There is a debate raging in HDTV over just what constitutes future proof HD signal capacity. One semiconductor firm we spoke to this week, Valens Semiconductor from Israel, reckons that upwards of 10 Gbps is a bare minimum and that 20 Gbps is just around the corner. It also happens to have both on its roadmap, 10Gbps now and 20 in its next generation chipset.
Which means that the ‘pipe’ wars we have been having, MoCa versus HPNA are only just the beginning. Other standards and technologies the are trying to compete for multiple HD streams include the HomeGrid Forum, which is promoting the ITU’s G.hn standard and as recently as February a new group called DiiVA (Device Interactive Interface for Video and Audio) touting support from LG, Panasonic, Samsung, Sharp, Agilent and MediaTek, said it expected to be ready by 2010.
The DiiVA system runs over CAT6 LAN cable and supports uncompressed video and audio, plus bi-directional high-speed data. It will aims for 13.5 Gbps for video and 2Gbps for bi-directional data. DiiVA made a point at announcement that it will simultaneously support HDCP 2.0 and DTCP-IP content protection systems.
All of this has come about because over the past 4 years TV technology has continued to fight a battle based visual quality. Back in 2006 HD really meant that a TV would show ‘most’ of the pixels of a 720p signal, now in 2009 it tends to mean it will show 1080i ‘probably’ while a handful, at the top end of the specifications lists show 1080p in all of its beauty. But still this is not enough.
In Asia we are seeing 2K and 4K permutations, double and four times the number of pixels along each side of the TV screen, with the total number of pixels needing to be transmitted, going up by an order of magnitude. Four times as many pixels along one side means 16 times as many total pixels in total, but then again, the data requirements of sending a pixel are also changing.
It’s pointless sending all that in 30 frames a second, so TV makers have doubled it to 60 and one or two already have 120 Hz devices in their sights. We have been told that 24 frames a second is all that the human eye can distinguish. How then is it that when Faultline is wandering around the halls of IBC and NAB, that all the really neat screens, with a phenomenal picture turn out to be a 60 Hz 1080p with 10 bit encoding?
Most broadcasters and pay TV services use 4:2:0 sampling, which means that the colours (chrominance) and brightness (luminance) only have those many bits to describe their exact measurement. In effect they are throwing away data about the original signal when it is placed into the Codec in order to get a best fit. Contribution systems use anything up to 10 bit 4:2:2, and are the best format for sending video around the world for later editing.
But already the big Consumer Electronics manufacturers are looking to use 36 bits to more accurately describe each pixel and some have roadmaps to go to 48 bits, which is called Full HD Deep Colour mode. Where will it end?
Well for a second let’s assume that it will end there or at least slow down there. In fact there is an argument to say that we are way beyond what the normal household will chase after, except for the number of people which upon seeing first level HD say how disappointed they are. It could be that all this effort is wasted and that although the TV industry will seek to use video quality as a differentiator, this will be in vain at the consumer level. But we don’t think so, and at least 1080p 60 HZ deep colour will become common before this war is over. Anything that is an edge, in what has become a 200 million device per annum market, is worth trying, especially to dominate the highly profitable top end.
To support 1080p 60 Hz deep colour, uncompressed video sent from a playing device to a TV screen will require 16 times the pixel density of today’s signals, 4 times the refresh rate and be described in 6 times the detail. That increases the bandwidth requirement for connecting these devices to 16 x 4 x 6, an increase of 384 times.
What this generation of devices will bring us is high definition, no blur slow motion replays. It will bring us photorealism and 3D, and perhaps the ability to zoom. We see it on TV spooks programs all the time, like 24, where a satellite image is zoomed and a clever computer works out how to de-pixellate the image, and suddenly all the spies can recognize the bad guy. That is science fiction, this is not. Imagine being able to do all that on a video of your favourite star. We’re not sure if this will kill the porn industry or take it to another level.
At present we can only dream of this in access networks, which are, for the most part limited to anything from 2 Mbps at the low end up to to 20 to 50 Mbps over the internet and 30 Mbps over a broadcast signal. But Blu-ray players and TV sets are not limited by delivery networks, since the customer ‘carries’ the video into the home on a disk and it is those which are driving this technology battle.
The Blu-ray is up to it, the TV is up to it. The only thing stopping consumers from playing at these incredibly specs are the cables which connects the two devices. It was always a secret that to play a DVD through an HDMI cable, improved the picture fantastically, virtually HD.
Currently cable tops out by HDMI 1.3 which operates at 10.2 Gbps. But it only reaches distances up to 3 or 4 meters (maybe 5 meters) although many systems have amplifiers, equalizers or repeaters which can increase this slightly to perhaps 50 meters, otherwise the signal gets unstable and the screen starts blinking. Of course these add to the cost, but in a world where video servers and set tops are being increasingly divorced from the playing devices they are driving (take for instance multi-room DVRs) this type of long distance, extremely high speed connection will be required.
We spoke to Micha Risling, VP Marketing at Valens Semiconductor who tried to put this all into perspective for us.
‘Right now, like so many other ideas in advanced CE technology, the HDMI route is expensive and only works in intra-room distances ‘ so 3 to 4 meters. Over 10 meters it can support 720p,’ says Risling. He also maintains that sending uncompressed video increasingly already requires more like 6 to 7 Gbps.
