All posts by Mike Dimmick
Below are all of Mike Dimmick's postings, with the most recent are at the bottom of the page.
Incidentally, the next Astra 2 satellite, 2E, was due to go up this Sunday. It also has a UK spot beam so should increase the capacity for FTA channels.
I said was. It's due to go up on a Proton Breeze M rocket, but the Russian government lost three navigation satellites (their version of GPS) in a failure two weeks ago. 2E is suspended until the investigation is complete. Federal Proton Launch | Failure Updates | Glonass | July 8 | International Launch Services
| link to this comment |
The reason for the low power levels is in part because Arqiva are going to reuse old assets as far as possible. The nominal ERPs match their latest Reference Offer. They're based on whatever the available antennas can handle, and what power the available transmitters can provide.
So, for example, Crystal Palace C33 will be from the PSB antenna at the top of the mast, using the transmitter previously used for pre-switchover Mux 2, while C35 will come from the COM antenna just below, and use the pre-DSO Mux B transmitter. Meanwhile, the former backup transmitters from Croydon will be travelling to Winter Hill, where they will drive the pre-switchover digital array, which has to be rotated so its null points in a different direction.
http://arqiva.com/documen….pdf for the details.
As far as which sites have been selected: "the list of 30 sites has been chosen based on indicative population coverage and ease of implementation." You have to remember just how many people the top few transmitters serve. The whole point of this is to get a cheap service serving as many people as is feasible, as soon as possible.
| link to this comment |
Briantist: 4G is ultimately supposed to have Voice-over-LTE - analogous to Voice-over-IP. It's fundamentally a data network but can carry voice data in packets, unlike 2G which is a voice network that can carry packet data if it's encapsulated properly (but the phone can't do a voice call and data at the same time).
Phones may well need updates to support VoLTE, though, so 2G and 3G will be around for a while yet.
(Incidentally, I've changed email address - planning to leave Demon since I'm moving, they don't do fibre and my new address is too far from the phone exchange for reasonable ADSL speeds, and they got bought by Vodafone. Any chance you can associate my new posts with my old ones?).
| link to this comment |
Briantist: Have you seen the 'Table of Broadcasting Stations for Multiplex L' (the Local TV multiplex)? It's appeared on Ofcom's licensing website, dated this Monday (29 July):
http://licensing.ofcom.or….pdf
Confirmed start dates, UHF channels, aerial heights, polarisation, ERP and radiation patterns.
The licences themselves have also appeared at Ofcom | Multiplex L
| link to this comment |
trevorjharris: Freesat and Sky's EPGs are totally independent streams. Nothing is combined into the channel itself. The video stream and audio stream have no internal metadata; each transponder does carry a small amount of metadata in the DVB Service Information tables that are multiplexed into the content. This metadata isn't enough to present a full EPG, though - it doesn't include channel numbers and region selection, it's just enough to do Now/Next on a generic free-to-air receiver.
The actual audio and video streams for any free-to-air channel are unencrypted, and they are broadcast once only. Take a look at Astra 1N / Astra 2A / Astra 2D / Astra 2F / Eutelsat 28B (28.2°E) - All transmissions - frequencies - KingOfSat , for example. Observe that BBC Four is listed once on the entire array of transponders in the cluster of satellites at 28.2°E, where UK dishes are pointing. The Freesat and Sky EPGs both simply say, 'when the user selects BBC Four, tune to 10773 MHz, select Horizontal polarization, decode it as DVB-S with symbol rate 22000 and FEC mode 5/6, then look up Service ID 6316'.
The other columns in that table say that Service ID 6316 points to Program Map Table (PMT) 259, which says to show the video stream with ID 5200 (the VPID), get the Program Clock Reference (PCR, used to synchronize video and audio) from that stream, get audio from stream ID 5201 or 5202, and teletext from stream 5203.
This is why we say that Sky, and Freesat, and generic receivers all use *exactly the same broadcasts*. A 'generic' receiver does not understand Sky's or Freesat's EPGs: the user has to do the set-up work of assigning channel numbers to Service IDs, but once they have set it up, the only inconvenience is if a channel changes its Service ID or which transponder it's using.
The broadcasters separately and independently send their programme metadata to Sky for inclusion in their guide, and to Freesat for inclusion in *their* guide. (And to Digital UK for inclusion in the Freeview EPG.)
