Signal distribution: The costs of compression

Corporate networks are struggling to cope with the demands of video traffic. Should you plump for high quality or low latency? 1GB or 10GB? Steve Montgomery explores the options.

Distribution  of  multichannel AV streams over IP networks is  becoming commonplace, however  limitations  in network  bandwidth  place restrictions  on  the  amount  of  video  that  any network can carry.

However, recent research report from Netgear found that most corporate businesses are rapidly migrating from 1GB to 10GB Ethernet networks. This is good news for vendors of AV equipment, allowing them to seize opportunities to provide equipment  to  distribute  audio  and  video  over these networks.

One  of  the  main  user  requirements  is  to provide  several  video  channels  at  once,  and this requires a considerable amount of available network bandwidth.

Compression technology is used  to  reduce  the  data  size  of  those  streams and squeeze more channels onto a network. 1GB and 10GB networks cannot carry multiple video streams without some form of compression.

However  compression  comes  at  a  cost in  latency  and  image  quality.  A  range  of
compression  techniques  has  been  developed over the years allowing users to select a specific type that best meets their requirements.

Samuel  Recine,  director  of  sales,  Americas and Asia Pacific at Matrox, believes that image quality of very low bitrate streams is good enough for most corporate applications: “Crestron and other brands manufacture exceptionally balanced products using JPEG 2000 compression. Other companies,  like  Matrox  have  achieved  4Kp60
4:4:4 quality and latency levels using H.264 that are good at bitrates even well below the 1 GB/s requirements of JPEG 2000. The majority of AV-over-IP applications can tolerate a small amount of latency in exchange for massive networks, a high degree of interoperability, ease of network set-up and maintenance and scalability. This is
the  crux  of  the  AV-over-IP  market.  No  codec competes with H.264 for the balance of quality, latency, bitrate and interoperability with millions of other devices.

“In general the relationship with latency also has to do with the quality and sophistication of coupling the transmitter products with receiver products  in  AV-over-IP  versus  not  just  the choice of codec.”

Compression  inevitably  delays  the  signal. However some manufacturers claim zero latency. Recine  believes  this  is  a  misunderstanding: “When  vendors  refer  to  zero  latency,  they  are most  often  referring  to  latency  levels  not detectable  in  the  application.  Most  end  users are  far  more  concerned  about  deploying  real standards and designing AV networks that are interoperable and scalable versus latency levels between excellent and better than excellent.”

As  Rob  Muddiman,  EMEA  sales  director, Zeevee, points out: “It is incredible how some vendors  claim  zero  latency  when  using  a JPEG2000 type compression, when in reality it introduces  20-30mS  latency.  ZeeVee  has  both compressed  and  uncompressed  solutions.    Our compressed  solution  uses  JPEG2000  and  has
only a 20mS latency.”

Justin Kennington, president, SDVoE Alliance, concurs:  “Companies  cannot  realistically  claim zero  latency  on  compressed  signals  and  quite often their technical publications indicate what their marketing fails to. High compression of the
type required to put 4K video onto a 1G network will  always  demand  compromises  in  both latency and image quality.  These technologies need  to  use  a  whole  frame  of  video  to  make a  reasonably  good  image  quality  in  the  low bandwidth environment.  Just the act of waiting on  a  whole  frame  (16  milliseconds  at  60  Hz)
takes time and in all cases the system delay is at least two frames.”

Crestron’s  DigitalMedia  NVX  product  is marketed as having no latency penalty, although this translates to an actual end-to-end latency of  30ms,  as  Rob  Carter,  technology  manager, DigitalMedia, Crestron explains: “When we talk about no latency we are referring to our NVX technology  that  combines  codec  and  scaling latencies. As a result, NVX has the same latency performance as our HDBaseT scaling products.

Some  companies  don’t  include  scaling,  which typically adds a minimum of 30 ms on top of the  quoted  numbers.  And,  according  to  our internal tests, they can be significantly higher.”

Given that compression is needed to distribute multiple  video  streams  over  a  single  network, the choice of compression technology is crucial.

“Choosing  the  right  one  is  critical,  since  it has a direct impact on end users,” says Carter. “JPEG  2000  can  offer  amazing  quality  with latency  low  enough  to  fit  within  our  scaler latencies.  Unfortunately  not  all  JPEG2000 implementations are created equal, so there are other  network  video  products  on  the  market
that  use  it  but  still  have  quality  and  latency problems. VC2 is another codec out there but it sacrifices quality for latency.”

