-
720/30p – 720/60p – 1080/30p Best for SD down convert?
In the BBC Tech Report, “Assessment and settings for Sony PMW-EX1 and
EX3” Alan Roberts writes:“There is no standard definition recording mode in the EX1. However, it records at 1280x720p in HQ mode (35Mb/s), and this mode is interesting because it represents the best way to get a standard definition picture from the camera. If recordings are made at 1080 interlaced, then the down-converter (external) will have to de-interlace in order to produce the output fields, while recording at 1080 progressive may not give the look the user wants (jerky motion).”
“Since the sensors are cmos and are scanned progressively, it could make sense to shoot in 720p with the view to down-conversion to SD, since there would then be no need for de-interlacing, and the down-conversion would have full frames to work with, at the output field rate.”
“Clearly, a subsequent down-conversion to standard definition from this format should be the best route to take. It is highly unusual to see such good downconversion in a camcorder.”
These comments don’t make it completely clear what frame rate is being assumed for the 720p. Since he mentions “jerky motion” in the 1080p case (which is maxed out at 24 or 30 fps), it seems implied that he is assuming a doubled frame rate (48 or 60 fps). But then the comment “full frames to work with” leaves me confused: what use will the mepeg compressor make of the two full frames from the doubled frame rate when ultimately it encodes it down to at 24 or 30 fps SD video?
On a related question: if I have no intention of using any slow motion effects, but perhaps I will use some minor zooms, stabilization, and centering in post – then which is the better format?
A. 1080p/30p (so the zoom resizing has more to work with)
B. 720/30p (so there are fewer compression artifacts from the 35mbs compression ??)
C. 720/60p (so the 16 gop encoding gives an I frame about every 1/4 second instead of every 1/2 second, thus helping the mpeg encoder ??)
I should mention that in my case I will be filming classical concerts. Lighting is often low. Large scale motion is low, pans are slow – but there is some extremely fast arm motion of the string players during fast passages.
Any insights will be appreciated – and apologies in advance if this ground has been covered definitively before.
-
18 Replies
-
I am not the most informed guy at this forum (for sure) but as you’ve not had a reply yet – sometimes a reply often triggers some of the senior hosts to post answers that set you straight. I will tell you this: 1920x1080p24 is a very nice format to make the best use of the full CMOS sensor, and give you a filmic look (24 progressive frames per second). This gives you the greatest resolution which always presents the most quality when down converting to SD. As for the decision (opinion) of 24 versus 30, well, I think that most DVD’s are suited to play 24 frames back nicely because most DVD’s that you watch are feature films that were shot in 24 frames a second – and also, you will thereby have more space if your symphonic program runs long.
1920×1080 also gives you the best DOF but there is a lot of things that go into making a decision like this. Is this a single camera shoot? Meaning are you going to be wide most of the time? or is there a second and or third camera for cutaways and CU’s. If it’s a static doc then I stand by my first choice. If it’s a multi cam shoot, you might make use of the EX1 shutter and overcrank settings to capture crisp fast motions that are tight since the primary establishing camewra is getting the audio feed and running at 24fps. I think that’s all I should say and you can be sure you will get better answers from some of the forum leaders now. -stephen
Stephen May
Keystone Media Productions
Freelance Videographer -
I believe that where Mr. Roberts speaks of having a “full frame to work with” he means that if, for example you’re in NTSC territory (like me) and you shoot 720p60 (actually 59.94) you have a full progressive frame from which to create each field in the downconvert. As far as the compression pros and cons I frankly think you’re thinking too much there. I think you should look at frame rates as more of an aesthetic issue. If it were me, I’d do as Mr. Roberts suggests, 720p60. I use it all the time and my downconverts are gorgeous.
Good Luck!
