Camera/Lens performance compared to Varicam and Panny 400.

[Matthew Greene]

I went to Scott Billups' website www.pixelmonger.com and downloaded the full rez captures of the DSC charts he's done with the Varicam and Panny 400, I put them side by side with the chart Shannon shot.

I noticed that the black and gamma levels on the H1 had to be brought down significantly to match them to the other cameras. This is probably because of the way Scott set up his cameras and it shouldn't be an issue to set up the H1 in this way.

Although this chart isn't the way to determine this, the H1 hinted that it has slightly less dynamic range.

However, the most significant find is that it looks like the camera/chips are very good performers but the lens fails miserably. The chromatic abberations severely hinder the performance the camera seems to be capable of.

We knew (and if not, it is obvious now) that producing a lens for a high def 1/3" imager is a significant challenge. I wonder if Canon is considering producing a higher quality (and cost) lens for the camera or if all the lenses will be this severely compromised. Aside from the 35mm lens adapters, does anyone have ideas regarding lens options for this camera? For example, could the Fujinon 13x lens for the JVC 100 be adapted to mount on the H1?

[Petr Marusek]

I would not take it too seriously. The guy also has Sony F900 chart:
http://www.pixelmonger.com/hd_assets/chrt_f900.jpg
With this camera Star Wars and other films were made and the resolution on this chart is about 550Hx400V. After compensation for wrong frame size it comes to about 700Hx500V, which is a lot less than on the Varicam, which is nonsense. I would not trust the "no chromatic aberation at all" on the Varicam. See the chromatic aberation on the Sony. All prism/lens combinations used on these cameras cause some chromatic aberation.

[Matthew Greene]

Yes, the shot from the F900 is trash, however, there's no visible chromatic abberations (where different frequencies of the color spectrum don't line up, most noticably in the corners.) in that still, it looks like it might be a field capture of a (blured) moving image (look at the aliasing and blur pattern). I can testify to the F900's imager's resolving power being superior to the Varicam's. However the charts from the other cameras seem to have been properly shot.

A high quality, well engineered lens would show no discernable chromatic abberations at most stops. The Digiprimes I've MTF tested have virtually no visible chromatic abberation and resolve pretty evenly from center to corner at just about any stop. Of course it's much easier to do when you have a 2/3" target size. Given it's geometry, the optical block prisim would introduce diffraction (chromatic abberations) throughout the whole image, not just the corners.

How can I not trust the stills from the Panasonic charts? There is no way to undo these lens defects in post, if the charts looked awful I wouldn't trust them (as with the F900 chart) but the Varicam and 400 charts look accurate based on my experience with the Varicam.

Of course, there's no science in how we're comparing here but the lens build differences aren't subtle either. I'd like to know what focal length and stop Shannon captured the chart at.

[Ash Greyson]

What lens was on the Varicam? It is not fair at all IMHO to expect the same from a $9K cam as a $125K+ cam... I am sure they COULD make a better 1/3" CCD Lens but at what cost? $15,000+?



ash =o)

[Matthew Greene]

From the behind the scenes photos looks like a HD Fuji Zoom (maybe $25K). The Varicam can be had for under $50K. The HDX400 for $26K.

Don't get me wrong, I'm blown away by the H1's camera section performance, it's much better than I expected. However, I haven't seen chromatic abberation this bad in any lens I've ever used (be it on film, stills or video), not even in cheap russian film lenses. I'm sure that it's mostly due to the optical compromises when engineering for a 1/3" target size.

However, it almost makes me want to use one of my Schneider or Angenieux lenses from one of my Beaulieu Super 8 cameras on it, they are great lenses and are engineered for a small target as well. From an HD telecine of Kodak 7201 (50 ISO) S8 stock they easily resolve close to 800 tvl without such a significant optical compromise.

I honestly have to say that I would and have paid more for a lens than the camera it's on. For as long as I can remember the cost of a great zoom lens has been close to or greater than the camera itself. A complete set of primes can easily cost way more than a motion picture camera.

[Petr Marusek]

There is no #400 Panasonic HD camera in the charts that you're referring to, plus that camera costs a lot more than $26K and Varicam costs a lot more than under $50K, unless the pricing recently changed drastically. The DVCPROHD camera that is on the site is an old discontinued 1080i camera with 1080i chips. That is the reason for the high resolution. The new #400 HD camera has Varicam's 720p chips, which it's using for 720p, 1080i, and in Europe also for 1080p.

[Matthew Greene]

Sorry, I didn't clarify that the HDX400 chart didn't come from Scott Billup's site although the HDC20 chart is also a good comparison candidate.

If you check Panasonic's site at: http://catalog2.panasonic.com/webapp...tGroupId=14616

There you'll see that list price for the HDX400 is $32,000 and $65,900 for the Varicam bodies. When you go through a dealer you can get 20%-30% discount on list prices. Plus right now Panasonic is offering up to $5,200 cash back on packages. A dealer quoted us under $50K for the Varicam early last year. So it can be had for under $50K.

