A new Canon contender -- the SX20
 

Much frustration with shooting video with MANY of the latest still cameras comes from the perceived downside of no control over shutter and iris when shooting. The fear stems from:

1) Inability to set shutter-speed when shooting 24p, 25p and 30p. Of course, even when a camera offers this ability, given the smaller chips, iris is typically limited to under f/8 so in bright light an ND filter is required.

2) Inability to set optimum F-stop, typically f/4 to f5.6. Of course, even when a camera offers this ability, given the smaller chips, shutter-speed will typically be too fast so in bright light an ND filter is required.

3) The ability to set both. Once again, in bright light an ND filter is required. Complicating these issues is that when in AE, a still camera has a program that is biased to keep shutter-speed as high as possible to eliminate camera shake artifacts.

So, one way or another, adding an ND filter is a must in bright light.

I've long wished for a camera that would switch programs when shooting video to keep shutter-speed as low as possible to eliminate strobing artifacts when shooting at 24p, 25p and 30p. The new Canon SX20 has exactly this function.

In an EV14 (ISO = 80) the "still" exposure is f/2.8 at 1/2000th. When shooting video -- you can see the current exposure by touching the exposure button -- the exposure changes to f/5.6 at 1/500th. "Still" mode is biased for a high-speed while the "video" mode is biased for smaller aperture.

Adding an ND8 filer brings the "video" exposure to f/5.6 at 1/60th. Very nice. Adding an ND1.2 filer would drop another stop, which could bring the exposure to f/5.6 at 1/30th. This would be perfect for 720p30, which the camera shoots.

The 1.2ND filter would allow a brighter scene of up to EV15 and still keep the shutter-speed under 1/60th. At a very bright EV16, shutter-speed would only be 1/120th which is OK.

========

The other great thing about the SX20 is that it uses a 1/2.3-inch (0.44-inch) 12Mpixel CCD. Yes, no jello!

Plus it has a very good Optical Image Stabilizer System. It also has a 2.5-inch TFT color articulating LCD. Best of all it as a large 0.44-inch viewfinder with 235,000-pixels.

It shoots 30.0p (not 29.97) H.264/AVC (Baseline@L4.1; no CABAC with 1 reference frame) at 24Mbps VBR with stereo PCM 16-bit audio captured at 44.1kHz. I see no motion artifacting.

And, of course, AF while shooting. But, you can Auto-Lock Focus and AE as you may need to do.


I've sent my Casio EX-F1 back.

This looks real interesting, I have been seriously thinking of getting a panasonic LX3 as my next point and shoot still/video camera, as I like the fact it goes to 24mm wide and shoots raw, but I will look into this one more..thanks for the lead. It appears you can buy an adapter for using threaded filters.. John

yes -- real threads for 52mm filters. Cheap too at under $400.

PowerShot SX20 IS Digital Camera

I compared it to the new Pana and every test I could find the Pana had better ois at full 20X zoom, but the Canon had far more acurate -- not bluish -- awb. This is important to me -- although you can manually set wb. The Pana also had more edge-enhancement which I hate when viewed on a 63" screen. The Canon looks softer, but I can live with that.

I'm surprised no one is talking about these new alternatives to the way too expensive GF1. I think Canon may sneak-up on everyone from above and below.

PS: stills are great too. No RAW, but for the internet I really could care less. I haven't made a print since the `80s.

well I ordered one, and the filter adapter from lensmate (they are 58mm not 52).From all the samples I could see they looked great, and the price is great as well. I got mine from newegg for 359.00. I know they have a firmware hack via CHDK for adding raw for the sx10, so hopefully there will be one out for this cam soon.

i own sx200is, and the video is really excellent. so film looking and sharp. ois is quite good - and only the lack of zoom and focus lock is the downside. i filmed interview in bar on the loud concert (just to see how it will sound) and it was impressing. the video on sx200 is wow.
i also own a630 and the raw from hacked firmware is not so impressive, not to say that is is quite useless.

