Analysis: DVB-T2 Australian Television Networks Test Transmission

As I’ve been quite busy lately with work and job hunting, I really haven’t had much time to watch TV or look at the signals on the air. In fact, I’ve not had enough time even to look at Twitter, not that I really use it much. Imagine my surprise when a message from @Shaun_R was waiting for me from 20th May, letting me know about a test transmission in Sydney with HEVC-encoded 1080p. I had to do some analysis – even if I’m now at a location with marginal signals from Sydney’s main transmitters at Artarmon/Gore Hill.

The Test

Looking into the past, it seems that the test itself was publicly announced on 21st March 2018 by Broadcast Australia in an article about partnering with Free TV to trial DVB-T2 technologies for eventual replacement of DVB-T and to bring 4K TV to consumers. At the time, it acknowledged the existence of lab tests in Chatswood with transmission from three sites from April to June with further information available on the Broadcast Australia website.

However, despite this, there hasn’t really been much fanfare or more information made available, possibly to avoid any needless enquiries from the general public. On 24th April, tvcl made a post on MediaSpy forums with the footage from the test transmission. I wasn’t aware of this until it was bought to my attention and began my hunt for the signal on 22nd May.

Knowing what I know now, it seems that the transmissions were licensed under the name of Free TV Australia Ltd. They have two licenses (10424871/1 and 10402649/1) covering three sites for scientific purposes, expiring 30th June 2018. The information tables from the license images are reproduced below, as I know ACMA tends to remove such information upon license expiry and it could be of academic interest in the future.

The first license covers two sites on frequency 536.500Mhz – one at Kings Cross and the other in Manly. The reported EIRP is 1.07kW and 1.64kW respectively which appears consistent with the powers used normally (EIRP 1066W at Kings Cross for 529.500Mhz, EIRP 1640W at Manly for 529.500Mhz). Because of the use of very much similar powers and geographically close location, I suspect these two sites have been chosen to test real-life single-frequency network (SFN) performance of DVB-T2 signals.

The second license covers the main transmitter on VHF at 212.500Mhz which should cover most of the Sydney basin from Gore Hill. It has a claimed EIRP of 82kW, identical to the EIRP for ABC on 226.500Mhz from the same location. This particular transmission in VHF probably will assess how well DVB-T2 signals cover wide areas and cope with propagation conditions in Sydney. This one would probably be quite good for drive-around tests across most of Sydney.

Both licenses seem to have been granted 16th April 2018, so transmission could have started at that time or soon afterward.

The Signals on the Air

In my present location, I get best service from Sydney North West over UHF, which is not part of the DVB-T2 trial sites (as above). As a result, I needed to get a signal from Gore Hill. The roof-top aerial hasn’t been in the best shape and as houses have cropped up on the side of the hill, they’ve obstructed our line-of-sight resulting in much diminished signal levels.

I started with a VHF/UHF combination log-periodic antenna propped above the highest cabinet in the upper storey, pointed upward to catch signal diffraction around the houses at the top of the hill. It was connected to a 19dB Kingray masthead amplifier straight into my TBS 5220SE running CrazyScan2.

In the present location, I can receive Gore Hill/Artarmon with rather limited SNR (20-28dB normally, varying quite frequently). I also receive Sydney South-West and Sydney North-West fill-in signals on UHF (from off-axis in the above scan). The only DVB-T2 carrier is the VHF one at 212.500Mhz, as expected, as the local fill-in at Kings Cross and Manly could not be expected to reach out to the west of Sydney. After some testing, I found my roof-top antenna with masthead amplifier had about 3dB more signal than my “antenna of desperation” above, so I swapped over for the remainder of the experiments.

I observed the signal on the air from 22/05/2018 through to 29/05/2018 to before writing this article. At this time, the test transmissions still have about one month remaining on-air.

Three modes of broadcast were identified in the trial – the transmitter cycles between these modes at semi-random times, probably under the command of testing engineers. Provided you have a DVB-T2 capable device and a decent signal, you will be able to receive the physical layer data conveyed by the transmissions at 212.500Mhz and 536.500Mhz. However, being able to receive the physical layer transmission may not allow you to “watch” the service due to the compression formats in use (see next section).

