Restack 2.0

[ron12]

With Malcolm Turnbull announcing his intentions to free up some of the TV spectrum for other uses, e.g. mobile broadband, I would like to put forward some ideas on how this could be done.

I have read elsewhere that MPEG4 would allow the same video to be encoded using half the bitrate, while DVB-T2 would allow the bitrate to be increased. On that basis, the number of channels required for any area would be reduced from 6 to 3 (already mentioned by others, e.g. here, This would free up 15 channels. Given the block structure of our channel allocation, this would mean freeing up three channels on VHF (could be used for expanding digital radio) and 12 on UHF (which could be used for freeing up 610-684 MHz for another digital dividend).

What about H.265, which I've read requires half the bitrate again compared to H.264? Would this reduce the channels required per area to 2? This could free up another 5 channels, e.g. one channel from VHF and four from UHF, or we could dispense with the VHF band altogether for TV, freeing up 2 UHF channels, or just use them as unallocated channels. H.265 may be awhile off from being a viable alternative but we could start planning for it, moving to MPEG4 first, then switching to H.265.

[Guest Malich]

I have read elsewhere that MPEG4 would allow the same video to be encoded using half the bitrate, while DVB-T2 would allow the bitrate to be increased. ... What about H.265, which I've read requires half the bitrate again compared to H.264?

All I'll say is this: be very very careful when assuming that MPEG-4 or H.265 halves the required bitrate or that DVB-T2 doubles the available digital bandwidth.

Most of those claims come from Alan Hughes, and therefore have little to no basis in reality.

For example, Alan says crap like:

• HVEC compression produces a data rate 1/4 compared to MPEG2.
• The DVB-T2 maximum data rate is up to twice DVB-T which we use.

... while the truth is:

• the maximum data rate allowed in DVB-T2 is just over 50Mbps*
• the maximum data rate allowed in DVB-T is just over 31.5Mbps*
• so the DVB-T2 maximum data rate is really only ~1.6 times that of DVB-T*

(* All figures taken from or based on ITU & ETSI DVB-T and DVB-T2 specification documents.)

I'll let someone else find officially-sanctioned figures and do the maths for MPEG-2, MPEG-4 and HEVC - but they disagree with Alan's claims to a similar degree. If you worked on the basis that H.264 required 65%-75% the bitrate of H.262 (MPEG-2) for the equivalent quality, and that H.265/HEVC required 65%-75% the bitrate of H.264 for the equivalent quality, then you wouldn't be far wrong...

Edited September 11, 2014 by Malich

[ron12]

Hi Malich,

The claim that MPEG4/H.264 halves the bitrate compared to MPEG2/H.262 was from an article I read about H.265 (which halves the bitrate again compared to H.264 according to the article). The only things I found on the internet are this forum post which says that MPEG4 halves the bitrate compared to MPEG2 according to the official mantra, and another thread on the same site that says that there is a 20-30% improvement.

The claim that H.265 halves the bitrate compared to H.264 is also mentionedin the Wikipedia article on H.265, where the results are just above or just below 50%.

I never said that DVB-T2 doubles the bitrate compared to DVB-T, only that it allows the bitrate to be increases.

Combining MPEG4/H.264 with DVB-T2 would hopefully lead to being able to deliver twice the video content at the same quality compared to DVB-T/MPEG2. Using MPEGH/H.265 would hopefully double that again.

[pgdownload]

The only things I found on the internet are this forum post which says that MPEG4 halves the bitrate compared to MPEG2 according to the official mantra, and another thread on the same site that says that there is a 20-30% improvement.

I'm no expert, but I'd be very dubious of any modern compression algorithm reducing data needed by 75%. (50% then 50% again).

Perhaps they're quoting peak efficiency (as proponents like to do) where the occasional ideal frame is massively compressed, but I'd think it much more likely a 20%-30% hit on each new iteration might be sustainably possible.

