community Digest, Vol 14, Issue 50

Martin Lefkowitz lefko at
Sun Feb 18 20:29:21 CET 2007

Broadcom no, unless you are talking about the SOC ap thing that Broadcom
and Atheros use a  MIPS for, and bascially has the same kind of
architecture we would have with the ARM920 on the Samsung.  Broadcom
downloads microcode into the device to start it up.  TI downloads
C/Assembly firmware to start it up.  Atheros is pure hardware.  The host
interface for the Broadcom and Atheros is as I described earlier.  The
Host interface for the TI is different.  A lot more is (or at least was)
done inside the chip on the ARM. 

Marvell is a different story.  They have an SDIO system that has (I
believe) an ARM inside.  They do even more inside the chip.  They do all
the MAC level functions inside the chip.  This is going to be a problem
for the company since more and more higher level functions are appear to
be using the 802.11 protocol (e.g. CCX, TGv and TGk).  The last time I
checked they were keeping up, but at some point I bet they are going to
run out of resources both machine and human to keep up.

In fact the complaints here should be about no SDIO, or CF interfaces,
but again they've bitten off enough.  Even with the features that
OpenMoko has it would have been insane only a few years ago to attempt
an opensource phone.  The GSM AT stuff is pretty interesting....


> Message: 4 Date: Sun, 18 Feb 2007 09:44:57 -0800 From: "Wolfgang S.
> Rupprecht" <wolfgang.rupprecht+gnus200702 at> Subject: Re: Wifi
> again To: community at Message-ID:
> <87k5yfmkbq.fsf at> Content-Type: text/plain;
> charset=us-ascii Marty, thanks for the interesting insight. Martin
> Lefkowitz writes:
>> > what both Atheros and Broadcom have done is have a binary that
>> > handles the interface for the chip that needs to be included in the
>> > opensource project.
> >From my limited understanding, it appears that both Broadcom and
> Marvell use an embedded ARM chip also.  An interesting blog entry by
> Jim Gettys (at ) mentions
> how Marvell helped free up the Marvel wifi driver.  He also talks
> about how the Marvell chip is crucial to the OLPC low power goals.
> (Although the 300mw estimate for the running chip strikes me as kind
> of high for use on a cell phone where one has maybe 3 watt-hours of
> battery all together.)
>> > BTW it's interesting that the HTC uses the ACX100.  I actually designed,
>> > and developed the first version of, that interface for TI.  
> I wonder how its power compares.  I'd certainly expect TI to have more
> engineering and manufacturing talent on hand for building a really low
> power wifi chip.
> -wolfgang
> -

More information about the community mailing list