
NXP Semiconductors
PNX15xx/952x Series
Volume 1 of 1
Chapter 11: QVCP
PNX15XX_PNX952X_SER_N_4
NXP B.V. 2007. All rights reserved.
Product data sheet
Rev. 4.0 — 03 December 2007
11-362
The UDTH unit, however, does more than just undithering. It does format conversion
as well. QVCP has its own data format protocol (10-bit signed data -- UP(Y/R), MI(U/
G), LO(V/B) --- and 8-bit unsigned alpha) and is designed to process data with a
nominal range of 9 bits; one extra bit is reserved for overshoots and undershoots.
Input data gets converted, from an external format to the native QVCP format, inside
the UDTH module.
Figure 3 shows the data-ow diagram of the UDTH unit. In the
gure, the nominal data range is shown for each conversion step for both 8- (bottom
part) and 10- (top part) bit inputs. Note that an 8-bit input to QVCP will be converted
to a 9+1 format by undithering or by left shifting. A 10-bit input, however, can be either
true 10 bits or 9+1 bits; for the former, there is not too much head-room left and so,
the nominal range of the various layers should be equalized at the VCBM unit.
The function of Alpha Processing unit is to insert a xed alpha or pre-multiply each
pixel with a per-pixel alpha. For a 10-bit input, only xed alpha is supported.
Figure 3:
Undithering and Pedestal Manipulation
Unsigned to
signed
conversion
Nominal
range
increasing
by 2
Sign
extension
of the MSB
bit
Pedestal
removing
8bu
8bs
9bs
10bs
format
Left shift by 1 add 0 to
LSB position
Append 2nd MBS to
LSB position
Undither
Increasing nominal range by 2
or
8bs
9bs
16~235
-112~107
-224~214
if pedestal substraction is on
-256~182
else
-224~214
nominal data
range for 8bits
input
Reconstruct
10bit data from
u/m/l path and
alpha path
8MSB
2LSB
u/m/l (8 bit)
alpha (2 bit)
clip
-512
~511
Alpha
processing
nominal data
range for 9+1
bits input
288~726
-224~214
if pedestal substraction is on
-256~182
else
-224~214
nominal data
range for 10
bits input
64~940
-448~428
if pedestal substraction is on
-512~364
else
-448~428
format
10bu
10bs
10 bits input
8 bits input
For true 10bits input, pedestal substraction is recommended off, the different nominal ranges for various layers
could be equalized at the VCBM unit.