[CWN-Summit] Thinking About SDR -- The Problem of a "Hot Cut" To SDR

[Harold Feld]

I posted this to a list where the nth iteration of "the FCC's rules are
holding back SDRs" was going on. It got me thinking about here in the U.S.,
and in other countries where we appear to be on the cusp of introducing
rules that could move us to the next generation of SDRs.

There is a problem we never talk about. How do we make the transition from
the current wireless structure, which depends heavily on exclusive
licensing, to a system that relies heavily on SDRs and unlicensed access?
Most folks in our movement argue, in essence, for a "hot cut." Lets just
authorize SDRs and that technology will naturally win. There is an
associated argument that the failure of regulatory authorities (the one I'm
most familiar with is the U.S. Federal Communications Commission) resist
expanding the rules for SDRs not from genuine engineering concern, but
because of the political influence of incumbents licensees, the mobile
bullies, the dinosaurs, who will fight to the death to protect their
business models and their irrational and baseless fear of interference.

Granted the later is true. But even leaving aside the economic problems, the
mobile bullies, etc., there is a very real engineering problem of
compatibility with embedded technologies. Bluntly put, I have come to doubt
the idea that we could suddenly make a "hot cut" from the existing world of
predominantly licensed to a new world of predominantly unlicensed spectrum
use. Indeed, I have come to view the problem here as remarkably similar to
the problem of IPv6 adoption. Those favoring an upgrade are completely
ignoring how human beings actually act and what is important to those who
view wireless merely as an input for doing cool stuff.

Before folks get the wrong idea, I continue to believe that the technology
will, ultimately, work just fine and in the predicted way. That we can
develop smart devices that remove the need for exclusive licensing and that
would therefore allow us to radically alter the way in which we construct
wireless networks and what we can expect them to do, and at what price. But
I do not believe it is enough simply to get the FCC to change its rules.
Worse, I have come to believe that the FCC should phase in these
technologies gradually. Not because I fear the licensed lobby, but because
there are 300 million people dependent on devices operating under
assumptions developed in the existing regime and the vast majority of these
devices are crappy pieces of hardware that could not survive a sudden,
radical shift in rule changes as they are likely to be implemented in the
real world.

What shook my faith was the current relatively minor incidences of
interference in FAA Doppler radar from unlicensed operation in the
neighboring 5.3 GHz band. Operation of unlicensed in 5.3, shared with the
U.S. military radar, is a triumph of engineering technology and proof of the
sensing concept. The military would not reveal any information about the
nature of its radar systems -- the wave form, the energy, or positions of
transmitters. It was necessary to build devices that could nevertheless
avoid interference with military radar yet still prove useful and economical
to build. After several years, this was achieved, and we have seen the
recent opening of the 5.3 GHz band for commercial use.

The problem is that, as  the equipment manufacture got ramped up and
economies of scale kicked in, production got just a smidge cheaper and
shoddier. Chips complied with sensing and avoidance in band, but became just
a bit noisier out of band. Shielding was reduced to make devices just that
half cent cheaper which can mean a difference between a profitable run and
an unprofitable run. All of which resulted in a device just a little bit
noisier than actually allowed.

This might not have been a problem had the neighboring service not been FAA
Doppler radar with _extremely_ sensitive double-array receivers. As it was,
however, the increase in the number of devices, combined with their
increased noisiness, began to cause interference problems. Happily, the
situation appears to be resolved.

But the process is entirely too reminiscent to me of how the 800 MHz band
got screwed up by Nextel. Any individual conversion of a system to a digital
two-way system was not problematic. But when you ramped up to several
million users, operating at about a zillion times greater frequency of use
that the previous set of users, it started to cause problems.

Lesson 1: Reality always plays out funny. Not because the underlying science
or engineering is wrong. Often it is from extraneous factors. It is
_predictable_ that manufacturers will develop cheap devices. It is known
that the current universe of devices are built as cheaply and inefficiently
as possible given the current set of rules assumptions. Change the
underlying rules and unanticipated things happen as the interplay of these
facts on the ground start manifesting themselves at hyper speed from
dramatic ramp up. Because if we are right about how much better these
technologies are in terms of cost savings, ramp up will be incredibly
dramatic.

Lesson 2: Any serious plan to alter the current wireless paradigm must take
into account the embedded technology and figure out how to change it,
gradually. _This_ requires considerable subtlety and a willingness to
consider taking the scenic route to the ultimate goal. This route must
detour through lots of apparently unrelated places such as overall
improvements to receiver standards to enhance robustness on the
receiver/embedded tech end, creating incentives for licensed services to
want better spectrum sharing and spectrum reuse technologies that do not
directly undermine their business models. It includes a couple of stops
designed to introduce new considerations in business model and encourage
people to want to own/deploy their own networks without worrying whether
they also "own"/hold the license to the spectrum.

In short, it is no longer merely a complicated technical problem. Like
conversion to IPv6, it is a complicated technical problem made worse by the
irrational and unrelated (and often contradictory) needs and desires of the
human beings actually using the legacy technology who do not respond well to
the argument of "look, just trust us and switch, you stupid non-technical
monkey boy." Unlike conversion to IPv6, however, there is no near term
looming crisis to drive people to the preferred engineering solution.

For those serious about engendering a social revolution based on
technological change, this requires a radical rethinking of tactics and
strategy. It also requires gaining a new perspective on FCC regulations.
Absolutely they are manipulated by incumbents. But after nearly 10 years
working with the Office of Engineering and Technology (OET) on spectrum
issues, I have developed a respect for their inherent conservatism, even as
I chafe under it and try to move them along a more aggressive path. They
_must_ consider what happens if we are wrong and the ramp up unexpectedly
screws up vital services in the embedded technology. It is not enough to
know that a solution is possible. It is not enough to blame others for
making cheap receivers for licensed services. Because if this stuff screws
up, real people are impacted. That's enough to give anyone charged with the
responsibility of making the decision a serious case of cautious
conservatism _without_ insulting them by claiming they are simply tools of
the existing licensed incumbents. At them same time, however, anyone who
thinks that the economic incentives of incumbents and the political power
they wield is not a significant factor weighting against change would be a
naive waif doomed to failure.

So how do we move forward? Excellent question! I am only now starting to
struggle with this. But the first stage is to recognize the enormous
complexity of the problem and the pressing weight not merely of economically
invested incumbents, but of the embedded technology base predicated on more
than 75 years of technological assumptions. My gut feeling tells me that
countries outside the U.S. willing to free themselves of the "Washington
consensus" that got everybody to auction spectrum should have an advantage
here, simply because the weight of embedded tech should bear less heavily on
them. But this may not prove correct, at least in the short term.

Harold
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.chambana.net/pipermail/cwn-summit/attachments/20100906/afffc35c/attachment.html>