I'm considering purchasing the Canon HF-S100. I'm especially interested, yet cautiously skeptical, in this camera's 30p mode (internal 60i-to-30p deinterlacing, right?).

I currently deinterlace my raw footage (HV10 60i mts files of mostly high-action sports with moderate panning) with Avisynth filter TempGaussMC. I prefer to deinterlace before editing. This deinterlacing step takes forever, but quality output is worth the time, IMO. Crisp results with no detectable scanlines, jaggies, or artifacting. It looks sweet. The S100's 30p, if the quality is on par, would save me a ton of time.

A few questions:

• If the original shooting is in 1920x1080, does the 60i to 30p conversion maintain 1920x1080 (but 30p) resolution, or is it downscaled to something like 1280x720?
• How's the deinterlacing quality? Does the 30p contain any noticeable scanlines, teeth, blurriness, or other artifacting?
• Anyone compared this camera's 30p output with any 30p from a software-based approach (see my test idea below)? Results?

Test idea: Take both 60i and 30p footage of a spinning windmill. Deinterlace the 60i in software. Compare that to the camera-deinterlaced 30p footage.

I've seen sample S100 footage at youtube and vimeo, and it looks nice, but not sure if that was from captured 60i, 24p, or 30p, then encoded for those sites.

Thanks for any feedback, objective or subjective, anecdotal or first hand.

Everything else being equal, shooting in "True" progressive mode is better than de-interlacing (unless your video has little or no movement). De-interlacing does not fix the "different time each frame was taken".


Rich

__________________
FireWire Survey

Thanks. So there's more than just deinterlacing going on to get 30p (or 24p).

And, just to clarify, I suppose you mean "True" progressive mode in the sense of enabling the 24p or 30p mode in the HF-S100, and not "native" progressive mode, which the camera doesn't appear to support.

All current (and last year's) VIXIA models have the ability to capture 24 and 30 progressive frames natively. The DiG!C image processing chip inside the camcorder then breaks up those progressive frames into fields and puts those fields into a 60i stream. For 30p, this is easy, as each frame produces two fields and those two are placed at the top and bottom in the 60i stream. For 24p, the chip does the 2:3 pulldown and some fields are duplicated according to the 2:3 telecine cadence. The end result is, regardless of the capturing frame rate (24p, 30p, 60i), the resulting video stream will always be 60i, but depending on how the frames were captured, it will contain different temporal information. Extracting true 30 progressive frames out of 60i stream can be done by many software utilities on both Mac and Windows, and some of them are even free. Extracting the 24p frames requires removing the 2:3 pull-down, which involves detecting the pulldown cadence from the interlaced stream and removing the duplicated fields. This feature is not available in the free software, but there are tools (such as Cineform's NeoScene, or After Effects CS4) that can automate the process.

Bottom line, long story short, the result from any current or recent VIXIA will produce MUCH better 30p video after de-interlacing than what you get when you de-interlace a 60i video that was captured at 60i.

For the life of me, I still don't understand, why does Canon insist on shoe-horning progressive frames (be it 24p, 25p or 30p) into a 60i (or 50i in EU models) stream. These are all HD camcorders. I don't think that ANYBODY who owns a HD camcorder is watching it on a SD TV. Therefore, every single target medium for this video will be able to support a progressive signal natively, whether HD TV set (via HDMI), or some computer software (Captured files, or AVCHD transfer). There is absolutely no point in putting progressive frames into an interlaced stream. All it does is forces those who consciously choose to shoot progressive to take the extra step of deinterlacing (and let's not even talk about pulldown removal) before they can even begin to edit.

This is just a colossal hassle and I'm still waiting for Canon to give a plausible explanation.

@vasic
So, if I record in 30p and then analyze the captured mts files (or do Match media in Sony Vegas project setup), it will show up as 60i? Then, I still have to instruct Vegas or some other renderer to de-interlace it to get the 30p I wanted in the first place?

If anyone can point me to one or a few sample 30p mts files for download from the HF-S100, I'd appreciate it. I just want to test out the workflow a bit on my home setup. That might win me over, or not.

Canon's 60i wrapper decision must have roots in some kind of marketing/up-selling/revenue/cost-cutting motivation. Like (my forecast only), "here are the new half-baked camcorders for 2009, then in 2010, after CES, we'll offer the same exact cameras, but with native progressive capture. Then you'll all buy those, too." Lame.

Yes. Most common workflow is to use some stand-alone deinterlacing software, such as JES Deinterlacer (on Mac) or something similar on Windows, transcoding the AVCHD stuff into AIC or ProRes (on Mac), or some AVI/WMV in Windows. This makes the editing easier than working with AVCHD directly, and the stream ends up 30p you wanted in the first place.


You're probably right, although I have a feeling that a possible excuse could be to make the video playable on Standard Def TVs via the built-in analogue composite output (NTSC in US, PAL in EU). I don't believe any one owner of ANY recent VIXIA (and iVIS, and Legria in EU and Japan) models had ever hooked up their camcorder to a Std-Def TV set using those analogue outputs. In other words, the only reason they're really there is probably the legacy from early days of HD, when SD compatibility was necessary. More than half of developed world already has HD screens at home, and those are the people who buy HD camcorders.

Let's wait for 2010 CES and the next generation of VIXIAs. Perhaps Canon will surprise us? I won't hold my breath...