Odd. I put what I thought was a 52. I'll double check. Can you describe the hack

the hack info is located here..
CHDK Wiki

its an ongoing collaboration to alter the firmware in many cameras to give you more options. I use it with my canon sd800IS to get raw files, timelapse,etc.

"... and the filter adapter from lensmate (they are 58mm not 52)."

My filter is definitely 52mm.

that is wierd because I just bought this adapter from these guys and it says 58mm, so go figure..but maybe the lens itself is 52mm, but I assumed it didnt have built in threads on the lens..either way, cool camera.

Lensmateonline - Canon PowerShot SX20/SX10 & SX1 IS

It has threads which is one of the nice things about it. B&H has a Tiffen 52mm 1.2ND that cuts light 4-stops.

You can twist a 52mm filter onto an SX1 or SX10 lens and it will stay firmly attached, even though Canon didn't intend for this to be possible. No adaptor is needed. The SX20 lens is probably the same.

Very odd.

My old Sony 52mm camcorder filters screw on fine.

My new 52mm Tiffen 1.2 ND does not -- although it will screw into any Sony 52mm filter.

So I've screwed the 52mm Tiffen 1.2 ND into a clear 52mm Sony.

What's the 58mm lensmate and how does it fit with the hood?

the lensmate adpater is a bayonet mount that goes on end of the sx20 lens, and has 58mm threads on the other end..

The footage from the Canon point and shoot cameras like the SX-1, SX-10 and S-X20 is at 30p rather than 29.97 and is at cRGB levels rather than the sRGB levels of a regular HD camcorder. Both these items are automatically corrected when you convert the .mov format native clips from the camera to .avi Cineform codec with any of the Cineform products like Neo Scene. The 30p will be slowed down slightly to 29.97 and the levels will be rescaled so that it won't look to contrasty and the highlights won't be blown out on an HD TV. I believe that the color correction is a playback function and not something that is actually changing the data. Anyway, without Cineform it is quite challenging to use the footage from one of these cameras. With Cineform it is extremely easy to get great results.

There's no real reason for CineForm.

While it's true that 29.97 isn't 30.00 -- with a typical clip used a project being about 5 seconds long -- there's no cause for concern. You can do the math to see how long a single clip in a project needs to be before a 1 frame error will occur.

Likewise, cRGB or sRGB is going to be converted to YUV by all PRO NLEs. If your NLE allows for SuperWhite (over 100IRE) then cRGB is simply going to give you are wider dynamic range -- up to about 108-110IRE which is perfectly normal. In fact, it's great!

Given the fact about 1 in 10000 TVs have black level set correctly ( at 0IRE) when the NTSC norm as been 8IRE since Day 1 -- the fact that the camera can record a few more dark shades which will get truncated going to YUV is not a significant issue. (Assuming the CMOS chip actually provides real information at such low light levels.)

Moreover, the fact is that HD cameras use 709 standard which is YUV and not sRGB.

The whole CineForm premise is that there is an advantage to not editing native formats. In terms of speed this was true 5 years ago with MPEG-2. And, yes, with H.264 it remains true until the quad-core laptops become the norm. On the other hand unlike CineForm, disk band width is VERY low with H.264. Especially 720p30!

Every other claim they make is false. No intermediate is needed with any any current pro editor. There can NOT be an advantage to converting one format to another -- quality can only lost. It can't be preserved nor can it be improved.

All modern NLE's have there own intermediate codecs to use IF AND WHEN RENDERING NEEDED is needed and all are real-time codecs -- which with FCP CineForm isn't. So, CineForm will be no better than native H.264. You'll have to render both!

Moreover, unless you have 10-bit data -- not bloodly likely -- only the internal NLE calculations need be done at 10-bits. FCP give you better than that. Avid gives you 16-bits.

Remember, any source is only decompressed ONCE when the final export is made. No intermediates (e.g., ProRes) are used. Everything pre-rendered is NOT used by FCP. And, with Avid, you can force renders to be discarded.

Renders are only used for pre-viewing.