> So how come BluRay disks are 5x the storage capacity, yet still have
> roughly the same run time? I thought it was because HD video requires
> more space to store.

On DVDs you'd be limited to under 2 hours if you used the maximum 
bitrate (around 10 MBit/s IIRC), so you are forced to use lower bitrates 
if you want to include menus and other various extras on the disc.  With 
BluRay, even if you used an average of 30 MBit/s (which is extremely 
high quality with h264, and likely never actually needed, it usually is 
around half that) you have about 4 hours run time.  This is why you can 
fit two versions of an entire film (eg 2D and 3D version) on a single 
disc, plus all the extras.  That would be impossible on DVD.

> I thought it was more or less the case that *all* codecs work by
> transforming the input, deciding how "important" each signal component
> is, and then keeping only the most important bits, according to what the
> requested bitrate was. I don't see anything there that makes a higher or
> lower bitrate change the amount of compute power required.

The whole point of video compression is trying to find patterns 
frame-to-frame to reduce the information needed to reconstruct the 
correct video (otherwise you'd just have a series of JPEG images).  The 
longer and more detailed you search for such patterns, the lower bitrate 
you will be able to achieve for a given level of quality, or 
equivalently higher quality for a given bitrate.

Post a reply to this message

Invisible wrote:
> Ah, but what's this I see? A "privacy filter"? Yes, it's a small piece 

> of plastic which is transparent only when viewed from certain 
> directions. In other words, you stick it to your screen to reduce the 
> viewing angle. WTF, people!? :-D

What part of "privacy" did you misunderstand?

> First of all, it's supposedly a video camera, and yet it costs less tha
n 
> £2,000. Actually, it costs less than £500. To be blunt, it co
sts £75. 
> This is highly suspicious.

Let me blow your mind:

http://www.dealextreme.com/p/usb-rechargeable-1-3mp-pin-hole-spy-av-camer
a-disguised-as-sunglasses-tf-slot-42736

> Quite why anybody would pay for an expensive 2MP photosensor and then 
> stick it behind a crappy 7mm lens I have no idea. 

I've never understood that, unless a 2MP photosensor is actually cheaper 
in 
bulk than a lower quality one.

-- 
Darren New, San Diego CA, USA (PST)
  "How did he die?"   "He got shot in the hand."
     "That was fatal?"
          "He was holding a live grenade at the time."

Post a reply to this message

Invisible wrote:
> OK, so you can change the frequency weightings and how often a keyframe 
> is put in, and so forth. I'm not seeing much that fundamentally changes 
> the whole algorithm such that it might require less computer power.

Making every frame a keyframe reduces the computing and memory you need. Not 
using any B frames significantly reduces the amount of memory you need. Not 
doing motion prediction reduces how much compute power you need. (At least 
for MPEG. I don't know H.264, but I'd assume it's similar.)

> (Although, as I say, I imagine the really hungry bits probably 
> have custom hardware acceleration in a device like this.)

Exactly.

-- 
Darren New, San Diego CA, USA (PST)
  "How did he die?"   "He got shot in the hand."
     "That was fatal?"
          "He was holding a live grenade at the time."

Post a reply to this message

Darren New <dne### [at] sanrrcom> wrote:
> > Quite why anybody would pay for an expensive 2MP photosensor and then 
> > stick it behind a crappy 7mm lens I have no idea. 

> I've never understood that, unless a 2MP photosensor is actually cheaper in 
> bulk than a lower quality one.

  Well, as technology advances it often quite ironically happens that
the older some technology gets, the more expensive it becomes to purchase.
The reason is simple: Lack of demand reduces supply, and low supply means
that the product becomes expensive to produce. Production lines get
upgraded for the newer technology (if not even dismantled outright), so
the older technology becomes more difficult to produce.

  Thus the price of most technology experiences a strong U-shaped behavior
as time passes: When the technology is brand new, it's extremely expensive
(both because production costs are still high due to the lack of cheap
production lines, and to capitalize on the eagerness of early adopters).
Then it quickly gets cheaper and cheaper, until it reaches the bottom of
the price valley (and depending on the product it can become almost
ridiculously cheap compared to its original price). Then it stays there
for some time, and when the technology gets more and more obsolete, and
demand decreases, it starts becoming once again more and more expensive
to produce. Depending on the product, it may at some point become impossible
to purchase anymore (because nobody is making it), or extremely expensive
(even more expensive than when it was brand new).

  This is the reason you eg. don't find half-gigabyte USB memory sticks
anymore, even though one could think that they would be very cheap to
produce. In fact, they might not be. They might actually be pretty
expensive to produce, which is why nobody does anymore.

  It may also explain why it's cheaper to produce a camera with a 2MP
sensor than one with eg. 1MP.