0.02

Peter Gillespie

[Guest Malich]

Hi Ron12,

As you can imagine, the figures floating around re: H.264 & H.265 efficiency improvements are very rubbery - there's the initial claims (which are mostly marketing / staking out ground), there's the results of 'objective' testing (e.g. PSNR/qPSNR, SSIM, etc) which aim to quantify the difference between original & encoded video (and, despite there being large differences in the types of image degradation between codec families & generations, get used [arguably, 'abused'] to compare codecs), and there's the results of 'subjective' testing which (in every meta-analysis I've ever seen) turns out to be almost as heavily influenced by source content as codec type. So there are no real hard-and-fast comparison figures...

But in both cases - H.264 vs H.262, and H.265 vs H.264 - the ITU and friends seem to have settled on the ~65%-75% figure (i.e. ~25%-35% reduction in bitrate) as a general expectation / rule of thumb.

(I'd link to some examples of that in the ITU/IEEE documentation, but they seem to be down - for me at least - at the moment.)

In the meantime, have a look at this fairly recent paper I have a link to handy:

Performance Comparison of H.265/MPEG-HEVC, VP9, and H.264/MPEG-AVC Encoders.

It obviously doesn't cover H.262/MPEG-2, but between H.264 and H.265 they find:

• on raw bit-rate over a number of scenes and perceived quality levels, H.265 results in an average 38.9% (23.6%-55.1%) reduction of bitrate compared to H.264
• using a weighted average (BD-BR), H.265 results in an average 39.3% (24.9%-47.9%) reduction of bitrate compared to H.264.

which is not terribly far off the rule-of-thumb range I mentioned.

Or in other words, as I see pgdownload has said much more succinctly since I started replying , proponents like to quote peak efficiency...

[MLXXX]

It is problematic to compare the performance of the video codecs Mpeg-4 and Mpeg-2 as the relative efficiency varies with the absolute bitrate. Mpeg-4 AVC can successfully compress to very small video files with remarkably little degradation to the subjective quality. It is streets ahead of Mpeg-2 if the goal is to produce an extremely small file.

However television broadcasting ordinarily uses higher bitrates. The goal is for the picture not to look like a marginal quality streaming download!

Another factor I recall reading about is encoder efficiency. Some high tech video encoders encoding on the fly (as is often needed for broadcasting) do a more efficient Mpeg-4 encoding job than others.

Another complicating factor is whether the source is progressive (e.g. 1280x720p59.94) or interlaced (e.g. 1440x1080i50). With interlaced content, evaluating the subjective quality is complicated by possible variation in the quality of the de-interlacing of the decoded video, some displays performing de-interlacing better than others.

A factor often not mentioned in these discussions is audio quality. With free to air SD transmissions, I often find myself hearing compression artefacts in the audio. I would hope that any move to reduce the video bitrate for H.264 [or H.265] would be accompanied by use of a sufficient bitrate for the audio codec.

Edited September 12, 2014 by MLXXX

[GoForMoe]

We've also made some massive improvements in the quality of MPEG-2 broadcasts over the years as the encoder technology has advanced.

Obviously you'd see something similar as HEVC reaches the peak of its life - but the gains at the moment would be comparatively small. But, there is no point making a move to an outdated system like MPEG-4 now is, particularly on DVB-T2. It would be like replacing a copper broadband network with fibre to the node instead of fibre to the premises. Absolutely no one sane would suggest that.

Edited September 12, 2014 by GoForMoe

[ron12]

Hi GoForMoe, if MPEG4 and DVB-T2 are outdated, what would you suggest is more up to date for each? I know MPEG-H/H.265 would be more up to date, is that what you would have instead of H.264? What would you replace DVB-T2 with?

[GoForMoe]

I meant that as in it would be a waste to move to DVB-T2 and not go straight for HEVC, when equipment replacement would be required anyway in most cases.

One of my few positive things to say about our digital radio rollout is that we at the very least waited for DAB+, despite the fact that it meant a smaller number of receivers on the market at launch and higher costs to consumers.

A MPEG-4 rollout on DVB-T might be attractive in terms of the larger installed equipment base, but when you force consumers to replace their equipment again, you should choose the most future proof option available at the time, not put us immediately back where we started quality wise once the Government starts to want its spectrum back.

