Your browser is not supported. Please update it.

29 November 2006

Doubts raised over Qualcomm claims to MediaFLO superiority

While talking to Thomson Grass Valley, the transmitter and mobile TV arm of Thomson, for our forthcoming mobile TV report, the company raised some objections to the way Qualcomm has been positioning MediaFLO as capable of double the number of channels compared to DVB-H. As far as Yann Courqueux, director of business development at Grass Valley is concerned, the two mobile TV systems are largely comparable both in bits per second delivered and cost per bit delivered.

The Grass Valley credentials in this area are pretty impeccable having delivered a variety of equipment ranging from head ends, IP encapsulators, DVB-H encoders, transmitters, service delivery platforms, DRM processors and handset EPG software across 20 mobile TV services and trials, around the world, including some to Qualcomm’s MediaFLO.

Grass Valley head ends are used in the 3 Italia live DVB-H service, and it has supplied separate head ends for each of Telecom Italia Mobile, Mediaset and Vodafone on the competing Mediaset DVB-H service, also in Italy. It is also working in France with TDF, as well as TPS and Orange, and the largest mobile TV trial at Multichoice in South Africa.

‘We like the MediaFLO technology very much, but when Qualcomm gives out its comparisons, it always seems to include the use of statistical multiplexing in MediaFLO, but not when it is assessing DVB-H. This is simply because it’s not part of the standard, but of course you can use statistical multiplexing in DVB-H, and you can dynamically adjust the burst sizes and get a gain of around 40% in bits delivered.

‘There’s a buffer in the phone so the device doesn’t mind receiving variable bit rates, and using stat mux makes the difference between an 8MHz channel allowing 25 TV channels and boosting it up to 35.’

MediaFLO this week contacted us to say that it was claiming 40 channels not in an 8MHz European TV channel as we quoted last week, but in a 10 MHz channel, which makes it roughly in line with the Thomson Grass Valley figures.

The idea behind using a statistical multiplexer is that when any part of any frame uses less than its full anticipated bit rate on a particular frame, because there’s inactivity at that point in the video stream or there’s a high prediction rate going on from frame to frame, the freed up capacity is made available to other channels in the system. Also when DVB-H sends a ‘burst’ of around 4 seconds of video it doesn’t have to do it in a specified amount of time. It can vary the sending time between the segments of each channel, as long as the total series of bursts adds up to no more than whatever can be delivered in the 4 second interval.

Pre-processing in the head end and the stat mux calculates all of this dynamically so that the best possible outcome in terms of video quality can be obtained.

‘It means that a signal that might be thought of as 350 Kbps can end up actually being delivered in under 200 Kbps,’ explained Courqueux.

‘You might even synchronize each burst so that it always includes an I-Frame at the beginning of the burst, and that way, a channel change can be rendered immediately and the maximum delay will be the burst length if you have just missed that channel’s part of the multiplex signal in the previous cycle. We are also experimenting with variable and long GOPs.’

A GOP is encoder slang for a Group of Pictures, and is the number of frames between one reference I-Frame and the next. Most IPTV systems have these set at 12 or 15 and fixed because they need to re-iterate an I-frame in order to support faster channel change. But if you can guarantee that an I-frame is delivered at the front of a burst, then you can extend what are called delta frames in between, which refer back to the I-Frame for rendering, for up to 2 seconds and 60 frames.

Delta frames are far smaller in bit size than I-Frames and a threshold can be set so that if there are too many changes from the previous I-Frame, resulting in the amount of data being sent rising, for instance if there’s a complete change of scene, then instead of sending another delta frame, the system can elect to send a fresh I-Frame at a varied interval.

Another trick that Thomson Grass Valley told us about is arranging the burst in the same order as the channels. So channel 1 is the first in the multiplex, so that users get into the idea that if they want to zap channels, they can go upwards (not downwards) and get a faster channel change.

Faultline is writing a mobile TV report for early in the New Year and will be publishing it in conjunction with Multimedia Research Group in the US.

Meanwhile Qualcomm has added T-Mobile to the list of companies trialing its MediaFLO service. T-Mobile only a few weeks back said it was going to do technical trials with HiWire, the DVB-H joint venture between SES Americom and Aloha Partners that owns UHF spectrum in the US.

T-Mobile is also understood to have been involved in the Pittsburgh Modeo trial for DVB-H in L band spectrum. Sprint Nextel is also known to be testing MediaFLO, but both Sprint and T-Mobile would have to wait even if they wanted to use the existing network, because Verizon has an unspecified period of exclusivity with the service, which is likely to be at least a year, though this period has not been made public.