dumb question on capture FPS

[friendly_jacek]

The NTSC format has near 60 interlaced fields (half frames) per second (exactly 50 in PAL). when we capture at near 30 frames per second in NTSC (25 in PAL), we merge 2 fields into one frame giving the interlacing artifacts visible on progressive display devices. even displaying 30 (25 in PAL) merged frames per second on interlaced display is not the same as showing 60 (50 in PAL) fields per second, as in the original analog source, is it?

Wouldn't capturing at twice the frame rate help with the interlacing artifacts?

[poisondeathray]

The NTSC format has near 60 interlaced fields (half frames) per second (exactly 50 in PAL). when we capture at near 30 frames per second in NTSC (25 in PAL), we merge 2 fields into one frame giving the interlacing artifacts visible on progressive display devices. even displaying 30 (25 in PAL) merged frames per second on interlaced display is not the same as showing 60 (50 in PAL) fields per second, as in the original analog source, is it?

Wouldn't capturing at twice the frame rate help with the interlacing artifacts?

No , because the NTSC source is transmitted at 29.97 (30000/1001) FPS (or 25 FPS for PAL) with the fields weaved . If you captured at twice the frame rate, each frame would be duplicated

[manono]

Wouldn't capturing at twice the frame rate help with the interlacing artifacts?

So, rather than capping 720x480 at 29.97fps, you're proposing capping 720x240 at 59.94fps? That's right, no more interlacing artifacts, but a pretty odd-looking picture. Or are you suggesting bobbing the fields while capping? If so, why, since you can do that, if you wish, afterwards, and using a better bobber than whatever is being used during the capture.

when we capture at near 30 frames per second in NTSC (25 in PAL), we merge 2 fields into one frame giving the interlacing artifacts visible on progressive display devices.

Which is what deinterlacers are for. For example, when playing an interlaced DVD to your progressive LCD Hi-Def television, you don't see any interlacing, do you?

[LMotlow]

The NTSC format has near 60 interlaced fields (half frames) per second (exactly 50 in PAL). when we capture at near 30 frames per second in NTSC (25 in PAL), we merge 2 fields into one frame giving the interlacing artifacts visible on progressive display devices.

Deinterlacing artifacts are visible in PC video editors -- but not in my PC media players, which deintelace the same way my tv and external players do. The "signature" of interlaced vido displayed as-is without deinterlacing is called "combing", which is not strictly considered an artifact unless your deinterlacer is doing something wrong.

[jagabo]

Analog NTSC video is transmitted one field at a time, 59.94 fields per second. When you watched that signal on a CRT TV that's exactly what you saw, one field at a time, never an entire frame. A 29.97 frame per second capture contains all those fields with pairs of fields packed into each digital frame. On a progressive display a simple bob deinterlace closely simulates what the analog video looked like on a CRT.

If you deinterlace before saving the video you will have twice as much data to save. And the artifacts of the deinterlacing will be locked into that digital video making them much harder to fix. All the extra work during capture will risk dropping frames. With modern computers that's not much of a problem but 10 or 20 years ago it wasn't possible.

[friendly_jacek]

The NTSC format has near 60 interlaced fields (half frames) per second (exactly 50 in PAL). when we capture at near 30 frames per second in NTSC (25 in PAL), we merge 2 fields into one frame giving the interlacing artifacts visible on progressive display devices. even displaying 30 (25 in PAL) merged frames per second on interlaced display is not the same as showing 60 (50 in PAL) fields per second, as in the original analog source, is it?

Wouldn't capturing at twice the frame rate help with the interlacing artifacts?

No , because the NTSC source is transmitted at 29.97 (30000/1001) FPS (or 25 FPS for PAL) with the fields weaved . If you captured at twice the frame rate, each frame would be duplicated

no read my post again. there is a difference between frame and field (frame = 2 fields). if i captured at 2x frame rate i would get each field separately and not duplicate frames.

[pandy]

no read my post again. there is a difference between frame and field (frame = 2 fields). if i captured at 2x frame rate i would get each field separately and not duplicate frames.

No, you capture frame twice - i.e. duplicate frame - btw remember that for static part there is no temporal difference between fields.
Fields have only half vertical resolution of the frame.

[friendly_jacek]

So, rather than capping 720x480 at 29.97fps, you're proposing capping 720x240 at 59.94fps? That's right, no more interlacing asrtifacts, but a pretty odd-looking picture. Or are you suggesting bobbing the fields while capping? If so, why, since you can do that, if you wish, afterwards, and using a better bobber than whatever is being used during the capture.

yes i was thinking something like line doubling, i guess you can call it bobbing. but it would be done at 59.94/s rate. after the video is captured at 29.97/s the results should be different.