However, the post is about options to the stock lens on the H1 which has severe defects (whether it's compared or not to the other cameras/lenses). I'd like to know if anyone has any practical ideas other than waiting for Canon to release a better lens.

[Petr Marusek]

The CA of the Canon lens will not be visible on normal moving color images. Have you seen the CA on some of the early HD zooms that cost tens of thousands of dollars? Have you seen the CA on the JVC HD-100 lens? Now, that is a serious problem. Canon H1 with the standard zoom lens is the best low end HD filmmaker's tool there is.

Panasonic does not have any resolution charts on their site so I wonder, where did you see the Panasonic Model 400 HD chart. I'd like to see it.

Normally you'd never be able to get more than 15% discount on this type of equipment.

The eye resolves luma (B/W) a lot better than chroma (color). The resolution is highest in the center of vision, so when you project normal motion footage from the Canon H1 on a theater screen, the CA will not really matter at all; the high resolution is so much more important. It works a little differently with smaller or less resolving screens but again the CA of the H1 lens will not be an issue in real life situations.

[A. J. deLange]

The lateral chromatic aberration for this lens appears to be about one or perhaps 2 pixels (uprezzed to 1920 x 1080) at the edges of the picture. Is this sufficient to "severely hinder" the performance of the camera? It is less than the color resolution in HDV mode certainly and is about at the level of color resolution in 4:2:2 SDI mode. Just for comparison a 35mm Nikkor (35 mm still camera with 35 mm focal length) shows every bit as much. Thus this lens (the Canon) is as good (in this one quicky test) as a prime intended for 35mm film (though this is not one of Nikons latest effots) WRT CA and is, thus, plenty good enough for me. Is anyone troubled by visible fringing attributable to CA with the XL-H1?

One further note: Lateral CA is correctable to some extent in post though it is probably too much of a PITA to consider practically. The video is mapped back to RGB and the red and blue images are radially scaled according to a mapping polynomial obtained via calibration of the lens. The scaled images are then transformed back to Y'CbCr.

[Edit] Thinking about this while driving in to work it occured to me that perhaps the 35 mm prime is NOT good enough for use with a 1/3" sensor. It cannot be stopped down much beyond f/8 before we start to run into the diffraction limit. Then it also occured to me that diffraction is wavelength dependent i.e. the point spread function for red light is about twice the size of the PSF for blue. OTOH you can't really make these lenses better without violating laws of physics. So the real question is "Are people actually troubled by CA with the XL-H1?" If so it's probably a condemnation of the 1/3" sensor rather than this particular lens.

[Ash Greyson]

Please tell me where you can get a new Varicam for under $50K... Also, the XLH/HVX come ready to use, no need to buy monitors, eyepieces, lenses, etc. etc. etc. A normal Varicam set-up with a decent lens is over $100K.....



ash =o)

[Alister Chapman]

I would like to know what f-stop the lens is at when the sample images are being produced. Could it simply be a case of the lens being stopped down too far, would extra ND's help?

The fact that you can clearly see the CA is telling me that yes, in the instances I have seen the lens is limiting the quality of the final image, don't know how anyone can suggest otherwise, but lets be realistic, a 20x zoom with image stabilisation at this price has to have had some corners cut.

[A. J. deLange]

The tests I did were at about f/2.8 (both the Nikkor and 20X). In neither case are we near the diffraction limit but the larger aperture puts more demand on the lens in terms of CA performance. ND filters would require still more aperture and potentially, therefore, more CA.

I for one cannot clearly see the CA degradation - that's why I have to look at the R, G and B images separately to detect it and that's why I am interested to know whether other users are seeing it clearly in their images. Remember that CA amounts to 2 pixels at the edge of the picture, 1 pixel about half way out and less than a pixel in the center part.

[Steve Rosen]

I, too, was troubled by the appearance of CA in my bench tests with this lens. I was being extremely critical and looking for it, and found it. I then went back and evaluated the footage that I've been shooting with both the 20X and the 16X manual. In real-world shooting the problem pretty much goes away, especially with the gain at -3 (not necessariy related, but it does help). One scene with a woman in a red sweater against the beach showed noticable fringing, but not much worse than some (very expensive) lenses I've used on 16mm cameras.

I would imagine that when shooting something like green screen (which I never do in my work) it could be an issue, but in normal shooting it's not something an audience would ever notice... Zoom lenses are always a design compromise, with some focal lengths performing better than others at certain aperatures... The lens isn't great, but it's better than okay in my experience...

[Matthew Greene]

Originally Posted by Petr Marusek

The CA of the Canon lens will not be visible on normal moving color images. Have you seen the CA on some of the early HD zooms... ...the JVC HD-100 lens?

Send it out to your HD monitor and check it out, looks almost like a bad convergence issue on the monitor. Early zooms? Back with Sony's HDVS tube cameras this wasn't as much of an issue, with the tubes you could adjust their individual geometry to compensate for most visible CA. Yes, I can see occasional Chromatic Abberations on HD broadcasts all the time and while in my opinion the H1 is the best low cost HD camera out there, I simply can't consider it without a better lens than that and am willing to invest in it. In my opinion the stock lens simply doesn't do justice to the sensors. The stock JVC HD-100 lens is not good either, the Fuji 13x seems to be a bit better.

