Another perspective on that "new" Ultra-High Definition TV picture

The press has been recently enamored by UHDTV – sometimes still called “4K TV” as the latest hot technology.  Indeed, as with past advancements in video displays - color, wide-screen, flat panel displays, stereo sound (and stereoscopic imaging) -- UHDTV is sure to make it into living rooms and home theaters eventually. 

But the benefits of high resolution imaging are not exactly new.

Here's an example:  take a look at this amazing high-resolution aerial photo of San Francisco -- right after the 1906 earthquake.  It was taken by George Lawrence, a prominent commercial photographer and inventor, who titled the photo "San Francisco in Ruins"

San Francisco may be in ruins, but this kite photograph taken in May 1906 has extraordinary resolution 
(Click HERE for the link to the high resolution file)

The photo shows the entire city on a single 17-by-48-inch celluloid-film contact plate, with amazing resolution.   It was taken from a kite flying 2,000 feet over the bay, in May 1906.  The camera is said to have weighed almost 50 pounds.  

I can make out that there were only a few houses in my Potrero Hill neighborhood back then.  They were built on solid Serpentine bedrock, so held up pretty well.

The combination of captivating content (the ruined city) and immersive image detail turned out to be a commercial success – netting Lawrence $15,000 in revenue from selling contact prints from the original negative.   That’s equivalent to around $380,000 in today’s dollars.   We can only hope that UHDTV commands the same monetary success.

Earlier, Lawrence built this gigantic camera used to photograph a train for the Paris Exhibition in 1900.  

This giant camera was built by George Lawrence to photograph a train in Chicago.
The resulting photo was a hit at the 1900 Paris Exhibition.

The camera weighed 1400 pounds and used a glass plate negative measuring 4½ feet by 8 ft.    It seems that several Camera Assistants were needed when using this puppy.  

If you could somehow manufacture a CMOS imager of the same size, and if such imager had the same pixel density as – say – a Nikon D90, you would have created a 120,000 Megapixel (that would be 120 Gigapixel) camera.   Now THAT would be a hit at the next CES.   The extraordinarily high resolution print created by this camera won “The Grand Prize of the World"  at the Paris Exhibition.   The Ultra-Ultra-Mega-REALLY high definition standard of 113 years ago.

Alas, we’re not likely to see 4 foot by 8 foot image sensors any time soon (thankfully).  But Canon has already made this 20cm x 20cm imager. 

Size Comparison: The Large Ultrahigh-Sensitivity CMOS Image Sensor
Alongside an EOS Rebel T3i (EOS 600D) Digital SLR Camera

Besides very high spatial resolution, this sensor makes possible the shooting of video at 60 frames per second with only 0.3 lux of illumination (approximately the same level of brightness as that generated by a full moon).

New UHDTV displays are very impressive – but one must keep these things in historic perspective.  It will be a while before we can experience images equivalent to those enjoyed in Paris over a century ago.

BTW:  Stop by and see me panel at the SMPTE Symposium “Next Generation Imaging Formats; More, Faster, and BetterPixels”  at the SMPTE conference, October 21, 2013 in Hollywood.  

4K TV is equivelent to Bay Bridge Lights. I can prove it.

My recent post on the new Bay Bridge Lights had several nice comments.   The lights are a new enhancement to our landmark San Francisco Bay Bridge, and have been it's been very entertaining to watch the modulated light patterns each evening.  One colleague on Linked In said (presumably tongue in cheek) "who in their right mind would want 4K in their living room when they can have 0.025K out their front window and get it commercial free!"

That got me thinking, and it turns out, it's a good point.   By some measures, the bridge and your new Ultra HD television provide the same visual performance!

Coming soon to your living room:  Ultra high definition TV,
with four times the resolution of 1080p.

In a previous post, I babbled on about the inevitability of 4K television, more accurately termed Ultra High Definition, with the "UHDTV" terminology likely driven by lawyers rather than engineers.  This new standard was prolific at the CES show in January, and several models will become available for purchase over the next few months, with prices likely to start below $4,000 for a 55-inch model.  UHDTV has four times the spatial resolution of the 1080p high definition that you are now used to.  Can you really tell the difference?  The answer is absolutely yes -- but only with proper content and viewing conditions.

