No, I have no experience with DSLR cameras but I believe you. In all the pictures I've taken with my 6MP digital camera or every digital picture I've looked at an Steve's Digicams for higher MP cameras - they all look soft to me. So yes, you are right in saying that they are all relatively "soft" and need some level of sharpening. I think that's just inherent in the physics of the hardware of digital imaging and capture.

But I guess it comes down to what the individual considers to be acceptable as being a sharp-detailed image or a soft-detailed image. And since I now realize that Cinemode has a real effect of softening the image on yours and Eugenia's HV20's (and perhaps even more so with mine), then I'm wanting to avoid using Cinemode to get back all those ever so subtle details that are being lost. Those kinds of details may not be noticable on distant subject matter, but for closeup work it could make for a significant difference.

Here's a 200% blow-up of the grid. Again, the only enhancement was that I auto-balanced the levels.

http://www.12south.com/video/hv20/24p_200percent.jpg

Give me another second, and I'll do the same with the board, which is where the "issue" really is visible.

Dude, look at the detail on my face in my first post ... you think any actress wants seen at that level of detail! :-)

Seriously, though, if you know what a good RAW DSLR image looks like, then you can understand how sharpening is a second step, whether it is done in camera or in post. There are pro's and cons to any decision as to when to do it, but all things being equal, you have more options in post. (Now, unfortunately, the HV20 does not seem to be treating some parts of the image equally, which is problematic.)

Cinemode is still noticable in those 200% pics. The Cinemode pic looks ever so slightly more blocky with a hair of fuzziness to it.

And yes, no actor/actress would ever want that much detail of their face. LOL!!! But I love to see super-fine details - I've been that way since I was a child and that is why I take notice of extreme subtle differences. (that probably explains why I do Quality Control for a living as well - having to search for audio and visual details). But I know that others here are seeing this too, so I'm not completely out in left-field here. LOL!!!

Ok, more tests. In my new test with more modes (everything else was left on defaults), I found that the TV mode is the *best* mode of all. It has less mpeg2 artifacts for some reason, possibly because no sharpening happens in that mode (and no softening either). The Aperture priority mode is as bad as Cinemode IMO. One other bad thing that Cinemode seems to expose is chromatic aberations (I saw it twice so far in various frames). The Program mode (auto), is good, but there are sharpening artifacts when you zoom in. The most *recoverable* detail in my opinion can be found on the TV shutter priority mode!

http://osnews.com/img/vegas/program.png
http://osnews.com/img/vegas/tv.png
http://osnews.com/img/vegas/ae.png
http://osnews.com/img/vegas/cine.png

For example, look at the flower in the glass left of the water bottle. It is only the TV mode that was able to exhibit details on that flower. This is the flower detail I am talking about:

http://osnews.com/img/vegas/comp.png

Here's the board behind the pages at 200%. (No correction at all this time.)

http://www.12south.com/video/hv20/24p_200board.jpg

Here, obviously, the difference in detail is much more noticable.

Eugenia,
Why the nasty stairstepping in those images? Are you shooting 60i and interpolating?

That would be very good news, indeed, since we could lock the shutter at 1/48th and work from there.

>Are you shooting 60i and interpolating?

Yes. When I use "blend" fields with Vegas it creates ghosting, so my only Vegas exporting option to a lossless codec that's acceptable is interpolating. As I explained earlier, I have to export before I can grab detailed screenshots, because Vegas won't give me 100% quality previews. BTW, please go back to my previous comment, I modified it and added one more picture to show the difference in a detail.

You're right, TV mode looks best of all those. And what to took notice of most was the "Crate&Barrel" wording on the wooden handle is the most sharp and detailed in the TV mode.

I made some contrast modifications to my earlier comparison picture in order to show under extreme stress which mode has the most artifacts or less clarity:
http://osnews.com/img/vegas/comp2.png
As you can see, in reality, the Auto mode and the TV/AE mode are not all that different. The only thing that differentiates them is the fact that the Auto and AE modes have in-camera sharpening which results to non-recoverable detail and artifacts. TV mode seems to be the most pristine of them all. Cinemode has the most chromatic aberrations and softening. I think I will switch to TV mode from now on...

It just occured to me (I'm not really much of a shooter, I'm a post guy) that aperture settings might affect focus. I know it definitely accounts for a shorter DOF, but isn't it also true that the more open the aperture, the softer the focus? I wonder how much of the difference in modes is attributable to differing f-stops?

Is everybody else also seeing that CINE footage seems to suffer the most visible MPEG-2 artifacts ... both color error noise and macro-blocking? Why in the world is this happening?

Conversely, though, CINE seems to have the most natural color response. Damn.

