What exactly does the Sharpness setting, as found in the Custom Preset menu, control? How does it do what it does?

Usually it controls edge detail (increases contrast of edges). But unfortunately noise contained within the image is sharpened too. I turn sharpness on NTSC video down most of the way, especially on monitors.

I turn the sharpness setting down two or three steps for a more natural look when using the built-in optics, but I turn it back up to the default when using a cheapo wide-angle adapter to compensate a bit for the softening from the low quality lens.

I agree with the sentiment here - I usually dial sharpness down two clicks because I find the factory default leaves rather obvious video edge effects.

Regards,

Julian

Personally I turn it all the way to the minimum on my PDX-10. I dislike that over-sharpened edge look. You will also note some color fringes under high contrast situations when using the default settings, and it makes noise more noticeable in low light situations.

But it really depends on your likes and dislikes. Shoot a few seconds using different settings, playback on a monitor and decide for yourself...

Thanks all! Much to explore...

The amount of sharpness, for me, depends on the type of shot. In well balanced lighting with little contrast, I often have sharpness turned all the way up. I just love the texture of details, as on plants, leaves, foliage, surfaces. But edges of things, trees, people, in relatively high contrast, gives away the effect being had on the image, and can make for a jaggy mess, especially with a good deal of distance from the subject. In closeups, I think it can be brilliant. Experimentation is experience is key!

I've played with the sharpness preset and I'd like to rename it as a softness control. Unlike you guys I find the softening left of centre unacceptable, but that could be because I've been shootiing a lot of buildings in testing the camera. The brick wall makes a very revealing subject in many ways.

tom.

It's a trade off between a too soft image and the enhanced noise of the sharpened image. I suspect there may a difference between PAL and NTSC, as to how the sharpened image looks. I find the noise much more objectionable than the soft image.

It's my belief that when you turn the sharpness all the way down you're seeing what the CCD's actually capture. The sharpening is a "filter" applied to that image in an attempt to enhance the detail. Please feel free to correct me if you have information to the contrary, but this seems to hold true for most of the digital still cameras I've used. I really don't think this control does any "softening" to the raw image.

I would prefer to get the actual CCD data onto my tape. If I think it looks too soft later (when I can view it at my leisure on a monitor back home) then it's a simple matter to apply a sharpening filter in my NLE. This will yield the same effect, possibly even with more subtle control.

But like I said before, it's all a matter of personal taste and style, and it's a good thing that we all do things a little differently or the world would be a pretty boring place ;-)

Boyd,

Do you know if this sharpening filter (looks like a high frequency shelf boost, to me) is at a neutral setting ( 0 dB gain) at the
middle setting or the minimum setting? On the Vx2000, the factory default is set to the middle of the range. I've seen the artifacts at max Sharpness, but I have not yet experimented with min Sharpness.

Thanks.

The software (NLE) filter gives you much more control. However, rendering time for some people could be a real problem. I think 0 dB would not be a middle setting but lower (minus 2 or the lowest setting). The middle setting would be what the manufacture thinks looks best. But sharpening is definitely added at the middle position. Noise is also being added or enhanced at the middle setting.

I like the look of the PDX10 with the sharpness at a minimum. Again, it's been said that if you want sharpness after teh fact you can add it in your NLE, but you can't take it off after it's been added in the camera.

I wonder also if the DV codec will not work better on an image that has less sharpness??

I tend to agree with what Jeff and Graeme say. I think the minimum sharpness setting represents the raw CCD data and you apply more of the sharpening filter as you increase it.

True, rendering time will be a factor. This seems pretty much like the same trade-offs you would experience with a physical glass filter in front of your lens. You might very well acheive the same effects in post, at the expense of rendering. But as Graeme points out, once you've recorded something filtered to tape then you don't have the option of changing your mind later.

I also think it's extremely hard to make judgements on things like sharpness based on the viewfinder and flip-out LCD. Unless you can judge an effect like this on an external monitor, or from experience with the camera, you are taking a bit of a chance. I formed my opinion on the sharpness after viewing tapes that I had made. Then to set my mind at rest, one day I hooked up an external 16:9 screen to the camera and just experimented around with the sharpness control while pointing the camera at different subjects. In pretty much all cases I liked the minimum setting the best. As you increase the sharpness you defiinitely start to see more noise, as Jeff points out. You also start to notice more fringes around the edges of high contrast objects.

