Thousands and Thousands of Hours of PVR TV

Thomas Hawk writes "Cory Doctorow is posting over at Boing Boing about some technology that he apparently saw this weekend at London's Open Tech conference. According to Cory, this new technology from Promise TV takes the form of a home-built PVR with lots of high-capacity hard drives and claims to be able to record every show on every channel being recorded in the UK for an entire month. 'Why program a TiVo to get certain shows for you when you can record every single show on the air, all at once, and then use recommendations, search, a grid, or any other means you care to name to figure out which of those thousands and thousands and thousands of hours of programming you want to watch.' The company seems somewhat cryptic with a simple website that appears to be collecting your email addresses for an announcement in August. "

Comments

A month later

[jcayer]

and there is still nothing to watch on TV!

Timing

[nmg196]

If they're going to be making an announcement in August, then why not wait until August to post the article?! There is no product and no information. It doesn't even say whether it records only the UK terrestrial TV channels (just 5) or the UK digitial ("Freeview") channels (MUCH more than 5).

I can understand how you could feasibly mock up a machine that recorded the 5 main channels to a RAID array or something, but I fail to belive that you can actually record "the entire UK channel multiplex" of ~30 digital channels in anything of a sensible size or price. It would have to save out 30 high quality(ish) feeds to very very large hard drives permanently. I can't see how you could do that with less than a few thousand pounds of disks and capture cards.

Re:Timing

[_Shorty-dammit]

well, with my PVR machine I record at 9Mbps for video and 384Kbps for audio, barely over 1MB/s. With two tuners, that's just over 2MB/s. Watching one of the previous recordings while recording two shows at the same time, that's just over 3MB/s. Even a mediocre HD can handle that no problem. Hell, while it's doing that it's also either scanning a show for commercial breaks or recompressing that 4GB/hour mpeg2 stream to a 1GB/hour mpeg4 stream, so there's a bit more workload, still doesn't break a sweat. So, one HD per recording is way overkill.

Re:Timing

[Goth Biker Babe]

Insightful? Hah!

Capture cards? 30 feeds? Don't be so analogue and old school. It only needs to save the multiplexes. Which on terrestrial digital is about eight including all the radio stations.

Right assuming it's digital only, it needs as many 'frontends' as there are multiplexes. Modern day silicon (non can/discrete component) tuners are pretty cheap and rather small. You'll also need the demodulators to go with them. All of which would fit easily on a single PCI card. Then you just process each of the multiplexes' transport stream enough to remove the redundant data such as the NITs and record the rest on to the harddrives as a stream.

Something like a Sky box already does this with two transport streams. One is recorded for the 'trick mode' pause live TV etc and one for recording a program. It will also play back a third stream from the disc. A more powerful PC based machine could easily cope.

Re:Timing

[TobascoKid]

If they're going to be making an announcement in August, then why not wait until August to post the article?

Because the system was demoed at OpenTech 2005 on Saturday.

I was there and I saw it. So here's a bit more info on how it works. I records digital terrestrial televison, not analogue. I suppose it could be changed to use satelite DVB instead of terrestrial DVB - but you can't get a DVB-S card that decode Sky's encryption, so there's not much point. It records an entire mutiplex off the DVB-T card. They only appear to have one card, so they were only recording the BBC multiplex. There are 6 multiplexes in the UK, so I suppose to record "all" DVB-T transmissions, you'd need multiple cards.

As for costs, while the DVB card was quite cheap (they said around 50 quid) and the PC is faily inexpensive, the storage costs are about the same as a plasma tv - but falling all the time.

Channel Hopping

[SimilarityEngine]

"...use recommendations, search, a grid, or any other means you care to name to figure out which of those thousands and thousands and thousands of hours of programming you want to watch..."

Those poor channel-hoppers, who can't watch a programme for more than 10 minutes without wondering what else might be on, will now have all the material from the past to choose from aswell. Lucky them!

Perfect /. article

[Overzeetop]

The summary is nearly as long as the actual article, and contains practically all the information. It can't get any better for /. readers - even those that don't RTFA have all the information available.

That said, this is about as useful as,well, nothing. A spam collector ad? At least the previous /. ads were for products. Wake me when there's news. And when DirecTV supports this.

Why stop there?

[rustbear]

Their next product: a home-built device that downloads the entire Internet for you to browse at your leisure...

Re:5 channels

[nmg196]

30 actually:

http://freeview.co.uk/whatson/index.html

I doubt you'd bother making something that recorded from an analogue source - too much CPU power.

There are hundreds of UK TV channels

[rklrkl]

This is utter bunkum because there are hundreds and hundreds of UK channels - 5 analogue terrestrial, about another 25 on digital terrestrial and about another 300 (!!) on digital satellite. Yes, with 5 analogue or digital tuners, they could record BBC 1, BBC 2, ITV1, Channel 4 and Channel 5, but let's face it, most of that's now rubbish and the better stuff is on digital satellite (which they will *not* be able to record massively in parallel - Sky who run it currently only have a twin tuner for example and that needs a dual LNB on your dish too !).