Mode 1: 16QAM, 1/2 FEC, 16k FFT, 1/32 GI, PP6

The least dense mode is 16QAM with 1/2 FEC rate. This mode offers much more robust reception where signal quality varies and seems to me to be a good mode for mobile reception or lower power coverage of wide areas where a single channel is to be conveyed. The mode tested offered only 2.56Mbit/s payload rate, using a 16k FFT mode with 1/32 guard interval which is half that of our present DVB-T. The transmission carried a regular transport stream but utilized Physical Layer Pipe (PLP) encoding which allows for different protection modes for the services carried on the multiplex. This particular transmission uses Pilot Pattern 6.

One thing that stood out to me was the rotated nature of the constellation. If you’re used to regular DVB-T and DVB-S/S2 transmissions, you’d expect the constellations to be “dead straight” unless you had some phase non-linearities or amplitude imbalances, similar to the DVB-T transmission below. You might also expect some fixed pilots as well.

As it turns out, the rotation is a feature of DVB-T2 which improves resilience to frequency-selective effects, allowing for better decoding under lower SNR conditions when combined with the more sophisticated LDPC encoding.

The figure above was taken from a paper titled “Performance of the Rotated Constellation in DVB-T2” by Ladislav Polak and Tomas Kratochvil from Brno University of Technology which explores this effect.

Mode 2: 256QAM, 2/3 FEC, 32k FFT, 1/16 GI, PP4

The second mode is a highly-dense 256QAM with 2/3 FEC rate. This is considerably more dense than the 64QAM used with DVB-T today, however, the improved error correction and signal acquisition strategies of DVB-T2 actually put up a surprisingly competitive performance. This offered similar or better performance than DVB-T at my location where the SNR is low enough to cause both of them to fault. It seems that DVB-T2 has a much steeper “digital cliff” where the signal loses lock or the data is garbled versus being just fine. Given my observation on reception likelihood, it’s also rather interesting to see that it delivers a payload rate of 32.22Mbit/s, almost 40% more than the 23.05Mbit/s from our regular DVB-T. It does use more complex 32k FFT, but has 1/16 guard interval which should provide similar multipath resilience compared to our present broadcast DVB-T. It uses Pilot Pattern 4, which (I suspect) rotates the pilots amongst the 16 bright “spots” in the constellation below.

This mode would probably be a decent replacement for the regular DVB-T multiplex for use carrying all the services from a single broadcaster, with the additional bitrate helpful for carrying more HD services or even UHD 4k services.

Mode 3: 256QAM, 3/4 FEC, 16k FFT, 1/32 GI, PP7

The third and final mode uses 256QAM like the above, but with reduced FEC to 3/4, thus requiring higher SNR. It reduces the FFT complexity to 16k, but also reduces the guard interval to 1/32 which is half of that of regular DVB-T transmissions today. This may reduce the resiliency to multipath conditions. The benefit, however, is an increase in payload bitrate to 38.63Mbit/s – around 20% more than the mode above and a whopping 68% more than the currently used DVB-T mode. This transmission uses Pilot Pattern 7, which results in a different set of “bright” spots on the constellation. Maybe they’re deciding just how far they can push the modulation parameters while maintaining a similar level of service to the present DVB-T, or even seeing if these changes in pilot pattern affect locking performance.

This mode would be useful in much the same way as Mode 2 – for carrying all the services from a given broadcaster in a single multiplex. However, given the available bitrate and improvements in compression afforded by H.264 (AVC) and H.264 (HEVC) services, there is the potential that a single multiplex could be shared amongst broadcasters to save spectrum and transmission costs in much the same way Digital Radio (DAB+) currently operates. Somehow, I hope this doesn’t happen.

The Content

In all three modes, the DVB-T2 multiplex carried a regular Transport Stream (TS). The contents of the TS varied depending on the mode.

Mode 1 transmissions provided significantly less bitrate and accordingly carried a different version of the test payload.

According to TSReader, the transmission consisted of a single program with LCN 402 and a name of “Test Transmission”. Around 30% of the stream was null packets.

