I didn't catch up with Junko Yoshida's reporting from the CeBIT show until March 21, the day I got back from Comdex Vancouver. While there, I chaired a panel on 2.5G and 3G trends that featured Kevin Chaplin from Sierra Wireless, Chuck Brown from Intel and Wanda Gass from Texas Instruments. It's remarkable how the message from all three panelists corresponded with what application developers were saying about handheld platforms at CeBIT in Hannover.
The issue is not merely one of keeping operating systems and binary environments open, but of disentangling the high-level control from the communications channel. Brown of Intel said that the convolution of text-based application with the circuit-switched wireless cellular channel was one of the factors that gave rise to lead times of as long as two years in developing even the simplest of applications.
TI's Gass agreed that one way to hasten that cycle is to provide clear distinctions between the control processor and the channel-which in the case of a handheld phone means a rather strict delineation between the control processor on the one hand and the baseband DSP and RF/IF chain on the other.
In the wireline world, this is old hat and is referred to as the segmentation between the control plane and the data plane or data path device. In any case, not only do the wireless developers have a model to work toward, but they also have a generalized integer platform on which to focus, namely the open licensable RISC core, in particular ARM and MIPS.
Brown said that though many assume an omnipresent Wintel conspiracy, Intel does not care if the development environment is a flavor of Windows CE, Java, Linux or one of the specialized handheld environments. It's more important to keep the hardware platform simple, low-power and capable of handling the advanced tasks demanded of 3G. To both Gass and Brown, that means ARM combined with a standard DSP and perhaps some specialized peripherals.
This will mean a shakeout among semiconductor players in handheld control, and probably DSP, though some value-add remains in RF/IF. But with handset and PDA system developers converging, wasn't that bound to happen anyway?
The core platform of most handheld devices of the future will be based on a narrow range of RISC and DSP processors, with differentiation reserved for application layers and special services of the communication channel at Layers 3 through 5. Only a PC-like development model can allow services and applications in 2.5 and 3G to blossom.