winter dreams

About a month ago I found myself at an airport waiting for a flight to Phoenix.  My father died a few hours earlier and I was deep in thought.  He loved classical music - particularly the 19^th century Russians.  Numbly browsing the iTunes store I found the same recording I remembered him buying when the family stereo was delivered.

 

It was a solid state Magnavox and probably a huge splurge for our family.   Home entertainment systems of the day had fine cabinetry and this was no exception, but the guts of the unit interested me - after all, the family transistor count just took a huge jump.

 

Any major piece of electronics came with a set of schematics - a shorthand blueprint of what made it tick.  I was fascinated by radio and could decipher the symbols and their relationships with each other.  It had a total of 31 transistors - a huge improvement from the six that were in my transistor radio.  Nerdy kids from earlier eras could tell you the family tube count and in the teens and twenties how many electric motors were in the house.¹  

 

We had a fair number of tubes in other family electronics, but transistors promised trouble free operation.  Tubes had to warm up for about thirty seconds and gave off a warm orangish glow.  This heat lead to failure and many stores had a tube testing machine.  If your radio or television wasn't working you would pull off the cover and look for the dark tube.  More often than not swapping it for a new tube would fix the unit, but radio repair shops were good businesses.

 

Accurate transistor counts today are a futile exercise - even estimates are very hard requiring specialized knowledge.²  So what to count?  It might be possible to count microprocessors, but even that is becoming difficult.³  Perhaps we should count computers, but that brings up a thorny issue.  

 

What do you mean by computer - what does it do?

 

This class of question is a good starting point for considering the emerging class of connected devices that talk to other machines - the so called Internet of Things.  In the past few years we've seen a big change of the perception of what the Internet is.  For most people it meant the web and what they could access with a browser, but mobile devices brought change.  Smartphones and tablets still have browsers, but they are also festooned with sensors and applications with connections to points on the Internet.  This is opening some major opportunities and causing a fair amount of scrambling.

 

Internet of Things is an unfortunate label as it lumps too many categories.  Nest thermostats are considers prime examples.  That class of device - a network connected control unit with a user interface - is quite expensive and difficult to get "right".  They will be important and visible and may even grow into hubs that connect groups of sensors and other processing units, but they will not represent more than a small percentage of IoT devices.

 

About twenty years ago I was spending a good deal of my time working on what were then high bandwidth networks.   One frustrating day I found myself late for something and having to turn back home because I couldn't remember if I had locked the door.  I built a key that would light an LED for ten minutes after it had been inserted in the lock, but the experience got me thinking about very low bandwidth opportunistic communications.  Specifically what is the byte of information that is very valuable?  The state of your house doors, refrigerator door, range element and so on...   The networking piece was mostly as anything that existed was expensive overkill.   Ultimately I wanted this on the Internet and readable by a portable device, but getting the information to the Internet in the first place was non-trivial.

 

Now there are many candidate networks that let you forget wiring your sensors.  The Nest uses WiFi, but that is overkill and too expensive (mostly in terms of power) for most of the IoT devices that will spring up.  For many shorter ranges and very low power consumption make sense.  BlueTooth LE is one of many network types that are appropriate - it may win early on as Apple has it in all of the iPhones produced in the last few years.  BLE exists on over a quarter of a billion devices ... devices that can interact with other inexpensive BLE radios.

 

People tend not to buy gadgets until they make sense.  Not many other than geeks and well heeled replacement customers will buy Nest thermostats.  On the other hand there is a good market for networked door locks.  It is easy to explain the value of the lock - even I could see the need twenty years ago.  BLE is very cheap and will soon be well under a dollar in added cost.

 

Anything with a processor will have a connection to a network.  LED light bulbs will change color and brightness from an app on your phone or tablet.  Your rice maker will talk to your phone and other appliances.  Some information will go to the Internet, but some will have decisions made locally on smartphones, PCs or even other IoT devices.  This piece is interesting - Apple and Google represent very different approaches to where and how the information from these devices will be analyzed and other approaches will emerge.  Questions that beget questions.  A fascinating and deep area to ponder.