More testing, more testing.

Yup, if we could get TV mode's no in-camera sharpening with Cinemode's blunt colors, we would be in heaven. ;-)

I don't think the problem is the focus in the AE mode btw, from the pictures I saw it's sharpening artifacts that caught my attention regarding that mode, just like in auto mode.

Here's the chart I'm going to shoot. I'm going to try to be thorough and accurate with every mode.

http://www.12south.com/video/hv20/chart.jpg

P.S. My system is calibrated end-to-end so the color, while lab perfect, will be quite good enough to judge the merits of each mode.

One last test from me:
http://osnews.com/img/vegas/comp3.png

The lower the shutter speed, the less noise exists it seems. I think shooting 24f with 1/48 TV mode is the best bet one has for *indoors* where the light is limited. For outdoor, the AE or P mode might be better. Under no circumstances I would go for Cinemode though.

As for color correction, I think that in this day and age we have tools to fix them, even if we are not shooting cinemode.

I had a long talk with my husband who is a prosumer photographer and has also studied some optics at the uni too. Apparently my tests, which indeed show that slower shutter has less noise, won't have the same results outdoors, or even if you... zoom in! In other words, in order to get the BEST quality out of an HV20, you need to spend an hour per shot going back and forth on your computer doing tests, because the camera does not let you control gain/aperture/exposure/shutterspeed all at the same time. I don't think that shooting and checking stuff all the time on a PC screen before deciding on the best settings is realistic.

Now, what's interesting is that while the user does not have access to the these controls, the *camera software* does.

And this is why shooting on P/Auto, which it might not *always* be the best, but it will be in most cases. If you don't like the contrasty video look, buy a $30 contrast filter, and you are done. Under no circumstances I would use Cinemode though.

Eugenia and Joseph, wow, thank you for your tests. I was a believer in CINE mode up till 5 minutes ago. My HV20 has been gone nearly a month now - warrenty (different problem) - when it returns, I promise I'll do my own tests, with an UNCOMPRESSED digital capture of the signal.

Regardless though, I think CINE mode is out for me.. I desire every ounce of resolution my HV20 can provide.

Thanks again!

I know this isn't completely scientific, but it works in the context of my video work (since I'm not a professional), but I took the nice color image Joseph provided and printed it out on 5x7 High Premium Glossy Epson Photo Paper. Then connected the Component Out (RGB) of the HV20 and connected that to my HDTV, then also had the firewire connected to the PC using HDVSplit for viewing on my monitor, then used the photo to compare what I saw on the HDTV to what I saw on my monitor and made a few minor color/setting adjustments to my pc monitor so that it came much closer to matching my HDTV.

Now before everyone jumps on this method, I am using the HDTV as my given standard to go by since I think that the Discovery HD Theater programs all look very well color-balanced when playing back as well as all the DVDs I play. So if I can match my footage to look like what I already think looks great on the HDTV, then my footage will be good for me. And having my pc monitor better match the HDTV now puts me closer in the ballpark of consistency. Anyway...this is all subject for another post.

So...I spent time going through all the different settings on the HV20 and came to these settings as looking best overall.

TV mode
AWB (or set it manually)
Custom Effect OFF (or)
Custom Effect:
Color Depth = 0
Sharpness = +1
Contrast = -1
Bright = 0

Now the interesting kicker is that when I switched the camera back to FULL AUTOMATIC - the results were very near identical. In other words, full auto actually does look suprisingly good for a set and forget setting. But I definitely came to even more conclusions that I DO NOT want to ever be using Cinemode ever again, it just softens the image too much. Now...you can get a very "slight" flavor of the Cinemode contrast if you use the NEUTRAL setting in the Image Effect. It's not anything drastic like Cinemode does to the contrast within an image, but it leans more in that direction.

Very interesting Nathan, thanks. However, you might want to draw a long cable and shoot with the HV20 outdoors while still watching the signal on your HDTV. My husband said that outdoors the TV mode *might* not be the best mode, but rather the P or AE. Although it indeed seems that the lower the shutter speed, the best the quality.

Thanks for that bit of additional insight and advice. I guess I should have mentioned that my settings were for indoors. Yes, I would think that once outdoors the equation lends itself to change, but if there is any definite in all that has been shared today amongst ourselves - it's that Cinemode is a NO GO MODE. :o)

True. Me and the entire film/television industry.

Interesting tests.

-Stu

I am a sucker too: http://eugenia.blogsome.com/2007/08/...ding-tutorial/
However, clarity of picture is every bit as important. ;)

Stu,
Care to share some thoughts on why you chose CINE over Tv on your recent shoot?