[alanh]

We currently use DVB-T and DVB-T2 doubles the maximum data rate available from a TV transmission channel. It has nothing to do with compression systems such as MPEG and HEVC. Compression systems reduce the data rate required to carry a program.

We should use MPEG-4 now because we could stop downscaling HD from 1920 to 1440 now. This would remove jagged edges on near vertical edges in the picture.

If you watch ABC I often see pixellation caused by too many programs and insufficient total data rate available. The problem is that no one knows how many viewers cannot view MPEG-4, however if they cannot, a set top box is available in a chain store for <$30.

Currently there is no widespread availability of HEVC equipped receivers or STBs.

Broadcasters will not use HEVC on all programming until all of the audience is equipped to view it. Surveys of viewers say that most viewers replace TVs within 7 years, so you will have to wait that long for HEVC unless you can get an HDMI plug receiver containing a DVB-T2 and HEVC. It needs to be controlled by the HDMI signals. This device needs to be nearly free if you wish to reduce the waiting time.

UHD or 4K requires DVB-T2 and HEVC as well.

Alanh

Edited September 13, 2014 by alanh

[CaptainObvious]

I will not go into the technicalities, as there are many, but coverage for coverage DVB-T2 will provide at least 30% more capacity. There is absolutely no guarantee, promise nor any other assurance that DVB-T2 will provide a 50% capacity gain and certainly no more so for a 100% gain for the same coverage and transmission levels in real world cases.

If we run a more aggressive DVB-T2 configuration, ie more prone to errors at the fringe, then indeed we can achieve higher capacities. However, the same can be said for DVB-T. Australian broadcasters typically operate with a configuration that yields about 23Mbit/sec. Sacrificing coverage this could be raised to about 28Mbit/sec.

Unfortunately the fog horn has started sounding in yet another thread.

*edit to fix formatting, not perform a fog horn

Edited September 13, 2014 by CaptainObvious

[Guest Malich]

We currently use DVB-T and DVB-T2 doubles the maximum data rate available from a TV transmission channel.

It's got me buggered why Alan keeps up this charade.

It's not like the theoretical improvement isn't impressive enough - by any fair measure, a 30-66% higher data rate for equivalent parameters is pretty damn good.

It's not as though multiple trials over the years haven't shown those theoretical numbers to be possible in practice - 67% is the best I'm aware of (in an SFN reconfigured to take full advantage of DVB-T2).

It's not even as though multiple real-world rollouts (including direct changeovers from DVB-T to DVB-T2) haven't matched the promise or the trials - a 40-50% increase in data rate for the same signal quality and coverage area as DVB-T is fairly common, with others getting up to 60-65% or more.

It's just that Alan Hughes can never admit he's wrong.

What's worse, rather than just shutting up, he tries to double-down by repeating the same rubbish over and over again. You can't stop him. Ask him to back up his claims, and he'll spout the same rubbish back at you. Ask him for solid evidence, and he'll tell you he doesn't need to provide evidence. Try to explain the truth yourself, and he'll demand evidence*. Go through the motions a few more times, and he starts trying to belittle you.

Alan Hughes is one hell of a bad example for someone who claims many years of industry experience at our national broadcaster, to have been a technical trainer, and who now claims to be a "technical author and broadcast industry consultant"...

(* Everything needed to support the figures I've given in this post can be found in ETSI EN 300 744, ETSI EN 302 755, ETSI TS 102 831, and many other related / cited documents.

Now let's see if Alan can provide something - anything reliable at all - to support his claim that "DVB-T2 doubles the maximum data rate"...)

(edit: forum died while posting; edited to fix partial post.)

Edited September 13, 2014 by Malich

[GoForMoe]

Allow me to cite the ACMA's own position on this - from 2012.