Originally Posted by Petr Marusek

...where did you see the Panasonic Model 400 HD chart.

It came from a Television engineer in Chile that was testing cameras for their network, I have the raw file, it's about 6MB, I'll be happy to ftp it if someone tells me where.

Originally Posted by Petr Marusek

Normally you'd never be able to get more than 15% discount on this type of equipment.

I guess, if normally means going to a dealer you don't have a relationship with and paid what they asked you for it.

Originally Posted by Petr Marusek

The eye resolves luma (B/W) a lot better than chroma (color). ...the CA of the H1 lens will not be an issue in real life situations.

Chroma vs. Luma detail is true only when the luma signal retains the detail and the chroma signal does not, but because this is not a recording format but the RG&B capture that produces the luma detail itself, it's noticable. Also, the Chromatic Abberations aren't alone, they cause other issues as well. Look at the charts and you'll notice that there's a severe drop in resolving power and contrast outside the center, this would affect the lens' MTF curve pretty drastically. Again, send it to an HD monitor and judge it for yourself.

Originally Posted by A. J. deLange

Is this sufficient to "severely hinder" the performance of the camera? It is less than the color resolution in HDV mode certainly and is about at the level of color resolution in 4:2:2 SDI mode. Just for comparison a 35mm Nikkor (35 mm still camera with 35 mm focal length) shows every bit as much.

Again, the Chromatic abberation not only affects chroma resolution, it drops resolving power, contrast and the MTF curve. I have personally never used a prime that had comparable defects, so I can't comment on that.

Originally Posted by A. J. deLange

One further note: Lateral CA is correctable to some extent in post ... The video is mapped back to RGB and the red and blue images are radially scaled according to a mapping polynomial obtained via calibration of the lens. The scaled images are then transformed back to Y'CbCr.)

Good theory, similar to adjusting tube cameras to compensate for these issues only that you'll never get the contrast and detail that was lost back. You would only correct fringing. And yes, this is the issue with smaller sensors, they place a great demand on the lens.

Originally Posted by Ash Greyson

Please tell me where you can get a new Varicam for under $50K... Also, the XLH/HVX come ready to use, no need to buy monitors, eyepieces, lenses, etc. etc. etc. A normal Varicam set-up with a decent lens is over $100K..... ash =o)

Well, yes, in this industry when you mention the price of a camera the only thing you're relating to is the head as everyone is going to need different accesories that can vary the price significantly. If I remember correctly the lowest Varicam price was quoted to us by mitmomo in Japan, www.mitomo.co.jp It was within the context of a large package and the head was itemized at under $50K early last year when converted from Yen. The full package with viewfinder, batteries and EFP lens was about $22K more.

Originally Posted by Alister Chapman

I would like to know what f-stop the lens is at when the sample images are being produced. Could it simply be a case of the lens being stopped down too far, would extra ND's help?

The fact that you can clearly see the CA is telling me that yes, in the instances I have seen the lens is limiting the quality of the final image, don't know how anyone can suggest otherwise, but lets be realistic, a 20x zoom with image stabilisation at this price has to have had some corners cut.

Of course that a cheap lens has to cut corners, that's why I started this thread, all I want to know what options may be available to not use a cheap lens.

I want to know what stop and focal length it was shot at too, if this is F2 I guess it's OK but if it's F8-11 that lens has serious issues that would only get worse at other stops. If you do find the stop with the least defects it is possible to use NDs to always use that stop in bright situations, the problem is that you have enough light to always shoot at it, the best stop is probably 8 or 11. This leaves you no DOF choices though.

Originally Posted by A. J. deLange

I for one cannot clearly see the CA degradation...

Look at the resolution wedges on the sides of the charts, compare it to another chart such as the ones previously mentioned... you can't miss it. Not only does the fringing stand out but there's just a blur on the wedge and the contrast is really low between the black lines and the white BG. This isn't the camera, which I think is great, it's the lens.

Originally Posted by Steve Rosen

I, too, was troubled by the appearance of CA in my bench tests with this lens... One scene with a woman in a red sweater against the beach showed noticable fringing, but not much worse than some (very expensive) lenses I've used on 16mm cameras.

What stops showed the least fringing? And what 16mm lenses are you comparing the fringing to? I just haven't experienced any really bad issues in 16mm but I've always used good lenses. Please note that as I mentioned before I'm not just dissapointed in the fringing but that the abberations affect sharpness, contrast and the MTF overall.

[Matthew Greene]

Ok, here's an image of the H1 split with the HDC20 (a discontinued panasonic 2.2 Mpixel 1080i camera).

This is by no means a scientific comparison, it's just a reference. Both images originated at 1920x1080 but the lens defects are clear even though they are zoomed in jpegs.

Of course this isn't a comparison of image quality, but with the other image as a comparison the lens issues should become apparent to those who haven't detected them.

The image, http://www.darrenromeo.com/assets/index.htm