What are the advantages of Side-by-Side versus Top-and-Bottom 3D Formats? (Part 3)

In the first installment we covered how interlaced formats affect 3D broadcasts.  In the second post we covered some of the technical reasons behind using side-by-side (SbS) and top-and-bottom (TaB). We were left with a perplexing problem – why did CableLabs specify 720p60 stereo 3D broadcasts to use TaB instead of SbS?  Looking at this from a pure image quality and resolution standpoint makes it clear this is not an optimal choice, as we covered last time.  We have to look beyond the purely technical and into strategic thinking of some of the key industry players, the installed base, plus inertia and momentum of the Market to guide us.

When one strays from the purely technical as is necessary here, some bit of crystal-ball-gazing will be required – in some sense I will be venturing guesses as to the collective state of mind of organizations (like CableLabs) and companies (like, say ESPN, or Comcast).  It is with this particularly in mind that I give the following disclaimer – any time I ascribe anything resembling opinion or any other anthropomorphic qualities (or, for that matter, facts unsubstantiated by references) to a third party company, they represent my own opinions.  They are merely what we in the technology industry would term SWAGs – a less polite way of saying “informed speculation”.

A Trip Down the Rabbit Hole (through the broadcast chain)

To begin, the 3D Masters golf tournament is a good case study to follow a 3D video signal all the way from the camera to the living room, and to see how the resolution and motion information is affected.

The cameras used here were pairs of Sony HDC-P1s and HDC-1500s — with the Pace Fusion rig linking them up stereoscopically.  The contribution was 1080i, two HD-SDI feeds per eye.  These were processed in NEP’s SuperShooter (which can originate either 720p or 1080i — pick your poison).  What isn’t clear from media reports, but that we can sort out pretty plainly, is that although the production was financed via ESPN (via Disney, and ultimately via Sony), the broadcast probably did not wend its way to ESPN’s ground station and master control in Bristol, CT, as Bristol has historically been a 720p facility.  At least if it did, it did not use the usual infrastructure.  In turn, its feeds to the rest of the World have, and continue to be, 720p60.  Why they built their programming and facilities around this format back in the day instead of the more typical 1080i is a whole other subject. 

Another wrinkle

A curiosity about this broadcast is that it is one of the few that the cable/MSOs such as Comcast and Cablevision had an exclusive on.  If transmitting 720p60 TaB was the goal of CableLabs, you could be sure that if Comcast had an exclusive on a first-time event such as this, they could have leaned on everyone to tow CableLabs’s line of reasoning.  They didn’t.  The broadcast landed in peoples homes as 1080i, side-by-side.  Or, another way of looking at it, their line of reasoning did…

So, it would seem we either have a 720p-only broadcaster who is very uncharacteristically originating content at 1080i for the first time, OR, we have the largest MSO, Comcast, immediately creating a scism with its own “brother-from-another-mother”, CableLabs.  

My belief is that the answer is mostly the first case, under duress.  The entire broadcast was produced, originated, contributed, and distributed in 1080i.  Just the way I suspect CableLabs and Comcast wanted it, to the opposite purposes of ESPN and the other 720p sports broadcasters.


There is plenty of consternation in the broader industry around CableLabs’ insistence upon TaB for 720p60.  It’s easy to accuse them of ignoring some issues, but the fact of the matter is that these are very intelligent people, and this decision is very likely not without reasons. Here are some possible reasons I came up with – and where the speculation begins:

Drive a stake into the heart of 720p via Sports Broadcast

CableLabs doesn’t particularly like 720p60, because every cable operator dislikes 720p60 (and CableLabs was founded by, is funded by, and exists for the purpose of benefitting MSOs). It’s a complication for settop boxes and other content production, contribution, and distribution in an otherwise harmonious world dominated by 1080i. There is an opportunity to drive a stake in the heart of this format, and CableLabs seems to have taken it.  ESPN has been “irksome”; primarily because of them, all MSOs have to support the entire universe of various ATSC formats most notably 720p.  What better way to “encourage” the House of Mouse and force them to 1080P (or 1080i) than by a Hobson’s choice?

Considering this more carefully, with a twist: 720p60 TaB Stereo 3D exposes the lack of resolution in this format to the point that even the most jaded sports fan is going to notice the lousy resolution (and therefore detail) in a 3D HD broadcast, while those using the more blessed 1080 SbS will only barely notice Comcast Media Center “gently stepping on” their otherwise HD quality with statmux and allegedly, requant.  TNT Sports shoots and contributes all their sportscasts in 1080i, and this difference will show in 3D like no other.  ESPN will be forced to play ball with Cablelabs, their content will simply be too awful to watch otherwise.  If you are a Programmer, and you want your 3D programs to have carriage on Comcast (or any other Head-End-In-the-SKY — HITS customer) AND you want to contribute 720p?  It has to be 720p60 TaB.  It’s a classic Hobson’s choice.  You could, but the quality will be so utterly hideous and revolting as to thoroughly eliminate it from consideration. Your only REAL choice is 1080i or 1080p if 720p60 SbS is not available as an option for contribution.

