The freeze in the IPTV market that has lasted most of last year is more about getting decent High Definition set tops than about the dominance of the Microsoft Alcatel axis, according to Matt Cuson VP of marketing at Minerva Networks.
Faultline was catching up with Minerva after a hectic IBC where we managed to miss them, but where it became obvious that Minerva had transformed both in terms of what it brings to the IPTV market and in terms of how successful it has been, through what has been a tough year for most IPTV companies.
‘There was a fair amount of pain during 2006 but that is really down to the delays in the Sigma and ST Micro MPEG 4 set tops for HD,’ said Cuson.
‘They were expected in June, but volumes really won’t start now until next year. Oh sure you can put a chip and a power supply inside a plastic box and call it a set top, but tightly integrating these devices with a software stack will still take time. Some vendors are ahead of the pack, and I know Amino and ADB are shipping product, but most set tops for H.264 will go to volume next year.’
Minerva spent a long time telling the world that it was an encoder maker before it go into the middleware business, and then it hit the jackpot. After fighting Tut Systems as a competitor in the US, it converted it into a partner. ‘Tut was reselling iMagic, then it very publicly signed a deal selling Myrio, but then we told them we have dropped encoders and found it was happy to work with us,’ said Cuson.
Much the same story has been in evidence with Motorola and Myrio. ‘Myrio has lost itself inside Siemens, and it no longer focuses on US independent operating companies (IOCs) and we have picked up perhaps 35 customers from there, while Motorola’s Next Level customer base is also converting across to our software.’ The result is that Minerva has some 85 commissioned deployments, most of them in the US, and many of them, as yet, very small.
Minerva claims that it has not yet lost a deal to Myrio in the US this year, but mostly doesn’t see it bidding. Myrio rocketed to fame when it landed both Belgacom and KPN in Belgium and the Netherlands in IPTV last year, just after it was acquired by Siemens and has focused on tier one telcos since then.
‘On the Sigma chips we are working with Motorola, Scientifica Atlanta, Tilgin and others. But when you pump a lot of bandwidth into a set top box there are performance issues, and it will take a while to shake them all out.
‘We have found that there is a difference between the Sigma chips and the ST Micro chips, with the ST Micro 7100 offering a better understood architecture. Some set top vendors just take the reference drivers that the chip supplier gives out, others that offer their own drivers have found that the 7100 is not so different from its predecessors and are able to respond faster.
‘Scientific Atlanta, ADB, Amino, Motorola and Thomson are all handling their own integrations and in general where they use an ST Micro chip the set top is more stable.
‘The reason that cable and satellite have had HD decoders for so much longer than IPTV is that it is much harder to write software for an IPTV set top than for a cable box. In IP the data arrives in Ethernet in IP packets, and that needs CPU time to unravel into an MPEG stream. If you have not yet optimised the drivers, and have your memory management performing tightly, then with 20 Mbps coming at it, it will crash,’ said Cuson.
The HD MPEG 4 set tops got delayed last year because the silicon had to be respun, and a new silicon design can take 6 months out of a process. Microsoft was thought to have requested changes in the silicon for AT&T, and that contract took priority and affected the non-Microsoft versions of the chips. ‘It was a security issue,’ explained Cuson, which seems to imply that the part of the chip that operates as a security processor, holding part of the decryption key in a tamper proof enclosure, might have been the delay.
‘That took us to the end of 2005 for chip samples and we all expected them to arrive in volume in Summer, but then it slipped again. There were enough boxes to go and design middleware for them, but no volumes. And if IPTV operators are going to offer DVRs, then it will be early Q1 and into Q2 next year before the software issues are stable.’
‘But the result then will be that HD devices will no longer be expensive,’ said Cuson, but there will be other issues. Channel changing with MPEG 4 (and VC1) is far slower than with MPEG 2 codecs. There are three types of frames, an I-frame which has an abbreviated description of what all the pixels in a frame are doing, and B frames and P frames which are rendered by making minor changes to the I frames that have come before or which are arriving next.
The big innovation with MPEG 4 (also known as H.264) is that the B frames and P frames can refer to just about any recent I frame and there is a special tag that says ‘From here on we can throw away old I-frames and start a new group.’
That means that if a channel change occurs in the middle of such a ‘cluster’ of I- frames, then ALL of the old I-frames that may get referred to have to be sent to the set top to sit in the buffer before any new frames can be rendered. In MPEG2 systems it was just the immediately prior I-frame.
‘There are two ways to attack that problem,’ said Cuson, you can opt for using more bandwidth (which defeats the object of having a very narrow bandwidth codec) or you can squeeze more complexity at the set top and the head end, which is what Microsoft chose to do. The funny thing is that you still need to overprovision the network for the bursts of I-frame data that are flying around on it.’
What Cuson says has been done is that not all the full capabilities of MPEG 4 have been used initially and that head ends are treating the new codec a little like MPEG 2 and making compromises in how the data is encoded. ‘If we can get a High Definition channel down from 20 Mbps to 10 Mbps, so that the next generation systems can work, we are happy. Later when set tops get even smarter or if we are dealing with on demand streams, then a lot more can be done with H.264 and we can get really low bandwidth’
Minerva insists that its operators have not had any trouble winning new clients this year, except for one type of customer. ‘The cable HD customer who has a DVR, remains resilient. After all, we have the same channels and all the same services, it is just the HD DVR. From Q2 next year a lot more will be attracted to IPTV.’
‘The service providers we work with have been winning clients by the value adds they can bring. They haven’t had to low-ball the price and most of them are shooting for a triple play at $100 a month. When cable comes up against those IOCs where this has been achieved, it drops its pricing to pretty much the same.’
Remember there are 10 million customers served by IOCs throughout the US and as they proliferate with IPTV and especially a HD based version, they are going to knock a savage hole in cable one way or another. Cable will either have less customers or it will have to drop its price in order to keep them. Either way it will hit mid year and third quarter figures for US cable operators in 2007.
‘The take rate at operators like SureWest is 60% to 70% of all new customers, taking IPTV. Smaller operators get better penetration and we reckon many of our customers now have over 30% total penetration with video, ‘ said Cuson.
Minerva says it was profitable in Q2 just gone, and has important systems integration contracts with Nortel, which helped it win the vital Telus deal it landed last month, and has bids out involving IBM and even Accenture. It has integrations with Latens, Widevine and Irdeto for conditional access and with Entone, Kasenna and Bitband for VoD servers and claims that it has yet another large tier 2 telco deal imminent.