Qualcomm wants to be a one stop shop for TV on mobiles for the US. Actually it wants to be the ONLY one stop shop for mobile TV. That’s the extent of the ambition built into its MediaFLO concept and product line and the only conclusion to draw after a long chat with its VP of Engineering, Rob Chandhok.
Chandhok and other Qualcomm executives have been abundantly clear in the press about why MediaFLO should be the clear technology choice, citing weaknesses in competitors, DVB-H in particular, because other systems were built, it claims, to solve the wrong problem.
You might argue that in fact MediaFLO solves the wrong problem, the problem of creating a highly efficient, but non-standard way of broadcasting digital TV to mobiles, when standardization is needed, but you wouldn’t get anyone at Qualcomm to agree with you.
Qualcomm argues that if the engineers in the DVB and DMB camps, and within Samsung and Nokia, had set out to solve the same problem that Qualcomm had set out to solve, then they probably would have gone about it the same way that Qualcomm did. That’s been the reprise Qualcomm has been putting forward constantly since it began its campaign to edge out DVB-H.
‘MediaFLO wasn’t a piece of technology that we had lying around that was looking for a use. The entire project came from a business case put up within Qualcomm two years ago and since then it has been about working out the right way to deploy video services, more profitably and with the right business model and the right system requirements,’ he said.
By extension this means that Chandhok thinks that the European and Korean efforts have been compromised by a number of early misconceptions. There has been a lot of talk about using the same broadcasting towers and the same receiver chips to capture DVB-T (T for terrestrial) and DVB-H (H for handheld).
The problem is, according to Chandhok, that as soon as you set up appropriate spectrum for DVB-H where it will offer a quality signal, right to the edge of its cells, then you find all the design compromises in making it as close to DVB-T, pointless. The two cannot occupy close enough spectrum to use the same towers and head ends as each other, so there was never any point in basing one on the other.
At least that’s his view, and the result is that MediaFLO makes a straightforward claim against DVB-H in terms of performance, and asks
to be judged on it. ‘In one 6 MHz channel we can get 20 Quarter VGA signals at 30 frames a second, plus 80 channels of clip casts, and 10 channels of enhanced AAC audio and other data services.
‘The DVB-H trials are getting just 8 TV channels in a 5 MHz spectrum channel,’ he says, referring specifically to the US trial that is being run by UK tower network Crown Castle and Nokia.
‘Channels can be changed in 1.5 seconds with MediaFLO, and it works in on a similar datacasting model to DVB-H, pushing content rapidly to the phone in short bursts, that sit in RAM, and no longer require the radio to be receiving while the video is playing out. But FLO can switch its radio on and off 4 times a second, DVB-H has to operate in bursts which are 4 to ten seconds apart.’
Clip casting channels are downloads that come down to the phone when the network is not too busy, used to drop pieces of video content that are 10 or 20 minutes long to subscription customers for playback later. The clip cast process is designed to fit in with the type of vertical subscriptions that most commentators think will be popular on mobile TV services, for instance to support a particular type of sports fan, by downloading goals or action or personality interviews.
Clip casting can be carried out overnight (when the TV broadcasting is slack) or during the day time depending on how operators want to configure their services. It drops back into a Personal or Digital Video Recorder (DVR) which maps onto the available storage within a given phone.
Additionally real time feeds, such as weather or stock quotes can be separately delivered to phones over the 10 MediaFLO data channels. These can also be suspended at night to make more room for overnight clip casts.
So is mobile TV going to only go down a single channel? Will its maximum capacity have to be 20 TV channels?
‘We wanted to be able to replace a single analog TV channel with enough TV capacity so that an operator would have a good business with just one channel. Of course, if anyone has more spectrum than that and can run multiple 6MHz channels, then MediaFLO can handle that,’ said Chandhok.
Chandhok continually makes comparison with DVB-H, ‘The FLO standard has better performance, it can offer 4 hours of TV viewing with standard 850 batteries. It can cover a city with just 2 or 3 towers, each with a 20 kilometer radius using towers operating with 50 KiloWatts of power. (DVB-H uses 2 MegaWatts of power over a larger area, he points out).
It can also work alongside various air interfaces, cellular standards, plus their data extensions EV-DO and HSDPA. By work alongside, this means not very much. Qualcomm’s vision of how TV works on a mobile is only one version of many possibilities.
Like the other competing systems (at the moment) a separate radio is installed in the handset and a separate frequency is used to broadcast mobile TV channels. The cellular frequencies that make up the core cellphone service are used for distribution of encryption key and for delivery of on-demand content that doesn’t fit into the broadcasting service type.
But Qualcomm thinks about MediaFLO and. FLO as two separate things. FLO is the technology in the handset and the broadcasting technology, MediaFLO is the entire concept with the support services, billing, content loading, encryption and conditional access, all built up as either a service or as a set of complete software tools that can be purchased by an operator to run their own services.
The FLO part of the equation is entirely proprietary Qualcomm technology, as we have come to expect from the owner of most of the global CDMA patents. It has developed its own conditional access system instead of using the almost universal OMA 2.0 DRM, it has tweaked H.264 video compression with some enhancements that make it non-standard, that it might (probably not) push back into the H.264 standard, and it ignores the compelling argument to use pure IP packeting to deliver its TV programs.
‘OMA has a lot of extra stuff that I don’t need,’ said Chandhok, ‘and it has licesnsing issues (see separate story) and so a simple conditional access system, which doesn’t describe my rights, but just has ‘play’, ‘not play’, ‘not copy’ and the controlled issuing of keys, is better for MediaFLO.’ Of course the operator has to pay for the OMA DRM anyway because it will need to protect any on-demand content it delivers separately, it’s just that Qualcomm is steering clear of the patent license debate.