 

That brings us to how rapidly this will grow.  If you only think about UI class objects like Nest thermometers, Fitbits and the like,  the answer is not many for quite awhile.  These tend to be expensive and most people don't have an immediate need.  Very few of us are interested in the "quantified self" for example.⁴

 

But start controlling lights, knowing which pills your 90 year old mother on the other side of the country has taken so far today, learning if you left the tea pot on when you're ten minutes from home on your way to work,  and looking for those missing keys and you'll see a need.  In very little time none of this will cost perceptively more than the things we own that lack a network connection.  BLE can be made to run for years on a watch battery and the basic electrics will disappear onto the silicon with existing microprocessors or be available as a standalone chip for ten cents or so.

 

I was recently asked how big the IoT will be by 2025.  Most estimates are in the 30 to 50 billion unit range (currently there are about 10 billion devices connected to the Internet), but that is far too conservative.  My semi-educated guess is based on the notion that almost every gadget will be assumed to have a network connection.  Most of these will be to local intranets within the home (I like to call this ground fog based computing rather than cloud based computing) with a much smaller number being directly connected.  I think we'll see a half billion households with at least 100 IoT devices in each and perhaps a similar number in business and government - so I'll put my guess at a trillion devices.  So ubiquitous and so necessary that the terminology Internet of Things will have completely faded.

 

Will it make our lives better?  Well - if I don't have to hunt for the keys a ferret stole, I'm a happy camper.  If my shirt sends information on my health to my insurance company I may not be quite as happy.  Additionally security will be a huge issue -- there far too much to go into now except that is is largely ignored at this point.

 

At this stage the most important point is our perception of what the Internet is will continue to move away from the browser.  It is clear Google and others know something big is afoot.  Search is becoming much less important.

 

Tchaikovsky's 1^st Symphony is known as Winter Dreams - how fitting on this cold January day as one dreams of the direction of sensors and the Internet.   A lovely piece of music.

 

__________

¹ Prior to the 30s electric motors were expensive and it wasn't uncommon to use one for multiple tasks.  

² Most of your transistor count is likely to be in flash memory.  Depending on design flash chips can store one, two or three bits of information per transistor.  DRAM stores one bit per transistor and SRAM usually requires six transistors per bit.  My 16 GB iPod Nano uses MLC NAND memory - two bits per transistor, so it has about eight billion transistors.  Sukie's 256 GB MacPro has a MLC NAND, so it has about 128 billion transistors just in the memory.   SLC NAND is used in high performance flash memory and has just a single transistor per bit, but it is more expensive.  (SLC is single level cell - MLC is multilevel cell - refering to the density of information per cell.  Higher densities correspond to slower performance and shorter lifetimes.) In any event it won't be long before average households have close to a trillion transistors.

³ Some cars have over 100 microprocessors ranging from very simple four and eight bit chips to large 32 and 64 bit microprocessors.

⁴ In a few years the objects on us will be providing us with what amounts to a stream of biometric selfies.  There is evidence that there will be very small - the people who act on quantified health measurements seem to be those who act on suggestions by their doctors during physicals.

__________

Recipe Corner

 

A friend lives in Cameroon.  Here is a vegan adaptation I made of a peanut soup. An excellent Winter soup. 

 

Peanut Soup

 

Ingredients

 

° about 2 tbl of finely chopped fresh ginger 

° 3 cloves of garlic

° 2 cans of unsalted tomatoes (I had 28 ounce cans) diced and keep the liquid

° 1 cup smooth peanut butter

° 2 tsp paprika  (cayenne or a hot sauce if you like spicy soup)

° 1 15 oz can black eyed peas (2 cups final if you soak your beans)

° salt

 

Technique

 

° Mince the ginger and garlic in a food processor (I'm lazy)

° add the tomatoes and  juice, pb, and paprika and process 'til smooth

° poor soup in a pot and cook over a medium heat for about 15 minutes

° add the beans and cook for another 10 or 15 minutes - salt to taste