My take on what's interesting and what's irritating.

Wednesday, May 10, 2006

Sensor Guide Number

I was trying to explain to a colleague that an 8MP point and shoot digital camera was no where near the image quality of an 8MP Digital SLR. Fact of the matter is that sensor size, noise filtering and in-camera processing don't register with most consumers. During this conversation, bells went off about an article I had read a few months back over at Luminous Landscape. In it, Michael Reichmann stated,
"More Megapixels aren't what most photographers need. We need better Megapixels, and the manufacturers seem to have realized this."
While the manufacturers know this, the general public still doesn't have a good, easy way to differentiate 8MP point and shoot garbage from 6MP high quality Digital SLR. It reminds me in a way of the MHz myth that Intel has only recently abandoned. There's more to performance than megahertz (or megapixels).

In the spirit of the flash Guide Number system, I'm proposing a similar system for digital cameras. While in no way perfect, this system at least gets the consumer in the right ballpark and lets them make more informed decisions about the image capture capability of any given camera.

It basically works like this:

Sensor Guide Number = Effective Sensor Megapixels * Sensor Diagnol (in mm)

The higher the Sensor Guide Number, the higher assumed image quality.

A quick comparison between apparent equals in terms of megapixels tells a different story when the Sensor GN is applied:

Canon EOS 20D 8.2MP; diag = 27.04mm; SGN = 221.74
Canon EOS 5D 12.7MP; diag = 43.27mm; SGN = 549.49
Nikon D200 10.0MP; diag = 28.40mm; SGN = 284.01
Pentax *ist DL2 6.0MP; diag = 28.26mm; SGN = 172.40
Pentax Optio A10 8.0MP; diag = 8.01mm; SGN = 64.04

As we can more clearly see, the 8MP Canon EOS20D and 8MP Pentax Optio A10 are not created equally. The 20D has a SGN of 222 while the Optio A10 has a SGN of 64. Even though both are 8MP cameras, the 20D SGN is 3.4x that of the Pentax. Even the two Pentax cameras - one a 6MP DSLR and one a 8MP Point and Shoot - show clear differences favoring the larger sensor size of the smaller MP digital SLR. A consumer might wonder why a 12.7MP digital camera with "50% more MP" than an 8MP point and shoot is so much more expensive. When you look at the SGN and understand it's 8.5x "better" then its a more realistic comparison of capability.

It would be easy to take this to the next level and have a Dollar Per Sensor Guide Number ($pSGN) metric where you can factor in the cost of the camera per SGN. The lower the $pSGN the better the "value" of the megapixels you are purchasing.

$pSGN = Price of Camera System / SGN

Continuing our 20D vs Optio A10 comparison, today's price at B&H Photo-Video lists a 20D with lens kit for $1129. The Optio A10 can be had for $299. The 20D has a $pSGN of $5.09 and the Pentax has a $pSGN of $4.67. Given the superior SGN of the 20D (3.4x of the Pentax), it's the better value. Whether or not you have >$1000 to spend on a digital SLR is another story!

I like that this method abstracts MP as the sole arbiter of camera quality, provides a way to compare disparate models and manufacturers and with the $pSGN value allows one to assess value.

This method is not without faults and doesn't encompass factors such as in-camera noise filtering or the quality of a given CMOS or CCD sensor, but it does provide a much more informed decision for the average consumer.

This same system could be used as a guide to consumers interested in translating "how much camera do I need to make 5x7 prints". If the answer is at least a GN of 50, then that makes a more informed decision as well.

What do you think? Let me know.


5/12/06 - Update - check out this post for another good analysis of the "MP myth".

Monday, September 12, 2005

Netflix On-Demand Idea

In my previous post, I talked about an idea for a "Netflix On Demand" service. After giving it a bit more thought, I think it could work in a couple of different ways.

  1. A download service owned and maintained by Netflix --OR--
  2. A torrent based service where everyone's downloaded queue is the content for the torrents.

Having Netflix own the distribution channel has the advantages of having them control the content and thus probably allaying the fears of the entertainment industry. Having a torrent service would significantly reduce the bandwidth requirements of Netflix.