Human vision has a remarkable ability to resolve detail.  The fovea of the human eye (the part of the retina that captures details in the center of your field of view) contains about 140,000 sensor cells per square millimeter.  This means that if two objects are projected with a separation distance of more than 4 microns on the fovea, a human with a normal visual acuity (20/20) can resolve them.  On the object side, this corresponds to 0.2mm at a distance of 1 meter, equaling one minute of arc.   This is the reason that the Snellen eye chart has special font characters, called optotypes, made from lines that are exactly thick enough to subtend an angle of one arc minute when viewed from 20-feet away.  That's where the term 20/20 comes from.  The optotype font is five line-widths high (notice the funny looking E at the top?).   As with all things, humans are a diverse lot, with many variables, so this is just a ballpark characterization.  Actual results may vary.

The Snellen eye chart Optotypes (line 8) subtend an
angle of five arc minutes vertically, when viewed from 20 feet

Based on this understanding of human vision, the International Telecommunications Union sector on Radiocommunication (ITU-R) recently issued a report on the present state of Ultra High Definition TV.  It concludes that the optimal viewing distance for UHDTV (3,840 x 2,160 resolution) is 1.5 times the screen height.   A 55-inch television, measured as diagonal screen size, has a picture that is about 27inchs, or 68.5cm high.  One and a half times this height means the optimum viewing distance is just a tiny bit over 1 meter.  Further away, a viewer with 20/20 vision would not benefit from such high detail, so why bother with the nearly 8 million pixel display?   Closer, and an astute viewer would start to see individual pixels, breaking the magic spell of Downton Abby's storyline with distracting dots.

Now, lets turn our attention to the Bay Bridge Lights, which use 250,000 LED lights grouped in 25,000 nodes, which we will consider pixels.  After some digging, I was able to determine the pixel spacing is 12-inches vertically.  But this is not a normal display -- the horizontal dot pitch is 30 feet, since the lights are affixed to the 306 northern support cable bands securing the bridge deck (roadway) to the huge suspension cables spanning over the towers.  In TV terms, this would imply a pixel width ratio of 30:1, far from the 1:1 "square pixel" that we are accustomed to.   As computer graphics wizard Alvy Ray Smith so eloquently puts it: "A pixel is not a little square".   This proves him right -- in this case, a pixel is really a huge rectangle!   (Yes, I know this is exactly the opposite of Alvy's point, so you don't need to remind me.  I just couldn't resist...)


The Bay Bridge has 612 cable bands supporting the deck from the suspension cable,
each with for wire ropes, for a total of over 42 miles of rope used in the 9,260-foot western span.

So, based on normal human visual acuity of 20/20, what is the optimal viewing distance of the bridge lights -- with a focus on the vertical resolution of pixels 12-inches apart?   The answer is one kilometer (actually, 0.9873 km, but who's counting).  That is to say, the vertical resolution perceived by a viewer 1 meter away from a 55-inch UHDTV (4k) television is the same as the vertical resolution perceived by a viewer 1 kilometer away from the bridge.  

Next, I had to find such a viewing location to check it out.  The Bay Lights Web site includes a map with recommend viewing locations.  Among the spot most used for photographs is Pier 7, just North of the Ferry Building.  Google maps teaches us that the end of Pier 7 is one kilometer from the second tower (W2 pier) of the bridge.   

It was a clear, cool evening here in San Francisco, so I paid a visit to Pier 7 this evening.  When viewing the bridge lights from that location, the vertical rows appear almost continuous, without visible pixelation.  This seems to reinforce the theory in calculating the optimal viewing distance.   Viewing the bridge from a closer location, such as Pier 1, could allow viewers to see individual lights more prominently.  Viewing from further away, perhaps from Coit tower or Pier 39, would look great, but not benefit from all 25,000 LED nodes -- the same effect would be achieve by fewer lights (and lower electric bills). 

Bay bridge lights viewed from Pier 7 achieve the limit of 20/20 visual acuity
(Photo: James Tensuan, The Chronicle)     

However, I'll admit it's not fully definitive.  First, when viewed from that distance, the lights appear to shimmer from atmospheric refraction.  Also, the pixel fill factor on the bridge is very low:  on a typical LCD panel, there is a very small space between the light emitting LCD cells, so that roughly 90% of the planar area of the panel is emitting light.  By contrast, the bridge has 1.25-inch LED nodes spaced on 12-inch centers, resulting in a very low pixel fill factor of 10%.  This probably contributes to imperfect viewing -- so it's not fully comparable to a $25,000 Ultra HD television.

Please consider this as you shop for your next television at Best Buy.   If you plan to watch your UHDTV from more than five feet away, you'd better spring for the 84-inch model to achieve that optimal viewing distance.

