Skip to Main Content
March 26, 2024

Leader & Grass Valley Flesh Out on JPEG XS

Leader America, in collaboration with its Phabrix business and Grass Valley made a webinar on a topic that we have been covering recently, JPEG XS used with ST 2110. You can access a recording of the one hour event here, but if you’d like to get a sense of it a little quicker, we watched it and summarized it for you.

After a brief introduction, the first speaker was Kevin Salvidge, Sales Engineering & Technical Marketing Manager and he was joined by Steve Holmes also at Leader and Kevin Hein of Grass Valley. Salvidge set out how the demand for higher quality in content is driving bandwidth requirements that are difficult to meet with SDI interfaces. However, a ‘single step’ switch from SDI to ST 2110 IP-based systems is not feasible for most broadcasters. That means that SDI and ST 2110 will need to co-exist for some time as content creation moves to 4K and then to 8K.

ST 2110-20, the first version of the standard, supported just uncompressed video, but quickly users looked for codecs. However, existing codecs used high complexity, with negative impacts on latency and device architectures. A quick audience poll on the webinar highlighted that 77% of viewers are not currently using JPEG-XS, which has been designed for low latency and device simplicity. The biggest user segment in the audience was in outside broadcast (13%) with studio production at 7.5% and delivery to cloud-based production at 2%.

How Often is JPEG XS Used?

A second question to viewers was the frequency of use of JPEG XS, with the most popular usage by those that do use the codec turning out to be ‘daily’. This surprised Holmes, who said that the widest use that he had seen was in live sports where it is an excellent choice.

Salvidge asked why the industry had stayed away from compression up to now? Holmes replied that there had been a feeling from creators of “Don’t mess with my video!”. Another factor was potential delay from the codec processes. In a live setting, Holmes said, “unpredictable or long delays can kill you”.

The multi-frame delays from typical MPEG or H.265 codecs cannot be handled in live sports but also in, for example, news programmes. There was concern also over a potential loss of quality. Holmes presented the following graphic which positions JPEG XS against distribution codecs and other intra-frame codecs such as JPEG2000.

The chart highlights that compared to intra-frame codecs, JPEG XS has many advantages while matching the quality.

The distribution codecs were optimsed for maximum compression with relatively little care about latency and some artefacts could be introduced by them. The Intra-frame codecs could deliver outstanding quality, Holmes explained, but the complexity was ‘pretty high’ and there was significant latency. JPEG XS, on the other hand has latency that is ‘phenomenally low’.

At NAB 2024, Grass Valley will show a camera with live analysis and comparison of the uncompressed SDI with compressed JPEG XS over 2110 on a single waveform monitor. The difference in timing on scene changes is very, very small, Holmes added. As Intopix explained recently to us, JPEG XS is also visually lossless and can even be mathematically lossless.

Critical to Separate Creation & Distribution

Hein said that it is critical to differentiate between creation and distribution. It used to be desirable to be uncompressed for creation but when UltraHD/4K arrived, the interfaces did not exist for cameras to output uncompressed video. Still, he said, you needed really low latency as you might have twenty cameras around a sports pitch and they need to be in sync. Once the content has been created, some latency might be acceptable, depending on the kind of production. There has been a shift after Covid to more remote production but there is not a global infrastructure for very high bandwidth transmission (although there might be in some countries). Even in studio environments, video might be uncompressed within the studio, but it has to be sent elsewhere for consumption.

JPEG XS can be used multiple times on the same content without visible artefacts (you might see them on the 10th application of the codec, Hein clarified).The incremental damage to quality was a significant reason for the reluctance to compress with earlier generations of codecs, he explained. He also said that Grass Valley has had IP support in its cameras for more than ten years, using the Intopix Tico codec, a pre-cursor to JPEG XS and using a 4:1 compression ratio. Given that JPEG XS and Tico is visually lossless at 10:1, 4:1 meant that quality was ‘absolutely fine’. Bandwidth requirements, on the other hand are drastically improved. Holmes supported Hein’s points especially the advantage of allowing multiple compression or expansion steps.

Low Complexity Means Easier Implementation

The low complexity of JPEG XS allowed Leader to add the codec to the existing FPGA in its analyzers, Holmes said. There was no need for a re-design to include the feature. Hein agreed with this and said that Grass Valley was able to add the codec to its camera without needing additional hardware. Salvidge added that the availability of JPEG XS is an enabler for moving more towards a software-based production architecture rather than needing dedicated hardware.

