While rarely mentioned over the decades, vertical filtering, which removes horizontal detail in your image, has been around as long as CRT has and is still used to this day in displays and source scaling. Even with the arrival of DTV in 1998, vertical filtering was commonly used in the video circuits of consumer 1080i HDTV CRT displays. For this article we are looking at vertical or horizontal filtering related to luminance response.
The first step with horizontal and vertical filtering is to understand that horizontal lines on your screen are based on vertical pixel / frequency response and the vertical lines on your screen are based on horizontal pixel / frequency response. For CRT, rear projection horizontal filtering was also used to prevent artifacts from appearing due to the combing effect that could be created when a 1920 vertical response of multiple lines is projected onto a lenticular screen that by its nature and function has its own set of very fine vertical lines. With the new 720p and later 1080p micro-display rear projection technology, such as DLP, LCD, the lenticular screen design was dropped using smooth outer screens for glare reduction instead since they could provide a high contrast response with high resolution, 1280 for 720p or 1920 with 1080p.
While HDTV enthusiasts across the country were crowing over the high resolution HDTV images delivered digitally to their CRT screens, those knowledgeable of imaging science were telling another story of how those same HDTVs weren't earning their HDTV label. Vertical and horizontal filtering was a big part of that story. By 2005, CRT technology had gone by the wayside in favor of lighter and smaller 720p rear projection micro-displays and flat panels. For numerous reasons, the 1920x1080 progressive (1080p) format was put on a pedestal as the holy grail of high definition and HDTV enthusiasts wanted true 1080p displays in their homes. For the 2005-2006 model season these same technologies had finally advanced to 1920x1080 pixel matrices capable of resolving a clear 1920 or 1080 line response. HDTV enthusiasts replaced their old CRT products with this new technology only to find out later that nearly all of the first generation 1080p displays were not actually providing such a response for a variety of reasons and vertical filtering was one of them. Indeed, this first generation of 1080p was all about perception rather than actual performance!
Silicon Optics provided test material for detecting vertical filtering and well known reviewer Gary Merson led the way in exposing this flaw of the new 1080p designs. It wasn't the 1080p display panel at fault but the design of the internal scaler that must convert interlaced 1080i into progressive for a 1080p display panel. Since then, most of the other problems related to achieving a true 1080p response have been tackled, but vertical filtering of 1080i remains with us. With the advent of the new HD disc formats, Blu-ray and HD DVD, anybody can test their display or HD disc player for vertical filtering using Silicon Optix HQV Benchmark, which is available in DVD, HD DVD and Blu-ray formats.
Failure of these tests due to vertical filtering does mean there will be a loss of detail. On the other hand, without prior knowledge of this attribute, it is unlikely you would have noticed the loss (recall the enthusiasm of early CRT HDTV adopters) and requires a deeper understanding of burst testing, what is going on and how it affects specific elements of an image.
A luminance burst is a series of alternating black and white lines. The thinner the lines, the higher the frequency of the video signal response. A common response artifact of legacy HDTV CRT rear projection displays was their inability to pass a horizontal 1920 vertical line burst creating the same type of response as the HQV Benchmark 1:1 pixel mapped vertical burst test of 1080 horizontal lines; the lines are missing and instead you have a gray box. By going down one step to two pixels per line, 540, you now see a burst response (although it lacks the contrast response of going down another step to three pixels per line, 270, and lower). An anecdotal conclusion would suggest that those displays failing such a test will not show a single pixel line but yet they do. The reason the 1920 burst won't pass but a 960 burst or single pixel 1920 line does is because the CRT display applies analog filtering somewhere in the circuit and for that filter to work requires a repetition of lines; in essence a video signal of continuous alternating frequency to which the filter is tuned. That does not mean a single pixel line is being delivered to its fullest potential either because the filter will at least soften that response, which is seen as a loss of contrast between white and black for CRT displays. For decades, the vast majority of consumer CRT displays have been using vertical filtering to hide artifacts from the source material, the NTSC system and in many cases a mediocre interlaced vertical sweep circuit design.
For fixed pixel digital displays (FPD), it gets a bit more complicated because in most cases this filter is digital instead of analog and pixel displays respond differently. I test for vertical filtering using a pattern generator that is based on 1:1 pixel mapping for the patterns and use the vertical and horizontal grid pattern of single pixel lines to detect vertical filtering. In this case the filter turns what should be a vertical response single pixel horizontal line into three; the middle line is slightly lower in peak light output than it should be while the other two lines of pixels above and below are greatly reduced further. While this is easily detected by getting close enough to the screen to see individual pixels the perceived effect at the seated viewing position is a single pixel vertical line, even at the recommended viewing distance of three screen heights for HD. While appearing to have the same brightness/contrast response as the vertical lines of the grid pattern, what you may pick up on is a reduction in a nice hard edge and/or subtle thickening of the line when compared to the horizontal response of vertical lines. Ultimately you are losing detail in whichever plane is being filtered, vertical or horizontal. There is yet another perceptual spin in all this; a great pixel response providing detail in one plane can offset the lack of detail in the other and still provide a perception of detail to your eyes. Making this issue even more complex is to clearly notice the detriments of filtering requires an image that has elements similar to a burst, alternating bright and dark lines at an even 1:1 pixel level. The simplistic finale for this complex technical content is that the artifacts created are specific to the material along with how often they occur. How much effort or money do you want to spend on an artifact that represents a very small portion of any content? How much will it matter to you that at chapter 13 and 1:41 minutes of XYZ movie you can clearly see an artifact for 30 seconds?
The stadium pan of HQV Benchmark provides an excellent example of this because it is the only video content on the disc that has detail similar to a vertical 1080 burst for an extended period of time due to the bleachers. The rest of the video content provided lacks that repetitious vertical response and appears highly detailed. If it were not for the horizontal lines of the bleachers the rest of the horizontal lines in the stadium test appear slightly thicker than a display or player that does pass. The high resolution of high definition muddies the waters even further compared to standard definition making perception of this anomaly far more difficult for the average viewer.
For casual viewers it is difficult to be concerned with vertical filtering.
If you aren't sitting close enough to the screen, then you are creating a perceptual filtering effect with your eyes already and vertical filtering concerns play a very small role, if any at all.
For those seeking the immersive experience of close viewing distances following the 3-4 screen heights rule of thumb, you will experience a subtle loss of detail due to vertical filtering.
If you are a performance enthusiast seeking every nuance of detail then vertical filtering is unacceptable for any 1080i source material from DTV to Blu-ray.
Only you can determine the role vertical filtering plays in your system and the content that you watch versus how much money you want to spend to remove the artifact. Please read the HQV Benchmark review for more details.
Posted by Richard Fisher, July 31, 2008 8:16 AM
About Richard FisherRichard Fisher is the President of Mastertech Repair Corporation, serving north east Atlanta, Georgia, and has been servicing, calibrating and reviewing audio video products since 1981. Tech Services USA, a division of Mastertech, creates sites, communities and libraries for consumers and professionals to share their technology knowledge and learn from each other. These include The ISF Forum and HD Library. HDTV Magazine exclusively publishes HD Library and Forum for Tech Services USA.
Richard is ISF and HAA certified providing calibration and A/V reproduction engineering services. Richard is a technical consultant and also provides performance ISF and HAA home theater systems and calibration via Custom HT. Mastertech Repair Corporation is a factory authorized service center for Hitachi, Mitsubishi and Toshiba and a member of the National Electronics Servicing Dealers Association, NESDA, and the Georgia Electronics Servicing Dealers Association, GESDA.