The NIT has a provider name of “Australian Television Networks” with a network ID of 8228.

Video
ID : 512 (0x200)
Menu ID : 4098 (0x1002)
Format : HEVC
Format/Info : High Efficiency Video Coding
Format profile : [email protected]@Main
Codec ID : 36
Maximum bit rate : 2 500 kb/s
Width : 960 pixels
Height : 540 pixels
Display aspect ratio : 16:9
Frame rate : 25.000 FPS
Color space : YUV
Chroma subsampling : 4:2:0
Bit depth : 8 bits
Color range : Limited
Color primaries : BT.709
Transfer characteristics : BT.709
Matrix coefficients : BT.709

Audio
ID : 144 (0x90)
Menu ID : 4098 (0x1002)
Format : AAC
Format/Info : Advanced Audio Codec
Muxing mode : LATM
Codec ID : 17
Maximum bit rate : 156 kb/s

MediaInfo reported the video stream to be encoded in 8-bit HEVC using [email protected]@Main profile. The stream has a resolution of 960 x 540 pixels at 25fps with 4:2:0 chroma subsampling in the BT.709 colour space. Audio was AAC encoded.

The above is a screenshot from the stream (click for 1:1) which shows an approximately 2m37s loop of scenes from around Shepherds Hill. The audio appears to be silent.

Mode 2 and Mode 3 transmissions were identical for content and carried a higher bitrate and resolution version of the content.

The transmission identifies itself as “Test Transmission” with an LCN of 401 (rather than 402 as above). A single service is carried in much the same way, instead, the video now occupies about 7.24Mbit/s rather than 1.55Mbit/s as above. The majority of the multiplex (77%) carries null packets.

The NIT ID remains identical.

Video
ID : 512 (0x200)
Menu ID : 4097 (0x1001)
Format : HEVC
Format/Info : High Efficiency Video Coding
Format profile : Main [email protected]@High
Codec ID : 36
Maximum bit rate : 7 500 kb/s
Width : 1 920 pixels
Height : 1 080 pixels
Display aspect ratio : 16:9
Frame rate : 50.000 FPS
Standard : Component
Color space : YUV
Chroma subsampling : 4:2:0 (Type 0)
Bit depth : 10 bits
Color range : Limited
Color primaries : BT.709
Transfer characteristics : BT.709
Matrix coefficients : BT.709

Audio
ID : 144 (0x90)
Menu ID : 4097 (0x1001)
Format : AAC
Format/Info : Advanced Audio Codec
Muxing mode : LATM
Codec ID : 17
Maximum bit rate : 156 kb/s

MediaInfo reports the stream to be encoded with 10-bit HEVC, using the MAIN [email protected]@High profile. It delivers a full-high-definition 1920 x 1080 resolution at 50 frames per second (progressive). Chroma subsampling is 4:2:0 with a colour space of BT.709 and AAC audio as before.

The above is a screenshot from the stream above (click for 1:1). The content of the broadcast is the same loop with no actual audio. Some more compressed screenshots follow below.

While watching the stream, it occurred to me that the software codec I was using does not handle corrupted HEVC streams as gracefully as I might have expected. Generally, the better MPEG error-concealment algorithms result in regions of errors staying “static” showing previous results or becoming blocky. Instead, damage to HEVC seems to render as flashing large grey segments, contrasty colourful blocky areas, etc.

On the upside, it did not crash the decoder, but this error-concealment performance may factor into whether HEVC itself is used in the future. This is especially pertinent given the patent royalties that may be involved with HEVC when similarly-performing royalty-free competitors may exist (e.g. VP9, AV1).

Conclusion

It’s rather exciting to see that, at last, Australia has some interest in DVB-T2. It’s taken a long time to move away from analog TV to DVB-T, which is unfortunate, as we have reached the limit of what DVB-T is capable of. The physical layer modulations and error correction code abilities can only deliver so much bit-rate, with most broadcasters settling for 23.05Mbit/s in a 7Mhz channel to deliver the balance of coverage versus payload.