Euginia,
Yeah, the AUTO mode and strong natural light don't mix all that well. Oversaturated colors and blown-out highlights are what I've noted. Maybe Tv mode with a contrast filter and and ND filter (or 2) will be the answer.

With any luck I'm going to do the more careful, comprehensive testing today. I'm going to use a lot of well balanced light.

I did, here.

-Stu

Thanks, Stu. yes, I had read your entry previously. I was curious if there was anything other than the gamma response that attracted you to CINE. What's your opinion on the apparent low pass filtering and poor compression that's going on?

My opinion is that I need to do some tests of my own. What it looks like to me is that the HDV compression, which is significant, is clobbering the high frequency detail in CINE MODE. It seems to me that Auto mode adds enough sharpening that the heavy compression has a harder time obliterating it.

An interesting test would be to compare the modes via HDMI out, so we could take HDV's gnarly compression out of the equation.

I'm not convinced that any softening is being added by CINE MODE, but that's just a hunch.

One downside of sharpening in post is that it happens after compression, wheras the camera's sharpening happens before.

I realize that tack-sharp images are or paramount importance to some people, but if I were one of them I a) wouldn't be shooting with a 35mm adapter and b) wouldn't be shooting to a highly compressed format with a sub-$1K consumer camera.

-Stu

I did 53 set-ups this morning, including a reference file from a DSLR. Gonna take me some time to edit and compile, but when done, I'll post here.

Alas, I could not record via HDMI, but I was able to monitor it via HDMI at 1080P, and I think that there might be something to Stu's theory about the compression sucking away the low contrast detail ... that's definitely a common side-effect of MPEG/JPEG style compression schemes.

Still, I think people will be able to garner alot about the different characteristics of the different modes, even with the HDV compression. I'll post as soon as I can, but I want to do this right. ;-)

It seems unlikely that the cinemode is really lowering resolution. Typically like cameras a lot of pp sharpening is applied to make the picture look sharper. But looking at Nathan's pictures that's not the case (there's definitely resolution loss).

So did anyone check if the gain is equal? If you crank up the gain I can see the CCD sensor would capture more resolution yet also more noise (and you can clearly see additional chroma noise in some of the posted pictures)?

So I'm curious how does the comparison look like in bright sunlight? Are the resolution results still that different?

Luc

Stu, please post your findings when you do your tests. Either here, or on your blog. Thanks.

I took the cinemode image you posted and adjusted the contrast and added some sharpness and it ended up looking much closer to the normal image.

I really do (at least from these samples listed above) think it is just a lack of electronic sharpness that is going on here. I will try some of the other images.

For what it's worth as a visual effects artist I have always shot video with the sharpness turned all the way down or off. To me it create the most raw image I can work with for FX work. I would rather have a soft raw image to key then one with a black or white ring around my subject.

I have to give props to Canon for at least having the vision to think of giving people the option of having as close to a raw mode as you can get with a HDV camera.

I have seen unprocessed film footage and it isn't pretty either. In fact even a lot of Hollywood SD DVD's will be sharpened because after the film transfer they still are not sharp enough. I remember similar threads with the JVC HD100 camera. In it's normal settings the images looked sharp but you could tell there was a lot of sharpening going on in the camera. With the electronic sharpness turned off the image became super soft almost out of focus looking. No the JVC camera was not softening the footage either. It is just how a chip readout looks like in it's raw form. The SONY Z1 has the same thing going on. Most people will tell you who have tried it that if you turn the sharpness off on the Z1 the footage becomes almost too soft to use. The fact that the HV20 still looks like a nice HD image even with the sharpness turned off says a lot about the level of raw detail from the chip.

>it ended up looking much closer to the normal image.

Well, upload the fixed image so we can conclude that ourselves too. IMO, this is lost detail, not recoverable, no matter what sharpen algorithm you use. Please, prove me wrong. :)

Unable to test my own cam, and rather using what I've seen here, I agree.

The HDV encoding happens AFTER the CINE mode (or after the footage would be sharpened, or not). RAW IMAGE > PROSESSED BY CUSTOM SETTINGS (incl Tv or CINE mode) > HDV compression. If it's not sharpened before it gets hit by HDV compression, it'll be gone - though it was there.

The HDV encoder is thinking "nah, there's not enough detail there.. they won't notice - lets make that area one large mass of blur to save data space in case I need it for the next frame of footage".

We need to find out if it's the HDV compression that's causing this! Someone needs to bypass HDV and do an HDMI LIVE capture for this to be proven!

Here's what I think.

