Ethernet, ATSC and MPEG-2

So last night we switched things up and debated technology stuff instead of ethics, politics and ideological systems.  Oddly even though this was an area in which the amount of knowledge around the table is significantly greater we flailed with the same random intensity as any other debate we have.   Funny.

First we talked about digital tv bandwidth, compression and problems.  This is a subject I’m quite interested in as it forms the basis of my current job, but it is one that I am just learning.  I didn’t provide information as authoritatively as I should have – so I did some research this morning…

(First protocols and standards – I’m using the words interchangeably although there is some difference.  Both imply a set of rules used to achieve a reproducible set of outcomes.  They are standardized by various bodies like IEEE, ITU, MPEG and the IETF.  The standards set by the IETF are known as protocols and define most of the Internet standards.  TCP would be an Internet protocols.  But Ethernet is an IEEE standard (802.3).  The main difference is the standards body that ratifies and keeps them.)

DTT – or digital terrestrial television is the signals that got out over the air.  Analog signals are governed by a standard called NTSC in North America (different elsewhere in the world).  That has been replaced with digital signals in the States and will be replaced in major urban centres in Canada this coming August.  The digital standard is ATSC.

Raw video data is not transmitted in ATSC.  It transmits encoded MPEG-2 transports streams.  Raw video would be about 1 Gb/s.  The max bitrate for ATSC encode MPEG-2 streams is 19.6 Mb/s.  Producers commonly encode their HD broadcasts between 15 and 18 Mb/s.

MPEG-2 is the same protocol used to encode video on DVDs.  It is also what is used to encode the Shaw cable cable signals.   A normal SD stream is around 7 mb/s.  HD is encoded at the same levels over cable as for DTT so between 15 mb/s and 18 Mb/s.  Some cable companies (Shaw among them) further compress the MPEG-2 stream received from the broadcasters to save bandwidth.This can mean an HD stream is carried in 12 – 13 Mb/s. (QAM or PSK is the equivalent of ATSC that dictates how the encoded stream moves through the media (ATSC= air, QAM/PSK = cable).

(Blueray disks don’t use MPEG-2 encoding, but MPEG-4 which offer greater compression.  IPTV systems might carry MPEG-2 streams for their SD channels and MPEG-4 streams for their HD channels – these are passed over the internet using  the RTP protocol.)

Many factors can cause artifacts in the video when watched.  The quality of a picture depends on many more factors than the final bitrate of the encoded stream.  Poor encoders, lossy media and decoder errors can all produce artifacting that presents as macroblocking, pixelization and other problems.

Obviously encoding an encoded stream a second time (by the broadcaster and by the cable company) is more likely to introduce such artifacts.

The second question was “What is Ethernet?”  We had a hard time giving this answer as we know quite a bit about the general subject of networking, but none of us are really Ethernet experts.  We got bogged down in symmantic arguments about the meaning of standards vs. protocols, the OSI model (OSI is defined by yet another standards body the ISO) and others.  I think in the end we provided several good answers.  I’ll try and summarize –

Ethernet (IEEE 802.3) is a set of standards used in a LAN.  It is the most common standard that defined how Internet traffic is passed along the wires in the common home and business network.  It defined how voltage should be transceived and received along the wires connecting your computer to your local switch, router or modem.  It can run over many wires including twisted pair, coax and fiber optics, but the most common is what people call CAT-5 cable with RJ-45 connectors (often just called ethernet cable).

So the standard takes about what cables and connectors are allowed, how to pull signal on and off the wire, how to deal with noise and signal loss, etc.  This is all data that composes what is called the layer 1 physical layer of the OSI model.  Ethernet also talks about addressing for devices on your local network – so that your switch has a different address than your computer.  These are known as physical addresses or MAC addresses (MAC = Media Access control).  This is activity at the layer 2 data link layer of the OSI model.

Now Ethernet is just one of many protocols.  Other LAN protocols might be Token Ring.  Equivalent protocols for sending wireless signals would be IEEE 802.1n (1a, 1b and others too).  WAN protocols for crossing larger distances are varied and could include frame relay, ISDN (both mostly obsolete) or ADSL.   (I don’t know much about WAN protocols).   Protocols for sending data to storage arrays (Dave’s specialty) would include ESCON and FICON and Fiber Channel.  (I know nothing about these).

As the traffic moves from one wire to another it would be bridged from one of these standards to another.  The neat thing about the OSI model is that higher levels of the network stack are masked from the changes of wire beneath them.  The level about 2 is 3 (surprise) the network layer.  This is where the IETF protocol IP (Internet Protocol) operates.  So the same IP packet moves across the network from my PC to the server which might be miles and miles away and across several different underlying wire technologies (or wifi or wireless or satellite, etc).  Above IP is four more layers in the model (which is not strictly adhered to by any real life network stack, but is very useful for reference). Layer 4 would be Transport and have TCP as an example.

Both Ethernet and IP defined addresses for network objects.  MAC addresses for Ethernet and IP addresses for IP.  A MAC address is like 00:1f:5b:d8:c1:0b and is used to tell one ethernet host apart from another on the same wire.  An IP address (like is used to tell one IP host apart from another across the Internet.  Ethernet only care about getting you from one end of a wire to another.  So it is addressing specific to that purpose.  IP cares about sending packets across the Internet so it needs a higher level address for that purpose.

(Here’s a place where the OSI model fails a little – there is a protocol called ARP (Address resolution protocol).  It is used to convert IP addresses to MAC addresses. It sort of exists between Ethernet and IP.)

Now to join the two discussions – say you were watching an HD IPTV show on your XBOX.  The full protocol stack as it crosses the wire between your Xbox and switch could be:

Ethernet-> multicast IP->UDP (an equivalent to TCP)-> RTP (real time protocol used for streaming video/audio of TCP/IP networks) -> Encryption -> MPEG-4.

Once it gets to your ADSL modem the next hop from the modem to your local DSLAM would be:

ADSL -> multicast IP->UDP (an equivalent to TCP)-> RTP (real time protocol used for streaming video/audio of TCP/IP networks) -> Encryption -> MPEG-4.

Any better?

Other protocols at the IP level would include IGMP and ICMP and IPv6.  TCP and UDP are the most common next protocols at level 4.  HTTP for web traffic is the most well known protocol at layer 7.  Others could include DHCP (dynamic IP acquisition), DNS (IP address to host name translation), SMTP (internet mail) and SSH (secure shell – used to make an interactive session or tunnel with a remote host).  Layer 7 is my balliwick although I’m not bad at layer 4 and I know layer 3 too.


6 thoughts on “Ethernet, ATSC and MPEG-2

  1. Craig says:

    “Above IP is four more layers in the model (which is not strictly adhered to by any real life network stack, but is very useful for reference).”

    IS-IS supposedly adheres to the 7layer cake, I don’t know of anyone outside of the old, big backbone providers (AT&T) that use it.

  2. Suellen says:

    Think of yourself as a well rounded geek who knows the difference between UUIDS and GUIDS. 😀

    • But not between a snapshot and a clone.

      • David Silvestri says:


        Snapshot only keeps track of delta tracks (those original protected tracks) in a separate pool.

        Clone is a full copy on completely separate disk.

        Failure of the source lun will pooch a snap. A clone is completely separate.

        thems the basics… lots of vendor specific differences… but that’s the important stuff.

        • Question I failed on my exam last week. That would have been may guess, but I wasn’t sure what the snap kept deltas from… so I started to second guess myself and decided to just leave it blank… made me mad because it was one I should have known.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s