Commercial  organisations  are  continually looking  for  ways  to  cut  expenditure.  Network infrastructure  is  a  primary  target,  particularly with the high cost of cable installation around buildings. An attractive option is to combine AV and data networks into one, effectively piggy-backing the video distribution onto an existing IP  data  network.  However,  this  is  not  always feasible. Muddiman: “Some vendors claim that
they work on a 1GB network, but in reality, they take  up  so  much  bandwidth  they  effectively have to sit on a dedicated stand-alone network.  A product that takes 800MB operating on a 1GB link leaves very little for IT traffic.  Our advice is, if you are putting in a separate network, do it properly and install 10GB.”

Carter agrees: “While true convergence makes the most efficient use of network resources, there are still significant advantages to network video even if you set up a separate network. Flexibility in design is probably the most important feature of network video. Network ports don’t need to be video inputs or outputs; they can be inputs or outputs or control systems or touch panels.

Given the port density of the network switches, the  system  can  be  easily  expanded  as  needs change  with  time.  Even  if  a  separate  network switch is used, users will still often want to take advantage of network runs that are already in the walls. Gigabit network video solutions can re-use  Cat5e  that’s  already  in  the  walls.  10G
Ethernet  requires  more  expensive  Cat6a  cable that’s not as readily available.”

“Running  networked  AV  systems  over the  same  wiring  and  IT  infrastructure  as
the  corporate  IT  networks  can  contribute to  reducing  the  installation  costs,”  explains Bertalan  Petrik,  sales  operation  manager, Lightware Visual Engineering. He does, however, point  to  a  potential  drawback:  “The  optimal Ethernet  switches  for  AV  over  IP  are  different from  standard,  corporate  network  switches, which could deteriorate the performance of an AV over IP system. On the other hand, using low
latency switches for emails would probably be overkill. To eliminate network security problems the separate AV network can be isolated from the corporate network, just like in the case of conventional  matrix-based  AV  systems.  Also, most  AV  over  IP  products  offer  multicasting, which is a problem on the corporate network.

Separating  them  can  solve  these  issues  too.  A separate  AV  network  enables  the  performance of the AV over IP products to be guaranteed.”

Martin  Bonsoir,  international  applications engineering  manager  for  Biamp  highlights another  reason  that  networks  are  kept  apart. “Organisations  separate  their  networks  for several  reasons,  although  cost-saving  is  rarely one of them. Arguably, the overall project has a higher cost when deploying two networks. But since those deployments are almost always done by different departments, each usually ends up
with  a  lower  cost  for  their  part  of  the  job.  It helps to step back and look at the overall cost.”

Once  the  AV  distribution  network  has  been installed,  control  and  management  generally becomes  the  responsibility  of  the  end  user.  

Manufacturers will provide a suite of software tools  to  set  up  and  configure  the  system,  but most  users  will  require  a  dedicated  control interface.  This  comes  at  a  cost.  Kennington says it’s always been this way: “In the past, the manufacturer of a matrix switch would provide the  software  necessary  to  commission  and operate the switch for free.  Yet no one ever put this software in front of the end user as the final
solution. Instead, a professional AV programmer was  brought  in  to  build  a  user  experience  on top of the control system.  Although billed as services, this person was really providing custom software for hire.  

“In the AV over IP world, the model is just the same.  If I buy encoders and decoders, I expect there  to  be  some  configurability  included  for free.  But I still expect that I will pay extra for software tailored to my needs.  This might come from a commercial software vendor, building on top of the API, or a traditional AV programmer, creating control software for a specific enterprise or other use case.”

With the wealth of products and technologies available, integrators need to consider the best solution for each project. These considerations must include the requirement for combination of  the  IT  and  AV  networks,  selection  of compression technology and whether to source components for single, or multiple, vendors.

It  is  sensible  for  organisations  to  plan  the network  infrastructure  with  a  view  on  future upgrades says Paul Harris of Aurora Multimedia

“Installing Cat6a cabling at the start saves cost as it will cost more later to pull new wire versus the cost difference. Fibre is a better choice as multimode will go about 1000ft (about 300m)and  at  40G  about  330ft.  To  be  8k60Hz  or 4K120Hz  ready  you  will  need  at  least  a  10G network. 40G will allow for minimal compression
of 8K when the time comes.”

Article Categories