Sunday, March 18, 2012

Know Your Limits

On June 22, 1941, Germany launched Operation Barbarossa, sending 3.9 million troops into the Soviet Union.  This was the largest invasion in the history of warfare, and was likely the most significant reason why the Axis powers lost World War II.

Untangling all the reasons why Germany lost along its Eastern front could keep a horde of rampaging historians busy for a lifetime.  However, very high among the top candidates would have to be that the Germans failed to pay proper attention to the most critical constraints.  They had plenty of weapons, ammunition, motorized vehicles and horses, which Hitler believed to be all that was necessary to win.  But, overconfident of his ability to achieve a decisive victory quickly, he made no provisions for the supplies that would be needed if the campaign stretched into November or later.

There are some constraints in warfare which are flexible.  Treaties can be bent or broken, and then rationalized later.  Troops who are tired or dispirited can be bribed or threatened to march just a little farther.  Recruitment standards can be lowered to expand the pool of available soldiers.

Winter in Moscow is not one of those constraints.  It makes no allowances for your failure to plan.  When temperatures hit -40 degrees Fahrenheit, guns jammed, fuel turned solid, and troops stuffed newspapers into their light summer uniforms in a futile attempt to avoid freezing to death.

Military planners have studied these battles carefully.  They have attempted to extrapolate the underlying principles of cause and effect, and apply these principles to as many other situations as possible.  But have the lessons really sunk in for everybody else?

On February 14, 2012, the FCC rescinded permission it had previously granted to Lightsquared Corporation to create a new broadband wireless network.  The effect of this decision was devastating.  Shortly thereafter, the CEO of Lightsquared resigned, half the employees were laid off, and Sprint cancelled a 15 year spectrum-hosting agreement.  It is unlikely the company will recover.

Should anybody care?

Let’s think about constraints from an engineering context.  Some constraints are more important than others.  When designing an airplane, the total weight is a critical constraint.  Engineers will spend thousands of hours and billions of dollars to find a new type of material that will make the airplane a tiny percentage lighter.  When designing a washing machine, weight isn't nearly as important.  It's not completely irrelevant, as you wouldn't want to build a two ton machine that couldn't easily be delivered, or which would fall through an average floor.  But in general, washing machine designers don't spend much time worrying about it.  A 10% difference, say between a 180 lb and a 200 lb machine, is largely irrelevant.

Some constraints tend to become less significant over time.  Computing power is a great example.  If you have an idea for a great video game that would be too complex to run on today's generation of computers, don't throw the idea away.  Given the rate of improvement, driven by Moore's law, it might just might run on the next generation.  In fact, given the lead time needed to develop most games these days, you might just want to start developing it right now, under the assumption that the hardware will catch up to you by the time you're done.

The size and cost of consumer electronics is similar (and closely related).  They are small and constantly getting smaller.  If you have a ten year plan to build a robot that needs a fully functioning Linux server the size of your thumbnail, don't worry that you can't find a server that size on the market today.  The Raspberry pi has already demonstrated a Linux machine the size of a credit card.  Ten years should be more than sufficient time to shrink it down to thumbnail size.

Other constraints aren't so easily resolved.  One of these is the amount of available electromagnetic spectrum.   Spectrum is needed any time you want to communicate without a connected medium (such as a wire).  Its uses include television, AM and FM radio, Wifi networks, cell phones, radar, microwave transmitters and garage door openers.  There is only only have a fixed amount of it, and there will never be any more.  Demand for spectrum is intense, and growing rapidly.  Companies pay tens of billions of dollars to license relatively small portions of the available spectrum.

It is more precious than gold.

With this in mind, a number of years ago, Lightsquared made a proposal to use advanced technology to deliver a new 4G LTE wireless network using a slice of unallocated spectrum bandwidth.  They would provide this network capacity to existing providers wholesale.  This would have increased capacity, exerted downward pressure on the currently rising prices, and been a boon for both consumers and corporations alike.  Everybody wins, right?

Well, not everyone.

It seems that GPS manufacturers were a bit concerned by this idea.  Lightsquared's plans would potentially interfere with some GPS devices.  In February, their concern made enough of an impact on the FCC to cause the FCC to put a halt to the venture.

If we left the story right there, then I'd probably side with the GPS manufacturers.  I like my GPS.  I don't want it to break.  If somebody has other ideas, well tough, because my GPS was here first, right?

Well, its not quite that simple.  You see, you need to remember what I said about the electromagnetic spectrum.  Because its so precious, it's all carefully defined and licensed out.  And as it turns out, Lightsquared was not infringing on the GPS wavelengths at all.  Instead, GPS's are infringing on spectrum that was licensed to Lightsquared.

Although we talk about spectrum in very precise, digital terms, its actually very analog.  If you broadcast on a specific frequency, unless you're very careful, it doesn't look like a sharp spike at that specific frequency.  Instead, it looks like a broad bell curve, centering around that frequency.  Try looking at a simple frequency analyzer (such as Soniqviewer for iOS) while you play a single note on a musical instrument.  You'll be amazed at how many sound frequencies are playing all at once.

The problem in this case isn't the transmissions from the GPS satellites, its the receivers.  GPS signals tend to be weak.  GPS devices are inexpensive.  To make the devices as cheap as possible, and get the best possible reception, GPS devices use receivers that are very imprecisely tuned, in an attempt to pick up as much of the signal as possible.  That means that GPS devices end up listening to a lot of spectrum they have no business listening to.  And, no surprise, they have the tendency to get confused when they hear some other signal on this other spectrum.

This is a fixable problem.  Lightsquared has demonstrated how a simple filter could be added to the GPS filter to clean up reception.  This would increase the cost of a GPS.  But cost isn't a significant constraint for consumer devices.  GPS chips are already such a commodity that they are being added into every model of smartphone on the market.  Garmin, the leader in standalone GPS devices, is desperately trying to find a new niche because it's product is becoming such a commodity.

So lets think about this from a constraint perspective:
  • GPS devices are cheap, and getting cheaper
  • Spectrum is expensive, and getting more expensive
  • The constraint the FCC wants to address is the current cost of GPS devices, rather than increasing utilization of spectrum
Does this make sense to anybody?

So far, weve managed to handle spectrum concerns.  Smart phones, which are among the biggest consumers of data, are still relatively new.  Weve been enjoying a pleasant summer of bandwidth availability.

But demand is growing fast.  Available spectrum is fixed.

Winter is coming.

No comments:

Post a Comment