Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!csd4.csd.uwm.edu!cs.utexas.edu!usc!apple!sun-barr!ames!lll-winken!ubvax!ardent!sim!sleat From: sleat@sim.ardent.com (Michael Sleator) Newsgroups: comp.graphics Subject: Re: Frame rate (was: Re: HDTV and ATV Glossary (TN32)) Summary: "Interlace" is not a dirty word. Message-ID: <7967@ardent.UUCP> Date: 25 Aug 89 22:40:31 GMT References: <13130@well.UUCP> <17400006@hpfcdj.HP.COM> Sender: news@ardent.UUCP Reply-To: sleat@ardent.com (Michael Sleator) Organization: Ardent Computer Corp., Sunnyvale, CA Lines: 109 In article <17400006@hpfcdj.HP.COM> myers@hpfcdj.HP.COM (Bob Myers) writes: >Interlace does NOT avoid flicker and motion artifacts; flicker (actually >a host of effects with various names, but let's just do what everybody >does and lump them together under "flicker) is MUCH worse in an >interlaced display over a non-interlaced display of the same size and >brightness at the same *frame* rate. You can't avoid it - the price you >pay for reducing the bandwidth required is the fact that each pixel gets >refreshed only half as often as before. This is particulrly annoying >in images containing narrow horizontal or near-horizontal lines, but even >a plain white raster, if interlaced, will be obviously worse than a >non-interlaced raster, all else being equal. I disagree. Have you ever compared a 60Hz progressive scan display with a 60Hz frame rate (120Hz field rate) 2:1 interlaced display? Under some circumstances I find the flicker from a 60Hz progressive scan display with a P4 or similar speed phosphor annoying. By comparison, a 60Hz (frame) 2:1 interlaced display looks rock solid. This is *especially* true with a plain white raster. There is something else confused in your paragraph above, in that at the same frame rate, an interlaced and a non-interlaced display require the same video bandwidth. Perhaps you meant to describe an interlaced field rate equal to a progressive frame rate??? One of my pet peeves in this area is that people, by and large, do not distinguish between interlace per se and refresh rate. (Witness the much- used but ill-defined term, "interlace flicker".) When most people hear "interlace", they think "30Hz frame rate". The two are not synonymous. For example, I can imagine circumstances (somewhat unusual, I'll admit) in which it would make sense to run at 60Hz frame, 180Hz field, 3:1 interlace. This might be appropriate in a display where you had a large number of scan lines, a very small spot size, and for some reason needed a very fast phosphor yet the screen had to be viewed by humans. This example points out the reason interlacing "works": loosely speaking (someone else can probably do a better job of speaking tightly to this point), your eye (brain?) integrates over a small spatial and temporal region. If the scan lines were very far apart, interlacing would not help. One of the frequent objections to interlaced displays is that one can easily construct pathological cases, such as fully illuminating all of the scan lines on one field with the other field completely dark, which look terrible. Note, however, that at equal *frame* rates, this case is indistinguishable in the interlaced and progressive formats. (See below re the practical realities of this.) So again, this is principally an objection to low refresh rates, not to interlace. So why would you want to interlace at all, if at the same frame rate you are not buying any bandwidth relief? Personally, at 50-60Hz frame rates, for more or less static displays (e.g., text), I would rather look at an interlaced display than a non-interlaced one. These rates are marginally acceptable in terms of flicker, and interlacing generally helps. It does have it's costs, however, and they can be high. The two biggest obstacles to making an acceptable interlaced CRT display are high-voltage regulation and vertical deflection stability. In order to maintain accurate geometery between fields, the electron beam acceleration voltage must be very tightly regulated. If the beam current (proportional to brightness) on one field is different from that on the other field, the high-voltage power supply will tend to droop, and there will be a geometrical expansion of the raster. Clearly, if the two fields are not exactly the same size, they won't interlace properly. In a non-interlaced display, it's not nearly so critical because the image tends to expand as a whole. The second obstacle, vertical deflection stability, is not hard to understand. In order to avoid line paring, the start of each vertical scan must be controlled to within a small part of one horizontal period. With say 600 lines per field, this amounts to a small fraction of a percent of the vertical period. This is mostly a matter of careful design and doesn't necessarily imply a parts cost penalty of the same order as the high-voltage regulation problem. There is a small cost in bandwidth, since more time is spent in vertical retrace. For 2:1 interlace, this generally amounts to less than five percent of the total frame time. Another problem is magnetic interference. Just as in the high-voltage problem, any distortion which affects one field differently than the other will be very noticeable. A constant magnetic field, such as the Earth's field, will not cause a problem. Changing fields, such as from transformers, adjacent displays, and power wiring, can cause vertically adjacent pixels, on different fields, to be deflected in opposite directions. This can be very objectionable. Even at equal frame rates, there will still be motion artifacts in an interlaced display. Or rather, the motion artifacts on an interlaced display will be different from those on a non-interlaced display. It's not clear to me that one is always better or worse than the other. I realize that most or all of these points have been made in this group at one time or another, but confusion still abounds. Despite its abuses, interlacing is a valid technique which can successfully applied to a variety of real-world needs. (Can you imagine watching European television, with its 25Hz frame rate, without interlacing?) If people were to look carefully at the real issues here, we might find that there are a varity of applications where 2:1 or higher ratio interlaced displays running at field rates of 100 to 200Hz provide better results than progressive scan displays. Interlace is not, or at least should not be, a dirty word. Michael Sleator Ardent Computer 880 W. Maude Ave. Sunnyvale, CA 94086 408-732-0400 {apple, decwrl, hplabs, ubvax, uunet}!ardent!sleat