This AVB Bandwidth Calculator is provided by Jeff Koftinoff <jeff.koftinoff@gmail.com>
See avb.statusbar.com for more information
The bandwidth allocation for AVB (AVB BW) defaults to 75%. At this setting the most bandwidth that AVB Time Sensitive Streams will take is 75% of the available bandwidth of each network link.
There are two audio stream formats used by AVB: AM824 and AAF.
AM824 supports the transport of 24 bit linear audio, iec60958 encoded audio (SPDIF and AES3), SMPTE Time Code, and MIDI.
AM824 has three options for the talker: "non-blocking (sync)", "non-blocking (aync)", and "blocking."
Regardless of the talker's AM824 packetization, an AVB listener is required to receive frames using any of the three options: AM824 non-blocking synchronous, AM824 non-blocking asynchronous, and AM824 blocking mode.
The typical AVB audio device transmits using “AM824 non-blocking, synchronous” mode. Non-Blocking means it sends one frame every observation period (125µs). Synchronous means it always sends the same number of samples per ethernet frame (6 @ 48 kHz, 12 @ 96 kHz).
“AM824 non-blocking, asynchronous” mode means that the transmitter may send an occasional ethernet frame with one extra or one less sample due to the packetizer being asynchronous to the transmitter's observation interval. Devices that handle more than one clock domain but have one packetizer typically use this format. Because it can send the occasional ethernet frame with 7 samples @ 48 kHz or 13 samples @ 96 kHz, it needs to reserve enough bandwidth to allow it to do so. The Apple Mac transmits AM824 non-blocking, asynchronous mode.
“AM824 blocking mode” is a mode that some firewire devices use, as it is easier to packetize and depacketize. At 48 kHz, each ethernet frame always contains 8 samples. at 96 kHz, each ethernet frame always contains 16 samples. The avtp_timestamp is always for the first sample in the ethernet frame and there is always an avtp_timestamp. Because it is sending this frame every 125µs, there are some times when it doesn’t need to send any ethernet frame at all. But when it does, it is always the same size packet. The drawback for this format is that you have to reserve bandwidth for the 8 sample frame @ 48 kHz or 16 sample frame at 96 kHz.
“AVTP Audio Format” (AAF) is a new packetization format that is defined in IEEE p1722a. It has lower overhead than AM824, requires each ethernet frame in a stream to be the same size and format, and allows 16 bits, 24 bits at 32 bits per sample, as well as a choice on the number of samples per ethernet frame. Every ethernet frame is always the same size and format.
All AVB stream formats can support 32 kHz, 44.1 kHz, 48 kHz, 88.2 kHz, 96 kHz, 176.4 kHz, and 192 kHz. AAF can also support NTSC pulled rates (* 1.001 or * 1/1.001).
AM824 formats only support 24 bits per sample. AAF formats allow packed 16 bits, 24 bits, and 32 bits per sample. For Floating point stream formats, choose 32 bits per sample.
AM824 packetization allows one ethernet frame per stream per observation interval (125µs). AAF format supports multiple ethernet frames per observation interval, allowing for more channels per stream.
The channel count.
IEEE p1722a supports AES-SIV encrypted AVB streams. When the ethernet frame is encrypted this adds some overhead to the ethernet frame.