Bay Bridge Lights - a 25 KiloPixel, two mile display

I’m very fortunate to live in the Bay Area,  with a decent view of the marvelous Bay Bridge from our house on a hill in San Francisco.  The Bay Bridge plays second fiddle to the Golden Gate bridge, but it is still an icon, connecting us with Oakland, not to mention the rest of the country.  Tonight was a very special occasion, with the launch of Bay Lights  the installation art project you’ve probably heard about.   Unfortunately, it was drizzly and overcast, so our view was limited.  But I watched the inauguration via web streaming with my family.  The telecast featured Mayor Ed Lee and serial politician Gavin Newsom.  While it was a nice event, the video streaming video quality sucked… but that’s a topic for a future post.



Artist Leo Villareal controls the Bay Lights from his laptop.  (Photo credit: Cy Musiker/KQED)

Being involved in display technology, I was naturally curious about the engineering behind Bay Lights.  The project was initiated by artist Leo Villareal , who (according to his web site) “is known internationally for his light sculptures and site-specific architectural works.”  Leo unquestionably did an amazing job at fundraising, and navigating a complex bureaucracy in order to attach 100,000 feet of cable to the bridge.  Before I forget, they still need $2 million bucks.  You can donate here.   

The overall system – over two miles long – provides and amazing showcase of geometric illuminated patterns.   The entire North side of the bridge appears to shimmer  with dynamic and unpredictable movement.  Villareal, whose office is right down the street from our house, programmed algorithms in the display drivers for the lighting sequence, which makes for an impressive site.   But I wondered about the display technology used. 

It took some digging, but I found that Bay Lights is based on LED technology from Philips Color Kinetics.  A team of engineers in Boston formed Color Kinetics in 1997, and it was acquired by Philips five years ago.  The Bay Lights project uses a product called eW Flex SLX.   125,000 white LED’s were used in the installation, grouped into 25,000 nodes, each strapped to the vertical rods serving as the suspension structure for the bridge.  The nodes measure about an inch and a quarter wide, and are spaced one foot apart.    Overall, the press reports say this covers a span of two miles wide and 500 feet high, on four towers.   Nearly 4.5 miles (24,000 feet) of LED light strips were installed.

The Color Kinetics architecture groups 50 nodes  into a string driven by one controller.  Each node is illuminated on an 8-bit (255 level) brightness through pulse width modulation.   This means that the LED is quickly switched on and off to create the perception of different light levels.  For example, a 50% duty cycle will cause the appearance of half brightness. 

 Control signalling is made over data connections on CAT 5 cable, using a proprietary protocol at 500 kbps.  The control data is processed by hubs, probably in the configuration shown here.

One of Color Kinetics proprietary technologies is the Chromastic chip, a low-cost and lower power consumption LED driver included in each node.  The Chromasitc is typically used for RGB color control of LED architectural lighting, but for the Bay Bridge it provided a convenient means for  Villareal to provide the 8-bit brightness control.

For the past month, I’ve occasionally seen tests of the lighting system in the middle of the night from the bedroom window. When I posted reports, some of my Facebook friends have questioned the environmental aspects of lighting up these zillion LED clusters, just for fun. Let’s explore this.

The Philips eW Flex SLX modules use for the bridge are emit white light at 4,200 degrees Kelvin correlated color temperature, and hit a maximum of 16.2 lumens per five-LED node. Each node consumes up to one Watt at full brightness, which means the luminous efficacy is 16.2 lm/W.

The Philips press release states that “the new LED lighting system uses 85 percent less energy than traditional lighting technologies” 

So I wondered:  what are “traditional lighting technologies” for an installation art light show??   It seems the luminous efficacy of a typical incandescent bulb is 16 lumens per watt, virtually identical toeW Flex SLX modules.  So this would NOT explain an energy cost savings of 85%.  What’s up with that? 

With 25,000 one-watt modules, the Bay Lights consumes 25 kilowatts of power a full illuminance.  The organizers say that the lights will be turned on from “dusk until 2:00am”.  Looking at sunset times, that equates to 7 hours 15 minutes per day on average, or a total of about 180 kW/hours per day.  A recent report  says that electricity costs in San Francisco are currently more than 60% higher than the national average, at 21.2 cents per kilowatt hour.  So if all these nodes were turned on, that’s over $14,000 in electrical consumption per year.  The organizers claim that power costs are $11,000 per year, so we’re in the right ball park  --- it’s doesn't seem like too much.

