Boiling down the TV Tuner - Signal Processing (cont.) Let's dive a little deeper...

Due to the inherent artifacts of interlaced scanned video, just about all decent TVs have comb filters, which are used to smooth out the jagged edges of feathering. There is less that can be done with line twitter, though one method is to filter out the flicker from fields, which some entertainment companies do for DVD creation and even some TV broadcasters in their conversion process of certain old TV programs and movies. Another method to counter line twitter is to use a vertical filter to reduce/smooth out the sharp contrast transitions. The problem is that this filter effectively smears the original image and affects the image contrast as a whole.

Just as TVs employ the use of combing and counter-line twitter filters, PVR cards do the same in the A/V processor, in the post analog to digital conversion. However, not all A/V processors do the filtering in the same way, which is why there can be even more quality differentiation between cards that use the same tuner and software suite. Also keep in mind that PVR card makers don't usually make their own tuner or A/V processor. Take the Hauppauge WinTV PVR-250BTV as an example; it uses a Philips tuner with a Conexant A/V processor (sometimes known as a broadcast decoder).

Comb Filtering 101:

There are two types of comb filtering: 2D and 3D. A good PVR card should have the ability to do both and know when to correctly use each. The ability to differentiate between the use of a 2D or 3D comb filter will depend heavily on the "adaptiveness" of its sorting algorithm. Keep in mind that a by product of comb filtering is the reduction or possible elimination of dot crawl (the borders between scanned lines). You can read a lot more about comb filters here, as we are going to do a quick crash course.

A 2D comb filter should be used when there is a lot of motion in the combination of complement fields. You can refer to this combination as a frame, but this is technically incorrect, since it isn't really a "frame" that interlaced scanned video is based. This filter will take multiple lines (usually 3 but it could be up to 5) and will mix the lines left to right along the borders of the scan lines. The percentage of "mixture" that occurs will depend on how well the filter logic was programmed to distinguish whether the line above or below was a better match.

A 3D adaptive comb filter is usually something seen on TV in the $500 price range or higher (which still have 2D comb filters). This type of comb filter effectively separates the Y/C (luminance ÿ Y, Chrominance ÿ C) components of analog video. Then it compares 3 or more horizontal scan lines with a pair of fields as well as the next up coming paired fields. The ability to look ahead is why this type of filtering is called 3D, as it deals with a multiple fields that comprise a video segment. Generally speaking, still and slow moving scenes will look better, crisper with less interlaced scanned artifacts, with a 3D adaptive comb filter over 2D.  This type of comb filter will not be applicable in a recording system if the Y and C are separated already, for example with S-Video and Component signal connections where the signals are brought out on their own feed.

Apparently, there are some companies that are trying to pass off their 2D comb filter as a "3 line digital comb filter." Of course, a poor logic 3D comb filter is just as bad, since you might as well only be looking at an image going through 2D comb filters. The problem with this is that it is deceiving and 2D comb filters cannot improve the quality of still objects. One of the big A/V processor makers made note that hardware developers and consumers should beware of low-quality comb filters, since it is possible manipulate a TV signal so that the end result has more artifacts (i.e. crawling dots). This has some validity to it, we have seen this occur on several "el cheapo" PVR cards pulled from the streets of Taiwan.

If the PVR card has a MPEG-2 encoder, the hardware encoding is done after the signal processing phase of the television signal. If it is a software encoding card, the burden will just be passed on to the CPU and GPU.