Embedded Systems Engineer's Dream - "PhoneBloks" The 100% Customizable Phone That LASTS - Polidicks

It'd be semi-doable although horrenderously expensive and thick if you stuck the CPU in the baseplate Otherwise good fucking luck

Its not impossible... the cpu block can detect where the other blocks and the pins are

[QUOTE=TNOMCat;42167926]Its not impossible... the cpu block can detect where the other blocks and the pins are[/QUOTE] But it shouldn't be the CPU doing it, it should be an independent bit of code running from the backplate. The backplate is effectively the mobo in this concept.

You could get the idea to work just in the way he wants it. It'd be more practical to just have it so you can swap out parts and replace them easily.

Nothing would ever be invented or revolutionized if it were up to the facepunch kids. Looks like a great idea to me

Can nobody see the pictures in my posts?

[QUOTE=S31-Syntax;42168528]Can nobody see the pictures in my posts?[/QUOTE] Dropbox sucks for hotlinking these days, use something else.

[QUOTE=itisjuly;42168696]Dropbox sucks for hotlinking these days, use something else.[/QUOTE] Imgur was borked in my browser where I am right now, I'll rehost when I get home.

Assuming it is possible (Not my area of expertise) it would be awesome if you could make changes fairly quickly. Like carrying different cameras, batteries, memory blocks, etc. in a case and swapping them out as you need them. As for the "nobody throws out phones anymore" argument, I know several people who go through 4-5 phones a year.

[QUOTE=TNOMCat;42167926]Its not impossible... the cpu block can detect where the other blocks and the pins are[/QUOTE] How will the board know when and where the CPU block is plugged in? It needs to provide the right amount of power or it risks destroying or upsetting the CPU through over or under voltage.

One change I would make is it could have casing over the blocks that you can detach to modify it.

[QUOTE=DrDevil;42167225]Do you guys have any electrical engineering background at all? Because there are very good solutions for modular pin configurations. Google Cypress PSoC for one of many examples.[/QUOTE] Yeah but none of them are compact enough for a phone application. [URL="http://facepunch.com/showthread.php?t=1306765&p=42161896&viewfull=1#post42161896"]Go back to my original post on the first page.[/URL] There are just way too many problems. It might be doable but it wouldn't be worth doing. You might be able to implement it in a very limited way, for example, by creating a new standard socket (just like AMD AM2/AM3+) and making new cpus with the same architecture. However, all other problems aside, the mere fact that every component has to be modular will add so much size to the phone that we might as well go back to these things: [t]http://1.bp.blogspot.com/-Nu5ptVmNbi8/UalG-csD2SI/AAAAAAAAHTU/9BNebAPXJ-Q/s1600/ancient1.jpg[/t]

There, I rehosted them [url]http://facepunch.com/showthread.php?t=1306765&p=42166626&viewfull=1#post42166626[/url]

No fucking way, I had this exact idea two years ago.

Nvm give me boxes

[QUOTE=SIRIUS;42170491]Guys, this is clearly past the idea stage, they wouldn't have gone this far if the problems were as obvious as you are saying.[/QUOTE] Why is it clearly past the ideas stage? The guy even mentions in the video that he needs engineers and other smart people to actually work this out for him, as he only has the idea and design.

[QUOTE=SIRIUS;42170491]Guys, this is clearly past the idea stage, they wouldn't have gone this far if the problems were as obvious as you are saying.[/QUOTE] No it's an idea and nothing more right now it's also really impractical. Sure at some level it could be figured out eventually, but who's to say that will be hard or easy? Well the engineers who do that work. And even just having a basic understanding of circuit boards and circuitry should tell you off the bat, creating a simple plug and play system for a lot of modular parts, is not really realistic.

You'd need to have a cpu slot and have the plugs all be clusters of four prongs. Basically you get a CPU with a whole bunch of tiny ass usb thigners.

