THS7372 Datasheet, PDF (37/49 Pages) Texas Instruments – 4-Channel Video Amplifier with One CVBS and Three Full-HD Filters with 6-dB Gain

THS7372
www.ti.com
SBOS578 – AUGUST 2011
LOW-PASS FILTER
Each channel of the THS7372 incorporates a sixth-order, low-pass filter. These video reconstruction filters
minimize DAC images from being passed onto the video receiver. Depending on the receiver design, failure to
eliminate these DAC images can cause picture quality problems because of aliasing of the ADC in the receiver.
Another benefit of the filter is to smooth out aberrations in the signal that some DACs can have if the internal
filtering is not very good. This benefit helps with picture quality and ensures that the signal meets video
bandwidth requirements.
Each filter has an associated Butterworth characteristic. The benefit of the Butterworth response is that the
frequency response is flat with a relatively steep initial attenuation at the corner frequency. The problem with this
characteristic is that the group delay rises near the corner frequency. Group delay is defined as the change in
phase (radians/second) divided by a change in frequency. An increase in group delay corresponds to a time
domain pulse response that has overshoot and some possible ringing associated with the overshoot.
The use of other type of filters, such as elliptic or chebyshev, are not recommended for video applications
because of the very large group delay variations near the corner frequency resulting in significant overshoot and
ringing. While these filters may help meet the video standard specifications with respect to amplitude attenuation,
the group delay is well beyond the standard specifications. Considering this delay with the fact that video can go
from a white pixel to a black pixel over and over again, it is easy to see that ringing can occur. Ringing typically
causes a display to have ghosting or fuzziness appear on the edges of a sharp transition. On the other hand, a
Bessel filter has ideal group delay response, but the rate of attenuation is typically too low for acceptable image
rejection. Thus, the Butterworth filter is an acceptable compromise for both attenuation and group delay.
The THS7372 SD filter has a nominal corner (–3-dB) frequency at 9.5 MHz and a –1-dB passband typically at
8.2 MHz. This 9.5-MHz filter is ideal for SD NTSC, PAL, and SECAM composite video (CVBS) signals. The
9.5-MHz, –3-dB corner frequency was designed to achieve 54 dB of attenuation at 27 MHz—a common sampling
frequency between the DAC/ADC second and third Nyquist zones found in many video systems. This
consideration is important because any signal that appears around this frequency can also appear in the
baseband as a result of aliasing effects of an ADC found in a receiver.
The THS7372 FHD filters have a nominal corner (–3-dB) frequency at 72 MHz and a –1-dB passband typically at
60 MHz. This 72-MHz filter is ideal for 1080p50 or 1080p60 component video. It is also ideal for oversampling
systems where the video DAC upsamples the video signal such as 720p or 1080i upsampled to 148.5 MHz. The
benefit is an extremely flat passband response along with almost no group delay within the HD video passband.
Keep in mind that images do not stop at the DAC sampling frequency, f_S (for example, 27 MHz for traditional SD
DACs); they continue around the sampling frequencies of 2x f_S, 3x f_S, 4x f_S, and so on (that is, 54 MHz, 81 MHz,
108 MHz, etc.). Because of these multiple images, an ADC can fold down into the baseband signal, meaning that
the low-pass filter must also eliminate these higher-order images. The THS7372 filters are Butterworth filters and,
as such, do not bounce at higher frequencies, thus maintaining good attenuation performance.
The filter frequencies were chosen to account for process variations in the THS7372. To ensure the required
video frequencies are effectively passed, the filter corner frequency must be high enough to allow component
variations. The other consideration is that the attenuation must be large enough to ensure the
anti-aliasing/reconstruction filtering is sufficient to meet the system demands. Thus, the selection of the filter
frequencies was not arbitrarily selected and is a good compromise that should meet the demands of most
systems.
BENEFITS OVER PASSIVE FILTERING
Two key benefits of using an integrated filter system, such as the THS7372, over a passive system are PCB area
and filter variations. The small TSSOP-14 package for four video channels is much smaller over a passive RLC
network, especially a six-pole passive network. Additionally, consider that inductors have at best ±10%
tolerances (normally, ±15% to ±20% is common) and capacitors typically have ±10% tolerances. Using a Monte
Carlo analysis shows that the filter corner frequency (–3 dB), flatness (–1 dB), Q factor (or peaking), and
channel-to-channel delay have wide variations. These variances can lead to potential performance and quality
issues in mass-production environments. The THS7372 solves most of these problems with the corner frequency
being essentially the only variable.
Copyright © 2011, Texas Instruments Incorporated
37