PDA

View Full Version : HD 24p Comparison Seminar Results (2/3/2009)



TVSpecialists
02-24-2009, 09:06 AM
2009 HD Camcorder Comparison Dynamic Range Tests
Ted Bollinger, TV Specialists, Inc.
This year we again thank all of the manufactures, Sony, Panasonic, JVC, and Apple and their representatives who participated and made presentations at this year’s 4th annual HD Camcorder Comparisons. This year we wanted to evaluate and explore the dynamic range of recently introduced HD camcorders. A test of this kind requires that we limit the number of cameras and manufactures that we evaluate and present. This year we limited the testing to 12 of the most popular progressive 24p and newly introduced models. (See 24p Comparison Chart)


Any discussion of video dynamic range must include an understanding and knowledge of knee adjustments (film stock characteristics refer to highlight handling as the shoulder) and black stretch and gamma adjustments (film stock type refers to shadow areas as toe). The ability to simultaneous capture both highlight detail and shadow detail from a scene is referred to as dynamic range. One of the key differences between film and video is the dynamic range in video is largely determined by the camera, its processing, settings and sensor characteristics whereas in film it is a function of the film stock type and its characteristics. Since video monitors have their own display characteristics and dynamic range it is very helpful to evaluate video dynamic range with both greyscale log charts as well as waveform displays to properly evaluate video dynamic range. (See Slope, Shoulder, Knee and Toe. Maximizing Your Dynamic Range chart.)


While the most prevalent train of thought would have us believe that it is the sensor ( camera chips ) size that determine the dynamic range we found there are actually three other variables that play an important part in determining total dynamic range in a given camcorder. They are pixel density, sensor type, and processing circuitry and their range and type of control. In general, it is usually safe to say that the larger the sensors in the camera the better the dynamic range however, it is certainly not the case in all the comparisons we did. We also found that because some manufacture’s “off” settings from preset scene files included some form of auto knee and also the the range of automatic knee and automatic dynamic range adjustments were so different in different levels of cameras it was impossible in the time we had allocated to evaluate these cameras ( both before and during the seminar ) to reach any definitive conclusions as to which cameras within each size held the most native dynamic range and also which cameras could be adjusted to achieve the absolute widest dynamic range. It also became very clear that without serious processing in Photo Shop or some other paint software no digital still camera including SLRs could equal what we were able to do with most of our HD camcorders live. (so photos do not tell the whole story.)
As a point of reference we also tested the Red camera under the same circumstances. Because of their use of raw files and down-converted HDSDI 720p live output we used post processed full-res files for evaluation. It is also important to note that any discussion of dynamic range stretching using various forms of gamma and black stretch adjustments or post processing techniques result in adding some level of noise to the image. It becomes a subjective matter of how much noise is acceptable in a given situation. That being said, here are the general findings of the twelve camcorders tested:
In general, the larger the camera sensor size the wider the dynamic range. The newest generation of 1/3” CCD sensors did better than expected but still not equal to ½” or most 2/3” sensors. The largest sensor tested, the 35mm size CMOS on the Red camera did not exceed but did equal the newest 2/3” CCD camcorders tested. We also found that the new ½” CMOS sensors in the EX-1 and EX-3 actually equaled and in some cases outperformed some of the 2/3” CCD cameras which is something that we did not expect to find.
The newer generation and higher bit processing (14 vs 12 bit ) camcorders usually outperformed and gave more control over dynamic range than earlier less sophisticated models tested. Some models gave the user considerably more control over the dynamic range and “look” of the image. This could reduce and in some cases eliminate the need for post processing to achieve the desired image.
Although some processing ( especially in the black or shadow areas ) used to achieve higher dynamic range does add some noise to the image, with careful attention to the adjustments on a good monitor during setup, it is possible to keep noise down to a minimum and avoid those extreme setting that add considerable noise. The handling of overexposure (clipping) varied in each camera and some tended towards yellow or some off-white color in overexposed areas of the image which tended to be very distracting to the overall image.

If we use the Red camera as a point of reference most tests and references to its dynamic range put it at or about 10 stops of dynamic range. If this is the case then our results would put all but the lowest cost 2/3” cameras tested at 10 stops as well. The ½” CMOS cameras tested at within ½ stop and most of the 1/3” cameras at 1 to 1.5 stops less than the 2/3” cameras. We also noted very large differences in image detail and visual noise between these cameras, which are also very important to consider in the overall performance of a given camera along with its codec, and actual off-media image quality. (subject of another upcoming seminar)
Conclusion: the absolute best way to evaluate dynamic range between two cameras is to actually test the cameras side by side in controlled lighting conditions and with good monitors with waveform displays. Test both the preset scene files and cine like gamma curves as well as the menu control over knee adjustments and black level control. The long-standing recommendation to expose for the highlights in HD is definitely worth repeating with the added caution that if shadow detail is important to your project do not use traditional video type scene files. If the cameras you are considering are different in sensor technology and pixel density we would strongly suggest that a comparison be made under the above conditions. It would be fair to say that when it comes to dynamic range you generally do get what you pay for, that is the more expensive and newer design of the camera the better its dynamic range, or at least it was in our tests. (Except for the very low cost or older 2/3” cameras). It is always better to compare to be sure.

jaket
02-24-2009, 10:59 AM
Interesting.

Leo Versola
02-25-2009, 10:23 AM
Yes, interesting discussion. Are the full test results, including comments about visual detail and noise levels available available anywhere?

Thanks!

TVSpecialists
02-25-2009, 10:34 AM
Not at this time. Yesterday we saw the new JVC cameras in Chicago. We hope to have it, the new Panasonic AG-HPX300, and any others within the next couple of months to do additional tests and comparisons.

DavidNJ
02-28-2009, 02:23 AM
Something to look forward to. On the chart you list the PMW-EX1/3 as SxS Pro cards. While true, I thought they also work fine with a Class 6 SDHC in an adapter.

Otis Grapsas
02-28-2009, 03:57 PM
The dynamic range of the camera depends on the lighting used, so no single comparison can evaluate it with precision. Each camera has a different spectrum response and uses different gain for each of the three colors. One camera could be close to 3400K lighting in its natural sensor output and another closer to 5600K or higher. Depending on the color correction decisions implemented, the range of a camera with superior color response (white precision and color range) could appear lower in an actual test. The cameras do not even use the standard HDTV gamma any more in most gamma modes, they implement cine gamma and overexposure orotection, so even setting an equivalent exposure besomes a challenge by itself. You can optimise the dynamic range of the camera if you know the lighting that will be used and gain more than 2 stops of dynamic range in some cases by using optical filtering in front of the lens. This allows all three colors to reach maximum saturation at the same point an minimises RGB gain and visile noise in the process, so it provides a true dynamic range boost (saturation level/noise level).

An example: Some cameras will saturate early under incadescent because of a powerful red response. Filtering the red lowers noise (blue and green gain required) and provides extra dynamic range. The same camera will perform better than another camera in daylight and colder lighting, when its red output becomes an advantage. The simple optical filters implemented inside the cameras do not do much to optimise all lighting situations. You have to precisely choose fiters for the target lighting.

Low end cameras optimise their low light sensitivity by compromising color quality, so it's even harder to compare these to others. Sometimes the better engineered camera appears less sensitive or having lower dynamic range in a given lighting, while under good use with filters it outperforms the other cameras in color quality, noise and dynamic range.