For Good Measure, a Silver Bullet for 720p via the 3DTV Market

Just in case the stake in its heart wasn’t enough, here’s a silver bullet.  Imagine a world where instead of shutter-glasses 3D with the attendant $200 per pair costs, a company like Vizio starts mass-producing polarized-glasses displays – wait, that happened at CES in January!  You can bet Comcast and Cablelabs did — these are people who touch the customer repeatedly, and intimately.  They have known for a long time that shutter glasses just will Not Work for a vast majority of households.  Period.  In spite of all the wishful thinking and considerable investment of Panasonic to the contrary.

Polarized glasses can be had for free by watching a RealD theatrical release.  Bought, they cost under a dollar, are disposable, and lend themselves to the “superbowl party at your house” scenario.  Thanks to compnaies like Oakley et. al. you can buy multi-use glasses that serve as both fashion sunglasses AND 3D polarized eyewear.  This will likely be the norm for the mass market, going forward.  Shutter-glasses displays will fail to produce significant mass-market adoption due to major ergonomic issues including flicker when you look away from the 3DTV, weight, dorkiness, and no chance of dual or multi-use.  Not to mention they’re all incompatible with each other.

All of the first generation 3DTV displays require active shutter glasses.  These are electronic devices in and of themselves, require precise engineering, and are very unforgiving to (frequent) design mistakes, and therefore expensive – the Panasonic plasma 3DTV displays, in concert with their glasses, show great 3D… if you can glue them to your face long enough to see 10 minutes of a movie.  The Panasonic glasses are frustrating – they are the best engineered glasses for a 3D experience, but they are very ungainly from an ergonomic point of view.  You simply can’t keep them on your face for the duration of a 90 minute movie, let alone a 4 hour baseball double header.

So how does this affect the SbS and TaB debate?

Given that passive displays will displace shutterglasses displays in very short order (my personal prediction), the Hobson’s choice if not chosen by the programmer — a suicidal choice — will be forced upon the consumer.  Most passive glasses displays divide the left and right eye views into interlaced lines as for “xPol” or line-polarized displays (with the very notable exception of Samsung, who are using RealD’s “zplate” for a full-res polarized solution — the right way to approach this, ultimately).

This means no matter the input signal format, say 1080P SbS, an xPol display will take HALF of the vertical resolution of the left eye, put it into the odd lines of the final display, and half the vertical resolution of the right eye and put it into the even lines of the final display.

Now imagine that a certain MSO’s settop boxes (and a goodly number of consumer electronics manufactuerer’s TV’s) use an xPol or other line-interlaced display technology.  You can imagine that in many circumstances, a 1280×720 video signal, carved up to 720p0 TaB becomes 1280×360 for each eye, then the display takes it and stomps it down to an effective 1280×180 resolution.

To see what this really means, lets take a small subrectangle of a 1280×720 image:

Now let’s fairly decimate it the way a viewer would see it in one eye, with a polarized display, according to 720p60 TaB transmission – assuming a good video processor in the TV:

Finally, let’s fairly decimate it the way a viewer would see it in one eye, with a polarized display, according to 720p60 TaB transmission – assuming a lousy video processor in the TV – and unfortunately, this is the norm (note the loss of resolution, plus the stair-step pixelation):


This is clearly both horrid and repulsive, and all but guarantees that the sports video market will abandon 720p if forced into 720p60 TaB 3D contribution by the MSOs and other distributors who will likely follow the lead of CableLabs.  The first insult would be the forcing of contribution via 720p60 which ensures 360-line effective resolution, even less than that of SD, and approaching VCR.  Second is the insult of the compression to jam 100 pounds of stuff into a 12-pound sack that the Comcasts, Verizons, and DirecTVs of the World use to deliver the video at allegedly-HD resolution, made even worse by 3D.  Finally, the third and final insult would be the likely, market-dominant display technology itself taking an already decimated vertical resolution and decimating it even further to barely-even-YouTube-resolution. 

… And VoD Buries the Evidence

Now, let’s take a different perspective.  Advertising inserts on ESPN and the like, minus the considerable carriage license fees are one thing.  ESPN is appointment-TV, however, and the prime demographic wants to be served what it wants, when it wants it, on the TFT panel it wants it on, be it TV, tablet, computer monitor, and to a lesser extent (this far, in North America) on small mobile, e.g. cellphone.  For my part, I’ve actually watched 3 of my 5 kids crowd around an iPod Nano to watch a television episode. (Seriously, an iPod NANO).  Pay-per-view, and on-demand across multiple platforms is the order of the day, and will be the order of the future. Movies (PPV) and Episodic TV (On-Demand, and Time-shifted via DVR) on any platform anywhere, anytime will be the way younger people consume this content in the present and future.

Let’s examine what this means for HD, and 3D formats in particular.