You may have also been wondering about total luminance flux.  With 25,000 modules at 16.2 lumens, that comes to  at total of over 400,000 lumens  -- equivilent to around 20 typical movie theater projectors.   (Except, of course, we're talking about a 10,000-foot by  500-foot irregular shaped "screen" which is emitting, not reflecting light.)

If you're interested in longevity, the Bay Lights project says they are committed to having this art installation functional for "at least" two years.  But the good news: the LED nodes are rated at about 40,000 hours, which should keep them running for 15 years, assuming dusk to 2am operations -- just over 7 hours -- each day.   I would guess that the atmospheric conditions and salt water will greatly shorten the time before something fails.

Bottom line:  this is a cool art project, but a bit expensive at $8 million capital cost.   The $11,000 per yer in electrical costs is at least particially subsidized by solar credits.  Pundits are predicting tourism benefits of nearly $100 million.   So... why not!   400,000 lumens of visual art is kind of nice, in our fair city.

Filling in the White Spaces

Yesterday, Google launched the public trial of their new spectrum database, and it's pretty interesting.     First, a bit of background.

For at least the last 10 years, there has been much chatter about better utilizing scarce radio frequency (RF) spectrum to accommodate the billions of phones, tablets, laptops and other mobile devises connecting to the web wirelessly.  It turns out the much of the prime spectrum, from 470 MHz to 700MHz, known as the Ultra High Frequency band, has been allocated for broadcast television since the 1950's.  Early on, this spectrum was barely used since few TV's were able to receive UHF channels.  That all changed after 1962, when the All Channel Receiver Act (ACRA) was passed by Congress, requiring all television sets to receive UHF channels 14 through 83.   Channel 37 was the only exception, since that frequency (centered at 611 MHz) turns out to be very useful for radioastronomy.  Immediately, the value of UHF spectrum shot up.  There are now over 1,500 UHF television stations, and thousands more translators and low-power stations filling this chuck of bandwidth.

UHF has the advantage of using relatively short antennae, with a quarter wavelength of four to five inches.  In addition, this 470-700 MHz band has favorable propagation characteristics, since it can pass through walls and other obstructions more easily than higher frequencies.  So the question is: how to take advantage of the "white spaces" in the spectrum between operating TV stations.




I searched for available spectrum in my neighborhood.  Slim pickings...

Google and nine other companies -- including Microsoft -- submitted proposals to operate databases to identify available frequencies for new, unlicensed services operating in the UHF band.  A few of them, including a company call Spectrum Bridge began trials of their database a few months ago.  The generally picky National Association of Broadcasters has already kicked the tires on the Spectrum Bridge database, and gave it generally favorable reviews. 

The Google database is fun to use, and has good explanations.  Their party line is "Google is doing our part to free up spectrum. One way we are doing this is through dynamic spectrum sharing, one of the most promising ways to make more spectrum available to the public for broadband access. As part of our effort to improve connectivity globally, Google.org is working to promote dynamic spectrum sharing through a TV white spaces database."   Sounds very altruistic, doesn't it?

I'm a bit skeptical of how accurate and up to date these ten synchronized databases will be.  Any errors will cause problems in wireless service quality, not to mention wireless microphones at the Taylor Swift concerts...

4KTV: Calculating Megapixels per Kilobuck

It’s been about ten months since the new Ultra High Definition television standard was announced by the ITU-R.   The press release explained that “UHDTV is an earth-shaking development in the world of television”.   The committee chair that completed the work is my buddy David Wood, who was quoted as saying “Some years will pass before we see these systems in our homes, but come they will.”  Of course, David’s right.  But how many years? ...and what exactly are we waiting for?

Already, UHDTV sets are being sold by Sony and others.  The first products to hit the market are large (84-inch) and expensive ($20k to $38k).   The picture quality can be amazing, but at that price, Ultra HD isn’t on my wish list yet, even if I could fit one in the house.   Smaller, and presumably less expensive sets were shown at CES in January, and are expected to start shipping next month.   Sony will have 55-inch and 65-inch models available.   

Sony's recently announced XBR-55X900A and XBR-65X900A are scheduled to ship in April 2013.

Respected pundit Pete Putnum recently observed that some Taiwanese and Chinese manufacturers (Westinghouse, Hisense, and TCL) are already floating aggressive prices on 4K TVs; about $50 – $60 per diagonal inch in sizes up to 65 inches.  I'll save you the math -- that's less than $3,300 for a 55-inch model, or a half-penny per dozen pixels, if you're counting.  Ultra HD sets have not only come to market quickly, but it appears the same price wars that have pummelled the HDTV business are already migrated to UHDTV.

