Shutter Speed Question

[Brad Tyrrell]

I might be asking one of those silly questions but....

Could someone explain a little about shutter speed?

I tend to grab stills from DV a lot on the XL1. It seems the recommended shutter speed is 1/60, but when I examine a frame or even half frame they seem often be slightly blurry if there's any movement at all.

I'm thinking that if I set the shutter speed higher it would help.

I just don't quite have it in my brain what the XL1 does if I set the shutter speed to 1/120. Does the camera take 2 pictures during the half frame scan? If so, it won't help much. If the camera takes a 1/120 th of a second picture every 1/30 th of a second it might.

Somehow it seems I should be setting the shutter speed to multiples of 1/60, maybe 1/30.

Could someone 'splain a bit?

Brad

[Don Palomaki]

The shutter speed indicates the amount of time the camcorder CCD gathers light for the field before it is read. Thus a faster shutter will result in less motion blurr in the field. Fields are still spaced 1/60 sec apart in time. (1/50 for PAL). Thus the relative motion between fields will remain the same (but may appear more pronounced) if you use a shutter speed faster than 1/60.

A shutter speed of 1/60 probably results in the smoothest lappearing motion even if the still grabs images are not quite as sharp. Ultimately use the speed that best gives you the image effects you are seeking.

[Jimmy McKenzie]

Don is correct. If you are looking to grab a still of somewhere at the bottom of the golf swing, you will need lots of outdoor light and a speed of 1/500th or above. Otherwise, when there is little motion the 1/60th standard will be fine.

Once you are in your image editing program, this is where you will make an immediate edit. In Photoshop the filter is video>deinterlace. This will remove the odd/even field and give you the progressive looking frame you desire.

[Brad Tyrrell]

OK, I got it. Thanks guys.

I hate twiddling knobs and hoping for the best when I don't really understand what I'm doing.

(I work with the Adobe video collection of programs and I don't think I'll ever really understand half of the stuff in them)

If I understand this correctly though ...

At 1/120 the camera takes a 1/120th of a second shot, then is effectively off for 1/120th then takes another 1/120th etc. The fields are scanned every 1/60th of a second. Each field a slightly different picture.

At 1/30 the camera takes a 1/30th of a second shot but scans it twice, every 1/60th so both fields of the frame are the same picture.

I suppose at other than multiples of 1/60 it would be possible to get split pictures, although it seems more likely the xl1 would use the last or closest full frame. Maybe not.


Brad

[Jimmy McKenzie]

Regardless of the shutter speed, which is simply an indication of time of exposure, the camera is still going to give you 60 interlaced fields in the form of 29.97 frames per second. How much light allowed for each exposure is the relationship of the electronic shutter "speed".

If elements has the de-interlace filter, you will see what I mean when you open the image up for editing.

If your final destination for these captures is for print, then this is the wrong device to be using in the first place.

Captured stills from dv have their place but usually in the spectrum of multimedia and not print.

[Brad Tyrrell]

>Captured stills from dv have their place but usually in the spectrum of multimedia and not print<

Well I use them for that a lot, but I do events and there are just some moments that only get caught on video.

DV stills are also great for websites and dvd labels.

Grabbing in Premiere Pro at 1296 x 960 with square pixels and exporting to Photoshop CS seems to work pretty well. And, from very recent experiments, a shutter speed of 1/120 on the ol' xl1 doesn't seem to give me any problems and is allowing slightly clearer frames for grabs.

I've experimented with the de-interlace feature in Photoshop and it helps if there's much movement. If not, most often the loss of detail isn't worth it.

I'm hoping that high definition cameras will make capturing stills from video more reasonable.

Brad

[Jimmy McKenzie]

I'm curious how you take the dv 720x480 image and export the frame out at 1296x960?

Where do the extra pixels come from?

[Brad Tyrrell]

>I'm curious how you take the dv 720x480 image and export the frame out at 1296x960?

Where do the extra pixels come from?<


Well.... I don't exactly know. That's another one of those questions I should figure out how to phrase properly for this forum. Another knob twiddling experiment in Premiere.

When you export frames from Premiere you can change the aspect ratio to square pixels but that leaves black at the sides of the grab. Soo... you normally change the 72 ppi output frame settings from 720x480 to 648x480. That takes care of the black.

Then, since I wanted higher resolution but couldn't find a way in Premiere (there is in Photoshop), I wondered what would happen if I doubled the size of the export frame. Resolution would be the same but on a blow-up of the original. I tried it and it worked. (There are probably help notes on this somewhere but I couldn't find them.)

The result is very similar, maybe the same, as the product of the resampling done when you increase the resolution in Photoshop.

I don't know how Adobe does this stuff, it just does.

Brad

[Rob Lohman]

Going from a lower resolution to a higher is called upscaling and is
done through a complicated process called interpolation (which
usually results in a softer image).

Let's say you wanted to go from 720 x 480 to 1440 x 920
(twice the size) and you have the following grayscale pixels:

12 14 19 21 28 34

The higher the number the brighter. So we have 6 pixels here that
go from dark to lighter (left to right). Now if we need to interpolate
this we have a problem:

12 .. 14 .. 19 .. 21 .. 28 .. 34

We are missing six pixels (where the dots are), so what we do
interpolate them (add the two neighbours and divide by two, a
very simple algorithm), so we get:

12 13 14 16.5 19 20 21 24.5 28 31 34 34

So we have created six new pixels:

13 16.5 20 24.5 31 34

Two problems already emerge here. I simply duplicated the last
pixel since I cannot really know what could be there. The other
problem is that I have two fractional numbers I need to round
(which introduces errors, basically).

The higher the upsampling the more pixels need to be interpolated
or (re)constructed, so you'll get more calculated pixels (and lines)
inbetween the real ones. The higher you go with this the worse
it will look. You look sharpness because such algorithm smooth
everything out where in fact there may have been harsh differences
(the more you upscale the more prominent this will become).

Ofcourse it all depends on the quality of the algorithms being used
as well and some programs have far better results due to this than
others! The little algorithm presented here can be vastly improved
in all sort of manors to get a better quality for example.