There is a philosophical argument here about whether anyone in their right mind wants to send uncompressed video around a home, or even across a room. But Risling will point you time and time again to the cabling requirements of big flat TV screens. Not only do they want to appear to work like magic ‘ look no wires ‘ which means either using a wireless technology or cutting a wire into a wall, but consumers want to put the set top or Blu-ray player in just one place, possibly out of site, in the basement, certainly across the room, for ease of control.
But when you have paid for a device to have the processing power and complexity to unravel the latest H.264 codec, the last thing you want to do is put the TV signal ‘back’ into a compressed format to send it across the house. And with the advent of multi-room DVRs, you would have to pay for that processing logic once for each TV you have taking a HD signal.
Not only that, but if you decide to do this in compressed video then every time you install one of these new super duper TV screens you have to cut a hole for your power cord, a hole for your HDMI cable and another to bring the internet to the back of the TV. Some people get sensitive about paying for that much wall work. So how to you get around this problem?
The Valens idea is that you take 10 Gb Ethernet components (that’s gigabits) and mess with them. Which is why it calls its product line HDbaseT (10baseT is an Ethernet term, HD is a video term ‘ geddit?)
We’re not sure when Valens started on this, but the 20 man Israeli company has had time to file 30 patents on just ‘how’ to mess with Ethernet techniques.
The company achieves this be making a complete change at the physical layer, using Pulse Amplitude Modulation or PAM using PAM 8 and PAM 16, and sends it over standard Cat5E cable. It has also managed to create its own packet size to accommodate different types of traffic, both HDMI and IP packets.
‘On every TV we saw at CES which had HDMI and internet delivered to the TV, they had two separate cables, one for the TV signal and one for the internet,’ said Risling
By creating a new physical level, the pipe simply carries a lot more, so it can support HDMI plus some, with room to spare for other services. The two extra services above plain video, which Valens has in mind are an internet back channel, for things like game play, playing internet TV and browsing the internet on the TV screen, and power. That’s right electrical power.
‘So far we have managed to keep our signal steady over 100 meters, while running a 50 watt signal over the cable,’ said Risling. You can see that the 100 meter claim solves the distance problem for remote devices operating across home networks, such as multi-room DVR. The power side of things is about ONLY cutting a grove for one cable, and using the power to power up the TV set.
‘It’s enough to power up a Blu-ray player,’ says Risling, ‘but not a top of the range TV set with a backlit LED display. A 30 inch screen which doesn’t use a backlit display will use about 30 watts. But we see the most likely outcome of putting power directly to the TV and then using the same cable to power the player or set top, in the same way we see a projector getting a mains source of power and then over this cable, using that to power a laptop, not the other way around.’ All of this is without the availability of OLED screens, which will drive down power considerably and put the power requirements within Valens reach.
But the Valens vision goes a lot further and Risling sees TV signals coming into the home and being uncompressed at the home gateway and its cabling starting right there, and taking it around the entire home.
‘When we first showed this to CE manufacturers, they just simply did not believe it was possible,’ he said. ‘Now we have regular meetings with them and every two weeks we have a new spec out of Japan or Asia asking for 48 bit support at 120 MHz.
‘We managed to get our HDCP certification in just three months,’ boasted Risling, which he explains is not because it’s tough to achieve, but due to political pressure from Hollywood, which has seen this technology and wants it in the field as soon as possible.
HDCP is a straightforward Intel licensed set of protocols for copy protection agreed by the studios to be the best available protection for HD content. It relies on a frequent (every 2 seconds) authentication and encryption based around secret keys. It uses a timing signal akin to a ping in internet speak, to measure how far servers are from TV sets.
‘If you take HDCP onto fiber, the distance that you can put the TV away from the server is over 1 kilometer,’ points out Risling. ‘But you can’t make people put their server next to their TV set any more, because many people who have such top of the range devices, want to bury all the electronics in the basement and have a simple flat screen with no apparent wires attached. So you do need that 100 meters.’
‘We will support a signal over 100 meters on cat5e cable, and we hope to get the price of the chips for that down to $5 to $10 in volume. Right now we are offering engineering samples of the termination chips in September, which will be around $20 a chip in 1,000s.’
‘Our next step is to start our own alliance,’ says Risling, ‘which we hope to get done in a month. We already have a number of content and manufacturers ready to join and it should be called the HDbaseT Alliance.’
This specification and approach puts it directly up against the recently announced DiiVA group and no doubt some of its supporters will be those same companies, LG, Panasonic, Samsung and Sharp ‘ but Valens should be more interesting to someone like Sony and Philips,, because it is a smaller technology company, one that Sony would have the potential to work alongside and potentially buy or control, rather than pay royalties to.
The spec of the two systems are very similar with 13.5Gbps of the DiiVA System dedicated to video and 2Gbps for bi-directional data, but the big differences are the power possibilities, the 100 meter reach and the arrival date of the chips, and it looks to us like Valens has an entire 12 month lead on DiiVA.
Whatever the outcome, those technologies which are shipping now, taking three or four compressed HD streams across perhaps a 200 Mbps capable cable, will have a finite period in which to get back their investment. After that, as these technologies go mainstream they will perhaps need to either lend their own endorsement to one of these systems for next generation transition or come up with their own technologies. Either way the cable wars are far from over.
Meanwhile the wireless technologies of WirelessHD, Amimon and WiGig (see elsewhere in this issue) will be fine for the current generation of HD, but clearly will run out of steam up at around 6 Gbps, if indeed they ever reach that far, meaning cables of some description will be with us for quite a while longer.