IP Technology for Media Production · camera-to-the-home stovepipe ... Developed by the IETF for...
Transcript of IP Technology for Media Production · camera-to-the-home stovepipe ... Developed by the IETF for...
The Last Workflow to Join
Graphics - File
NLE, NLW - File
DVB, VOD - Packets
Ingest - Packets
IPTV - Packets
OTT - Packets
Where we have been
Color television workflows embodied a brittle,
camera-to-the-home stovepipe
Production was Directly attached to the home TV
Every business model ROI was based on 1 to many
Kanabus, et al findings
• Over 75% correct
responses when
• Visual A to Visual B
stimuli are separated
by 80ms
• Auditory to Visual
stimuli (not shown on
graph) separated by
5ms
(Inter-Stimulus-Interval)
75%
Correct
Responses
80 ms
Mike Cronk/Chuck Meyer, Grass Valley – VSF October, 2016
Where we are Now
Digital television ( HD ) broke the strict connection
between the camera and the home
The requirement for nano-seconds is gone
Video production became line based
IP – Faster than real time –
Many new ROI models are proposed, leveraging IP
technology
Workflow Timing Model
SDI Still Used Latency Stratum Network
Latency Human Factor
Absolute Real-
Time ~10 us Imperceptible
Pseudo Real-
Time < 1 ms
Edge of
Perceptible
Fast Non-Real-
Time 100ms Noticeable
Non-Real-Time 1s PVR
Workflow Timing Model
Latency Stratum Network
Latency Workflow
Absolute Real-
Time ~10 us Computation
Pseudo Real-
Time < 1 ms Live Production
Fast Non-Real-
Time 100ms
Near On-
Line/Streaming
Non-Real-Time 1s File Based
SDI Still Used
Time Relationships
Sampling time
Processing Samples and Pixels
Homogeneous
Media Time
Relative relationship between media
Audio->Image, 3D
Lip Sync
Time of Day
Live Production Technology Inflection Point
Wire Speed Routing
$ per signal per physical network
segment
Transparent, low latency CODEC
Wire Speed Latency Update
Network Speed Delay Lines 2014 2017 2020
2K/60 4K/60 4k/120 8K/120
Line Time usec 14.8 7.4 3.7 1.9
1 GbE 37 2.5 5 10 20
10 GbE 3.7 0.25 0.5 1 2
25 GbE 1.5 0.1 0.2 0.4 0.8
40 GbE 0.9 0.06 0.12 0.24 0.5
Technology is simply a tool
"I suppose it is tempting, if the only
tool you have is a hammer, to treat
everything as if it were a nail."
The Law of the Instrument. Abraham
Maslow, 1966
Why IP?
IP is Internet Protocol, it is just a tool
Are we trying to replace SDI with IP?
Yes, but not quite……
We are trying to leverage Internet Technology
for new business models which effectively
compete in the modern connected world
OSI Protocol Stack
L7 = SMPTE 274
L6 = 2022-6 Header
L5 = RTP ( for example )
L4 = UDP ( for example )
L3 = Internet Protocol
L2 = Ethernet
L1 = Copper, Fiber, etc.
The argument for OSI
Audio, Video, metadata and time are all data
Strict compliance to the OSI stack ensures that
ALL media can transit over ANY physical layer
WiFi
802.XYZ ( copper, fiber )
SONET, OTN
DVB – XYZ 2
ATSC 3.0
Virtualization and Abstraction
Layers 3 and 4 isolate layer 1 from layer 7
The application does not depend on the physical layer
The infrastructure is abstracted
A property of the network is bandwidth
The application may run multiple instances provided
there is sufficient bandwidth
The infrastructure enables virtualization of applications
Content is Production’s end goal
Neither the fastener nor its tool are as
important as what is being built
New Content Factories must exploit Internet
Technology to simultaneously provide
immediacy and choice
News
Episodic
Documentary
Long Segment
Sports
Reference Architecture Model
JT-NM RA 1.0 Published at IBC
2015
http://www.jt-nm.org/
Cloud Architecture
JT-NM RA 1.0 Published at IBC
2015
http://www.jt-nm.org/
Client
SaaS
PaaS
IaaS
Production Video over IP
Primary Requirements for the Value Proposition COTS router optimized
Low latency, no dropped packets
Compressed or uncompressed
Time aware
Manage bundles easily with session protocols
– Everything is in-band
Technical metadata
It is all just data
Return
RTP
Real-time Transport Protocol
Developed by the IETF for real time transport of audio and video
data.
