
PRODUCT SPECIFICATION
TMC22x5y
49
P
Generation of the Comb Fail Signals
Luma Error Signals
The signals from the 3 low pass filters, 0HL, 1HL, and 2HL 
are subtracted from one another to produce an error signal 
proportional to the luma comb fail. The resulting signals 
(0HL - 1HL), produces 
LYE,
 and either (1HL - 2HL) in 
NTSC or (0HL - 2HL) in PAL produces 
UYE
. The 
LYE
 and 
UYE
 luma error signals are rectified if negative. In cases 
where the luminance component is constant, the error will be 
zero. Where the luminance goes from black to white over 2 
lines, the error signal will go to its maximum value.
The luma error signals can be doubled to facilitate inputs 
with low picture levels by setting the YESG register bit 
HIGH. The resulting signal is clipped to ensure no overflow 
occurs
Hue and Saturation Error Signals
In the past, comb decoders have relied upon comparing the 
difference in chroma magnitude between two lines to deter-
mine a comb fail. In fact, this chroma signal is normally the 
output of the high-pass or band-pass filter, and therefore con-
tains all the high frequency luminance information as well. 
As this signal was never demodulated, the sign bit was 
immaterial and was used only to rectify the chroma signal. 
This allowed chroma signals which where equal in magni-
tude but opposite in phase, and high frequency luminance 
signals, to fool the comb fail circuit.
The TMC22x5y uses a new, innovative approach to over-
come this problem. To detect comb failures in the high-
frequency portion of the video signal the outputs from the 
three high-pass filters, 0HH, 1HH, and 2HH, are passed 
through simple demodulators. The outputs from which 
provide the phase and magnitude of the in-phase and quadra-
ture components of the high frequency data. These compo-
nents are compared to determine the difference in phase and 
magnitude between 0H & 1H in all configurations, 
LME
 and
LPE
, and between 1H & 2H in NTSC or 0H & 2H in PAL, 
UME 
and 
UPE
. The magnitude error signals can be doubled 
to facilitate inputs with low picture levels by setting the 
CESG register bit HIGH. The doubled magnitude error 
signals are limited to ensure no overflow occurs.
The algorithm used to separate the quadrature components 
depends upon the relationship between the normalized sub-
carrier frequency and the number of pixels per line. This 
algorithm is preset for either a NTSC/M or PAL/I subcarrier 
frequency and a pixel data rate of 13.5MHz. It is therefore 
necessary to compensate for other pixel data rates by select-
ing the appropriate default using the CEST[1:0] register bits.
Picture Correlation
The degree of picture correlation depends upon the differ-
ences between the 
UYE, UME, 
and 
UPE
 upper error signals 
and the 
LYE, LME,
 and 
LPE 
lower error signals, and is mea-
sured as a percentage of full scale error. In flat fields of color 
you would have 0% error in picture correlation, however in 
sharp vertical transitions say between yellow and blue you 
would have large % errors between 
UYE
 and
 LYE 
and 
between 
UPE
 and 
LPE
, while there would be 0% error 
between 
UME
 and 
LME
.
Adapting the Comb Filter
In NTSC it is possible to switch from a 3 line comb to a 2 
line comb, and then to a simple decoder output. The 3 line 
comb to 2 line comb switch can be disabled, forcing the 3 
Figure 18. Generation of Upper and Lower Comb Fail Signals 
Luma
Comparison
UYE
LYE
UPE
LPE
UME
LME
YESG
YWBY
CESG
CSETBY
CEST[1:0]
OHH
OHL
1HL
2HL
65-22x57-59
1HH
2HH
Hue
Comparison
Saturation
Comparison
Chroma
Demodulation
&
Rectangular
to Polar
Conversion