However, I'm sure there may be situations where more sharpness can work in your favor, as Shawn describes. And for that matter, other people just see things with different eyes. But regardless, I think you will find any time you spend with a monitor connected to the camera well spent. This is a great way to get an understanding of what all the custom preset items do.

>Again, it's been said that if you want sharpness after teh fact >you can add it in your NLE, but you can't take it off after it's >been added in the camera.

Actually, you can take it off after it's been added. While it may be difficult to determine the exact filter parameters, the whole effect seems to a high frequency shelf filter which boosts frequencies above a specified frequency. The slope from the filter frequency to the shelf edge frequency will almost definitely be 6 dB/octave.
To reduce the effect, apply the opposite, which is a high frequency shelf filter with attenuation. Sure, the full signal can't be completely recovered, but most of the sharpening can be undone.

Someday, someone (yeah right) may dial into a camcorder designer for the exact specifications of the sharpening filter and how they correspond to the values in the camcorder menu.
If anyone is willing to take a picture of a black on white vertical
bar that is one CCD element wide for ten or so settings of the sharpness filter, I'll do the frequency domain analysis.

> I wonder also if the DV codec will not work better on an image
> that has less sharpness??

I think it should. Perhaps especially for chroma which is the DV codec's main weakness. But first we would need to know if the high frequency enhancment affects chroma or only luma. Although on some posts it has been mentioned that noise increases with higher 'sharpness' settings, my tests sugges not, and this would be consistent with the edge enhancement algorithm not affecting chroma, since most of the noise I see with low light & gain up seems to be more in chroma than in luma. Of course this is all very subjective and non-scientific so I could be wrong. Anybody care to comment on this theory?

Not so good with the theoretical stuff myself, but I suggest you hook up a monitor and just have a look. Use a dimly like subject and turn the exposure wheel so you get the iris wide open and maximum gain boost. Now try turning the sharpness dial. When I did this myself I noticed less grain with sharpness set to the minimum than I did when it was set to the default.

>When I did this myself I noticed less grain with sharpness set to >the minimum than I did when it was set to the default.

Boyd, this would still be a tough test. If the negative settings (those less than 0) actually put the Sharpness filter into a Low Shelf Cut mode, then those settings would be blurring the image,
of course, resulting in less noise. The best method would be to ask the designer. Second that, we can grab impulse responses
of the Sharpening filter at various settings by capturing pictures of single CCD line vertical bars (black on white, or white on black). Then, a Fourier transform would yield the frequency response of the filter.

Gentlemen,

I have been reading this forum for a few weeks as I learned about my TRV-950. Regarding sharpness:

When I first began utilizing the TRV-950 I was struck by how sharp the image was relative to my preceding video experience.

As a long time still imaging guy I quickly tested all the functions of the TRV with still images using all manner of test methods.....the easiest of which is the simple digital "blow up".

Below is a link to three images from the TRV-950. Minimum sharpening (full left), default (middle) and max sharpening (full right) on CUSTOM PRESET.

http://home.att.net/~colorgamut/TRV950/CompositeSharpTest.tif

The leftmost position, as one can see from the smooth gradations in tonal quality, is the native CCD response (no low pass filter is applied as far as I can tell). The middle sharpening position of factory default is riddled with noise (I will come back to the exposure settings and how they affect this in a moment). The max sharpening is too gross to discuss.

In the above image acquisition I hand held the TRV at full telephoto and acquired a still at max res and SFN. Please note that I should have used a tripod to norm out the possibility of my hand moving and motion blur. But, I was really only interested in noise. The gain up on the auto gain was +6dB.

When outdoors with good lighting the experiment is different. When there is no gain-up I use one step up from min sharpening with almost no detectable noise. When indoors (always have gain up) then I use min sharpening.