Developed in the full context of the OSI protocol stack
First published as RFC 1889 in 1996
Same age, but ATM, while deployed, frequently with SONET, is
far outstripped by the combination of IP and Ethernet today even
in carrier networks.
IP and Ethernet now map directly, based on G.709 and OTN,
(Optical Transport Network) to replace ATM
RTP
As Internet Network Protocol and Ethernet data link layer fabrics
became dominant, and bandwidth increased, it became more
efficient to map video directly into the RTP payload without the
additional overhead of TS.
RTP/UDP is 50% lighter than TS
Mappings for RTP into virtually any other Network Protocol
already exist.
Provides the ability to abstract the network for modern cloud
architectures
Transition Comments
IP is seen as a business enabler for live production:
SaaS and Client applications have the most leverage
IaaS should be hidden, or transparent, yet easily scaled
Based on contemporary OSI stack
Ethernet, IP, RTP
BW gains and cost declines based on Moore’s Law, or Gartner Curve
– More affordable distributed systems
– Ever decreasing latency
44
Open Standards for Interoperability
AIMS
Active in both marketing and technical efforts
JT-NM, SMPTE 2110, VSF and AMWA
Actively contributing to the standards, specifications and engineering
guidelines which blaze the trail of the AIMS roadmap
Active with the JT-NM which established the vision
Actively orchestrating industry aligned interoperability events for
the suite of technologies
45
SMPTE - Transport
SMPTE 2022-1 to -7 all originally intended for transport and
contribution
2022-6
RTP based transport
Includes VANC, HANC and video
All current IP full bandwidth video deployments use 2022-6
Pro: It is a field provide, interoperable standard
Con: It has the same complexity as SDI. Embedding audio and other data
makes signal processing awkward.
46
SMPTE – Transport
2022-5
FEC ( Forward Error Correction ) for SMPTE 2022-6
Protect a single network segment
Robust recovery for long, continuous data drop outs
Downside is delays exceeding 20 msec, based on protection time span
Excessive delay is detrimental to live production
Delay cannot be undone
Delay all signals to the time of the last arriving signal
Push late arriving signals to the next arrival interval
47
SMPTE – 2022-7
RTP based datagram hitless switching
Applies to all RTP transports
– 2022-6, RFC-4175, AES-67, etc.
Differential path delay must be managed
Class 1 defines <= 10 msec
This delay is additive
– Caution with multi-pass
Grass Valley will propose a 100 usec Class to SMPTE for
consideration in the upcoming upgrade to 2022-7
48
SMPTE 2059 – 1/2
Interoperability testing is increasing in frequency and expanding
in product scope.