[poisondeathray]

The NTSC format has near 60 interlaced fields (half frames) per second (exactly 50 in PAL). when we capture at near 30 frames per second in NTSC (25 in PAL), we merge 2 fields into one frame giving the interlacing artifacts visible on progressive display devices. even displaying 30 (25 in PAL) merged frames per second on interlaced display is not the same as showing 60 (50 in PAL) fields per second, as in the original analog source, is it?

Wouldn't capturing at twice the frame rate help with the interlacing artifacts?

No , because the NTSC source is transmitted at 29.97 (30000/1001) FPS (or 25 FPS for PAL) with the fields weaved . If you captured at twice the frame rate, each frame would be duplicated

no read my post again. there is a difference between frame and field (frame = 2 fields). if i captured at 2x frame rate i would get each field separately and not duplicate frames.

No read my post again.

The fields are weaved in the original signal into a single frame. If you capture the original signal at 59.94p you will get duplicate frames

But if you alter the original signal (ie. bob deinterlace) then the fields are separated temporally, no longer weaved into a single frame, then yes, capturing at 59.94p will make a difference, because that is the new signal

For the reasons mentioned above, there are more negatives than positives. Additionally, a simple bob deinterlace in real time will yield much lower quality than non real time (post processing) deinterlacing algorithms

[friendly_jacek]

On a progressive display a simple bob deinterlace closely simulates what the analog video looked like on a CRT.

this is the part that i'm not sure and hence asked a question. i see how bob deinterlace can nicely handle each field and make a frame from it to produce near 60 frames per second or smooth motion. this is why progressive DVD player replaced the inferior early interlaced ones. i don't see how after you capture at 29.97/s, and have the frames with sandwiched 2 fields already, the bob principle applies anymore. please explain that.

[friendly_jacek]

If you capture the original signal at 59.94p you will get duplicate frames

we have to disagree on this. should be easy to test though on fast moving objects.

The NTSC interlaced system shows 60 unique images (fields) per second, and each field is an independent snapshot in time. So we don't want to think of interlaced televisions in terms of frames but rather in terms of fields, interlaced fields, and 60 of them per second.

http://www.hometheaterhifi.com/volume_7_4/dvd-benchmark-part-5-progressive-10-2000.html

[Skiller]

Well, if you capture the signal at 240p 59.94p you basically capture the same data but store it differently. But it doesn't really make sense to store fields individually because they belong in a frame where a bob-deinterlacer would have to interpolate every other line that is missing.

[poisondeathray]

If you capture the original signal at 59.94p you will get duplicate frames

we have to disagree on this. should be easy to test though on fast moving objects.

The NTSC interlaced system shows 60 unique images (fields) per second, and each field is an independent snapshot in time. So we don't want to think of interlaced televisions in terms of frames but rather in terms of fields, interlaced fields, and 60 of them per second.

http://www.hometheaterhifi.com/volume_7_4/dvd-benchmark-part-5-progressive-10-2000.html

Yes, 29.97 frames per second interlaced is 59.94 fields per second (there is 59.94 moments in time represented per second). But the transmitted signal is interlaced (each field pair is weaved into a single frame - that's why you see what people call "interlaced artifacts" during motion in the first place) . If you just set the recorder to 59.94 frames per second without applying a deinterlace filter (bob deinterlacing it returns 59.94 progressive frames - essentially each field is separated and resized to a full frame dimension) - you will get duplicates - this is 100% fact. Go ahead and test it yourself

Here is another way of thinking of it: 59.94 frames per second is is 119.88 fields per second

[friendly_jacek]

Yes, 29.97 fields per second interlaced is 59.94 frames per second. But the transmitted signal is interlaced (each field pair is weaved into a single frame - that's why you see what people call "interlaced artifacts" in the first place) . If you just set the recorder to 59.94 frames per second without applying a deinterlace filter (bob deinterlacing it returns 59.94 progressive frames - essentially each field is separated and resized to a full frame dimension) - you will get duplicates - this is 100% fact. Go ahead and test it yourself

what you fail to understand is that the "weaved" fields are separated in time, all 59.94 of them in each second. the "weaving" never happens on a CRT display, except for inertia of phosphorus.

since you are so sure, did you do that test yourself?

[friendly_jacek]

Well, if you capture the signal at 240p 59.94p you basically capture the same data but store it differently. But it doesn't really make sense to store fields individually because they belong in a frame where a bob-deinterlacer would have to interpolate every other line that is missing.

yes, this was exactly my point, but i was not eloquent enough to express in my OP. my concern is that forcing the fields together into the frames when capturing at 29.97/s introduces some artifacts beyond the usual analog TV limitations.