Netflix and TiVo announced about a year ago a partnership that was pretty short on details. The implied assumption was that you could download movies from Netflix to watch on your TiVo. While I think this is a good start, I'd rather use Microsoft's Windows Media Center since it more easily supports external storage (try taking your current Series2 TiVo to 1TB!) and could more easily accomodate my download the queue idea.

Windows Media Center has a limited capability with the Movielink and Cinemanow sites integrated into the 10' UI interface. However, Movielink has several limitations I'd like to see Netflix overcome:

  1. Watching a movie more than once. Movielink has the annoying requirement that once you *start* a movie you have 24 hours to finish watching it. I'd like to see a model where, just like the current Netflix, you can watch any move you've got 'rented' as many times and keep it as long as you want.
  2. Only one movie at a time. I want the ability to have multiple rented movies that I can view. If I've got the 3-at-a-time plan, I want the ability to have 3 downloaded movies that I can watch at any time. When I want to watch another movie in my queue (that's already been downloaded) I would have to 'return' (delete) a currently rented to unlock the next movie.
  3. Waiting to watch while a movie downloads. Although both Movielink and Cinemanow offer buffering to start watching movies as quickly as 30 seconds after purchase, the reality is network hiccups can often disrupt the viewing experience. With Netflix, after the first movie is downloaded, the others would immediatly begin downloading. Every movie after the first one would be immediately viewable out of the queue.
  4. Ability to buy movies. Netflix customers can buy 'previously viewed movies'. Instead of buying used copies, I'd like the ability to burn a copy of one of the rented DVD for the same purchase price as the other Netflix customers. I might even buy the DVD case and artwork for a small fee.
  5. Continue the x at a time, queue based system. Continue to offer the current model but apply it to the download model. Having a pre-downloaded queue is the key to the instant gratification and high customer satisfaction.

Advantages for Netflix:

  • Same rental revenue but no rental physical media to deal with.
  • Begin phasing out regional warehouses in lieu of a centralized datacenter as adoption rates go up.
  • If content can be delivered for less than the cost of postage, profits go up.
  • Minimal bandwidth costs in the torrent model.
  • Ability to sell the downloaded movies as an incremental revenue stream.
  • Highly satisfied customers through the instant gratification of watching the movies in their queue.

Disadvantages for Netflix:

  • Cost of bandwidth for distributing 30,000 titles to millions of customers. Apple does it on a smaller scale with the iTMS but this is obviously a LOT more data. Still, with some sort of download management system, movies farther down the queue could literally trickle in over weeks. The bandwidth problem is significantly mitigated by the torrent method.
  • Potential for customers to 'rent' and thus turnover significantly more movies per month. If Netflix pays the studios fees when movies are rented, their costs could go up significantly.
  • Waiting around on TiVo while Microsoft and Movielink figure it out in spite of Netflix.

If I were a development manager at Netflix, I'd kick this project off by doing something like this:

  1. Cost-Benefit Analysis of the Torrent method vs the distribute it yourself method
  2. Get the Netflix CEO to assuage the fears of the movie industry that this won't be another Napster free-for-all
  3. License Janus DRM from Microsoft and work with TiVo to develop an equivalent
  4. Start developing a download manager, MCE plug-ins, TiVo interface and DRM wrappers
  5. Build out datacenter for distributing content (either torrents or content)
  6. Start ripping an arse-load of DVDs
  7. Test it
  8. Release it

As a Netflix customer and future Media Center owner, I'd move to this model rather than the snail-mail DVD system in a heartbeat.



What's wrong with DRM?

At some point, we all come to the realization that it's just too hard to enjoy the content you already own. Nothing's quite as frustrating as realizing that frightened media executives who can't yet figure out how to monetize the world of digital media are arbitrarily limiting your ability to enjoy the content you ALREADY OWN.

My realization has been formulating over the last month or so as I work to put together my new home theater system.

There's some great technology out there today - the iPod, TiVo, the latest versions of Windows Media Center, Netflix, etc. Wouldn't it be nice if they all worked together? They don't and there's no immediate solution on the horizon.

