System-on-a-chip?

Rayman

I was exploring some of the SoC (system on a chip) options out there and it's pretty amazing what's available...

Anyway, think the Prop qualifies as a SoC?
http://en.wikipedia.org/wiki/System-on-a-chip

I guess the main things possibly missing are flash and RTC...

Maybe Parallax will include these in Prop2 ?

Still, I think one could epoxy together a package with a Prop, usb controller, i2c RTC, and eeprom and call it a SiP (system in a package)...

Mike Green

I don't think the Prop qualifies as a system on a chip since it's missing persistent application memory (flash).

With a 1Hz frequency reference, the Prop could do the rest of the RTC stuff while it runs off its RCSLOW clock,
is mostly asleep, and switches to a crystal clock to "wake up".

Parallax has already made clear that the way the Propeller is manufactured is incompatible with any on-chip flash memory.

Oldbitcollector (Jeff)

Wouldn't the Prop qualify with it's non-volatile storage in either EEPROM or SD media?
Then it's just a matter of code...

OBC

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Just getting started with Propeller?

Propeller Cookbook

PropDOS

Mike Green

Rayman was talking about single chip systems where everything needed is on one chip (other than a clock source and power).

By the way, the SpinStamp and PropStick would qualify if you expand the concept of "chip" to include DIP modules.

David Cuthbert

Heh... I got so fed up with SoC hype around 99 or 2000. From what I've seen, a SoC is any chip successfully spun by your marketing department as such to the industry rags (EETimes, EDN, etc.)!

I was hoping to see three milestones hit by the SoC "revolution":

• All of the silicon components of a typical low-end PC (processor, I/O bridges, memory controller, video controller) on a single die, with dynamic RAM provided as a separate die in the same package.
• The same plus dynamic RAM on a single die.
• The same plus magnetic storage (using MEMS) in a single package.

The first milestone minus the video controller was met by some of the offerings I saw. But things never went much further. Ironically, the Propeller does have the video controller.

The magnetic storage milestone may be moot now with the rapid increase in flash memory densities as of late. Back when I came up with this list, it seemed they had resigned themselves to niche status and weren't going to go head-to-head with the magnetic storage vendors. Times have changed...

OwenS

The first will be more expensive than a seperate DRAM chip. Putting two dies on one carrier is expensive, as standardized packages with two dies on them are rare. Additionally, the die bonding time has just gone up, since you now need to bond the chips to each other. This is generally reserved for really high performance cases.


It's not possible to do DRAM effectively on a standard digital CMOS process. As I understand it, capacitors can be implemented, but they end up wildly innacurate and huge in terms of consumed silicon, so the second is out. Doing the reverse is also infeasible, partially because DRAM processes tend to lag behind digital processes, and likely because of other reasons of which I am not aware.


Magnetic storage on the same package has happened, to an extent, in the 8-Bit world, if you count FRAM as magnetic storage. Ramtron's Versa 8051 microcontrollers have on board FRAM memory.

1T-SRam could solve the RAM issue in the future, however. The Gamecube/Wii are already using it. It's not as dense as DRAM yet, but i'm sure that it will reach that capability with the dropping of silicon feature sizes, and will likely have benefits over DRAM when you consider that smaller DRAMs become more susceptable to external noise

deSilva

The Etrax100 from Axis (used on FOX boards e.g.)

David Cuthbert

Intel has been doing the stacked die with RAM for some time now on their I/O processors. But given that they have their own fabs and can play with things like this, this makes your point: it *is* expensive. (But, then, so is everything else in low volume...)

It's also possible to do DRAM on a CMOS process, but I'll agree that the methods so far haven't seen anything effective for mass-production. Capacitors are actually not too bad in terms of precision (we did them regularly for analog), but they do occupy large swaths of area and/or inject a lot of noise into the substrate (if you're doing poly/bulk caps).

I don't count FRAM in the same boat as the magnetic storage I was thinking of. It's much less dense than something disk based and is less of a challenge from a fabrication process perspective. At Seagate, we were talking about trying to take a slab of rectangular media and fabricate a matrix of probe-based heads below it. MEMS drivers would slide the media along the X/Y axes. I don't know if this ever got beyond the musing stage; I doubt it.

Anyway, the hurdle for heads-and-media magnetic storage with CMOS is again the fabrication process. The high temperatures required to anneal magnetic materials will cause dopants to diffuse from silicon -- your PN junctions become PN mush.

Paul Baker

The MEMS X/Y probe method is used in PRAM, so the MEMS structure has been created.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Paul Baker
Propeller Applications Engineer

Parallax, Inc.

David Cuthbert

Yep; the MEMS structure in question is fairly straightforward, as is the GMR probe (though this isn't/wasn't available in volume -- research labs only, last I checked).