[poisondeathray]

Yes, 29.97 fields per second interlaced is 59.94 frames per second. But the transmitted signal is interlaced (each field pair is weaved into a single frame - that's why you see what people call "interlaced artifacts" in the first place) . If you just set the recorder to 59.94 frames per second without applying a deinterlace filter (bob deinterlacing it returns 59.94 progressive frames - essentially each field is separated and resized to a full frame dimension) - you will get duplicates - this is 100% fact. Go ahead and test it yourself

what you fail to understand is that the "weaved" fields are separated in time, all 59.94 of them in each second. the "weaving" never happens on a CRT display, except for inertia of phosphorus.

since you are so sure, did you do that test yourself?

I'm 100% sure. Yes I've tested it with an external recorder. When you capture the original signal at "60p" you will get 59.94p or 119.98 fields per second . Ie. duplicates

Yes, the individual fields represent individual moments in time. But what you fail to understand is they are not stored separately. That's what interlace means. 2 field pairs are weaved together into a single frame - that's why you see the "interlace" artifacts and combing . When you deinterlace, you separate the individual fields and resize them to full frames. That's what bob deinterlacing is - you get full 59.94p

There are different methods of deinterlacing and interpolating the missing information - this is a very large topic, but doing it real time with a simple bob (simple field resize) in real time will always yield lower quality results . You will "bake" in the low quality if you apply a deinterlace filter before capture . That's why you want to capture the original signal - not only is it lower bandwidth, you have many processing options later

[jagabo]

i don't see how after you capture at 29.97/s, and have the frames with sandwiched 2 fields already, the bob principle applies anymore. please explain that.

It's just the opposite of weaving two fields together into a frame. At playback you separate the two fields and display them individually. Every knows how to do this. It's only poor software on PCs that don't do this. Or when encoded improperly so that the player can't tell the video is interlaced.

[manono]

i see how bob deinterlace can nicely handle each field and make a frame from it to produce near 60 frames per second or smooth motion. this is why progressive DVD player replaced the inferior early interlaced ones.

No, it isn't. Modern progressive television sets have their own deinterlacers built in and can handle old-style DVD players outputting fields easily. What can make a progressive-scan DVD player worthwhile is its ability to take telecined DVDs and not deinterlace them, but rejoin and realign the fields and get rid of the duplicate fields and output the original unfiltered frames. This on-the-fly IVTC should produce a better picture than one that's been deinterlaced, no matter how good the deinterlacer. And no bobbing takes place.

[friendly_jacek]

I'm 100% sure. Yes I've tested it with an external recorder.

you're right! i just tested that too. 50 fps (PAL signal) just duplicated the frames. looks like the capture card automatically de-interlaces the fields by weaving on capture. i learned a lot from this discussion. that puts the whole dogma "don't de-interlace for DVD encoding" in question, doesn't it?

[jagabo]

I'm 100% sure. Yes I've tested it with an external recorder.

you're right! i just tested that too. 50 fps (PAL signal) just duplicated the frames. looks like the capture card automatically de-interlaces the fields by weaving on capture. i learned a lot from this discussion. that puts the whole dogma "don't de-interlace for DVD encoding" in question, doesn't it?

That's only for film based sources. For live video like sports and news you will have 50 unique pictures per second.

[poisondeathray]

I'm 100% sure. Yes I've tested it with an external recorder.

you're right! i just tested that too. 50 fps (PAL signal) just duplicated the frames. looks like the capture card automatically de-interlaces the fields by weaving on capture. i learned a lot from this discussion. that puts the whole dogma "don't de-interlace for DVD encoding" in question, doesn't it?

actually no - if you have your capture setup to record 50p, assuming it was true interlaced content - it shouldn't be deinterlacing unless you have specified deinterlace somewhere (hopefully "bob", not single rate) - in that case it would capture true 50p without duplicates . If it was single rate deinterlacing, it would be 50p with duplicates (25p content, half motion sampled)

If you just recorded 50p without any processing (as-is, and your card isn't doing any funky business), then you will get 25i with duplicates (it will be essentially the same thing as a 25i capture, but with duplicate frames)

[friendly_jacek]

actually no - if you have your capture setup to record 50p, assuming it was true interlaced content - it shouldn't be deinterlacing unless you have specified deinterlace somewhere (hopefully "bob", not single rate) - in that case it would capture true 50p without duplicates . If it was single rate deinterlacing, it would be 50p with duplicates (25p content, half motion sampled)

If you just recorded 50p without any processing (as-is, and your card isn't doing any funky business), then you will get 25i with duplicates (it will be essentially the same thing as a 25i capture, but with duplicate frames)

i think we're running into semantics. i have no way to record 50p (whatever that is?), just captured 50 frames per second. it was interlaced signal alright, from my old PAL analog camera. the fact that the frames came with both fields in them, that deinterlacing by definition, right?