Can you please cite some sources on this as it seems to be a lot of confusion. From CCIs HF-S11 review:
http://www.camcorderinfo.com/conten...Performance.htm
From the review of the yet-to-be-available Canon HF-S21:
http://www.camcorderinfo.com/conten...Performance.htm
If this is true, then I would expect that the (on average) best results could be obtained by:
1. Capturing at 1080@60i
2. Using the best available deinterlacing to 1080@60p (keeping both smooth movement and high static resolution of 60i e.g. avisynth filters)
3. Combating noise by temporal filtering (in practice dropping framerate, but doing so only when needed)

-k

CCI reviews were written in a very confusing way. In the discussion thread below the reviews, there is some clarification on this. However, I have confirmed this personally, with my HF-100, as well as with some borrowed HF-200 and HF-S100, testing them myself (comparing 30p and 60i footage and pushing it through a simple deinterlacer). 60i footage clearly had jagged edges, while 30p was clear and sharp.

The images are captured at 30 (or 24) full, progressive frames per second (as I explained above).

Since the capturing was done progressively, there will be no temporal shifting and all associated artifacts that any subsequent deinterlacer would need to deal with, as it does when the original material was captured at 60 interlaced fields. However, since all this is being encoded in a 60i stream, proper deinterlacing is still necessary, but is much faster, since there isn't any need for pixel-pushing. All deinterlacer needs to do is reassemble those interlaced fields into progressive frames, which is fairly effortless.

How does this proove that these cams grabs at 30p native?

If :
*the sensor does 60i native, and
*the camera has a very good deinterlacer (for 30p),
*your software deinterlacer was very simple
I would expect just the kind of results that you are describing.

I recommend this forum for help on deinterlacing:
http://forum.doom9.org/forumdisplay.php?f=33

The safest method is probably to film a moving scene, then inspect one field at a time using VirtualDub or something similar. If every field shows movement, then it is a true 60i camera.

-k

That's precisely what I'm talking about. In addition to the test mentioned above, I have checked every field of a 30p inside a 60i on an HF-100. It is very clear that the two subsequent fields are two halves of the same image, without any motion between them, and when you put them together, you get a complete image. Same goes for 24p. Using NeoScene, I had removed the pulldown from the stream and it re-assembled a clear, clean, pure 24p stream, without any interlacing artifacts. By the way, since these Vixias don't insert pull-down flags anywhere, NeoScene doesn't know the cadence of a clip and tries to figure it out by analyzing footage itself for motion artifacts, so when it removes the pulldown and deinterlaces a 24p footage out of a 60i stream, first few frames are always improperly handled, and you can clearly see temporal shift between adjacent fields. After 5th or 6th frame, the cadence is established, and frames are clear and clean, without interlacing artifacts. This is explained by NeoScene support on DVInfo forums.

So, yes, I think we can safely trust Canon's statement that the 30p is a true 30p, chopped up and shoved into a 60i stream.

When you think about it, it would make no sense. Why would Canon (or any other manufacturer, for that matter) bother to develop a deinterlacer, when it would be much easier just to capture progressive images, chop them up and stick them into 60i? They already do this chopping up anyway; the only difference is in the timing of the capture (whether to do a continuous single progressive scan of the CMOS chip, or two consecutive scans, then discard odd/even rows). It seems totally unnecessary and counter-intuitive to deinterlace an interlaced image, only to re-interlace it into the stream. Why would anyone do this?

Sorry, I think I have misinterpreted your original post. I thought you said that the sensor grabbed at 30p/24p at all times, when what you said was that it was capable of grabbing at natively 60i, 30p and 24p?

The CCI review is very misleading on this topic. Even worse, the review of the brand-new HF-S21 is equally misleading.

What is interesting is:
1. How is the video natively captured (60i, 50i, 30p, 25p, 24p)
2. In what format is it transported through e.g. AVCHD (60i)

http://www.camcorderinfo.com/conten...eview-37609.htm
Let me see if I can rephrase CCI using human language :-)

• Motion performance was excellent, and the addition of a native 24p video container is very exciting.
• The HF S21 also has its PF24 and PF30 modes, but they are still embedded inside a 60i stream.

Canon HF S21:

Code:

Format  Sensor  Stored as
"60i"    60i      60i
"PF30"   30p      60i (pull-down)
"PF24"   24p      60i (pull-down)
"24p"    24p      24p

Canon HF S11:

Code:

Format  Sensor  Stored as
"60i"    60i      60i
"PF30"   30p      60i (pull-down)
"PF24"   24p      60i (pull-down)

The Canon HF S11 also has a number of alternate frame rates for recording video, but all are embedded inside a 60i video stream. Canon's PF24 and PF30 modes capture video at 24p/30p, then stores it as 60i video using pull-down techniques. This means that the user (equipped with the right tools) can extract the original 24p/30p video with very little loss. A similar process is done with Panasonic's 24p mode, which is called Digital Cinema recording.

-k

Exactly right. Although technically, when you capture at 30p and encode inside 60i, I don't think it's called 'pull-down', but 'interlacing'. If I understand correctly, pull-down refers to breaking up 24p stream and duplicating some fields in order to stretch that 24p stream across 60 interlaced fields. With 30p to 60i, no additional fields are necessary; just breaking up original progressive frames into fields and interlacing those into 60i stream.

http://en.wikipedia.org/wiki/Telecine#2:2_pulldown
-k