The broadcasters *separately* arrange their uplinks. Each transponder's signal is uplinked separately. The transponder simply receives on one frequency and transmits on another, there's no intelligence in there - which means that you can change which standard is being broadcast, as has been shown by some transponders changing from DVB-S to DVB-S2, and even direct from FM-PAL to DVB-S2.
Only those broadcasters which are actually sharing a transponder need to make any arrangements, and they generally do so between themselves. Sky only get involved because they have taken out a number of long-term leases on transponders, and then they sub-let that capacity to broadcasters and arrange multiplexing. It's usually a condition of carriage that the channel is exclusive to Sky for some period, and usually under the subscription.
However, as I said, the PSBs, for the most part, lease their own transponders directly from SES.
The big myth is that Sky is heavily involved in broadcasting from satellites and is vital to the process. It is not. All the work is being done by the broadcasters themselves. Sky is a gatekeeper to the subscription platform, nothing more - that is, they get involved in taking money from subscribers, and distributing some of it to the channels that are in the paid business.
Sky only have a free-to-air offer because they were made to. They continue to offer it because it allows them to effectively advertise all the subscription content in-line, and makes it easy for people to take out a subscription - they can simply activate the subscription virtually overnight. The subscriber boxes are basically subsidised by the 'Platform Contribution Charge' that forms the majority of the payments made by broadcasters, and which the PSBs are objecting to.
If the BBC were to leave Sky, they would save money. They would not be paying Sky for a service that they don't want and which undermines their multi-platform strategy. No other platform asks the broadcasters to contribute to subsidising boxes for subscribers to *other content*, and Sky can only get away with it because they have significant market power. While viewers can use the Other Channels feature of a Sky box, those channels aren't assigned channel numbers, and don't appear in the main channel list; nor do they show programme information and the Accurate Recording and Series Recording features of Sky+ don't work. Those are the only reasons to appear on Sky - but since not doing that effectively drops 40% of main set viewing, the BBC will never do it.
| link to this comment |
Tim: No, that's completely wrong.
The signal/noise/interference profile is governed by the low-level modulation. You can trade off the multiplex capacity in megabits versus the signal-to-noise+interference ratio required (effectively the coverage). The PSB multiplexes have approximately the same required SNR for reliable performance: PSB1 and PSB2 use the same DVB-T mode and have the same SNR requirement, PSB3 (HD) uses DVB-T2 and has slightly different requirements, but less than 1 dB difference in the most difficult conditions. The COM multiplexes started out using the same mode as PSB1 and PSB2, but chose to trade off a little less coverage for a little more capacity. The rollout of that change completed last year.
Changes to compression efficiency only affect the round-trip accuracy of the result. MPEG-2 Visual and MPEG-4 AVC get most of their compression from predicting where a block is going to move from one frame to the next. (MPEG-2 Visual can only encode moves in whole-pixel units; AVC can encode moves in quarter-pixel units.) Where motion prediction is used, all that needs to be sent to the receiver is the list of block movements, plus the differences between the resulting reshuffled old picture and the actual new one. The better the prediction, the less difference data needs to be sent. With more computing power available, more effort can be expended on the prediction in the same amount of time. The encoders have to work in real-time - i.e. a fixed, small duration from the frame arriving at the encoder to the encoded content coming out - because many of the channels are live, or can have live programmes, and even if a multiplex only contains pre-recorded programmes, the effort to co-ordinate all the channels to do the encoding off-line is impossible.
Different content requires a greater or lesser amount of multiplex capacity to encode with equal quality (and equal computing capacity). In order to keep quality consistent over a range of content and allow extra capacity momentarily where needed, the multiplexes use 'statistical multiplexing' - another device, the 'statmuxer', looks at the bitrates that the various channels on that multiplex are currently using, and what they're asking for, and makes decisions about reducing the rate on one in order to increase it on another. Again, the better the computing power, the better decisions it should make.
The other source of compression is how a full picture, and the differences between prediction and actual, are encoded. This is the same as JPEG: the picture is divided into 8x8 pixel blocks, then the brightness and colour are converted from a sequence of values (scanned diagonally across the block) into frequencies, using the Discrete Cosine Transform. The encoder then uses a quantizer to say how large the contribution from each frequency is. The more compression is required, the fewer bits are allocated to each frequency, which means fewer values are available. The quantizer selects the nearest available value to the true value, which means it's a little bit wrong. The fewer levels available, the more wrong it will be. But if it's fairly close to the 'right' value, the viewer won't notice.