Bobhttp://www.lightstreamassociates.com
Mac Pro 2.66 – 8GB memory – Mac OS 10.5.5 – Quicktime 7.5.5 –
Mac Book Pro 2.33 Duo –
FC Studio 2 (Final Cut 6.0.4) – Kona Lhe
Adobe Production Suite CS3 –
Sony Multiformat 14″ – Panasonic 42″ Plasma –
Ikegami HLDV7 – PVW EX-1 -
Bob thanks for jumping in here! -can you help me get that (it’s making sense but I don’t know if I get it completely):
So you shoot 1280x720p60 and you cut in a sequence that assumes that res. in other words you’re editing in the exact same res as your shot media 720p60, and how do you go from there to arrive at your downconvert to SD? do you export the sequence to compressor? Do you mind sharing this?
Stephen May
Keystone Media Productions
Freelance Videographer -
Stephen,
Don’t mind at all. I edit in a 720p59.94 timeline and either use the Kona card’s downconvert to master to tape or export a self-contained movie (still 720p) and use compressor to whatever format is needed (usually dvds). The question of codec is more controversial. I’ve had good luck sticking with xdcam sequence setting (with the option for “render with prores” selected). If you’re doing lots of compositing and such you might want to either capture your footage to pro res, DVCPRO HD or uncompressed. I recently did a chroma key project using EX1 footage rendered with After Effects Keylight and the xdcam worked as well as any DVCPRO HD footage I’ve keyed with.There’s lots of other opinions out there, but that’s what’s worked for me.
Good luck,
Bobhttp://www.lightstreamassociates.com
Mac Pro 2.66 – 8GB memory – Mac OS 10.5.5 – Quicktime 7.5.5 –
Mac Book Pro 2.33 Duo –
FC Studio 2 (Final Cut 6.0.4) – Kona Lhe
Adobe Production Suite CS3 –
Sony Multiformat 14″ – Panasonic 42″ Plasma –
Ikegami HLDV7 – PVW EX-1 -
Thanks for the response – and you probably won’t be surprised that I’m still confused.
Assuming that no slow motion effects will be applied, what is gained by shooting 60p instead of 30p?
As you point out, 60p gives a full frame available as source for each interlaced output frame (NTSC), but assuming that the desired output is going to be progressive anyway, these interlaced frames will be merged back together – so it seems like you would want to be pulling the interlaced half frames from the same source frame rather than adjacent source frames.
In other words – II don’t understand how the extra information available in a 60p source can give any significant benefit when it is transcoded to a 30p output. Or, relating my question back to Alan Roberts analysis for BBC: I don’t understand how the “jerky motion” that he sites as a characteristic of 1080/30p can be avoided by shooting at 720/60p when both will be transcoded to 30p for final viewing.
-
I think the way Mr. Roberts phrased it is a bit confusing. I think what he means is that progressive frames make for cleaner downconverts, but the the tradeoff is that at 1080 you can only have 30p, which is half the temporal resolution, and so it can result in the studdering (jerky motion) that some of us object to. If you go with 720, you can shoot 60 progressive frames which make for very clean downconverts without the jerkyness. Many perceptual studies have shown that it’s pretty difficult to perceive the resolution difference between 720 and 1080 – especially on screens smaller than 42″, and so makes a great choice if your deliverable is standard def.
That’s why I’m such a fan of 720p60. Do some tests and see what you think. There are no wrong answers, just creative choices. Isn’t this stuff fun??
Bobhttp://www.lightstreamassociates.com
Mac Pro 2.66 – 8GB memory – Mac OS 10.5.5 – Quicktime 7.5.5 –
Mac Book Pro 2.33 Duo –
FC Studio 2 (Final Cut 6.0.4) – Kona Lhe
Adobe Production Suite CS3 –
Sony Multiformat 14″ – Panasonic 42″ Plasma –
Ikegami HLDV7 – PVW EX-1 -
Bob, thanks for the dialog.
Here is my test:
1) Use Quicktime Conversion to export a 15 second clip from the XDCAM 720x60p source. Save as uncompressed 59.94 fps (same frame rate as the source).
2) Use Quicktime Conversion to export this same clip uncompressed, but this time at half the frame rate, 29.97.
3) Bring these two exported clips back into FCP and insert both into an uncompressed 59.94 fps timeline.
As far as I can see, these steps yield the expected results:
1a) Comparing the exported 59,94fps clip to the original XDCAM source – I can’t see any difference.