First, most movie features, whether intended for 3D theatrical, 2D theatrical, direct-to-DVD or direct-to-BluRay, are shot 24P with either digital cameras or actual film cameras.  Many popular episodic TV dramas are shot 24P this way as well (old habits die hard).

Then some variation of what is usually an awful process converts this 24P footage to a 59.94 interlaced 1080 format suitable for broadcast.  The details of this are too gory to go into among mixed company, but suffice to say with few exceptions, it takes smooth motion and makes it various degrees of ugly. Insult upon injury, doing this conversion is not cheap (or, at least doing it well, is not cheap).  In any case, converting this stuff requires labor and capital input something over zero, so it is a prime tendency therefore to scan films only ONCE. Why do it twice if it’s already digitized?

In the meantime, we have to consider 3D — is SbS or TaB better?  Since the original master is progressive, 24fps,  one would think that the simple analysis in the first couple posts (and this one) would apply.  Not so fast.  If you have thousands of VoD programs that have already been digitized and made broadcast-ready via conversion to 1080i, why would you upset the apple cart?  The result is that most films scanned and converted (badly) to 1080i (or 480i) will likely remain that way.  Conversion from interlaced formats to TaB is bad, as covered in the first installment.

For distributors and MSOs, it’s better to stack the deck in favor of content that’s already been mangled to fit the 1080i broadcast chain.  By mandating 720p60 be contributed as TaB, this not only kills the format for broadcast 3D, but utterly ensures its extinction by making Films converted to 1080i and 480i via the industry standard processes for decades, utterly unusable if the Programmer wants to use 720p60.  A Programmer could re-scan the film assets to 720p60, but why? 


In summary — my opinion is that the 720p60 TaB choice by CableLabs is simply this: a solution that is not optimal for the viewer, but is optimal for the MSO along several dimensions — 1) control and leverage over programmers generally so they can control their own intrinsic contribution, distribution, and customer-premise-equipment costs, 2) another lever against Hollywood and other originators, especially Disney and her broadcast properties; and finally 3) a way to give their own VoD content a leg-up cost wise against other libraries that have not been digitized and otherwise would be restrictive economically to re-scan or reprocess.  Particularly clever is that they are using several Consumer Electronic companies as accomplices (whether witting or unwitting).

Nothing is quite as cut and dried as it might seem at first blush.


~ by opticalflow on February 18, 2011.

4 Responses to “What are the advantages of Side-by-Side versus Top-and-Bottom 3D Formats? (Part 3)”

  1. I don’t understand why you say the vertical resolution halves twice with xpol? The resolution is halved for left/right eye, but that is the same halving of the 3D image in TaB form?

  2. In the case I described, the transmission method halves the vertical resolution the first time. Then, depending on the video processing hardware in the xPol set, there is a smart way to re-sample the lines, and a really naive way. The really bad way is illustrated in the last picture. I’ve seen many 3D xpol displays not be particularly clever about observing parity or the assumption that the inbound signal may already have been decimated.

  3. Well for me your conclusion of halving twice really doesn’t make sense. I am not an expert but I just bought a polarized 3d-tv set and I actually think the opposite. The vertical resolution per eye is already halved for that technology when using 3d, so I actually have at full hd only 540 lines, a maximum resolution of 1920×540 per eye. If using half sbs, this would result in 960×540 per eye, which is a result of the maximum horizontal resolution of the material the TV set gets (Half SBS means 980×1080 per eye) and the maximum vertical resolution the TV can display.
    So the problem here is that the TV set gets a lower vertical resolution than it could display, which results in loss of detail due to SBS, and on the other hand gets a higher horizontal resolution than it can display and therefore needs to discard half of it.

    But if the TV would get Top/Bottom content it would get its maximum resolution, which is 1920 x 540 for each eye, and it would vertically stretch the images without further loss of resolution!

    I was looking for Half Top/Bottom content to verify that assumption when I stumpled on your blog.

    It does not make any sense to half a resolution when it can be displayed, and as there is no interlacing at 720p you really would just get the 360 lines per eye and never go down to 180 because the display will have more available lines available per eye than it gets (540 vs. 360). It will upscale and stretch the image vertically and not discard half of the lines! That would just be plain stupid.

  4. Well, while you claim not to be an expert, you ask the right questions and seem to be on top of things!

    I fully agree with you that the scenario I outlined is not ideal, but it seems to be common. I have seen actual polarized TV sets that make this mistake. I suspect that it is because of typical low-end deinterlacing processing. Most over-the-air and cable/satellite 1080 HDTV content is interlaced, and all LCD displays are progressive, so this interlaced content needs to be converted to progressive. “Lousy Video Processing” as I described usually consists of essentially throwing away most of the vertical resolution with an aggressive antialising filter (fancy wording for a guassian “blurring” in the vertical direction) in order to reduce interlacing artifacts.

    Recall that I did single out the difference between a “good video processor in the TV” and a “lousy” one. Also, I would recommend reading Part 1 for a complete treatise on how interlaced formats interact with both SbS and TaB 3D formats.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s

%d bloggers like this: