Datasheet

3−27
If G is set to 0 and O is set to 0, loudness compensation is disabled. If G is set to 0 and O is set to 1, the biquad-filtered
audio is directly added to the volume-level-adjusted audio. Typically, LG and LO are used to derive the desired
loudness compensation function, G is used to turn loudness compensation on and off, and O is used to enable and
disable the biquad filter output when automatic volume tracking is turned off.
3.6.3 Time Alignment and Reverb Delay Processing
The TAS3103A provides delay line facilities at two locations in the TAS3103A—in the 3D effects block (reverb delay),
and at the output mix block (delay). There are three reverb delay blocks, one for each monaural channel, and these
delay elements are typically used in implementing sound spatiality, a single-mix reverb, and other sound effects.
There are also three delay blocks, again one for each monaural channel, and these delay elements are typically used
for temporal channel alignment. The delay line facilities are implemented using a single 4K (4096) × 16-bit RAM
resource. Each delay element implemented provides a one-sample delay (1/Fs). The size of each delay line can be
programmed via the I
2
C bus, or set by the EEPROM download in the I
2
C master mode. The only restriction is that
the total delay line resources programmed cannot exceed the capacity of the 4K × 16-bit memory bank.
Figure 3−19 illustrates how delay line structures are established within the 4K RAM memory. As seen in Figure 3−19,
each delay line immediately begins where the previously implemented delay line leaves off. The actual placement
of the pointers is computed by the resident microprocessor; the user needs only to enter the delay value required.
However, consider the following four points when programming the lengths of the delay lines.
1. Each delay line of length L requires L+1 memory sample spaces.
2. Reverb delay lines require three memory words (48 bits) to implement a single delay element, as the reverb
delay line operates in the 48-bit word structure of the digital audio processor (DAP).
3. Delay lines require two memory words (32 bits) to implement a single delay element, as the delay lines
operate on the mixer outputs after 32-bit truncation has been applied.
4. There are five words of reserved memory space that must be preserved.
In Figure 3−19, the terms P
CHx
refer to delay-line size assignments for the delay lines. The terms P
Rx
see the delay
line size assignments for the reverb delay lines. For the example shown in Figure 3−19, the delay for channel 3 is
set to 0 and the reverb delay for channel 2 is set to 0. From these zero-valued settings, it is seen that a delay of 0
requires the use of one delay element. For the case of a delay of 0, the write transaction into the single delay element
takes place before the read from the single delay element, thereby achieving a net delay of 0.
In making the delay-line length assignments, the only restriction is that the 4K memory resource not be exceeded.
Figure 3−20 illustrates the computations required to determine the maximum delay-line length obtainable for five
cases:
1. One reverb delay line
2. One delay line
3. Three equal-length reverb delay lines but no delay lines
4. Three equal-length delay lines but no reverb delay lines
5. Three equal-length delay lines and three equal-length reverb delay lines
All input crossbar mixers use signed 5.23-format mixer gain coefficients, and all are programmable via the I
2
C bus.
The 5.23 format provides a range of gain adjustment from 2
−23
(−138 dB) to 2
4
– 1 (23.5 dB).