Handbook for Sound Engineers

Digital Audio Interfacing and Networking 1465

Again the data is transmitted in the sequence shown
from left to right.
The parts shown as X, Y, and Z above represent the
three versions of the preamble portion of each
subframe. When version Z is used, it indicates the start
of a block of 192 frames. When version X or Z is used,
it indicates that the channel data to follow is from
channel one. When version Y is used, it indicates that
the channel data to follow is from channel two.
Blocks are used to organize the transmission of
channel status data, Fig. 39-4.

39.2.3 Channel Coding

AES3 needs to be able to be transmitted through trans-
formers. Transformers can’t pass direct current (dc).
Ordinary binary data can stay at 1 bit level for any arbi-
trary length of time, and thus by its nature can contain a
dc component. Therefore a coding scheme is needed
that eliminates this possibility.
We must also be able to recover the sampling rate
clock from the AES3 signal itself. It was desired not to
have to rely on a separate connection to carry the
sampling rate clock. Since ordinary binary can stay at a
given bit level for any arbitrary length of time, it is not
possible to extract the clock from such a signal.
It was also desired to make AES3 insensitive to
polarity reversals in the transmission media.
To meet these three requirements, all of the data
except the preambles is coded using a technique called
biphase-mark.
The binary data shown in the source coding portion
of the diagram above has the sequence 100110.
The clock marks shown are at twice the bit rate of
the binary source coding, and specify a time called the
unit interval (UI), Fig. 39-5.
The channel coded data sequence has a transition at
every boundary between bits of the original source
coding, whether or not the source coding has such a
transition. This allows extraction of the original clock
rate from the received signal since there always is a
transition at every source bit boundary.

If the source coding data is a one, the channel coding

will insert a transition in the middle of the source

coding bit time. If the source coding data is a zero, the

channel coding will not insert any additional transition.

The combination of these channel coding character-

istics provides the desired features. There is no dc

component, so the signal may be transmitted through

transformers. The sampling rate clock may be extracted

from the signal. The signal is insensitive to polarity

reversals since the source data state is carried by the

presence or absence of an additional signal transition

rather than the coded data state itself.

39.2.4 Preambles

The single portion of the subframe that is not encoded

using biphase-mark coding is the preamble. In fact the

preambles are deliberately designed to violate the

biphase-mark rules. This is done to allow easy identifi-

cation of the preamble and to avoid any possibility that

some data pattern could by chance duplicate a

preamble.

This also allows the receiver to identify the preamble

and synchronize itself to the incoming audio within one

sample period. This makes for a robust reliable trans-

mission scheme.

As mentioned in the Frame Format section above,

there are three different possible preambles. Each

preamble is sent at a clock rate equal to twice the bit

rate of the source coding. Thus the eight states of each

preamble are sent in 4 bit time slots at the beginning of

each subframe.

Figure 39-4. AES3 frame format. Note that the subframes are numbered 1 and 2, but frames are numbered starting with

frame 0.

X Channel 1 Y Channel 2 Z Channel 1 Y Channel 2 X Channel 1 Y Channel 2

Subframe

1

Subframe

2

Frame 191 Frame 0

Start of block

Frame 1

Figure 39-5. AES3 channel coding. The time between clock

pulses is called the unit interval (UI).

1

0

1

0

Clock

(2 times bit rate)

Source coding

Channel coding

(biphase mark)