Power-savvy GSM network (was: gsm322.c Bug Fix Patch)

Kurtis Heimerl kheimerl at cs.berkeley.edu
Fri Feb 3 08:32:12 UTC 2012

Comments in line!

On Thu, Feb 2, 2012 at 9:17 PM, Alexander Chemeris
<alexander.chemeris at gmail.com> wrote:
> Hi Kurtis,
> (I'm changing subject to the actual topic of this discussion)

Good idea.

> Yes, I quickly looked through your code. Looks like a big hack right
> now, but I guess it's meant to be a hack at this point :)

And it will remain a big hack into the near future!

> So, your idea is to manipulate Tx power of a BTS to cover only
> actually working handsets, if I understand correctly. It's not
> possible to do with a normal GSM phone, because phone does not
> transmit until it sees a beacon. So you want to create a "wake up BTS"
> channel to get around. Am I following your logic correctly?

Ah... sorta not really, unless we're using different terminology. We
use the existing channel (that the BTS is listening on) and send a
RACH over that channel to the BTS. Though the handset can't hear the
beacon, the radio is capable of transmitting at that ARFCN without a
beacon (thanks Sylvain!) if we tell it to do so. We do so ("wakeup"),
causing the handset to RACH at a specific ARFCN and the BTS to hear
some discernible noise (however not a RACH as the clocks are not
synced) and turn the PA on. Following this, the handset camps normally
and makes a call.

In writing that, it is awfully hard to describe. I hope that made some
sense. I can tell you that it's working in our lab, and it's pretty

> It does look like an interesting idea if could be done dirt-cheap. But
> how do you plan to do paging in this system? I see the only way - to
> use the same "wake up channel", but in other direction. So basically
> you have a network-specific "default" ARFCN which is used when no
> active BTS is in range. All communications are first tried on this
> ARFCN and only then on other ones. Right?

Paging is actually quite simple, as we own the BTS. When we receive a
call, we hold it ("please wait for connection") and turn the PA on. We
then wait for the HS to camp, and if they do, page them once
available. A little delay (much better with SMS), but I think most
users would be fine with it. We haven't written this yet.

> On Thu, Feb 2, 2012 at 22:43, Kurtis Heimerl <kheimerl at cs.berkeley.edu> wrote:
>> This is part of my thesis work at Cal, yes. Range is not in any way involved.
>> That's roughly the use case, areas where there are too few users to
>> keep a BTS in constant use. Our designs allow the power usage to scale
>> with the number of users, rather than sit at a fixed output, as they
>> do now. The BTS side is simple; the osmocom side is complicated. We
>> have a handset that can wakeup a sleeping tower, or a "wakeup button"
>> device which only transmits when a button is pressed. That thing is ~5
>> bucks and can probably be attached to the back of a legacy phone in
>> case we can't convince a large manufacturer to incorporate our changes
>> into the baseband.
>> Anyhow, the code is online if you're interested in looking at our
>> progress (https://github.com/kheimerl). It's nowhere near ready, but
>> you're a very knowledgeable guy, and I'd be happy to hear your
>> opinions on any of our designs.
>> On Thu, Feb 2, 2012 at 10:37 AM, Alexander Chemeris
>> <alexander.chemeris at gmail.com> wrote:
>>> Is it a part of your TIER university work?
>>> I wonder about use cases for this.
>>> One use case I see is when you have a BTS which is rarely used, like
>>> in a desert and you don't want it to work all the time. What use cases
>>> do you plan to use it for?
>>> On Thu, Feb 2, 2012 at 22:03, Kurtis Heimerl <kheimerl at cs.berkeley.edu> wrote:
>>>> Yeah, and we have that working.
>>>> On Thu, Feb 2, 2012 at 12:56 AM, Alexander Chemeris
>>>> <alexander.chemeris at gmail.com> wrote:
>>>> > On Thu, Feb 2, 2012 at 12:08, Kurtis Heimerl <kheimerl at cs.berkeley.edu> wrote:
>>>> >> I think I looked at that... I'll give you some context.
>>>> >>
>>>> >> We've modified osmocom to "wakeup" a specific tower at a specific
>>>> >> ARFCN.
>>>> >
>>>> > Interesting. Do you mean you send some packet to a "sleeping" BTS to wake it up?
>>>> >
>>>> > --
>>>> > Regards,
>>>> > Alexander Chemeris.
>>> --
>>> Regards,
>>> Alexander Chemeris.
> --
> Regards,
> Alexander Chemeris.

More information about the baseband-devel mailing list