DVB-T2 is an emerging platform standard, and is the second generation of the DVB-T terrestrial television system. Many late adopter countries (countries that did not make the early move to DVB-T) are now choosing DVB-T2 for terrestrial services. DVB-T2 can offer substantial transmission efficiencies of 30–50 per cent in its use of spectrum compared to DVB-T. The current most commonly used operating capacity of a DVB-T system in Australia is around 23 Mbit/s. Operating in enhanced modes, DVB-T2 can be used to provide data rates of around 38.5 Mbit/s with a signal of very similar robustness...

So basically - if you take the 'very similar robustness' approach - DVB-T2 delivers 115.5Mbps in 3 channels, compared to DVB-T delivering 115Mbps in 5 channels. Which is pretty much exactly what would be the ideal situation for a restack per what I suggested in the other thread, restacking the existing A-E blocks to be 6/7/8, 10/11/12, 28/29/30, 31/32/33, 34/35/36 - which then frees up the entire 600MHz band.

As the Government has flagged making the 6th channel available for assisting the transition - that makes it simple enough - start up a single DVB-T2 multiplex, you could conceivably carry all 5 main channels in HD on it. Otherwise, the Government could push the ABC and SBS to go SD only on one of their multiplexes and use the other for a trial multiplex.

As I'm sure Alan would be quick to note in all other situations - Italy and Germany have already announced plans for a DVB-T2 with HEVC rollout - with the former dumping a DVB-T with MPEG-4 service to do so - is Australia so backwards that we would transition to something that is already being transitioned away from?

Once we do the second restack - the third 'restack' is dumping terrestrial television - maybe a legacy single channel service on VHF at best. There would be no room to manoeuvre - we couldn't set up temporary channels for new broadcast standards, and we couldn't reshuffle the existing services onto fewer multiplexes without significant quality reductions.

[digitalj]

The ultimate restack would be to have the FTA networks signals sent over future FTTP networks separately from internet data in the same way as TV is carried over current HFC networks without touching the internet. People then simply tune into this IPTV service on their TV's.

Remote communities would get their signals from satellite until such a time as a FTTP rollout is feasible for those communities.

[MLXXX]

We currently use DVB-T and DVB-T2 doubles the maximum data rate available from a TV transmission channel.

Alanh, here is an opportunity for you to withdraw/clarify the above incorrect statement. I am cataloguing this matter as alanh reviewopportunity 014. I would refer you in particular to posts #11, #12, and #13, above.

[timj]

As the Government has flagged making the 6th channel available for assisting the transition - that makes it simple enough - start up a single DVB-T2 multiplex, you could conceivably carry all 5 main channels in HD on it. Otherwise, the Government could push the ABC and SBS to go SD only on one of their multiplexes and use the other for a trial multiplex.

In the UK a single multiplex is used for HD transmission (from different broadcasters) using DVB-T2. The SD transmissions are DVB-T.

It can be done.

[Soap]

In the UK a single multiplex is used for HD transmission (from different broadcasters) using DVB-T2. The SD transmissions are DVB-T.

It can be done.

And I honestly hope it is done in Australia. C'mon Malcolm, you can do it. And don't even think about allocating any shopping channels into that multiplex.

[CaptainObvious]

In the UK a single multiplex is used for HD transmission (from different broadcasters) using DVB-T2. The SD transmissions are DVB-T.

It can be done.

Techincally yes, but in the Australian context there are many technical hurdles and let us not forget the financial side of the equation either.

Australia has metropolitan commercial broadcasters, 7 9 and 10, which are not necessarily the same company, and regional afiliates that may carry content sourced and time shifted from the three metro commercial broadcasters as well as their own, admittedly limited amount of, internally played out content. Each broadcaster has its own playout centre, possibly including hubbed regional playout centres fed from central playout.

Whilst it is conceivable that the metropolitan broadcasters could sort out delivering their HD service to a central, no doubt joint owned, facility to encode and stat mux the HD services into a single transport stream for delivery to a 4th broadcast network given their limited number of licence areas, imagine trying to co-ordinate that from the regional broadcaster's point of view. Their playout centres may be thousands of kilometres away from the transmission sites, and each other for that matter. Remember that at this stage each licence zone potentially has different programming, even if in most cases that is only region specific advertising so it isn't practical to supply a 'one size fits all' generic broadcast. Pulling in each and every licence zone's HD services to a central point to encode and then deliver to each transmitter would be a task of mammoth proportions, not forgetting the cost of setting up a myriad of new transmitters in the first place.