I'd rather see some effort made to allow broadband users to download TV shows (even a small fee for this would be acceptable - a few pounds a month) from the time they are aired on normal TV for, say, up to 2 months afterwards. Now this would be *far* more useful, especially now that 2Mbit/s is starting to become the normal for UK broadband.

Is this really a feasible home appliance?

[ErpLand]

Let's make some calculations assuming that they're going to record all the DVB-T ("Freeview") content in the UK. I watch DVB-T in Spain using a MythTV box but the numbers should be roughly the same as for the UK.

45 mins recording of one channel = 1401390703 bytes
=> 1 hour = 1868520937 bytes
=> x 24 hours/day x 30.5 days/month = 1.37 TB per month per channel

Now there are about 30 freeview channels so we would need 41 TB of storage .... that's 82 500GB hard disks in RAID0! Which would occupy something like half a rack and use about 1kW of power ...

Even to record the 5 main channels would be nearly 7 TB - still a lot of noisy spinning hard disks to stick under the TV. This doesn't sound like a feasible idea with the size of today's hard disks.

Simple Math

[WarwickRyan]

As this sounds like pure marketing, we can make some assumptions:

a) Number of channels included will be the minimum available to all.
b) It'll be "VHS quality" recording.

There are 5 terrestial TV channels in UK:
BBC1
BBC2
ITV (commercial)
Channel 4 (commercial)
Channel 5 (commercial)

We've about 50 via digital TV, and loads more via cable or satellite.

However there are only 5 available right now.

So, that's 5 channels * 24hrs * 28 days = 3360 hours of recording.

Lets assume a VCD bitrate of 1300kbit/s video 128kbit/s. Total 1428kbit/s.

Number of seconds in 3360 hours
= (3360*60)*60
= 12,096,000

So, for all that video we'll need
= 1428 * 12,096,000
= 17,273,088,000 kbit
= 17,687,642,112,000 bits
= 2,210,955,264,000 bytes
= 2,159,136,000 kilobyte
= 2,108,531 megabytes
= 2,059 gigabytes

So that's like 4 * 500gb drives plus 1 * 120gb drive to correct for the drive maker's marketing departments.

I'm using VCD/MPEG as a basis for this, they'll invariably be using a better codec, probably with far stronger compression.

Why not? Here's why...

[Junior J. Junior III]

'Why program a TiVo to get certain shows for you when you can record every single show on the air, all at once, and then use recommendations, search, a grid, or any other means you care to name to figure out which of those thousands and thousands and thousands of hours of programming you want to watch.'

Because the amount of overhead involved is ludicrous?

Downloading every show broadcast in a month would be like downloading the entire internet and then running searches on your local server for the information that interests you.

Imagine duplicating this in EVERY household in the country. The impact to our energy grid would be sickening. We should be looking to lessen the amount of power we are sucking down, not increase it.

Moreover, there's no need -- TV listings are announced, you know what's going to be on, you can narrow down significantly what you know is highly unlikely to be of any interest to you. You don't want to capture something and then have to sift through it all. Finding that one good show or moment in a month of crap content will be like finding a needle in a haystack, unless you can find a way to dope the captured video stream with some metadata that you can use to aid your search.

There might be the occasional oddball thing that no one predicted would happen on TV that you might miss, but (and this is the true beauty of the internet) if that happens, there's sure to be SOMEONE who captured it, and it will be hosted on the internet somewhere (copyright laws be damned). It's just a matter of finding it. Google can make that reasonably easy. Friends and family forwarding links that they found interesting to your email can take up any slack.

Re:Seriously Doubt

[jcsehak]

I mean, just look at a standard Tivo box. 40G hard drive gives you about 35 hours of recording time.

That's only if you record at crappy quality. If you record at "good" (not "best"), you get around 15. Which goes real fast, let me tell you. What's worse is that there's no way to find out how much space you've used up or is available.
[/gripe]

How it might work, and some calculations

[threeturn]

So let's think about how this might work. Looking at BoingBoing it looks like it's based on the UK's DVB-T system. Simplest way to implement what's described would be to just decode each multiplex in a particular area and pump all the data on to disk with some time markers.

According to http://erg.abdn.ac.uk/research/future-net/digital- video/dvb-trans.html each DVB multiplex runs at 24Mb/s.

So, storing one multiplex for a month needs
(24/8)*60*60*24*31 Mbytes of storage = 8 Terra Bytes

So 8TB per multiplex per month just about doable at the state of the art, but not very likely.

I haven't checked how many muxes in use for different channels. I think it's about 3, so say 24TB all in. That's a lot of disks!

Re:You just need one hardcore nerd per block...

[Cervantes]

Before you trot out all your legal objections, just let me say that you now have a legitimate reason to talk with the cute girl three doors over you've never met.

Dude, if I have my own T1 line and a cute girl 3 doors down, I'm not going to talk to her, I'm going to use my T1 to stream hidden webcams from her house and charge $29.95/month for membership to the site.

Talking to a cute girl := 1% chance of something that could be called success.
Selling pics of a cute girl to pervs and collecting $$$ := 100% chance of buying a russian bride.

Re:You just need one hardcore nerd per block...

[Anonymous Coward]

Only on /. would this get modded insightful.