I can't wait until they come out with 3D printers with outputs in FR-4, bulk copper (thin enough in resolution for conventional tracks, plus buildup), and ferrite loaded and dielectric loaded materials.
Who needs bypass caps? Fab your supply planes around barium titanate composite. Not only will it have a much lower impedance, the lossiness will keep high frequency squigglies down.
Sensitive RF sections? Pull your vias up through a ring of MnZn ferrite composite. Heck, weave traces in and out of the stuff, get as much impedance as you like.
RFI protection? Alternate ferrite and dielectric sections to build filters 'til your heart's content. Twenty pole Bessel? No BOM cost. Delay line? Sure, how many *micro*seconds ya need? Shock lines? Entirely possible.
Obligatory party pooping:
- Yes, the mu/k of composites will suck.
- Bla bla, there's no way it'll be cheap enough or good enough to work.
Sure. Should be trivial to put nichrome feedstock into the machine instead of copper. For certain values of "machine".
You can even make your supposedly fancy zero-inductance resistor pattern for every damn resistor you want. :)
Likewise, OLEDs need only a suitable combination of coatings; those could be applied today, with the caveat that you pay for extra layers and chemicals (and you need a board house that has them..). And you need the moisture tolerance (epoxy overcoat??) to make them last worth a damn. So, lots of layers. Would suck for indicators, but making a few million semi-graphical display boards, it would be worth it.
Leads to the next question, semiconductors? You can do shitty transistors in amorphous silicon, for instance on glass, but I don't think those can be done without vacuum, so that's an even more expensive proposition than mere 3D printing. (So, why not run the whole 3D printer in vacuuo...??!!)
A really thin copper ribbon that gets stuck down to a substrate, with lots of them in parallel for big ground planes or wide tracks?
A selection of copper ribbons in different widths, perhaps with some ability for the machine to trim wide ones before sticking them down?
A piece of copper foil as big as the board that gets machined out in a separate process and then stuck down to a substrate?
Copper particles in some kind of goop that can be squirted onto the board? The goop either then goes away or turns into glue. Probably hard to get a low enough resistance doing this, but might work for some "signal" traces.
Molten copper and some kind of ceramic nozzle to squirt it where you want it? Hopefully the substrate won't melt.
How the hell do you keep the people at the board house from stealing the copper ribbon / foil / ingots from the feed hoppers and selling them for scrap? :) (Okay, this is a solved problem: the machines that do gold-plated edge connectors have the same issue.)
Okay, the can of composite goop has a lot number, guaranteed analysis, conflict minerals certificate, and everything. How do you prove that once it's in the board? Not every application will care about this level of traceability, though. It doesn't have much to do with it
*working*, either, but manglement sometimes insists on it.
Is it repeatable, or will the board house start to water down the composite so that either the capacitance or the withstand voltage goes down? How can you tell?
It will be priced by the board house to be some fraction of what the discrete-component filter would cost, but not zero.
I am thinking you might end up with a 2" x 4" board that has, like, $5 of BOM for a few discrete components and connectors. But the board itself costs $2000.
It would be the ultimate in non-repairability. But if the 3D printer could crank out boards cheap enough, you might not care about that.
Somebody would eventually try something like using a special proprietary mix of composite goop to make an in-board resistor or capacitor (or network) with weird properties. At bootup, the onboard D/A and A/D measure these components, and if they don't have the right signature, the whole thing shuts down. An attempt to slow down the cloners a little bit. (At least until they disassemble the code enough to JMP around the offending test.)
You can pretty much do this trick today, but: exactly the same firmware and CPU in all the boards, but the Model 100 has a 100K in-board resistor from every other CPU pin to ground, while the model 200 has
200K instead. At boot, you decide which model you are, and disable or enable features accordingly.
I wonder how you'd prove things like standoff distances or voltage withstand levels for line-connected power supplies. (Go look at the
17 different agency logos on your average X or Y capacitor.)
In some aspects, this already exists, as long as you want zillions of them and don't have to handle great heaping gobs of voltage or current
- you get somebody to design a chip for you. But you're talking about lower quantities than that, I think.
I think it'd be a cool thing to have, but there are a lot of challenges.
I'll drop a dog turd in this punch bowl. Sorry to spoil the party, but this e-printing fantasy is impractical. Consider what it took to develop the specific manufacturing methods to make 8gb usb drives for $10. The transistor was invented in 1950 era, and the usb drives took 50 years to develop. That's say 25,000 engineers working 50 years on this ONE SPECIFIC manufacturing problem. That's a MILLION MAN YEARS of manufacturing engineers or maybe 100b $. The sad truth is that manufacturing is a hassle. All the sins of the forefathers come to roost in the front line trenches of manufacturing. j PS - There is no Santa Claus. I'd like to be positive about this e-manufacturing, but it violates my "optimism budget," ie be practical about your hope and change. Corollary: In design, be skeptical about gee-whiz or complex products that might be difficult to manufacture. Try to design around them.
You're thinking in 2D. This is 3D, so building a 3D PCB is possible. Think of it as the ultimate buried via. You could do in layers, but I suspect free form would be better. Be the first to build a spherical PCB.