Remarkable developments -- and investment -- in display technology are contributing to these rapid price declines.  Consider the pixel density require for Ultra HD displays.  The UHDTV-1 standard calls for four times the resolution of HD, which comes to 3,840 x 2,160.   Applying Pythagoras' theorem, we can figure out that in a 55-inch diagonal panel, the screen width is 48 inches.  Fitting 3,840 pixels in 48 linear inches requires a pixel pitch of 317 micrometers, or 80 pixels per inch (ppi).

That sounds pretty impressive until you realize that the iPhone 5 display is four times higher pixel density, at 326 ppi.  With charactoristic hubris, Apple claims "The pixel density is so high that the human eye is unable to distinguish individual pixels."  That’s arguable.  But the new HTC One phone even beats that, with 468 ppi display supporting full 1920 x 1080 high definition on a 4.7-inch display.  Still not impressed?  How about these glasses, which display 1920 x 1080 on a 0.74-inch panel.  That works out to nearly 3,000 ppi – over 40 times the pixel density of the fancy new 55-inch Ultra HD set.   But that's not the limit -- Sony has released a 4K projector with 0.74 LCOS (Liquid Crystal on Silicon) panels, with an amazing 6,257 pixels per inch.    Perhaps this sheds some light on the further progress to be expected in LCD panel developments.

The 4K Sony VPL-GT100 has three 0.74 inch
LCOS panels with 4 micrometer pixel pitch

 So if UHDTV televisions are available, and will be affordable before long, what can go wrong?   Let me think about that, and get back to you...

4K or not 2K - Is that the question?

If you've been even vaguely following the latest consumer television trends, you are already familiar with the recent introduction of "4K" television, also known as Ultra High Definition (UHDTV) or Quad HD (QHD).  The 4K name is derived from the horizontal pixel count of 3,840 -- which is darn close to 4,000.  As with all new technologies, there is a fair share of confusing nomenclature.   In the world of digital cinema, the term 4K refers to a matrix of 4,096 (wide) by 2,160 (high) pixels, so the 4K name seems a bit more appropriate – there are even 96 bonus pixels, if you're counting.

However, for home television, the 16:9 aspect ratio is preferred.  By doubling both the width and height of the old fashioned 1,920 x 1,080 HDTV raster, you arrive at 3,840 by 2,160, which provides amazing spacial resolution, even though it’s about 7% less than 4K digital cinema, as if that really mattered.   Professional cinematographers routinely refer to this “way beyond HD” in terms like 4K or 5K.  But the consumer electronics industry seems a bit more skittish.  Last October, the CEA announced that “the next generation of so-called 4K displays… will be called ‘Ultra High Definition’ or ‘Ultra HD’”.   Here’s my theory as to why.  

Only a few years ago, consumer TV’s were defined as “19-inch” or “42-inch” in reference to their diagonal screen dimension.  Then, around 2008, retail ads starting using the term “19-inch class” or “42-inch class”.  Class?  What the heck does that mean?   Perhaps the answer comes from the 2010 regulatory action in San Diego and six other California counties against six leading TV manufacturers.  The injunction clarified how screen size measurements are to be represented in the future.  In the past, most TV makers rounded up the size to their screens, measured diagonally, to the neared inch.  But the California Division of Measurement Standards and the local Weights and Measures Departments felt that this violated California law, and brought their concerns to the attend of county prosecutors.  The CE guys settled by agreeing to donate a million bucks worth of TV’s to schools (and to pay the lawyers another million).

In my view:  silly.  The difference between a 41.5 inch and 42 inch panel is insignificant to the consumer, especially considering bezel design, and the letter-boxing common in much programming.  But this past injunction may have made the consumer electronics folks a bit more cautious in calling a panel with 3,840 pixels “4K”.   (Where are my 160 missing pixels??)  So there you have it:  It’s not 4K, its UHDTV.    Whatever.  After all, professional digital cinema has a standard called 2K, representing 2,160 x 1080 pixels… just a tad more than high definition at 1,920 x 1080.  But nobody talks about 2K for home; it’s consistently called HD.  So why not Quad HD / Ultra HD for the next step in home theater, while leaving 4K for the multiplex?

Yes, I have opinions on how important this format is at home.  But I’ll save that for later.