
1996 Sep 27
24
Philips Semiconductors
Product specification
Digital Video Encoder (DENC)
GENLOCK-capable
SAA7199B
Accuracy of matrix
Evaluation of quantization errors.
The RGB to YUV matrix is achieved in accordance with
the following algorithm:
Y = INT
[(NINT(R
×
2
×
0.299) + NINT(G
×
2
×
0.587) + NINT(B
×
2
×
0.114) / 2]
U = NINT [(B
Y)
×
0.57722]
V = NINT [(R
Y)
×
0.72955].
Errors can occur in the calculation of Y, which as a result
influence the U and V outputs. The greatest positive error
occurs, if in all of the three for Y calculation used ROMs the
values are rounded up to 0.5 LSB, and no truncation error
of 0.5 LSB is generated after summation:
= +0.75 LSB;
with truncation “error”:
2
0.5 LSB = +0.25 LSB.
The greatest negative error occurs at rounding off in all the
three ROMs and by consecutive truncation:
0.5 LSB =
1.25 LSB.
As a result, the matrix error can be
±
1 digit, which
corresponds to approximately
±
0.5% differential
non-linearity.
Estimation of noise by quantization
The sum of all sqared quantization errors is SS normalized
to 220
3
input combinations (3-dimensional colour scale).
SS = 0.187545 LSB
2
.
Compared with noise energy for ideal quantization,
SSI =
1
12
LSB
2
results in a deterioration by the conversion
matrix of:
D = 10 log (0.187545
×
12) = 3.5 dB (equals 0.5 bit).
If SS is the sum of all squared quantization errors,
normalized to 220 input combinations of a grey-scale
(R = G = B), then:
SS = 0.12273 LSB
2
.
Compared with noise energy for ideal quantization,
SSI =
1
12
LSB
2
results in a deterioration by the conversion
matrix of:
D = 10 log (0.12273
×
12) = 1.7 dB (equals 0.25 bit).
3
2
0.5 LSB
×
3
0.5 LSB
×
3
--0.2
×
Normalizing amplifiers in the luminance channel
The absolute amplification error for 50 Hz non-set-up
signals is 0.375%; differential non-linearity is
0.333%
(equals
1 LSB).
The absolute amplification error for 60 Hz set-up signals is
1.5%; differential non-linearity is
0.365%
(equals
1 LSB).
Normalizing amplifiers in the chrominance channel
The absolute amplification error is approximately
±
0.5%
with a truncation error of
0.5 LSB.
The subcarrier amplitude for standards with luminance
set-up is the same as for the standards without luminance
set-up.
Modulator
The absolute amplification error is
0.39%; there is no
truncation error.
Functional timing
(see Fig.10)
GENLOCK
MODE
The encoded signal can be generated earlier with respect
to CVBS7 to CVBS0 bits (offset t
ofs
set by GDC-bits;
index 05). The HSN output signal can be generated early
by PSO-bits (index 07) with respect to CB to compensate
for pipelining delay t
Rint
of the RAM interface (valid also in
stand alone mode).
The horizontal timing is independent of active video at data
inputs PDn(7 to 0). The line blanking period on the outputs
is set to approximately 12
μ
s in 50 Hz standards (11
μ
s in
60 Hz standards).
S
LAVE MODE
HSN pin is used as an input. The active edge of the input
signal is assumed to fit to the incoming CB signal.
Deviations can be compensated in the range of the
GCD-bits (index 05).
The t
enc
time is the total delay from data input to analog
CVBS output; it is 55 pixel clock periods long (PIXCLK)
plus the propagation delay of the LDV input register
regardless of mode and colour standard.
The key input signal is delay compensated with respect to
PDn(7 to 0) data input. The generated vertical field and
burst blanking sequences are shown in Fig.11 (50 Hz
PAL) and Fig.12 (60 Hz NTSC).