NEX video - MPEG4 probably the best choice

[David Kilpatrick]

Today I shot comparison videos using AVCHD (1080/50i), MPEG4 (1440 x 1080, 25p but with stretched capture pixels, so the result is still HD shape) and 720/25p on the NEX-3. MPEG compression was set to Fine for both the NEX-5 1440 x 1080, and the NEX-3 720. The NEX-5 will not shoot 720 at all, it's either a 1080 format, or VGA.

Results:

NEX-3 720p is compressed to the limit. It's very low quality 720p, lacking the sharpness and detail of the similar format on the Nikon D5000 for example. The compression maybe combines with NR, and totally obliterates detail in the shadows of many shots. The NEX-3 is a very poor choice for video, and those who suggested it would equal the 1080i of the NEX-5 will be disappointed.

NEX-5 - it is so hard to tell the difference between 1080 x 1920 (50i) and 1080 x 1440 (25p, interpolated to 1920) that some clips looked better using the MPEG4 encoding, some looked better using AVCHD. The colour and contrast seemed superior with AVCHD, but the sharpness and detail were if anything more consistently superior using MPEG4, especially when viewing freeze frames.

MPEG4, the format used by the NEX-3 and by the NEX-5 only for its 1440 and VGA settings, is a much more accessible format for editing than AVCHD and can also be uploaded directly to video sharing sites. I would compare AVCHD to using raw put through a standard maker's converter, and MPEG4 to using Fine JPEG out of the camera.

The 1080 high NEX-5 results packed in much more useful visible detail, every time. It's clearly worth the extra £100 over the NEX-3 even if you later reduce the video to 720p for sharing. The biggest difference is the level of detail blurring and smoothing present in the NEX-3, it sacrifices almost all trace of textures when contrast is low and the subject matter is dark. However, in bright conditions it can be sharp enough to throw up the occasional moiré effect from patterned subjects (of course, so can the NEX-5).

To sum up - don't believe those who suggest that because 1080/50i is interlaced, and 1440 x 1080p is an odd format with stretched pixels, these modes offer no gain over the NEX-3 720p. They offer an easily visible improvement in detail and texture making the NEX-5 a far better choice for videos.

David

[Vidgamer]

Be careful looking at single frames --if the machine/software doesn't de-interlace first, you will get a lower-res still. But if it properly de-interlaces, you should end up with a higher-res still with the interlaced pic. I wouldn't be surprised that it's hard to tell between the two high-res formats -- it's hard to tell the difference between 720p and 1080i/p on most displays. The display has to be large and you have to sit closely to really take advantage.

Well, I feel better about paying the extra $100, I guess! I took some video clips -- I'm eager to try out the Sony software that creates a Blu-ray compatible DVD disc and see what it looks like. OH, and I'll test out David's assertions about the built-in mics -- I taped a live drum performance that was loud enough to feel the boom in your chest.

[Vidgamer]

I have only tried AVCHD so far. Just reviewing the video on the computer (in PMB), one drum recording distorted -- you can hear some "crunching" with the loudest beats. I guess the db level was just too high for it. It's not severe distortion, but enough for me to notice it right away using speakers.

On the other drum recording, it sounds fine. On yet a different recording where there were a lot of voices, the cat freaked out. Good sound quality when it fools the animals.

The video quality is nice too. I love the detail. Focusing was accurate except when light was very low; in that case, I found the wider end of the zoom to work much better. Zooming in, it sometimes worked, and sometimes hunted, so it was right on the edge. I have similar problems with my dedicated video recorder, but I seem to be able to zoom quite a bit further with that one in a theater before it hunts, but without being in exactly the same theater, it's hard to compare the two.

[David Kilpatrick]

I've got some distortion which is unpleasant with full volume vocals, causing other instruments to drop back (the auto gain and compression work very well). But it's just acceptable to the ear, unlike the result from Nikon's built-in mics, which turns into a violent clipped crackle you just can't use at all.

I have done one 'walk round' vid moving through the whole of the pub during a performance and right up to the speakers, using the 16mm lens. It's an interesting social record and the sound is surprisingly good. I have had two occasions where the stereo failed and the left channel dropped out on the NEX-3. It's not happened so far on the NEX-5 and I would be thinking 'camera faulty'.

Uploading it now to YouTube, will post link.

Link now here: http://www.youtube.com/watch?v=v1URJaOU7nE

David

[agorabasta]

The overall appearance of the video with this and many other amateur cams depends greatly on the lighting conditions. Video appears much smoother if shot in dim light simply because the cam makes use of a longer part of frame period, and then in bright light it may degrade into sequence of short-exposure stills that don't capture the motion blur.
But Nex is really good in low light compared to its competition. So this advantage may be used to make videos shot in bright light much smoother - just use ND filters till enough motion blur is captured, and it really can be controlled through the back LCD image motion smoothness. (And you also get some indirect control for aperture...)

[RubberDials]

NEX-5 - it is so hard to tell the difference between 1080 x 1920 (50i) and 1080 x 1440 (25p, interpolated to 1920) that some clips looked better using the MPEG4 encoding, some looked better using AVCHD. The colour and contrast seemed superior with AVCHD, but the sharpness and detail were if anything more consistently superior using MPEG4, especially when viewing freeze frames.
David

David, a single frame of 1080i video contains only one field (ie half the resolution). It will never look as 'sharp' as a single frame of 1080p which has no fields and is a full-resolution image.

'De-interlacing' 50i to 25p does not create genuine 25p any more than 2 photographs shot at 1/125 overlaid on top of each other give you an image shot at 1/250th. Good results can be achieved but is is always better to have the video in a progressive form in the first place.

Incidentally AVCHD is MPEG4 (H.264) so it's not surprising that they are hard to tell apart.

[agorabasta]

The 50/60i of the Nex is encoded from the 25/30p sensor output, as evidenced from the 1440x1080 25/30p mode.

Any decent player, when paused, displays a full frame and not a single field of an interlaced frame nor a mix of two fields of adjacent frames.

[InTheSky]

15 years ago when some company started selling more Progressive TV I was never expecting that after all that time we will start again this debate ... For me Progressive is the way of recording video and having good quality, interlaced will never be has good as real picture from each frame. Yes they are playing with double the frame rate to make this better in interlaced, but it is still visible by eyes. I now equip in my basement with a 1080p projector with a 120 inches wide screen and you can see the huge difference on that setup vs 1080p and 720p and also interlaced video.

regards,

Frank

[agorabasta]

....you can see the huge difference on that setup vs 1080p and 720p and also interlaced video.

You sure can, if the video was captured by interlaced fields at the very source/sensor.
Then if the source is progressive 25p, and the display is progressive 50p, all that you see is exactly 25p progressive. It simply doesn't matter if the signal has been transmitted in between as two fields per frame (in fact it does just a little bit - the MPEG4 compression is less stressed in the two fields per frame than it otherwise would be in one full frame).

[RubberDials]

The 50/60i of the Nex is encoded from the 25/30p sensor output, as evidenced from the 1440x1080 25/30p mode.

Any decent player, when paused, displays a full frame and not a single field of an interlaced frame nor a mix of two fields of adjacent frames.

Okay, so PSF. I didn't realise this. The camera produces 'fake interlaced' rather 'fake progressive.' The A and B fields come off the same frame so no lost vertical detail when recombining etc.