2a) Comparing the clip exported at 29.97 to the original 59.94 source – I can’t see any difference when comparing the export to every other frame from the original (e.g. the export seems to have simply dropped every other frame as expected).
Note that the when both clips are brought into the 59.94 uncompressed timeline, the 29.94 clip needs rendering because every other frame is missing. After rendering, I was a little surprised to see that the missing frames were just filled in with a repeat of the previous frame – there was no interpolation.
Now the crucial test: use compressor to export my uncompressed timeline that is composed of exactly the same clip at the two different frame rates. Will the compressor make any use of the extra information in the 59.94 fps portion when it does the conversion to standard mpeg-2 NTSC video at 29.97fps? To my eyes, no!
The portion of the mpeg-2 movie that was compressed down from 59.94fps looks the same as the portion that was compressed down from 29.97. Some of the fast motion portions are blurred by the lower resolution of NTSC and by the mpeg-2 compression. It looks significantly worse than the 720p source but I can’t see any difference between the portions coming from the two frame rates.
My tentative conclusion: if the ultimate format is going to be 29.97 fps and there will be no slow-mo or effects that change speed then the extra “temporal resolution” of shooting at 720x60p is not used in the conversion to 29.97. So its better to allocate the 35Mbs data limit of HQ XDCAM for spatial resolution. That means either 1080x30p or 720x30p. And if there are going to be digital zooms, pans, centering, etc. in post then it seems like the 1080x30p would be the best starting point.
Are my conclusions wrong – I don’t have any real experience to back them up. Just a few experiments and trying to understand how all this stuff probably works. Note: my experimental clip is string players with bows whipping around in fast passages. Maybe a similar experiment using a clip where the camera is panning (so the motion is full frame instead of localized) would show very different results? Anyone done similar tests for panned footage?
-
I just did some rudimentary tests with an EX3 for a percussion project (fast moving hands and mallets), and my thoughts are that 1080 30P has twice the exposure time as 720 60P, and therefore exhibits more pronounced blurring during fast movement. I suppose this can be referred to as “temporal resolution,” and it seems that 720 60P has more of it. I just filmed myself clapping my hands at various frame-rates, and 720 60P had the least amount of motion blur. It’s not a definitive test, but you may want to consider it. I think I’m going to shoot my project using 720 60P.
-
But if you are controlling the shutter speed manually (instead of turning off the shutter or letting the automatic exposure control it), then can’t you control the amount of smear or blur based on the shutter?
With shutter turned off, it seems that 60p capture is going to show ‘half’ the smear of 30p. But setting the shutter to 1/60 or faster for both 60p and 30p should smear about the same. Note: setting to 180 degrees would give 1/120 (for 60fps) versus 1/60 (for 30fps) – not the same as setting both to the same shutter speed.
And then with equal shutter speeds, we are back to the question of what happens when you down convert to 30fps mepeg-2. My test seems to say the doubled number of frames in 60p doesn’t give any benefit after down conversion.
-
Denis, did you check your dvd on a crt? On a computer, lcd or plasma you won’t see the interlacing, which is where those extra frames come into play. I haven’t done a 30p downconvert yet, but I’ve done lots of 60p downconverts and I’m positive that I’m seeing interlacing (coming out of the Kona).
The issue of shutter speed is yet another aesthetic consideration. 60p has a minimum shutter speed of 1/60, whereas 30p is 1/40. The slower shutter gives a bit more motion blur, which some say gets you closer to film look.
If you have access to a camera that would probably be the best way to test your theory.
Bobhttp://www.lightstreamassociates.com
Mac Pro 2.66 – 8GB memory – Mac OS 10.5.5 – Quicktime 7.5.5 –
Mac Book Pro 2.33 Duo –
FC Studio 2 (Final Cut 6.0.4) – Kona Lhe
Adobe Production Suite CS3 –
Sony Multiformat 14″ – Panasonic 42″ Plasma –
Ikegami HLDV7 – PVW EX-1
Reply to this Discussion! Login or Sign Up