Datasheet

3−50
When updating multiple TAS3103s with dither seeds, timing should be taken into account. The recommended seed
update process is to load all TAS3103s with their seed values in less time then the minimum LFSR cycle time of
496.5 ms, and use the same set of seeds for all TAS3103s. Each TAS3103 immediately begins running, starting at
the state set by the new seed, upon receiving the new seed. The sequential delivery, in time, of the seeds to the
multiple TAS3103s assures the TAS3103s do not all start with their new seeds at the same time, and the completion
of the process in less than 496.5 ms assures that previously programmed TAS3103s are not repeating the cycle at
the same time another TAS3103 is being programmed with the same seed set—causing correlation between the
dither data streams of the two TAS3103s.
CAUTION: The state of the digital audio processor may prevent the loading of a new I
2
C
—commanded dither seed value. Anytime a new seed is loaded into the TAS3103 via an
I
2
C write transaction to address 0xC7, it must be followed by an I
2
C read transaction to
address 0xC7 to verify that the new seed value was accepted. If the new seed was not
accepted, the write-read sequence must be repeated.
3.9.2 Dither Mix Options
In Figure 3−29 it is seen that the two LFSRs are logically combined to produce a triangular probability distribution.
This distribution is then sampled at six different points in time, by the DAP processing clock, to create six statistically
independent data streams. Sampling the LSFR outputs at different points in time within an audio sample period
(1/LRCLK) assures the six dither data streams are uncorrelated. The six uncorrelated dither data streams are then
routed through mixers. Each pair of mixer outputs is then applied to a summation block. It is noted in Figure 3−29
that each of the three mixers pairs has the same set of coefficients (set by I
2
C subaddress 0xC6). If the coefficients
for both mixers are set to 1.0, a quadratic distribution is obtained. If either coefficient is set to 0.0, a triangular
distribution is obtained. If both coefficients are set to 0.0, dither is disabled.
3.9.3 Dither Gain Mixers
Figure 3−29 shows the peak magnitude of the triangular distribution to be ±16 in the 48-bit DAP word (25.23 format).
The peak magnitude of the quadratic dither data is twice this, or ±32. Figure 3−30 shows the position of this peak
magnitude value in reference to the DAP 48-bit data word. Table 3−8 lists the mixer gains required to position the
dither data stream at the LSB of the output data word for different data sample word sizes.
Table 3−8. Mixer Gain Setting for LSB Dither Data Insertion
DISTRIBUTION
MIXER GAIN COEFFICIENT
DISTRIBUTION
32-BIT SAMPLE 24-BIT SAMPLE 20-BIT SAMPLE 18-BIT SAMPLE 16-BIT SAMPLE
Triangular 2
−19
2
−11
2
−7
2
−5
2
−3
Quadratic 2
−20
2
−12
2
−8
2
−6
2
−4