[QUOTE=hexpunK;42170521]Why is it clearly past the ideas stage? The guy even mentions in the video that he needs engineers and other smart people to actually work this out for him, as he only has the idea and design.[/QUOTE] wow, never mind I'm completely stupid... and apparently so is he.

[QUOTE=aydin690;42169775]Yeah but none of them are compact enough for a phone application. [URL="http://facepunch.com/showthread.php?t=1306765&p=42161896&viewfull=1#post42161896"]Go back to my original post on the first page.[/URL] There are just way too many problems. It might be doable but it wouldn't be worth doing. You might be able to implement it in a very limited way, for example, by creating a new standard socket (just like AMD AM2/AM3+) and making new cpus with the same architecture. However, all other problems aside, the mere fact that every component has to be modular will add so much size to the phone that we might as well go back to these things: [t]http://1.bp.blogspot.com/-Nu5ptVmNbi8/UalG-csD2SI/AAAAAAAAHTU/9BNebAPXJ-Q/s1600/ancient1.jpg[/t][/QUOTE] Like Samsung aren't already doing that [t]http://i.i.cbsi.com/cnwk.1d/i/tim/2013/03/14/whitesmoke_35627724_11.jpg[/t]

The people who are claiming it's impossible all seem to be stuck on the idea that the concept will never change from what's displayed in the video. Which is incredibly silly. Only four pins is bullshit, we know this much already. But we can add more. And if the designer made the configuration of the modules on the back a fixed arrangement (and not the arbitrary thing he has going now) I'm certain the ability to upgrade bits as you go could be rather doable. Doing that would mean you'd have to have a certain set of features, so you couldn't forgo camera and intensive graphics for battery and storage or whatever, but it still seems sensible. Hell, if that guy on the last page has it right, the arbitrary bit is equally doable, but I don't know enough about this sort of thing to be certain.

I can work on fleshing it out a bit more later, at least on the physical side. I can't even [I]begin[/I] to imagine how the backboard's internals would work, nor the firmware. My design still needs an effective way of holding the blocks in place on the backboard. The direction peg could serve to be better, and everything else can't really be done in CAD.

I was thinking about a method to hold the blocks in place and took inspiration from CPU sockets, where there is that plastic layer that slides into the pins to grip them. When the screws are screwed in, a small part of their screwing distance could pull a plate across the entire backboard (it'd probably have to be exposed with the pins under it) to latch in to the blocks on some way, using your CAD design S31-Syntax, it could latch on to the directional peg. Though this still doesn't solve the problem of the board overvolting the CPU because it can't tell it was a CPU plugged in, or the pin configuration not making any reasonable sense electronically (trying to pass all the different signals every component gives over a set of standardised pins just doesn't seem to add up), or the latency between components thanks to everything being so detached.

Saying "impossible" is pretty much standard response for shit like this, specially because the guy is clearly a designer If you're head isnt reverbating with "SHUT THE FUCK UP FOR THE MILLINOTH TIME YOU CANT DO THAT" every time a designer speaks, you're not an engineer I reconsidered the design a bit and it would possibly be doable: CPU + RAM in baseplate 8 bit wide data bus for each "slot", just directly connected into a main device bus Only one block communicating at a time 12V, 5V and 3.3V power connectors for each "slot" Battery delivers power over the 12V line, which should have a larger connection on every second/third slot (since no battery will be smaller than that) That said it'd probably be pretty expensive to do, and it would get thick [editline]13th September 2013[/editline] Not to mention creating those tiny blocks and tracing the base board would be a fate worse than death