-- 
                                                          - Warp

Post a reply to this message

Warp wrote:
>   Well, as technology advances it often quite ironically happens that
> the older some technology gets, the more expensive it becomes to purchase.

That was what I had intended to imply, yes. :-)

> (both because production costs are still high due to the lack of cheap
> production lines, and to capitalize on the eagerness of early adopters).

It's more that the technology requires a shake-down period. Yields are 
tremendously low until every one of the hundreds of precise steps gets 
adjusted just right.  A 10% yield on early runs is a good number.

You're also paying off the design. I have a friend who designed RISC chips 
for one of the big CPU companies before ARM bought them out, and he told me 
about half the cost of a chip over its lifetime is design costs. That is, if 
you sell a total of $100million worth of CPUs over its lifetime, it costs 
about $50million to design them and $50million to manufacture and distribute 
them.

+=+=+=+

Oh, and if anyone cares, here's what the video from those glasses looks like:

http://darren.s3.amazonaws.com/PICT0001.AVI
http://darren.s3.amazonaws.com/PICT0003.AVI


-- 
Darren New, San Diego CA, USA (PST)
  "How did he die?"   "He got shot in the hand."
     "That was fatal?"
          "He was holding a live grenade at the time."

Post a reply to this message

On 25/01/2011 07:26 PM, Warp wrote:
> Darren New<dne### [at] sanrrcom>  wrote:
>>> Quite why anybody would pay for an expensive 2MP photosensor and then
>>> stick it behind a crappy 7mm lens I have no idea.
>
>> I've never understood that, unless a 2MP photosensor is actually cheaper in
>> bulk than a lower quality one.
>
>    Well, as technology advances it often quite ironically happens that
> the older some technology gets, the more expensive it becomes to purchase.
> The reason is simple: Lack of demand reduces supply.

I don't recall the details now, but the other day I discovered that 
something like 128MB of PC100 RAM is more expensive than 1GB of PC3200. 
Or similar. Which is quite bizarre, really...

Alternatively, if you're actually buying PC100, it means you have 
something really critical that you need to keep working. (Rather than, 
say, just replace the damned PC.) And if it's that critical, they can 
charge you "sucker fees" (i.e., whatever they charge, you just have to 
pay it, because you need it that badly).

In general though, I was under the impression that most technologies 
don't dissappear, they just move down the food chain. For example, the 
Z80 is still in production. It's extremely popular, apparently. And 
rather cheap too. (I have no idea what the original sale price was, but 
currently it's a few quid.)

-- 
http://blog.orphi.me.uk/
http://www.zazzle.com/MathematicalOrchid*

Post a reply to this message

Orchid XP v8 <voi### [at] devnull> wrote:
> In general though, I was under the impression that most technologies 
> don't dissappear, they just move down the food chain. For example, the 
> Z80 is still in production. It's extremely popular, apparently. And 
> rather cheap too. (I have no idea what the original sale price was, but 
> currently it's a few quid.)

  Intel stopped producing the 80386 a few years ago, so it does happen
with some products.

-- 
                                                          - Warp

Post a reply to this message

On 25/01/2011 09:22 PM, Warp wrote:
> Orchid XP v8<voi### [at] devnull>  wrote:
>> In general though, I was under the impression that most technologies
>> don't dissappear, they just move down the food chain. For example, the
>> Z80 is still in production. It's extremely popular, apparently. And
>> rather cheap too. (I have no idea what the original sale price was, but
>> currently it's a few quid.)
>
>    Intel stopped producing the 80386 a few years ago, so it does happen
> with some products.

I've seen printers and so forth that use old x86 CPUs inside. I wonder 
if Intel stopped manufacturing the 386, or just licenced the design to 
somebody else?

-- 
http://blog.orphi.me.uk/
http://www.zazzle.com/MathematicalOrchid*

Post a reply to this message

Orchid XP v8 wrote:
> It's extremely popular, apparently.

The Z80 is the CPU of choice for things that need a low-power CPU. Like, 
digital voice recorders, toasters, remote controls, etc.

-- 
Darren New, San Diego CA, USA (PST)
  "How did he die?"   "He got shot in the hand."
     "That was fatal?"
          "He was holding a live grenade at the time."

Post a reply to this message

On 25/01/2011 06:28 PM, Darren New wrote:
> Invisible wrote:
>> OK, so you can change the frequency weightings and how often a
>> keyframe is put in, and so forth. I'm not seeing much that
>> fundamentally changes the whole algorithm such that it might require
>> less computer power.
>
> Making every frame a keyframe reduces the computing and memory you need.

Presumably the bitrate is going to skyrocket if you do that though.

> Not doing motion prediction reduces how much compute power you need.

Does the standard actually allow that?

(Now that I think about it, the decoder algorithm is usually far more 
tightly specified than the encoder...)

Post a reply to this message