In some sense, Napster ruined it for everyone. Not that it's all Napster's fault. If it wasn't them it would have been someone else. MP3 was fun and widely ignored until Napster's P2P network scared the crap out of the music industry - so much so that the labels started suing their own customers. Here we are SIX YEARS later and outside of the iTunes Music Store, there is no truly successful online music model.

Bought a song at the iTunes Music Store? Want to play it over your Windows Media Center in your living room? Sorry. Apple's FairPlay DRM doesn't work with Microsoft's Windows Media Player. Yes, you can burn protected AAC to CD, then rip back to MP3. And yes, tools like the Hymn Project make it possible but it's still way too difficult - needlessly so, IMO.

It sure would be nice to purchase a song from the iTMS in either the format of your choice, or better yet, if FairPlay were supported on non-Apple devices like my TiVo or Windows Media Center.

I've taken to using AAC on my iPod (both protected songs from the iTMS and unprotected songs converted from my existing CDs) since the quality is very good. I've also recently begun the task of converting all my CDs to FLAC. I like that FLAC is open source and lossless. I've set aside 100GB for archiving my CDs (I'm getting about 2:1 compression) and I'm going to put them in a box in a closet somewhere. There are lots of good audiophile-grade FLAC tools to be found and it's pretty easy to transcode FLAC to AAC or MP3 for listening on my iPod (which doesn't support FLAC - surprise, surprise).

Netflix has a great selection. Much better than my (or yours!) local Blockbuster. I just wish it was integratred into Media Center. Yeah, there are plugins like the one here but I want an even higher level of integration. Imagine this scenario:

As a Netflix customer, I've signed up for the 3-at-a-time $17.99 package. Instead of waiting for CDs in the mail, why not download my queue in the background? This should run as a background service and be integrated in Windows Media Player interface. When my connection is idle, use my bandwidth to be sending the next 4.7GB movie. I reguarly keep 40 or so movies in my Netflix queue. Allow me to download them all or allow me to specify the amount of storage to set aside. Use DRM to only allow me to unlock my 3 movies at a time. And unlike MovieLink, allow me to watch these same 3 movies as many times in a row. When I'm ready for my next movie - force me to delete one of my unlocked movies. Instant gratification and no DVDs to mail!

Proprietary camera RAW has got to go. Every camera maker should be using Adobe DNG as a standard - I get NO value from another proprietary RAW format. I'm paying for an image sensor, lens/flash integration and a camera system, not crappy software from a camera maker.

Due to the nature of HD, being about 10GB/hour, this one has been a little less painful. Intel's HDCP standard is at least that - a standard - that everyone's gotten behind so we're on the right track there. I feel for the folks that bought into HD at very high prices before DVI-D or HDMI was widely adopted. I've also heard the next version of the Windows Media Center extenders will support HDCP.

HDCP is supposed to protect the content providers to keep HD from getting out in the wild on P2P networks. Since HD is still (relatively, from an adoption standpoint) in it's infancy, this is a workable solution for the most part. Also, since HDCP is built-into HDMI you'll see a big push for HDMI in the consumer electronics world.

Interestingly, the technology world is not on board with HDMI nor HDCP. Perhaps since the PC OEMs know best what standards mean, they're not going to be paying the HDMI consortium (Phillps, et al) nor HDCP (Intel, Silicon Image) but are instead proposing a different standard for PCs. Sigh. Be on the lookout for DisplayPort - an open interface standard based at the physical layer on PCI Express but with a different, new DRM technology. History and politics say this is not going to be a good thing from an interoperability perspective.

There are no easy solutions in the near term. Apple's FairPlay is no better than Microsoft's Janus - both are proprietary, closed formats. HDCP is at least a little more forward looking in terms of ability to move content around a local network. There are a few open source DRM projects out there but none have any serious momentum, especially comapred to FairPlay or Janus.

Any DRM can and will be cracked. FairPlay has been cracked. DVD CSS has been cracked. Janus will be. The industry knows this and is aiming for a pain threshold where most people just won't bother.

I just wish the 'industry' would focus more on interoperability and standars rather than re-inventing the next wave of DRM that is 10 minutes from being cracked.