24p vs. NTSC video
Conversion of 24p to NTSC-based frame/field rates Working with 24p material via video equipment working at NTSC frame rates has many of the same attributes as the 24 frame/s workflow, but is more complicated by the NTSC-rate practice of using
telecine pull-down rather than the PAL practice of transferring 24 frame/s material at 25 frame/s. At 525 lines analog NTSC video rates (30000/1001 frames per second) a full "interlaced" frame, unlike a progressive frame, is nearly 1/30th of a second and is composed of two separate "fields", each field nearly 1/60 second. The first field (the odd field) contains visible scan lines 21-263 and the second field (the even field) contains visible scan lines 283–525 (though lines 263 and 283 are half-lines). What is seen onscreen is two of these fields, "interlaced" together, to produce a single full frame. This comes from the proper longhand designation being vertical resolution, followed by the interlaced/progressive notation, and then the frame rate. So typical DV video is correctly listed as 480i/30. The long hand for 24p is 480p/24. Often the resolution is dropped and the i/p designation moved after the frame rate for shorthand. 24p cameras do not, as NTSC video cameras do, shoot 30 interlaced frames per second (60 fields); they shoot 24 full progressive frames per second. 24p material can be recorded directly into formats that support the framerate. Some of high definition formats support the 24p framerate in addition to 60i and 50i (PAL). Previously, few formats supported 24p and the industry used workarounds to work with 24p footage with 60i equipment. To record 24p material onto a 60i format (i.e. any NTSC-based format), pulldown is typically added to 'pad' the 24 frames into 60 fields. This is done by taking every frame and splitting it into two fields. Then, every second frame has one of its fields duplicated, resulting in three fields. The fields are then played back in that pattern – 2-3-2-3-2-3-2-3-2-3-2-3-2-3 ... and so on. The resulting video becomes a 60i stream and can be displayed on NTSC monitors. However, the aesthetic of 24p motion is retained and the footage does not have the motion of typical 60i video. This 3:2 pulldown is the same process that is used when transferring
film into video. Any editing application which supports NTSC video can be used to edit footage employing the 3:2 pulldown scheme. It can be captured as a standard 60i file and edited like footage from any other camera, while still retaining its 24p aesthetic. There can be issues when editing the footage as 60i, however, including choppiness in short transitions or fades, and also a mismatch in the motion characteristics of the footage and any graphics which may be added to it, such as text or logos. So, while 24p footage can be edited as 60i, it is usually better to edit the footage on a 24p timeline with the pulldown removed. Most current
prosumer-level editing applications which edit native 24p can remove the 3:2 pulldown for editing in native 24p, although some cannot. However, this is not ideal; the removal of the 3:2 pulldown involves reconstruction of every fourth frame from two different field groups, which can cause a generational loss and some banding problems if the application doesn't interpret the footage properly. Therefore, using the 3:2 pulldown scheme is not ideal when planning to edit on a 24p timeline. Note: "3:2 pulldown" has a cadence of 2-3-2-3-2-3..., but in the industry is called "3:2 pulldown", even though the cadence is 2–3. Some people use the term "2:3 pulldown", which corresponds to the cadence, but is not normally used in the industry for the technique.
Advanced pulldown Another pulldown pattern is the "advanced pulldown" ("24pA") pattern, first implemented in the
Panasonic AG-DVX100 camcorder. Instead of padding the frames into a repeating 3:2 pattern, the frames are padded into a 2:3:3:2 pattern. This pattern is specific to the
NTSC DV format, and would serve no purpose in native 24p formats. It converts the first frame into two fields, the second into three fields, the third into three fields, and the fourth into two fields. It then repeats this pattern for every group of four frames that follows. This pulldown pattern is used to avoid segmenting a 24p frame into two different 60i fields that exist in two different 60i frames. When a 24p frame is split up and recorded into separate 60i fields, interlacing artifacts can exist in the 60i "frames" (i.e. two fields). These artifacts decrease the compression efficiency of
DV and can result in cycles of efficient compression followed by less-efficient compression. The advanced pulldown scheme avoids this as every 24p frame can be found intact within the resulting sequence of 60i frames, yet the compression efficiency remains the same as with 3:2 pulldown. When editing 24pA footage, conversion from 60i back to the original 24p is very efficient. It only requires blending the fields made from the frames back into full frames. Then, only every fifth frame will be made up of fields from two different frames, and that frame can be discarded, leaving only the other four full frames. In order for this to work properly, the DVX100 camera records video in chunks of five video frames. This ensures that each clip has regular and predictable cadence. Because the 2:3:3:2 scheme was devised for efficient pulldown removal for editing, and because 24p editing applications more universally support its removal, it should always be used when planning to edit in native 24p. Editing systems need specific support for the 24pA format to be able to detect and remove the pulldown properly so that the 24p frames can be edited in a 24p timeline. Many but not all prosumer and professional-level non-linear editing systems are able to recognize and remove this advanced pulldown scheme. However, among the editing applications able to remove pulldown and edit in native 24p, it is more common for them to have support for 24pA 2:3:3:2 pulldown than for standard 24p 3:2 pulldown removal. Still other editing applications have the option for editing on a 24p timeline, and will accept footage where the pulldown has already been removed in another application. Remember that although computer editing systems may refer to "24p", usually the frame rate is 23.976 frame/s. To add to confusion, the popular editing program Final Cut Pro refers to 23.976 as "23.98" in menus and dialogs, even though it correctly works with the footage at the 23.976 frame rate. 23.976 is also not precise though, as the real frame rate is 24000 ÷ 1001, so 23.98 is also a correct approximation. Also because the 2:3:3:2 pulldown scheme was devised in order to make pulldown removal for editing in native 24p more efficient, the pulldown arrangement is not ideal for watching footage. There can be exaggerated stutters in motion, because the frames which are split into three fields are not only onscreen for 50% longer than the other frames, they are back-to-back. As such, 2:3:3:2 pulldown should be used only when a native 24p edit is planned, and not for final viewing. This includes when shooting the footage initially, and also when printing back to tape from an NLE.