DVB-T2 brings a number of innovations, including more dense modulations, more complex waveforms for better real-world performance and better error correction codes to squeeze out the last few bits from the air. The modes tested included a robust mode enough for a single SD channel in HEVC and two modes more suited for network multiplex carriage applications, boosting the bitrate over present DVB-T multiplexes by 40-68%. This should provide the additional bitrate necessary to carry more services or higher resolution services.

This trial also demonstrated the use of HEVC over DVB-T2 transport streams, a rather new video compression format claimed to be twice-as-efficient as AVC while being somewhat more computationally complex. This may eventually mean that multiplexes could be shared between broadcasters to save on spectrum even further. This may, sadly, be necessary especially if we are to transition to DVB-T2 over time in parallel (similarly to how the UK is doing it).

It is a shame, but despite observing for a number of days, there was no test of broadcasting 4k content. While it would only mean a small change to the input TS to carry it, it would demonstrate the full ability of DVB-T2 stream. Owing to a lack of time, if they change the broadcast content now, I probably wouldn’t notice!

The use of three transmitters seems to imply that testing was also done to see how well DVB-T2 modes handle SFN applications, which is important given the present-day fill-in transmitter frequency co-ordination.

There’s no need to worry though – while older TV sets won’t be DVB-T2 capable and some newer TV sets may be DVB-T2 capable but not HEVC capable (in the broadcasting profile used), it seems that the road to DVB-T2 is all but assured. For one, it will take close to a decade to make the change (assuming we choose to and that broadcast TV survives that long). Another is because of the complexity of managing spectrum – there aren’t that many “free” channels to run new multiplexes in parallel for a “phase-over”. Eventually, when it does happen, older sets may need DVB-T2 and HEVC capable set-top boxes to watch the new services, but there’s no point in grabbing one right now since we don’t know exactly what codecs and profiles might be adopted in the future as the standard. That being said, it seems that a PC-based tuner is probably the safest bet for now.

Bonus: Amateur DVB-T from VK2RTS

While scanning around the band looking for services, I managed to come across a transmission in the amateur radio band at 446.500Mhz.

The transmission originates from VK2RTS, an amateur digital TV repeater stationed at Lawson, NSW (in the Blue Mountains). I’ve never received an amateur digital TV broadcast and I suspect a majority of people have not either. Many TV tuners are programmed with the Australian band plan (174 – 230Mhz + 526 – 820Mhz) and would miss the amateur band entirely without using a manual entry (if it was allowed). The other reason is that it’s not particularly strong and doesn’t broadcast continually.

Regardless, it seems that VK2RTS tries its best to cover the Sydney basin by using a combination of power (40W), antenna gain (11dB), location and modulation mode. As a result, it uses QPSK which is more akin to satellite transmissions rather than higher order 16QAM or above. It also uses a very high level of forward error correction – 1/2 rate. This allows reception down into around 6dB SNR at the cost of delivering “only” 5.44Mbit/s of payload rate. This is enough for a basic SD service in MPEG-2 though.

The program identifies its name as “VK2RTS LawsonNSW”, using a Network ID of 0x5555 and a transport stream ID of 0xAAAA.

An analysis of the null packets inside the transport stream suggests they are using a MainTech GmbH DVB modulator, probably the MiniMod.

As is customary, Vivid is in Sydney at this time, so there was some imagery of this in the stream at the time I tuned in in the evening of 28th May. It seems the video source is an analog PAL source.

Later in the evening, I caught VK2ATU coming in via VK2RTS. Given the technical challenges of making it all work, it was amazing to see it on the air myself and receive a solid image despite the antenna facing Gore Hill rather than the Blue Mountains. It’s really something also to see VK2ATU come into the repeater (presumably) via S-band being received by a satellite decoder box.

From the settings it seems that the frequency tuned on the box is 11027Mhz with a symbol rate of 3000Ks/s. Assuming that the tuner is set to “universal” LNB, it would assume it is connected to an LNB with a 9750Mhz L.O., thus 11027Mhz set on the unit would correspond to tuning on the IF at 1277Mhz which sits right inside the ATV Channel 2 (1274-1292Mhz) part of the WIA Bandplan although not quite their recommended centred on 1283Mhz unless the offset is to account for inaccuracies in the receiver alignment.