David Burch
Forum Replies Created
-
Not sure if the two are related, but I would double check all your settings on the Kona and on the camera for your HDSDI out. Is the component out that you say works sending SD video or HD? If it’s SD, you need to switch a menu setting before the SDI will send HD video. Also, make sure the formats you are working with match. For instance, if your camera is in 1080p24 HQ, but you don’t have the 1080 mode set to PSF, the camera is actually sending out 1080i video with a 3:2 pulldown, so the Kona won’t recognize the video properly if it’s expecting 24p PSF or progressive.
-
John is correct on this. I have gotten in to the bad habit of calling any frame rate that’s not an integer “drop frame”. *sigh* Video was much more simple when all editing was done tape to tape 😀
-
I was not aware that broadcast specs included 60p…if so I stand corrected on that aspect 🙂 I do know that the stations around here only broadcast in 1080i or SD, and that has been the extent of my experience with broadcast.
By 30i I was implying 29.97 interlaced; I’m well aware there’s no such thing as 30i 🙂 My point was that I can see no advantages from a compression or bandwidth standpoint by shooting progressive over interlaced given the same amount of frames per second. By shooting 30p, you are essentially trading temporal resolution for image resolution.
I disagree completely with your claim that 30p looks smoother than 24p with a 3:2 pulldown, and here’s why: 30p uses no pulldown to play back in an interlaced format. Essentially, it is displayed as a progressive segmented frame. The result is a progressive-looking image on an interlaced screen, because there is no temporal difference between fields of the same frame. The 3:2 pulldown required for 24p, however, creates the illusion of higher temporal resolution because half the resulting frames are a combination of two temporally different source frames. In other words, 2 out of four frames contain fields that are temporally different, creating a smoother look. The cadence is as follows:
AA BB BA AB BB AA BB BA… ad infinitum.
The 3rd and 4th frames use fields that are temporally separated. Compare this with 30p:
AA BB AA BB AA…
None of the frames are temporally segmented, so the image appears to have less temporal resolution than the 2:3 pulldown of 24p, meaning the motion looks less smooth overall.
IMO, the only time 30p will look smoother than 24p is when both are viewed back to back in their native progressive format. I have to disagree with your claim about web video. Web video is almost always progressive, and rarely uses a full 30fps. The frame rate can be nearly anything, but most video I have seen is typically either 12 or 15 fps, in order to save on bandwidth. If you want to put something online at its full frame rate, 24p is an option, as is 30p. There would be no advantage to adding a pulldown to 24p for web video, as interlaced material generally gets de-interlaced anyway. But then, perhaps we are talking about different forms of delivery? I am essentially referring to things such as YouTube and Apple TV, both of which support progressive, 24fps video (to my knowledge).
One last thing; it is true that DVD players apply the pulldown when outputting to an interlaced monitor, but most DVD players these days support progressive scan output, which ONLY works when 24p material is being played back. As far as I know DVDs do not support a true 30p format, but since every frame in 30p contains no temporal segmenting it really doesn’t matter. An interlaced monitor will look progressive when displaying 30p, and a progressive monitor will de-interlace the video perfectly. Progressive scan DVD players, however, are capable of sending a true progressive (or possibly PSF) image to a screen that is capable of displaying it (LCD monitors, for instance). This is another advantage to shooting 24p, besides the fact that the limited bandwidth of DVD compression produces better results.
-
Yep…good point to remember 🙂 However, I don’t believe 23.98 is considered “pulled down” but rather drop frame. Pulldown is when fields are repeated in order to conform one frame rate to another; I don’t believe any frames or fields are being repeated in a 23.98 timeline, although it probably is necessary to repeat a frame every couple of thousand or so when transferring to film. I’ve never done a film-out so I’m not sure how that conversion works.
-
Don’t forget pulldown. IMO, footage shot at 24p with a 2:3 pulldown looks much better than footage shot at 30p, especially when seen on an interlaced monitor. Since broadcast is an interlaced medium, any footage shot at 24p needs a 2:3 pulldown added before it can be submitted. For this reason, 24p still results in a far better “film” look than 30p. IMO, any time I see 30p on the air it looks cheap, like they are cutting corners. Really, I can see no benefit of 30p over 30i in terms of bandwidth, as an interlaced and progressive frame both have the same number of pixels. However, for web distribution 24p does have the distinct advantage of having 20% less frames per second, and 24p looks just fine on a progressive scan computer monitor. Honestly, I really have a hard time understanding the desire for 30p in any situation.
