I previously wrote about how the National Broadband Network (NBN) seems more suited to porn delivery than regular Internet use [1]. It doesn’t seem to be of much use really. In a particularly insightful comment John Hughes suggested that the real purpose would be TV delivery.
The ABC is currently delivering 640*360 resolution MPEG4 files via iView for it’s TV content. To use iView on Linux you need Jeremy Visser’s Python-iView program [2] in conjunction with Luke Kenneth Casson Leighton’s rtmp program [3]. Note that to get this working on Debian/Lenny you need to install the python-simplejson package as well. On any version of Debian you need to install the python-beautifulsoup package.
For the current ABC service it requires 251652 of data for a 3167.6 second Torchwood episode, that averages to 79.4KB/s of data. In contrast a video of Dan Gilbert from TED.com at high quality was in 640*480 resolution and required 301652KB of data for 1276.5 seconds which averaged out to 236.3KB/s. My ADSL2+ connection is theoretically capable of something over 1MB/s and occasionally gets such speeds for unusual download situations (such as downloading multiple large files at the same time). But generally I can’t rely can’t rely on sustained transfer rates of more than about 200KB/s. So I could watch streaming TED talks at reasonable quality, but for the best results I have to download them and watch them from disk. Assuming the same ratio of compressed data to raw pixels used for HDTV as used for TED talks a 1920*1080 HDTV resolution MPEG4 with the same quality as a TED talk would take 1600KB/s. It is possible to vary the compression level and possibly the usage of a TV stream would permit better compression than the usage of a TED talk for the number of pixels. But it seems reasonable to assume that something like 1600KB/s is needed for best HDTV, that is more than the vast majority of ADSL2+ installations can be relied on to sustain. But 100Mb/s would allow at least two 1920*1080 HDTV transmissions to be viewed at the same time, and maybe three or four – as few homes have only one TV this should be of great interest.
Now the question is how much people would pay for this. Currently there are two main pay TV companies in Australia, Foxtel which charges $916 for a 12 month plan [4] and Optus who’s web site is so awful that I gave up before discovering what they cost (I will assume that they are competitive with Foxtel). Now I think it’s reasonable to assume that $916 is at the high end of what potential customers are prepared to pay, as they are competing with free to air TV, DVD sales, Youtube, and downloads of pirate content. The current bank interest rate on term deposits from the Commonwealth bank is 6.3%, the interest rate on raising the finance would have to be greater than that – let’s assume 7%. So the $5000 per household will require an interest payment of $350 per annum. If a household signs up to pay TV services at a cost of $1200 per annum that might be enough money to pay slightly more than $350 to the NBN plus make a profit for the pay TV company. So if every household in Australia signed up for pay TV over the NBN it should be profitable. But that seems unlikely.
The majority of the Australian population (IE the majority of the city population) is used to paying not much more than $240 per annum for a basic phone service and not much more than $360 per annum for a basic broadband Internet service. A bundled deal of $1600 per annum for phone, Internet and pay TV should allow the pay TV company and the NBN to be barely profitable if everyone accepts the deal (unless of course interest rates rise). If half the population aren’t interested then the bundled deal would have to cost $2000 to have the potential for being profitable. If three quarters of the population aren’t interested then it would have to cost $2300 or more!
The Australian Bureau of Statistics documents that in 2008 the median household disposable income in capital cities was $593 per week [5]. So it seems that the idea of the NBN being profitable is based on plan to have more than three weeks of household disposable income per year being spent on Internet/phone/TV services.
But fortunately for the pay TV companies and the content companies there is no requirement for the NBN to be profitable, it’s being paid by tax money so if it loses money then we can just pay more tax! It’s the ideal “public private partnership”, the citizens take all the risk and the corporations reap all the profits!
James Purser describes how Stephen Conroy gave the TV networks something between $250,000,000 and $500,000,000 a year for the next two years [6]. It was claimed that this huge gift would be conditional on the production of more Australian content but they ended up not putting any conditions on it. In a strange coincidence Stephen Conroy did this a month after having a meeting with Kerry Stokes (head of Channel 7). Based on this I don’t believe that there was ever a serious plan to make the NBN profitable. I think that the plan was just to take our tax money and spend it on things that benefit friendly corporations. Really it’s better for us if the government just hands out $500,000,000 at a time instead of spending $43,000,000,000 on a project that has the same aim of giving a few hundred million to cronies.
- [1] http://etbe.coker.com.au/2010/02/16/national-broadband-porn-network/
- [2] http://jeremy.visser.name/2009/08/30/python-iview/
- [3] http://lkcl.net/rtmp/
- [4] http://www.foxtel.com.au/
- [5] http://www.abs.gov.au/Ausstats/ABS%40.nsf/0/C5B82D434B4BB5F0CA2573D20010E735?opendocument
- [6] http://jamespurser.com.au/blog/Two_Ministers_and_a_Shadow
Your video bitrate estimates are a little high, you can get better-than-iview quality video streams in 500kbit/sec if you use more modern formats (h264 & aac in particular) which is reasonable considering how far into the future the NBN is. Of course, you can also argue your estimates are a little low, if you think the future involves Theora!
nine: You’re really looking for a fight there! ;)
The best argument for my estimates being low would be based on higher resolutions than current HDTV – Dell sells monitors with higher resolution than 1920*1080 at prices that are much lower than what many people were spending when Plasma TVs were trendy. The next argument would be that someone desires coding that is less lossy than the TED talk I randomly selected as an example, or the possibility that someone might want high quality video of people running around in tie-died shirts – or some other pathological case that performs poorly with the typical encoding methods.
I think that the case for Theora files being overly large is widely overstated. But if you have a good reference to compare the efficiencies of the formats then I would be happy to read it.
While it’s true that monitor resolutions will increase, TV/movie content will still be released in a maximum of a 1080p format for many years. New codecs are constantly in development aiming for lower bitrates for the same quality. H.265 is already scheduled for 2012 – http://en.wikipedia.org/wiki/High-performance_Video_Coding
Let’s take an example of high-quality 720p content released legally online (such as Caprica episodes in the US) these streams/files are 1.3GB for an ‘hour long’ (44min) show. This is around 550KB/sec or around 5mbit/sec and likely up to 8-15mbit/sec for 1080p – http://en.wikipedia.org/wiki/Bit_rate#Video – one could stream this on ADSL2+, but for interruption free playback, download is almost always required.
As for the price point of NBN plans? The $43bn is (hopefully) a ‘worst-case’ figure and the actual network cost should be half to one-quarter the price.
Lukian: Firstly thanks for the references, it’s good to have quality links to background information.
Now you seem to be presuming that compression algorithms can be upgraded more easily than display resolutions. When integrated TV systems are used (presumably a TV with an Ethernet port will be released shortly) then probably upgrading the compression algorithm will be equally difficult as upgrading the screen resolution as it will require replacing the TV (I don’t think that TVs which automatically upgrade themselves will take off). When computers are used upgrading the resolution and algorithm will also be equally difficult – IE not very difficult.
The bit-rates Wikipedia plan describes HD-DVD as 29.4Mbit/s – which clearly needs something significantly faster than current ADSL2+ or cable net access.
Finally you suggest that $43,000,000,000 might be a worst-case figure, can you cite a reference for this? I have the impression that the money will be provided by the government and will then be spent, I expect that if the cost estimate turns out to be incorrect then it will be due to significant budget overruns… Why would a private company want to save the tax-payer money?