If the HV20 (and maybe other HDV cams) work with the same principle as Wiki describes for DV cameras, then the following is true:

"To avoid aliasing, optical low pass filtering is necessary (although not necessarily implemented in all camera designs). Essentially, blurry glass is used to add a small blur to the image. This prevents high-frequency information from getting through and causing aliasing. Low-quality lenses can also be considered a form of optical low pass filtering.

Sharpening is often used to counteract the effect of optical low pass filtering. Sharpening can be implemented via finite impulse response filters."


This would mean that the 'in-camera sharpening' is actually necessary to get a decent image. HV20's CMOS is bigger than 1920x1080 so a soft image from a higher resolution that is sharpened as compensation for its bluriness and then downscaled to 1920x1080 would actually preserve detail that would otherwise be lost by downscaling an image that's slightly blurred and low contrast in the first place.

Here's what i ended up with simple curves adjustment and smart sharpen filter in Photoshop:

http://eki.pp.fi/temp/Eki/HV20/Cinem...nd_sharpen.png

Not ALL of the detail is as sharp, probably due to the compression. But i'd say it's close enough.

Personally, my eyes hurt when watching overly sharpened imagery, i'd be perfectly happy with much softer images than cinemode without sharpening, as long as i can keep my highlights intact for color correction, and my edges fringe free ;-)

To me, cinemode was the number one reason for getting this cam, and i haven't changed my mind yet...

It was rather easy to make this Cine Mode shot look similar to the TV mode shot. But it would have been simply impossible to do it the other way - one can't get back the detail from blown out highlights, and getting rid of the sharpening artifacts (fringe, ringing) in the edges is very hard if not impossible too.

So, it's cinemode all the way for me ,-)

That is true - like you, i'm a visual effects guy, and i always shoot as low-contrast, sharpening free stuff as i can.

Except with Z1.

The footage is SO soft without sharpening that it's practically impossible to restore it decently after HDV compression.

So, i shoot with sharpening at 6 or 7 with that cam - luckily Sony's sharpening algorithm is really good, and there's not too much artifacting going on at this setting.

Erik, I honestly don't see the point. Not only you indeed NOT recover any lost detail, but you lost MORE detail when you applied your contrast/sharpening. Here is my proof. Original cinemode image:
http://osnews.com/img/vegas/2.png
Your modified cinemode image:
http://eki.pp.fi/temp/Eki/HV20/Cinem...nd_sharpen.png
Canon's Auto image:
http://osnews.com/img/vegas/1.png

Notice the two windows on the building. The windows' details are very visible in cinemode because that shooting mode allows for such details (and that is the only good thing it has). But when you applied all the filters, all you ended up with is:
1. less detail in general, compared to auto
2. The same loss of detail on windows as with Auto mode.

So, I really don't see the point of Cinemode, sorry. You ended up with the same blow outs as in Auto, plus less detail in general.

As I wrote earlier, just buy a Contrast filter and an ND filter, and shoot either AUTO or TV mode -- depending on the situation. This will give you better quality than cinemode and it won't blow up the colors so you will still have the ability later to color grade with ease.

Alright guys and gals, here it is:

www.12south.com/video/hv20/modes/index.html

Even though I've formed some strong opinions on what's going on after this test, for the moment, I'm going to hold my tongue and let everyone draw their own conclusions.

Any questions, let me know. If you really, really want a PNG of a particular set-up, let me know via PM.

Enjoy!

Joseph, thanks for all the hard work. However, I really don't like you saved them as JPEG though. It just won't help us make an evaluation, because JPEG's artifacts are very close to sharpening/mpeg2 artifacts in look. I am sorry for being anal about it since the beginning of the thread, but it's true. If you didn't have enough bandwidth/storage online, I could have helped. I also hope that you actually cropped the pictures rather than resizing them. Because if photoshop actually resized them instead, then the whole test is invalid, because this is not real frames we are seeing, but photoshop's algorithms.

My PNG pictures are unedited btw, grabbed from a lossless codec playback screen.

From what you have though, I still like the TV mode better.

- Images are cropped. NO RESIZING. Exact pixels. Promise. I purposely didn't fill the frame with the chart so as to stay in the sweet spot. (The last set, the exposures, do use more of the frame, though.)

- The JPEG compression is VERY light. (If you like, I can post a difference image to show just how light.) The massive size increase to use PNG's just isn't worth it for a comparison of the same shot on the same camera. Because every take is compressed exactly the same way, minor JPEG artifacts or not, you can draw a valid conclusion between the any of the images as long as you are comparing within this group.

- If you want some particular shots as PNG's to study, just let me know. I can even make the transport stream available if you really want.

Here's the difference. Negligible.

http://www.12south.com/video/hv20/modes/difference.png