Primarily, when not enough bits are available for a channel, the block edges become visible. This is because the quantization in one block doesn't match that in another, so a continuous tone across two blocks ends up as one level in one block, but a different level in its neighbour. The decoder has a 'deblocking filter' to correct this, but the bigger the difference, the less effective it is.
Yes, one way to increase the compression ratio is to increase the length of the Group of Pictures (GoP). A GoP contains one full picture (the 'Intra' or I frame) and then any number of either Predicted (P) frames (motion-prediction from the I frame only) or Bi-Predicted (B) frames (motion "prediction" from an I frame and a future P frame, or between two P frames - the frames can be sent out-of-order and the receiver puts them back in the right order). As you say, a longer GoP also means it takes longer to start displaying the video stream after switching channels, and if you're directly cutting the data stream, it gives you fewer cut points, since you can only start a cut at the start of a GoP, and only end it after an I or P frame (since the B frames don't make sense if you remove the P frame they depend on).
| link to this comment |
Charles Stuart: The Croydon transmitter will provide coverage to the south-east, while Crystal Palace's service won't cover that area. Presumably that's easier to build than an antenna array serving the whole area off Crystal Palace itself - it may be down to some other tenants on CP being in the way, given that the Croydon aerial is so much further down the mast.
The Croydon and Crystal Palace services have been assigned the same frequency, so they will be running as a single frequency network.
There was certainly criticism from the candidates for the London service that it wouldn't actually cover London all that well, this may be the best compromise that Comux and Arqiva could come up with.
| link to this comment |
Briantist: I've just noticed that the Mux L table lists aerial height above ground, not height above ordnance datum as used in the other multiplex licences.
| link to this comment |
Ian: A red button service from where? Each service occupies some capacity, either taking capacity from some other service or requiring additional multiplex capacity. A red button service actually requires more capacity than a standard service appearing in the EPG, because one or more normal services have to host a small program to display the 'press red' popup and switch to the other service.
One of the reasons that these local services don't have as much reach is that the TV spectrum is very crowded. This set of stations are what was left over after the three PSB multiplexes had reached their coverage goal, plus the three COMs had had their chance for as much coverage as feasible without taking on new sites.
So the point is to provide a local TV service for those towns and cities where coverage is feasible at a reasonably-affordable price. Not to cover all towns regardless of population, and not to create a full regional network.
Waltham's local TV service on C26 has to avoid interfering with D3&4 from Crystal Palace, BBC A from the Bromsgrove/Lark Stoke/The Wrekin group, two of Nottingham's relays in the Derby area (Little Eaton and Eastwood), four of Tacolneston's relays in Norfolk and BBC A from Bilsdale.
Users find 'red button' far less discoverable than real channels - they are less likely to be aware that the red button service exists. That's why the BBC changed BBC Three's hours for the Olympics, rather than just adding a new red button service. (They added that too, but it was a lot of work to reorganize the channels' run times which wasn't strictly necessary.)
Unless, of course, you mean a 'connected' red button - the program that handles the button-pressed event merely tells the box to switch to an IPTV service. Sadly, this feature is only guaranteed to be on Freeview HD newer than a certain date. On those boxes, however, you can also select a service from the EPG which just loads and runs an program that redirects to an IPTV service. See 'Connect TV' etc in the 200+ range of channel numbers.
| link to this comment |
Wednesday 17 July 2013 1:05AM
MikeB: I don't know why you're assuming satellite=Sky. The *exact same* transmissions are used for Freesat and Sky for all BBC services now, there's no reason to think that will be different for the new HD services.
The only difference is the payments to the platform operators to appear in their EPGs. Each box is looking in a different place for a different format of EPG data. Freesat charge slightly more for that than Sky do, but Sky *also* apply a 'platform contribution charge' based on viewing figures, so overall being in the Sky EPG is far more expensive. Unfortunately, that's where the viewers are.
Carriage on a UK-footprint transponder - allowing free-to-air transmissions - is also more expensive than a Europe-wide transponder and Sky's encryption fee. Most commercial channels have decided that they will make more profit if they go behind the paywall, despite Sky's outrageous charges, than relying on advertising alone and paying those extra transmission costs.