60i to 24p conversions Another method of achieving the 24p aesthetic is to capture 60i footage and convert it into 24p. Various techniques can be used to perform this conversion. A simple scheme would blend the fields together. This can result in motion artifacts where comb-like jagged artifacts appear in areas of high motion.
Deinterlacing can remove these artifacts, but certain methods will cause up to half the footage's vertical resolution to be lost. Adaptive deinterlacing schemes only deinterlace areas of high motion, hence preserving resolution in stationary areas. More advanced techniques can be used to mitigate problems such as aliasing from the temporal displacement between the 60i fields.
The Optical Flow Method This is currently the highest quality method of converting 60i footage to 24p. It involves using
optical flow to extrapolate 24 frames of information from 60 frames while compensating for the time displacement between the two. For example, in one second of 60i footage, each image is captured at 1/60 second, which does not perfectly align with images that would have been captured 24 times per second. Simply "cherry picking" 24 images out of 60 does not present 24 frames with perfect temporal consistency, since more or less time may have elapsed between frames. The result is a slightly jittery picture, which appears to jitter in a cyclic fashion. Optical flow algorithms will analyze the footage and make corrections to the picture in order to better "fit" each frame into the new 24 frame sequence. The resulting footage is much smoother because it simulates equal exposure time between frames. For best results, footage should be deinterlaced and frame-doubled to 60p. This preserves all of the footage's temporal information, which is key in determining what the "missing" points in time should look like when converting to 24 frame/s. The last step is to compensate for the lack of
motion blur in the 60i footage. Since the images were captured at 1/60 second, there is less motion blur between images than there would have been if shot at 24 frame/s with a 180° shutter (i.e. 1/48 second exposure time). Optical flow is used to introduce motion blur between frames, mimicking the motion blur present when shooting the standard 180° shutter angle. This method of creating motion blur is far more realistic than simple frame blending, which is simple to implement and usually a standard feature in most non-linear editing programs. The optical flow method also works with 30p footage and is currently the best option for 30p to 24p conversion.
60i to 24p slow-motion conversions Using Adobe After Effects This method requires the use of
Adobe After Effects and applies to any interlaced material. It uses all of the temporal information in 50i or 60i footage to create the equivalent of a slow motion sequence shot at 50 or 60 frames per second, respectively. It also does not require multiple render passes to achieve the effect, avoiding generation loss from multiple compression cycles.
Using VirtualDub + AviSynth VirtualDub, along with
AviSynth, can be used to perform a 60i to 24p conversion in a similar way to After Effects. AviSynth performs the deinterlacing, then
frameserves the 60p half-resolution result to VirtualDub for further processing (specifically, adjusting field height using the "field bob" filter, resizing back to full resolution and then outputting at 24 frame/s). The reason AviSynth must be used is because VirtualDub cannot split the fields into a 60p sequence on its own, and this technique requires 60p input.
Displaying 24p material With NTSC equipment, it is impossible to display a 24p signal directly as the monitors only support the 60i framerate. Hence, pulldown must be added to the 24p material to be displayed. Most editing systems will either add 3:2 pulldown or 2:2:2:4 pulldown. In the 2:2:2:4 pulldown scheme, used as a choice primarily by Apple's
Final Cut Pro v7 and earlier, every fourth frame is repeated. This scheme is easier for slower hardware to implement as it requires less processing, but it introduces significant judder due to frame duplication. In
HD production, the HD-SDI interface supports the 24p framerate in addition to the 60i and 50i framerates. Many HD monitors are able to receive a 24p signal (not a 60i signal with pulldown added) and can display the 24p material directly. For end-user viewing of HD material, many digital formats are offering 24p support. Computer formats such as Windows Media, QuickTime, and
RealVideo can play 24p video directly on a computer monitor. Many early NTSC plasma and LCD monitors operated at 60 Hz and only supported 1080i (60i) or 720p (60p) content sources, requiring input signals 24p to be converted by the external source. Later 60 Hz monitors could accept 1080p24 content but employed a 3:2 pulldown to display 24p content, creating judder. Many monitors now support signal processing at 120 Hz or higher, allowing 24p content to be displayed without judder by showing each frame for a fixed number of refresh cycles. For example, a 120 Hz display can show each 24p frame for exactly 5 display frames. This capability is independent of the
motion interpolation features that are often associated with 120 Hz+ televisions.
24p compared to 30p As
Charles Poynton explains, the 24 frame/s rate is not just a cinema standard, it is also "uniquely suited to conversion to both 50 Hz systems (through 2:2 pulldown, 4% fast) and 59.94 Hz systems (through 2:3 pulldown, 0.1% slow). Choosing a rate other than 24 frame/s would compromise this widely accepted method of conversion, and make it difficult for film producers to access international markets". == Native 24p ==