In fact, once I discovered this little SONY sharpening trick at default and the relative softness of the image without beating on it with a high pass filter I was disappointed in the resolution of the video quality relative to my expectations based on the price of the thing (the TRV cost 4X what my old Canon ES-970 cost).

Anyway....in summary, in the presence of CCD gain up I am forced to use no sharpening to minimize noise resulting in video and still softness that is apparent. The final vido resolution, when playing side by side against my ancient 8mm Canon is better. But, it is not 4X better (as was the price difference between that old Canon and my new TRV-950).

Since my first, and last, post in this thread, I have come to agree with you, Mike, and Boyd. Sharpness is insultingly fake resolution, no ifs ands or buts, and it is messy, likewise. Any amount serves only to ruin images.

These are some great observations Mike! Looking at your images, it's a no brainer for me.... I much prefer the one on the left. But I think you should have used a tripod to be just a bit more scientific :-) What was the shutter speed? From your description I assume these are still images as opposed to video frames?

Now I think it's a little unfair to expect a direct correlation between price and image quality. For starters, DV has some severe limitations. Look at the quality of a still photo from your camera and compare to a frame of video. But I suspect the tape transport and other innards of the PDX-10/TRV-950 are built to a higher (prosumer) quality standard than your other camera, which I also assume is a 1-chipper.

Would you expect a $400,000 house to have four times as many rooms as a $100,000 house? Would a $40,000 car to get 4x better mileage than a $10,000 car? Maybe. But generally that isn't the way things seem to work. ;-) Anyway, I do understand your general point, and the frustration with how expensive this stuff can be! I'm very happy with my PDX-10 due to its 16:9 mode which the TRV-950 lacks. For 4:3, you might be happier with the VX-2000, but that costs even more!

Boyd,

Thanks for the comments and reply. You are right....it is not really fair to expect linear mapping of price and resolution. The TRV950, because of its price and my very, very mortal bank balance, and because of many very positive reviews.....came to my door with my expectations set on high. Some of those expectations are not being met. But....others are.....

The shutter speed was 1/60, aperture was F1.6, and as noted I was at full telephoto......a compliment to the image stabilization system used by the TRV950.

I am posting a question about the autofocus system on a new thread to assess if I have a unit problem......if you have a chance let me know what you think on that thread.

<<<-- Originally posted by Mike Sanchez : The shutter speed was 1/60, aperture was F1.6, and as noted I was at full telephoto -->>>

As Tom has noted elsewhere, the data code tends to lie about this sort of thing. I don't think it's possible to have a F1.6 aperture at full telephoto, is it?. Browse back through the forum a bit for a discussion of the "double secret" internal ND filters and questionable data codes...

You're right Boyd, the 'replay' display is good as useless, and my PDX10 often said I'd shot at f1.6 at full telephoto. As we all know, the maximum aperture of the 12x zoom at full tele is 1 1/2 stops less than this, at f2.8. Surely it's not beyond Sony's command of the DV language to fix this incorrect read/write? After all, their cams generally supply more information in the 'display' mode than, say, Canon do.

tom.

Boyd and Tom,

I will check the files for the max aperture and am embarrassed that I did not think to question the F1.6 displayed on the screen in data code mode at full telephoto. Not one of my SLR zoom lenses maintains its maximum aperature at full tele so why should my TRV950? Anyway.....thanks.

And......I was very happy to see the data written to the screen. I think lots of the information may be accurate. I've checked the settings many times in low light and in outdoor settings to assess the auto gain-up. Seems like when I think it should be at max +18db gain it is and vice versa. In fact, I've used the display to kind of get a sense of when the aperture changes and when auto-gain up kicks in......

As Tom notes......my old 8mm Canon told me precisely nothing about the settings. I could tell when gain-up was maximum by the visual SCI FI look of large and small scale noise streaming across the screen on playback.

I would like to note: The noise level of the TRV-950 is so far below the noise level of my old Canon that I have not thought about noise since beginning to use it, except to turn off sharpening. One that is off, noise is no issue on the TRV....even in firelight......

Relative to my old Canon anyway. If I had shot with a VX2000 for the last few years I might think a bit differently.