Its conditional access system supports key revocation in case a key is being used for piracy, uses Public Key Infrastructure, 128 bit Advanced Encryption Standard, separate key delivery and renewability of keys and software.
‘I have strong feelings about not re-inventing protocols for the sake of it, for instance we use IPv6 addressing in MediaFLO, but DVB-H has bizarre layering and IP layering that just introduces overhead because it is designed to be as close to DVB-T, the terrestrial TV network, as possible. The last project that I worked on was SIP and Push to Talk, and on a wireless network using pure IP just doesn’t perform well enough.
‘Our air interface is a very simple bit stream interface, mapped tightly onto the physical layer,’ said Chandhok.
To Qualcomm, wasting cellular bandwidth has become a true sin, and although it is operating here in a broadcasting space, it still retains those instincts which fight standardization in favour of performance.
‘We send the video over the air as deeply optimized as we can,’ he emphasized, ‘because spectrum is a scarce resource which we, and our clients, care very much about. Pure IP applications don’t perform well enough, the latency is too high.’
The truth is that this has been said about video signals for the past ten years, and yet gradually most wired triple plays are moving inexorably to pure IP, because it introduces component and service commoditization and despite spectrum issues, makes service provision cheaper. Eventually new generic routing protocols will catch up with the mobile world and IP will reign, in Faultline’s view, but this is certainly not Qualcomm’s position, at least not right now.
‘You can treat what we do as simply compressed IP. I can take a video signal as IP, convert it, send it over FLO and then turn it back to IP if you are worried about it’ says Chandhok, but right now, in his view, his customers would rather have to pay a little more for their equipment and get as many channels as they can.
The reason that Qualcomm thinks of MediaFLO as much as a service as a set of products, is that it has purchased Channel 55 UHF throughout the US, 716 to 722 MHz, most of it for the extremely cheap price of $38 million, and Qualcomm is spending $800 million to build out and operate a 20 channel TV network that it plans to rent to US cellular operators.
‘We don’t plan to have any direct consumer relationships on this network,’ explains Chandhok, ‘and in other markets, such as Asia, we expect to see wireless operators, broadcasters or consortia take the lead to do something similar.’
The Qualcomm view of mobile triple play is not entirely borrowed from the US cellular carriers, but it certainly overlaps. Chandhok sees quite a variety of monthly packages that are designed to build ARPU at the mobile operator, bundling certain amounts of programming with data and voice units. The cost of the addition of a DVR in the phone can either be passed on to the customer, or built into the bundle cost, but today even the short form video used on mobiles will need around 250 Megabytes of data to store a full length feature film, so this service model, including the DVR, assumes the delivery of hard disk drive phones within a year or so.
So far the only hard disk drive phone that has been introduced has been in Korea by Samsung, but within a few months every major handset maker is expected to add hard disks, and the price premium won’t be too high, with 2 GB and 5GB devices initially costing something like $80 each and rapidly falling to $25. Real estate is being freed up all the time on mobile phone chips (Texas has reduced its chip footprint by around 50% this year alone), so there should soon be room for plenty of DVR storage capacity.
However this service model is certainly not the only one possible. Operators using WiMAX raises the stakes considerably for Qualcomm. When asked about the carrier for FLO, Chandhok says ‘yes, it’s OFDM,’ but that’s not WiMAX, because WiMAX is designed as a two way service.
But Samsung definitively disagrees with Qualcom here, and it has just put forward the idea of marrying its Wi-Bro broadband wireless that is being integrated into WiMAX, with its DMB implementation (See separate article). The critical step here is that no separate network needs to be built to carry television signals to mobiles, but instead the WiMAX base stations reserve a proportion of their channels for broadcasting video.
The underlying WiMAX technology has been used to get more Megabits per second out of the airwaves, making WiMAX spectrally efficient, but after that the network usage can be worked out with less regard to efficiency because there’s so much bandwidth available when WiMAX is used.
That’s OK in Korea perhaps where there is more free, WiMAX class spectrum, but it may not suit the US so well, and perhaps not Europe either.
The argument here will linger long and be debated loudly. In the end the Qualcom vision of mitigating on-demand use of the cellular channel by offering DVR software for the phone, and clip casting, can be seen by TV purists as a bit of a fudge. This is old school television, supported by advertising, or bought on a monthly charge. But this is quite simply because mobile telephony really doesn’t yet offer enough spectrum so that every phone owner can watch a different on demand program at the same time. This was tried once by South Korea Telecom, and its cellphone network ground to a halt. But again, WiMAX might make this possible and it’s certainly becoming the model in IPTV over wires.
‘We don’t see this as an advertising market,’ said Chandhok, ‘but some of the carriers do, and we have to support that. We see advertising working in this environment more like product placement.’
‘But MediaFLO supports localized advertising insertion, should any operator want to use it,’ said Chandhok. ‘We also support market blackouts, for sports events that are being offered pay per view in some markets and not others.
So is Qualcomm going to pick the 20 TV stations that run on its US MediaFLO network, or are the US cellular carriers? ‘We will contract directly to put on the most popular content and then allow a certain amount of differentiation between the operators,’ said Chandhok.
Does that mean that Sony Pictures or Disney could offer its own film network, and market it directly to the cellular operators? ‘We would look at each proposal on its merits economically,’ concluded Chandhok, which means that he would if there was enough money in it for Qualcomm.
Differentiation though of the separate Cellular networks is going to be down to the on-demand programming and the clip cast content that each operator buys into, and mobile TV, in Qualcomm’s world is just an additional $10 or so a month added to the Average Revenue Per User, shared between Qualcomm and the cellco.
Somehow we think that Cellular TV will need to offer more than that, in the end, to keep the attention of most cellcos.