Salvidge said that he is often asked what level of compression should be used. Hein replied that he has the same questions. He explained that the EBU, BSkyB and others had tested the codec on UltraHD/4K and had found that 11:1 is ‘the edge’ but 10:1 is ‘perfectly fine’. A 4K 50/60 signal could come down to 0.95 to 1.1 Gbps although interlaced formats are more challenging and probably should not go beyond 6:1 to 7:1. 1080P can take up to 8:1 and even with a very large monitor, you will hardly see anything.

Degradation Means Fewer Artifacts

Unlike the block artifacts of MPEG, JPEG XS is wavelet-based, so the main degradation if you do push the compression levels is in the loss of high frequency content. You also do not get ‘cascaded errors’ as you might do on a GOP-based codec. Typical compression/decompression times are just 16 or 17 lines, which Holmes described as ‘phenomenal’.

Salvidge said that his firm had seen some issues where not all parts of a full production signal chain can handle the JPEG XS content well. Therefore his firm has made a test pattern with various levels of bits per pixel that can be used to test whether the signal is being correctly re-transmitted. It can also be used to check the quality of productions if signal paths are not good enough.

JPEG XS Standardization Helps

Hein said that JPEG2000 is still used a lot in remote content creation especially using wireless but Grass Valley actually started using Tico/JPEG XS very early but there was a lot of resistance to using any compression when SDI was still being used, although it could be used. Now things have been standardized but it is important to check technical levels (although Hein was very complimentary about the wide support in the Leader products).

He went on to say that he doesn’t like the phrase ‘cloud production’ but prefers ‘remote productiion’ and when you go that way you need compression and that can introduce latency. Hein introduced the VSF TR-07 recommendation for the transport of JPEG XS video in MPEG-2 transport streams over IP. This helps to eliminate the need for synchronization in some production schemes. It is ‘on the horizon’ and some equipment can already accept these streams. More will be revealed at NAB 2024.

Holmes explained that MPEG streams are ‘self clocking’ with everything synchronized at the receiver. The industry is very comfortable with that technology and Salvidge said that there is a lot of interest in using the TR-08 idea. Having said that, Hein said that distributed Precision Time Protocol (PTP) is no longer the challenge that it was for larger productions, but when you have more distributed productions, for example, ‘all over the world’ it can still be tricky. In that case, the self-clocking nature of MPEG stream technology is handy.

JPEG XS also has headers to specify the type of content that is being transported. That is important as that information is not in the stream itself. Data flow, color, video and other information is included to allow optimum decoding. That information helps with the ease of implementation.

Holmes pointed out that another advantage is that because of the low latency, JPEG XS streams can be used alongside uncompressed streams in 2110. In a stadium, cameras on the side of the production could be uncompressed, while those across the other side are compressed. Hein agreed that this is the ultimate goal. There had been a lot of speculation about how compression will come into 2110 over many years. Now there is a clear standard for the next number of years (with views of at least 5-10 years of life for JPEG XS among the panel).

Compression without quality loss can enable remote production.

Hein said that content creators are under huge pressure to produce more and more content at the highest quality, but often with declining budgets. Pressure comes from the intense competition. Creators need to be more agile and move from site to site very quickly. As a result, more often production is being centralized rather than being on-site in OB vans. That does mean that less equipment may be needed, so it can be of higher quality. Fewer monitors, intercoms and routers means better equipment can be purchased. Staff see fewer equipment changes so that they are more familiar with what they are using, improving productivity.

The only barrier to this centralization is the transport of the video and this is where JPEG XS can be the only solution. Hein does not see another compression technology on the horizon at the moment, if you really need low latency. He said that NAP in the Netherlands used some equipment to record ice skating in China but the whole production was actually controlled from Hilversum, near Amsterdam, to China. Bi-directional audio delay was just 200ms.

JPEG XS is a complement to ST2110-20

The last question posed to the listeners was whether after the discussion, viewers saw JPEG XS as a replacement for ST2110-20 or as complement, 58% thought that it would. be a complement, while 30% were still undecided.

The webinar finished with some audience questions and Hein used the opportunity to show the range of Grass Valley products with JPEG XS support. He made a plea for standardization and highlighted challenges with HDR, audio and color spaces where there are so many options which makes life difficult. Holmes also supports standardization to ensure interoperability.

0 0 votes
Article Rating
Subscribe
Notify of
guest
0 Comments
Inline Feedbacks
View all comments
0
Would love your thoughts, please comment.x
()
x