[QUOTE=hexpunK;42177442]I was thinking about a method to hold the blocks in place and took inspiration from CPU sockets, where there is that plastic layer that slides into the pins to grip them. When the screws are screwed in, a small part of their screwing distance could pull a plate across the entire backboard (it'd probably have to be exposed with the pins under it) to latch in to the blocks on some way, using your CAD design S31-Syntax, it could latch on to the directional peg. Though this still doesn't solve the problem of the board overvolting the CPU because it can't tell it was a CPU plugged in, or the pin configuration not making any reasonable sense electronically (trying to pass all the different signals every component gives over a set of standardised pins just doesn't seem to add up), or the latency between components thanks to everything being so detached.[/QUOTE] I like your latch idea, thats nice. The backboard's firmware should be central in assigning pathways and configuring voltages. Best way I can think of to allow for this is a effectively a multi-layer network with a crapton of firmware controlled gates. Pins that aren't used at all on a block are totally isolated from the network. Effectively a 3 dimensional mesh with gates at every possible intersection, all controlled from the backboard's firmware. You get your backboard, your blocks, you arrange them on your backboard, and you turn on the phone. From there, the firmware gets power from a dedicated set of leads from the battery, and it boots up. After a self test, it runs a checklist for block discovery based on block priority. At this point, all backboard block networking gates should be totally off and not allowing communication between any blocks whatsoever. Firmware then identifies the battery, make, model, and specs, and keeps those details logged. Then it plays with the gates, opening and closing them until it finds the CPU. It identifies it, specs, requirements, etc. It repeats this process until it has identified all blocks attached. From there it should then start the routing process, again based on priority and importance of a module. High requirement/high importance modules like the CPU, GPU, display, and RAM should get the highest priority. If latency is less of a concern for a block, it is routed later in the list. The firmware then determines the best voltages and pathways and configures block-to-block communication using the gates it has. Errors are reported in any way it can, saving a log to storage if available, showing it on the screen, etc. Of course, backboard firmware updates should be possible, and shouldn't be that hard to pull off. SIM cards should be inserted into the cell radio block, if they're used at all. storage blocks should have a redundant USB data jack on them so that it can be mounted as a mass storage device in the event the built in USB jack on the phone is broken or otherwise unavailable. This would allow for data recovery even in any event.

I'd avoid any special routing, I'd just slap all 3 usual voltage levels on the connector and a shared bus, as in, each socket would be wired to the exact same thing as the other If the CPU is always the bus master with a special interrupt line (which could be separate for each socket, but hopefully isnt even necessary) it shouldent be a problem

I would see this working without the "plug anywhere you want" bit. Have a dedicated camera socket, battery socket, screen socket, et cetera. That way instead of the nightmare of having to figure out which pins do what, just get a battery that fits where you want it to. It'd still be really inelegant and probably fuckoff huge for a phone, but it could actually work.

[QUOTE=Tobba;42178298]I'd avoid any special routing, I'd just slap all 3 usual voltage levels on the connector and a shared bus, as in, each socket would be wired to the exact same thing as the other If the CPU is always the bus master with a special interrupt line (which could be separate for each socket, but hopefully isnt even necessary) it shouldent be a problem[/QUOTE] That works. I just specialize in basic concepts, the nitty gritty of anything other than physical structure is gonna leave me derp-eyed.

[QUOTE=Tobba;42178298]I'd avoid any special routing, I'd just slap all 3 usual voltage levels on the connector and a shared bus, as in, each socket would be wired to the exact same thing as the other If the CPU is always the bus master with a special interrupt line (which could be separate for each socket, but hopefully isnt even necessary) it shouldent be a problem[/QUOTE] Didn't think of having each set of pins pass all the possible voltages, which could work, if a block doesn't need a voltage just don't include that pin.

Not necessarily, it'd be more advantageous to make the backboard as simple as possible when it comes to the base lines (Akin to a breadboard), have just one VCC/GND pair to pass power around and include power management circuitry to each block that requires a special voltage. This is already true with most computer designs, where the CPU has a power management IC to step the voltage down from 3.3/5V down to the typical ~1.2-0.9 required for most CPUs. The nice thing about this is it would reduce some problems when it comes to compatibility, instead of having to make sure the blocks and the baseboard are of the same series (Series 1 blocks only go with Series 1 baseboards, etc). In my opinion, the baseboard should be the one with the most compatible design with future proofing in mind. I'll have some routing designs done later today.