[friendly_jacek]

That's only for film based sources. For live video like sports and news you will have 50 unique pictures per second.

i used a live video camera feed. this is as live as it comes.

[poisondeathray]

actually no - if you have your capture setup to record 50p, assuming it was true interlaced content - it shouldn't be deinterlacing unless you have specified deinterlace somewhere (hopefully "bob", not single rate) - in that case it would capture true 50p without duplicates . If it was single rate deinterlacing, it would be 50p with duplicates (25p content, half motion sampled)

If you just recorded 50p without any processing (as-is, and your card isn't doing any funky business), then you will get 25i with duplicates (it will be essentially the same thing as a 25i capture, but with duplicate frames)

i think we're running into semantics. i have no way to record 50p (whatever that is?), just captured 50 frames per second. it was interlaced signal alright, from my old PAL analog camera. the fact that the frames came with both fields in them, that deinterlacing by definition, right?

Yes, by 50p I mean 50 progressive frames per second . I think we are on the same page with that.

But there is a distinction between content and recording. You can have interlaced content recorded progressively, or vice versa. The type of recording can affect how devices and software handle it. For example, if you have interlaced content recorded progressively, then the target HW or SW might not know to bob deinterlace it and you will see combing. If you record it "normally" ie. 25fps interlaced content, encoded and flagged interlaced - there is a higher chance that SW or HW will handle it properly automatcially

Different capture hardware and/or software might be doing other processing "behind the scenes." Some automatically adjust or sync to the signal, others can override in software. So if it "sees" a 25fps / 50 fields per second signal, your HW might be adjusting to capturing that - it's difficult to say unless you provide a sample or more information on the setup

A pure unprocessed 25frames/s (50fields/s) signal recording from a live feed recorded at 50 frames per second would have duplicate interlaced frames because 50 frames/s is 100 fields/s . If you are seeing something else, it might be your viewing/analyzing method isn't ideal, or maybe your capture setup is doing something differently - again tough to say unless you provide more info

[friendly_jacek]

Update: i did some experiments with recently downloaded neat video filter. even though VD shows weaved fields on the monitor screen, neat video sees separate fields in each frame. furthermore, the BT878 documentation (easy to find in pdf format) clearly states that each field is outputted separately in memory (ie interlaced). mystery solved and thanks to everyone who helped to steer me in the right direction.

i have to admit the video thing is hard to wrap the mind around. i thought i knew well how TV (and radio) worked as a child, but the capture thing is challenging that. i have a renewed regard for all those analog electronics engineers.

[jagabo]

i have to admit the video thing is hard to wrap the mind around. i thought i knew well how TV (and radio) worked as a child, but the capture thing is challenging that.

Why? It's trival. The analog signal consists of a sequence of fields. When captured the fields are paired together into frames. On playback the frames are separated back into fields for display in the same order as the original analog signal.

[poisondeathray]

If you loaded a file as-is into vdub, it will not deinterlace unless you add a filter - so if you see "combing" during motion, it will be interlaced - that's how it should be. If you see no combing artifacts , not even during motion - then something went wrong with your capture, it was deinterlaced somewhere, or your decoder is applying a deinterlace

In neat video, you can process interlaced material by selecting "video scan type : interlaced" (or progressive by selecting progressive)

However, temporal filtering effectiveness is greatly diminished by the presence of duplicates . If you recorded at "50 frames per second", you should have duplicates unless it was bob-deinterlaced somewhere. I would decimate duplicates before filtering

[friendly_jacek]

When captured the fields are paired together into frames. On playback the frames are separated back into fields for display in the same order as the original analog signal.

well, i didn't know that separation part, i kind of suspected it was a dumb question from the beginning though. i reread the whole thread and everyone was trying to explain it from the beginning, but it took a while and some experimentation for me to sink in. now it it crystal clear.

thanks again. this website is awesome after all.

[jagabo]

In theory one could record at 50 frames per second where each frame is really a field. But then you have to keep track of which frames are top fields and which are bottom fields. Their position when viewed on the screen is different. The top field is drawn on scanlines 0,2,4,6... the bottom field is drawn on scanlines 1,3,5,7...

If you want to save your videos this way you can do so in AviSynth with SeparateFields().