-
Try following your subjects as much as possible, keeping the depth of field small, and avoiding medium-speed pans. These are all shooting techniques that can keep the “jittery” look down. Yes, 24p viewed natively will have just as much “jitter” as 30p, but so does film. The difference here is that 24p can have a 2:3 pulldown added that significantly reduces this. This is what happens when you play a movie from a DVD player on a non-progressive screen. The player is essentially repeating a field on every other frame, resulting in 2 progressive frames followed by 2 interlaced frames. This is not possible to do with 30p, and as a result 30p will look far worse on an interlaced screen.
As for 60p, I assume what Avid is doing is using the frames to create true interlaced 29.97 SD video. Remember, NTSC 29.97 is actually 59.94 fields per second. In other words, for every frame there are two exposures that are temporally separated and interlaced together. This is what gives 29.97 interlaced footage its smooth look, and represents the difference between 29.97i and 30p. 30p only has one exposure for every frame, so while the image will be sharper while objects in the screen are in motion the motion itself is less smooth. Another way of saying this is that 30p has less temporal resolution than 29.97i.
60p gives you the benefit of both temporal resolution and image resolution. Every frame you see has only one exposure, but there are twice as many frames per second, so the image is just as smooth as 29.97i yet just as sharp as 30p. This is the real advantage of 720p60, as the vertical resolution you have per frame is actually greater than in 1080i.
Back to your question about Avid, what is happening is that it is taking two frames and interlacing them together to create a single frame that is indistinguishable from one that was shot interlaced to begin with. This is why 60p makes a great format for downconversion; the result is a clean, true interlaced image that benefits from the source being progressive (and therefore easier to downconvert cleanly).
In other words, Avid is not “pulling out” any frames; it is using all of them to create 29.97i, which it could not do if you started with 30p.
-
I agree completely…content is in no way overrated. I’m simply pointing out that there’s more to the technical side than just working out problems, and that while content is the most important thing it isn’t the ONLY thing. Knowing exactly what setting to use in a specific situation is part of the art as well. Video and film are visual mediums, and everything that affects how the final product looks is part of the overall creative process, be it lighting, shooting, editing, or deciding on a frame rate. 🙂
-
I agree with Alan; for many things, the “film” look of 24p can actually add to the aesthetic, as long as it’s shot correctly.
As for workflow, I highly recommend staying in 24p throughout and simply adding pulldown when going from the computer to your final tape. The AJA Kona does a wonderful job of both adding pulldown and downconverting (if you need that) in real time, so no software frame rate conversion is necessary.
-
I have found I get the best results using our company’s Kona LHe to do the downconversion, but typically this isn’t practical. I used to use Compressor, and before FCS 3.0 came out it did a fantastic job. Since then, however, I have been getting very poor results with it, and have stopped using it for anything that needs downconversion.
Our company just started to produce Blu Ray discs as well as DVD, so lately I’ve been authoring in Adobe Encore and using Adobe’s media convertor to handle my downconversion. It seems a tad softer than the results I used to get with compressor, but overall the image quality is good enough for DVD. One thing you might try is MPEG Streamclip, which can be found free online. I have had very good results with this program across the board, with everything from ripping DVDs to capturing still frames, downconverting to de-interlacing.
Overall, it can be very difficult getting downconverted footage to look like true, broadcast-grade SD, and if that is what you need then I’d say a hardware downconverter will be your best bet.
-
“I have found on the EX it is noisier at F1.9 at -3 db gain than at F4 with 6 db of gain.”
Really? That surprises me. I normally try to keep my iris as open as possible because, in my opinion, the stock lens on the EX-1 and EX-3 looks much better the more open it is. I have found that anything more closed than F8 tends to start looking soft. Also, I typically shoot for as shallow a depth of field as possible.
I have actually been very pleased with how these cameras handle low light, and have not found them to be excessively noisy at all, except for in certain conditions. One thing I have found to be helpful in minimizing noise is to try to shoot in light that has a higher color temperature. Typically, in sub 3000K lighting the camera has to compensate by boosting the blues, which is where a lot of your noise can come from. There are some good PP settings floating around on the web that help minimize this, but if you have control of it try to keep the color temp up if possible.
Gamma settings can play a huge role in noise as well, as they can essentially be applying gain to the image even when the camera is set to 0dB.
If it’s true that the camera is noisier at F1.9 in -3 than at F4 in +6, then something strange is going on. The lens aperture should have nothing to do with noise, and if an image is correctly exposed at F1.9, the sensor should have enough light to minimize noise. What matters is how much light is hitting the sensor, not how big the aperture is.
That said, I never use auto iris, and a lot of the scenes that I shoot would probably cause the auto iris to overexpose. (I rarely have any control over lighting, as most of what my company shoots is live events.) Typically. my rule of thumb is to try to keep the iris as open as possible, which usually means my gain is set to either 0 or -3, and so far I have been very happy with the results 🙂