Consider that SC10 doesn't even employ stat-muxing at this point, no doubt for cost reasons.

[ron12]

I will not go into the technicalities, as there are many, but coverage for coverage DVB-T2 will provide at least 30% more capacity. There is absolutely no guarantee, promise nor any other assurance that DVB-T2 will provide a 50% capacity gain and certainly no more so for a 100% gain for the same coverage and transmission levels in real world cases.

But in both cases - H.264 vs H.262, and H.265 vs H.264 - the ITU and friends seem to have settled on the ~65%-75% figure (i.e. ~25%-35% reduction in bitrate) as a general expectation / rule of thumb.

OK, let's say that DVB-T2 provides a 30% bitrate increase over DVB-T with similar quality of reception, and let's say that H.264 requires a 25-35% lower bitrate than H.262 and that H.265 based video requires a 25-35% lower bitrate again, all while maintaining a similar quality picture.

In that case, we would only require the following percentages of current capacity:

DVB-T2+H.262- 76.9%

DVB-T2+H.264- 50%-57.7%

DVB-T2+H.265 - 32.5%-43.3%

However, the above assumes that we start off with capacity for six channels, each with the ability to transmit at 23 Mbit/s, then adjust for the change to DVB-T2, then Mpet4 or MPEG5. With the end to community broadcasting, we could assume that we only need to start with five channels, or 83.3% of curernt capacity. In this case, the percentages become:

DVB-T2+H.262- 64.1%

DVB-T2+H.264- 41.7%-48.1%

DVB-T2+H.265 - 27.1%-36.1%

So from the above, moving to DVB-T2+MPEG4 requires less than 50% of current capacity.

If we were to move to DVB-T2+MPEG5 we would have heaps of free capacity (13.9%-22.9% of current capacity).

EDIT: The above does not take into account that the reductions in bitrate required for audio and other material may be less than for video (or none at all) while the video technology is being updated.

EDIT: Corrected MPEG5 to H.265 and using H.264 and H.262.

Edited December 2, 2014 by Ron12

[Slattery]

It took 13 yrs to turn off analogue tv after digital began and we are still watching tv in SD.

it will be years before the above technologies are launched here lol

[alanh]

Ron,

There will be no more VHF/UHF spectrum available for TV in the future. So to enable improvements we should not force our existing services into a smaller bandwidth and the find we can never have a cheap way of improving TV eg UHD etc. Who wants to pay for Pay TV to provide it at a higher cost than terrestrial transmission.

You should be looking at 4K UHD TV transmissions which is one of the reasons for the development of H265 compression standard. This is achievable using DVB-T2/H265 complete with additional programs at SD but H265 compressed to fit into our 7 MHz TV channels. Remember also that the USA Free to Air TV channel is only 6 MHz wide. They transmit 1920 x 1080 x 29.94 HD and an 720 x 480 x 29.94 SD channel using ATSC and MPEG-2 compression. Their maximum data rate is 19 Mbit/s compared to our current 23 Mbit/s.

Candidate ATSC standard V2. November 2014

"Other more fundamental developments in video technology, incorporated in ATSC 2.0, significantly improve the operation of TV systems. Since the first widely-deployed video compression technology, MPEG-2, was introduced decades ago, the AVC (H.264) video codec is now widely used for high-quality high-efficiency video distribution; it is found in Internet

streaming services and also in the ATSC Mobile DTV standard. New high efficiency audio codecs are also found in the most modern video services. ATSC 2.0 is designed to be in-band backward compatible, which means that, although ATSC 2.0 services are not expected to run on current ATSC receivers, the inclusion of ATSC 2.0 services in a transmission are designed to be compatible with current ATSC receivers’ ability to receive current ATSC services in that transmission."