One nice thing about 3D is that you can build in impossible to machine cavities and channels. Use those for liquid cooling channels. With some creativity, you might be able to build a plastic cooling pump into the 3D PCB.
There's probably no way we're going to be able to extrude material suitable for making components directly onto the PCB. Well, maybe borrow some of the techniques from the 2D printed inkjet circuits people. Meanwhile, it will be just another form of pick and place with the added bonus of vertical constructs. For traces, I would be tempted to suggest stitch wire or what is now called Hitachi Multiwire. Basically, a sewing machine for copper wire.
Building a complete PCB with a 3D printer is only the beginning. There's little to prevent growing the technology into printing the entire product. Just add more plastic goo to the cube shaped PCB and you have a module or an enclosure.
None of those are very important. What will cause problems is construction speed, or rather the lack of speed. Pick and place can do 10 components per second. 3D printing might average 1 component per minute. Something must be done to improve speed. I don't know if anyone is doing it, but using UV cured acrylics might be the answer. The plastic is "spit" from the nozzle and onto the work, instead of physically "printed" by direct contact in order to prevent nozzle clogging. A flash of UV and the dot is hardened. Or, a layer of liquid acrylic can be printed onto the work, the head moved out of the way, and the entire layer is exposed to UV.
If the 3D printer is going to make repetitive identical structures, I see no reason why they need to be built one layer at a time with a universal nozzle. A dedicated "resistor goo" nozzle could be used on some kind of turret.
The real problem is whether there's a need for 3D printing. Right now, it's a technology looking for a problem to solve. I can see some immediate application, such as making plastic keys or copying unobtainable parts, where there is currently no competing technology. However, for manufacturing, it's going to be a long hard battle to make it competitive. The trick will be to have it make products that can't be made any other way. Unless you're building vertical 3D PCB's, that's not going to be easy.
Good luck and don't stop dreaming.
Jeff Liebermann firstname.lastname@example.org
150 Felker St #D http://www.LearnByDestroying.com
It's a clear pattern, from building bridges to writing web code, that the easier it is to iterate a design, the less care the designer will put into getting it right the first time.
PCBs iterate slowly, so people tend to check them fairly well. If PCBs get easier to iterate, less checking will get done and more iterations will become the norm. So maybe no time will be saved on average, and more bugs will be shipped.
John Larkin Highland Technology Inc
www.highlandtechnology.com jlarkin at highlandtechnology dot com
On a sunny day (Thu, 13 Mar 2014 09:36:59 -0700) it happened John Larkin wrote in :
That is only partly true.
Modern tools allow for much more complex designs in a short itme, think PCB layout programs, compilers etc.
Many more new things are created, and yes, most if not all have bugs, or rather 'limitations'. The definition of 'professional' (AV) equipment in my time was equipment that had its limitations exactly specified.
That is a design issue. You cannot agree with Jim on the simplest thing, so even if it was right you would call it wrong and vice versa.
I bought a LG robotic vacuum cleaner. been playing with it today, needed to change and tape all wiring in the living room, it throws over plants, gets stuck on carpet, and pulls wires out of the speakers.... It seems to be open source, so I contacted LG via email, see what they come up with (some of their stuff seems to run Linux). I can think of a zilion(tm) improvements in the software, and hardware too actually.
So, here is a new product (tested here as best, but not by me) that is buggy as hell, or rather has severe limitations, that would not be there if not for cad-cam and what have you, it talks too (muted that shit right away), these things are expensive you know, a simple 3D printer is cheaper. So that is the sate of life the days, OTOH my house has never been so clean... its re-charging it Lipos now.... But using the vacuum cleaner is faster, and still needed. :-)
On a sunny day (Fri, 14 Mar 2014 07:20:20 -0700) it happened John Larkin wrote in :
The UJT idea was not bad, thought of it myself. Modern high efficiency blue LEDs (I have one in use here for signalling) are clearly visible even at low currents, few ms is not much however. I would put a LED in series with the HV, chances are that will light up even if there is a minimal constant load. Easy to protect against higher currents with a few si diodes in parallel.
This is already possible with 2-d methods (seen a pole/zero cost breakdown on a SAW filter lately?). The most common delay lines, for some years, were for PAL TV receivers, and in mass production, milliseconds were cheap.
No idea -- never used em. Good point: I expect those amount to, like, metal-printed quartz or PZT or something like that? So instead of some fraction of the speed of light, you get the speed of sound (of a particular surface propagation mode), which makes it a hell of a lot smaller than RF microstrip stuff in the GHz, which has the same fabrication advantage.
So like that, but at even lower frequencies and higher powers. (What kind of power densities do SAW filters support, anyway? Would/could you use one on a 1kW cell tower transmitter -- assuming you needed skirts that tight on a power output?)
Like Mike said, those were 64us, and my understanding was, they weren't terrifically cheap, at least a first? A big single item cost, at least back when PAL was introduced?
Though they might not've been so bad if they were transistor or hybrid.
I don't know too much about those particulars, not having grown up when NTSC was new, let alone with any sort of PAL device! I have seen plenty of tube sets though, and -- properly adjusted, they seem pretty stable actually. Probably helps that all the old electrolytics and waxed paper caps have to get replaced before viewing is even possible...