Recently, AES and SMPTE Interop was tested
AES-R16-2016
Tables 5.3.1 and 5.3.2 define the necessary parameters
5.3.2 is likely more suited to our industry needs
– Excludes the IEEE 1588 Default
49
AES-67
This is the linear PCM audio replacement
AES-3, AES-10 ( MADI )
The rate of adoption is increasing
Current interoperability is based on 1 msec and 8 (10) sample packet
construction
AIMS ( Grass Valley ) proposed adopting the 125 usec operating point with
up to 80 channels as a second operating point
– Lower latency
– Higher Channel count, 16 up to 80
– Under consideration within SMPTE 2110
50
VSF TR-04
SMPTE 2022-6
Video, Audio and VANC/HANC meta data
AES-67 for Audio
8 channels, likely more to come
AES-R16 for Interoperability
SMPTE 2059 -1/2 for Timing
Table 5.2.3 of AES-R16 defines the mutual operating point
Increased number of sync messages
51
VSF TR-03
Essence based transport
RFC-4175 transport construction for video
– Referred to the correct SMPTE standards for video formats
AES-67 for Audio
SMPTE 2059 -1/2 for Timing
– AES-R16 for interop
IETF – ANC_291 meta data
– Encapsulate SMPTE 291 meta data into RTP
– On Track for publication this September
52
VSF TR-03 Essence based transport
All essence packetization is RTP
Each essence type can be optimized
To use 1990 terminology
There is a layer for each signal type
To use 2016 terminology
All layers can traverse the facility in one common fabric
Much easier to parse and process then embedded SDI
53
SMPTE 2110
A standard designed for multiple parts
-10 is the System
-20 is baseband video
-30 is PCM audio
-40 is ancillary data
-50 is 2022-6 as an essence
More to come
54
SMPTE 2110 - 10
The System
The Network basics
– RTP, IP, UDP, etc
System timing
– PTP, SMPTE 2059, RTP
– Reference to AES-R16 place holder
SDP
– Use of session description protocol
55
SMPTE 2110 - 20
Professional video version of RFC 4175
Clarifications of video formats
Clarifications of video sampling structures
Latency considerations
Payload construction considerations
Sources of Latency (2 of 6)
• Redundancy (2022-7) Buffering: R x nR • Class A/Low Skew Path Differential <= 10ms
– per SMPTE ST 2022-7:2013
Mike Cronk/Chuck Meyer, Grass Valley – VSF October, 2016
57
SMPTE 2110 - 30
PCM Audio
There is also a proposal under consideration for wrapping
AES-3 into RTP for non-PCM audio
4 Levels of compliance
– Latency and channel count
– Lower latency level can have higher channel count
– Covers the AES-67 Table 3 options
Preferred audio channel order mapping
Use of SDP
– Alignment with AES-R16
58
SMPTE 2110 - 40
Ancillary data essence
Just started
Based on
https://datatracker.ietf.org/doc/draft-ietf-payload-rtp-
ancillary/ IETF ANC 291
59
SMPTE 2110 - 50
SMPTE 2022-6 as essence
Bridge for 2022-6 into 2110
SMPTE 2022-5
Look for frame aligned FEC
There is also focus on 2022-7 to ensure 2110
essence definitions do not break it
Short delay required
Audio systems, Ravenna and Dante, rely on very short
delay RTP datagram protection
60
AMWA NMOS
Networked Media Open Specifications
Establish the necessary criteria for PaaS
Essential to virtualizing the network infrastructure ( IaaS )
Specification IS-04
Discovery and Registration
Utilizes classic Internet Identity methods
Demonstrated ability to operate as peer-to-peer, yet
scale up to LLDP size
Test on modest size systems, now increasing the scale
63
AMWA NMOS
Finished review of RA 1.0 with focus on control
Continue Development of Proposed methods and
stack build up
65
Why?
Create an abstraction for media construction
Encoded
Object
Computer generated
Create an extensible abstraction for how media
is carried
ATM, SONET, Ethernet
66
AMWA NMOS
Continue Efforts in Connection Management
Once the Connection is made, Capability and Capacity for
interoperability must be shared
Clean On-Air Switch in IP
Clean Switch
Edge Switch
End-Point Switch
IGMP Leave and Join
Program
Pre-set
100+ Inputs
69
AMWA NMOS
New Developments for Network API
Northbound interface for the COTS Router
Broadcaster OF
3 Possible Strategies to Switch
Source Changing source ports, or IP addresses is new control paradigm
In the fabric
Mimics broadcast. Solutions will be coming to market.
End-Point
The easiest to implement
Replicates current up-stream pre-select model
72
Library of References
Reference Architecture
http://jt-nm.org/RA-1.0/index.shtml
Videos Services Forum TR-03 and TR-04 http://www.videoservicesforum.org/download/technical_recommendations/VSF_TR-
03_DRAFT_2015-10-19.pdf
http://www.videoservicesforum.org/download/technical_recommendations/VSF_TR-
04_2015-11-12.pdf
AMWA NMOS ( Networked Media Open Specifications ) IS-04
https://github.com/AMWA-TV/nmos-discovery-registration
73
Library of References
SMPTE
http://www.smpte.org
AES
http://www.aes.org