ATSC Candidate Standard: 3DTV Terrestrial Broadcasting, Part 4 – Service Compatible Hybrid Delivery 3D using Broadband or ATSC NRT

ATSC NRT uses the standard 6 MHz TV channel and uses a standard MPEG-2 HD signal for the main signal and an additional view using MPEG-4 compression which is added to the signal. This produces a signal all viewers can watch and those with a compatible TV can watch in 3-D. We did not do this when we trialed 3-D TV with Rugby League and World Cup Soccer.

Advanced Television Systems Committee Invites Audio System Proposals for ATSC 3.0

Next Generation of Television to Offer Expanded Audio Functionality for Wide Variety of Devices

WASHINGTON, Dec. 9, 2014 – As part of the move to a next-generation TV standard, the Advanced Television Systems Committee (ATSC) today issued a Call for Proposals for a comprehensive audio system for the emerging ATSC 3.0 standard, offering a more immersive audio experience for living room and mobile viewers. Initial responses are due in January 2015.

The audio subsystem for ATSC 3.0 is expected to provide an enhanced feature set, improving upon the capabilities of the current broadcast digital television audio system. In doing so, this new system will provide the listener with both a personalized and an immersive experience, according to ATSC President Mark Richer.

"Personalization includes enhancement to the control of dialog, use of alternate audio tracks and mixing of assistive audio services, other-language dialog, special commentary, and music and effects. Plus, the system will support both the normalization of content loudness and contouring of dynamic range, based on the specific capabilities of a user’s fixed or mobile device and its unique sound environment," Richer explained.

ATSC 3.0 audio is expected to work with home theater AV systems, with television sets (both with and without "soundbar" audio systems), and also with personal audio systems such as tablets, smartphones and other handheld devices used both with and without headphones. The level of the immersive audio experience may vary depending on the platform in use"

Version 2.0 recommends HE AAC2 audio compression to drastically reduce the data rate in current use.

https://www.dvb.org/news/dvb-t2-trial-in-the-usa

ATSC Receives 10 Initial Physical Layer Proposals for Next-Generation TV Broadcasting Technologies

ATSC 3.0 Transmission Requirements Include Higher System Capacity, Robust Performance, Improved Efficiency for Future Broadcast TV System

WASHINGTON, Aug. 27, 2013 – The Advanced Television Systems Committee (ATSC) has received 10 initial proposals from 19 organizations for the Physical Layer of the new “ATSC 3.0” broadcast television standard. With higher capacity to deliver Ultra High-Definition services, robust reception on mobile devices and improved efficiency, the new ATSC 3.0 standard is expected to redefine TV broadcasting for decades to come.

The strong response to ATSC’s call for proposals, issued on March 27, reflects a high level of enthusiasm in the industry for defining the modulation and error coding technologies in the transmission system that will provide a foundation for the next-generation terrestrial broadcast standard, according to ATSC President Mark Richer.

Last Updated on Wednesday, 17 December 2014 09:39

Alanh

[Guest Malich]

Well, that's certainly relevant...

(Looking forward to the usual AlanH Crazy Christmas Post(s). Did we get one last year, or was the 'unallocated channels / community TV' stupidity of 2012 the last time Alan gave out humour and frustration as Christmas gifts?)

[MLXXX]

Remember also that the USA Free to Air TV channel is only 6 MHz wide. They transmit 1920 x 1080 x 29.94 HD and an 720 x 480 x 29.94 SD channel using ATSC and MPEG-2 compression.

I'm not sure why alanh wished to refer to practice in the United States. However, to be precise the typical free to air frame rate in the USA is 29.97fps, in an interlaced format of 59.94 fields per second.

(A list of high definition channels in the United States can be found at http://en.wikipedia.org/wiki/High-definition_television_in_the_United_States#List_of_current_high-definition_channels . It will be noted that the video format 1280x720 p59.94 is commonly used for cable channels. This high progressive frame rate is well suited to the fast action of sport.)