPU will gain a lot of design wins, based on what my clients in the past have indicated.
I am not merely a "consumer" or a "taxpayer". I am a Citizen of the State of Texas
I put BGA's, LLP's, and SON's on boards by hand all the time. I use a hotplate that goes to 260C rather than a toaster oven because I can place it and then look at the edge with a boom-type microscope set at 45 degrees. With a BGA you can see slightly under the package to make sure the balls are on the pads; with an LLP or SON you can see the edges of the chip leads and make sure they're aligned with the pads. Check all four sides, because it's not uncommon to have one side lovely, the two adjacent sides slightly twisted, and the opposite side almost one whole pad off. If you're using leaded solder it'll tend to self-align, but you can't trust that. Bolting insulative material (we used old FR4 boards) to each side of the hotplate prevents you burning your hand from touching it by mistake and provides you a place to steady your hand while adjusting the rotation/placement with a dental pick. We place 0.5mm pitch BGA's (which we call microSMD's) this way on a regular basis, when our BGA rework station is busy or down.
AND! it is possible, although painful, to check continuity to the chip from the board, in many cases, without an x-ray. Check for a diode drop from each node to the substrate.
Nostalgia's not what it used to be.
Have you ever reflowed a BGA package that sits on top of another BGA package, with 0.3mm pitch? It wasn't stuck down a tiny vibration moving it in to your toaster oven could mis-align it.
This is totally incorrect. You only need an oven, stencil, and fresh solder paste if you're doing large quantities of boards. Doing small quantities of boards by hand is fairly easy with some practice. All you need is a syringe with SP, a basic ($100) soldering station, a cheap ($100) hot air rework station, some solder wick, and a flux pen and you can solder just about anything.
I can comfortably and reliably solder parts down to 0.5mm pitch using these tools and I do it all the time with great success. I actually think it is easier than stuffing through-hole parts and generally try to stick with SMDs in my designs. And the smaller the better.
I have a big ass cartridge of Kester No-Clean 63/37 paste that has been sitting in the bottom of my refrigerator for the past year and a half. I fill syringes with it and it still worked just fine the other day. I often need to touch up some of the finer pitched parts by hand, but that isn't much of a problem in small quantities. Having to hand solder a few thousand boards wouldn't be acceptable, and that's where you need that oven, stencil, and fresh SP.
Would I want a kit of the Pi? Probably not. While I enjoy stuffing PCBs and find it quite relaxing, a good chunk of that enjoyment comes from it being my own design and having to do my own programming and troubleshooting to get it to work. The point is I think the suggestion that SMD assembly is outside the reach of hobbyists is total ignorant bullshit.
RGB to YUV is lossless in both directions.
Only if you're working in infinite-precision floating point. In the real world, this is a lossy conversion.