THE BIG PICTURE

Ronald Surprenant

I'm afraid it's more complicated than that. The picture quality of 1080i and 1080p can be nearly identical, depending on the circumstances, which include how the footage was shot, stored, transmitted, and displayed.

You are correct that a 1080i signal transmits interlaced fields at a rate of 60 per second, which is equivalent to a frame rate of 30 per second. A 1080p signal can also convey 30 frames per second (fps), but it rarely does in practice. Instead, it normally conveys 24 or 60fps; these are denoted 1080p/24 and 1080p/60, respectively. Movies are stored on Blu-ray at 1080p/24, and most modern players can send exactly that to the display.

There are no consumer sources that record or transmit 1080p/60 natively, but 1080p/24 is usually converted to 1080p/60 in the display by repeating frames in a 3:2 sequence. That is, one frame is repeated three times, the next one twice, the next one three times, the next one twice, and so on.

However, when 1080i is deinterlaced, the resulting signal is 1080p/60, not 1080p/30 as you surmise. Why? Because 1080p/30 looks noticeably juddery. BTW, 1080p/24 also looks very juddery if it's displayed at 24fps, which is why it's normally converted to 1080p/60. In a few plasmas and projectors, 1080p/24 is displayed at 72 or 96fps by repeating each frame three or four times instead of using the 3:2 sequence needed to achieve 1080p/60. In 120Hz and 240Hz LCD TVs, new frames are created between each frame in the input signal to reach the desired frame rate; this process is called interpolation.

How is 1080i converted to 1080p/60? That depends on whether the signal originated as 1080i video or was interlaced from a 1080p/24 film source. If it originated as 1080i video—which is common in HD newscasts and some concert videos on Blu-ray—it is converted to 1080p/60 by repeating fields. For example, let's say the incoming fields in a 1080i video signal are labeled a, b, c, d, e, f, and so on. The deinterlacer combines them into frames in the following pattern:

ab bc cd de ef etc.

Because the image was captured at 1080i, each field is 1/60th of a second apart, so there are no frames per se to reconstruct, and repeating and pairing fields in this manner poses no problem.

If a 1080p/24 film source is interlaced to 1080i—which is common when HDTV channels broadcast movies—some fields are repeated so the resulting signal has 60 fields per second; this process is called telecine. For example, let's say the original film frames are labeled A, B, C, D, E, and so on. When each frame is divided into fields 1 and 2, the sequence goes like this:

A1 A2 B1 B2 B1 C1 C2 D1 D2 D1 etc.

Notice that fields B1 and D1 are repeated in this sequence. This is the famous "3:2 pulldown cadence." Ideally, the resulting 1080i signal includes special markers or "flags" that identify which fields are repeated, but this is not always the case.

Deinterlacing a film-originated 1080i signal to 1080p is tricky. The best way is to apply a process called inverse telecine, which reassembles the original film frames and discards the repeated fields, resulting in the original 1080p/24. This process depends on the presence of those embedded flags, but not all film-originated 1080i signals have them. Some video processors can detect the 3:2 cadence and use that information to perform inverse telecine, but the process can break down if there's a bad edit or other anomaly in the signal.

More commonly, the fields are simply paired up as they are, resulting in some frames with mismatched fields. Looking again at the previous example, the frames would have the following structure (see also the diagram at the top of this blog):

A1A2 B1B2 B1C1 C2D1 D2D1 etc.

As you can see, two out of every five frames have mismatched fields. How does a video processor compensate for this? By analyzing the amount of motion in each field, pairing fields with no motion, discarding fields with high motion, and creating new motion fields using interpolation. In this process, fields are repeated as needed to reach 60 frames per second. Of course, some processors do this better than others, which is one of the things we look at when reviewing the video processing in TVs and AVRs.

The bottom line is this: A native 1080p source will generally look better than a 1080i source, but if the 1080i signal is deinterlaced properly, there should be little visible difference between them.

If you have an audio/video question for me, please send it to scott.wilkinson@sorc.com.