jcs

Just shot an actress for an audition using the stock PP6. Did quick experiments with saturation and curves, then used the footage as is, no changes. Looks pretty good- fast turn-around. Used a Canon 24-105 F4L at F6.3 (indicated), MB Speed Booster, ISO 1600.

For 1920x1080 pixels, we need at least 3840x2160 photosites for R, G, and B (444). For 422 and 420 targets, we can get by with 1/2h and 1/2hx1/2v R&B respectively. Anything less will result in artifacts (Nyquist). Additionally, we need a relatively strong cut-off filter for sampling frequencies above the photosite spacing (low-pass filter AKA anti-alias filter), else we'll have aliasing artifacts. Right now the camera that provides sufficient resolution with the least aliasing and artifacting, along with wide dynamic range is the ARRI Alexa (with only a 3.5K sensor: would be better with a 3.8K+ sensor in terms of resolution). The C300&C100 do really well in resolution due to the 3840 G photosites and simple color processing (averaging vs. complex de-Bayering).   The RED and BMC look pretty good for real-world content as the aliasing and artifacting can blend in with the noise.

Dropped the F55, F5, and FS700 "4K" down to 2.xK estimated based on sensor photosite count and Nyquist (resolution before aliasing). 8.9M photosites means 3978x2237 photosites*, or max pixel resolution before aliasing of 1989x1118. While the F65 comes closest to 4K, it's still not there: http://www.dvxuser.com/V6/showthread.php?296935-F5-F55-Sensor-resolution-and-bayer-pattern/page8 *Computed using Wolfram Alpha: http://www.wolframalpha.com/input/?i=x%2Fy+%3D+16%2F9%2C+x*y+%3D+%288.9*1000*1000%29

http://reduser.net/forum/showthread.php?47529-EPIC-and-Alexa Here we have a comparison of a 5K RED sensor compared to a 3.5K ARRI sensor. At first glance one might think the RED is sharper and more detailed, and is thus better. The RED sensor is indeed sharper and more detailed: 5K pixels can sample 2.5K resolution before aliasing. The 3.5K ARRI sensor stops producing contrast changes (resolution) right where we'd expect from Nyquist- right around 3.5/2 = 1.8K. After that, the ARRI goes flat, producing almost no detail. If we look at the ring charts, we see zero aliasing for the ARRI, and lots of aliasing with the RED. Thus, ARRI does an excellent job of cutting off frequencies above the sampling resolution of their sensor- this is very important to eliminate aliasing and moire. Their sensor is pretty trick as well (16-bit HDR): http://www.arri.com/camera/digital_cameras/technology/arri_imaging_technology/alexas_sensor.html Which is better in practice? For moving images, aliasing is a major give-away for a digital vs. film experience. When cost is no object, it appears ARRI is used the most. It's also the most expensive camera in this list, despite not providing the highest sampling resolution.

Looks good- didn't see any issues.

Right- and the fact that they got started by digitizing film and retaining the film look before creating a digital camera explains the 'Hollywood popular' image.

http://www.arri.com/camera/digital_cameras/technology/arri_imaging_technology/alexas_sensor.html   That explains a lot- very cool.

I will test at some point, but I thought the Cine gammas control the gamma curve completely, and that the Knee (% determines where it starts) would not do anything when a Cine gamma was enabled (read from a blog post- haven't verified it by testing).   Also note that Color Depth works the opposite as one might think: + values lower output levels and - values raise them.

Higher level image compression has been a long time goal in the industry. A few years ago fractal compression was thought to be promising, however it proved to be hard to tune and generalize. Wavelets were also promising, but so far only work reasonably well for static frames. Using textured vector contours sounds interesting. Looking at the PDF paper there were limited examples- hopefully it does indeed generalize and could prove useful in moving image compression technology forward. For video, I didn't see anything in the paper discussing bitrates and efficiency (sounded like it was all I-frame so far (no interpolation between frames)). Without interpolation frames, it won't be able to compete with the tried and true motion-estimated DCT methods (which effectively move textured blocks around- quite efficient).   Also note that this technology won't allow scaling an image up without it becoming soft. The contours will be detailed, but the internal texture (pixels) will get soft. Someday it will be possible to analyze and generalize texture detail, so it can be synthesized with very little data, at any desired resolution. This would of course be an iterated system, using DNA/fractal-like generators (the DNA code for a human being can be compressed down to the size of a single JPEG image, for example- the ultimate compression known to man at this time).   "You are data: 4MB to be exact" http://www.stemcellcite.com/Texas/bioinformatics

The FS700 does have aliasing issues in certain cases. It appears to be due to the the sensor/sampling design. It provides ~1000 lines of vertical resolution but only 800 lines of horizontal resolution before aliasing (and 1000+ lines diagonally). The C100 for example has less aliasing issues and full 1000+ h and v resolution, however it does not provide slow motion. In most cases the FS700 aliasing is very minimal, and even when visible most people won't notice (I see tons of content with aliasing on TV/BluRay/Netflix, etc.). I notice for a sec then ignore it. The somewhat low resolution 5D3 has pretty much zero aliasing, but resolution is 1600h pixels at best (and no slow motion except 720p60).   The good news is that spot aliasing is easy to fix in post, especially for slow motion shots (just track the bad spot and apply suitable tapered Gaussian blur and it's gone. I tested this with your glasses example and it was easy to do and worked great (in Premiere Pro CS6 it all ran in real-time (I have AE but won't use it unless I have no choice- not real-time)).   I would expect the next upline cameras to do better, such as the F5 and F55: I would expect little or no aliasing in normal speed or slomo at those price points. The manner in which the FS100/FS700 is gimped in horizontal resolution and aliasing could be due to business reasons (instead of technical ones). When comparing cost, the FS700 is right where it needs to be for features and performance...

Updated top post for recent discoveries with FS100/FS700 and C100/C300.

I'm learning as I go- recently purchased the camera and Speed Booster. Shot a bit today with a custom profile with Detail set to -7 and sharpened in post (recorded internally as well as to a Nanoflash). With color correction and grain, I'm getting filmic looks with much more detail than my 5D3, especially wide shots. No deal breaker issues so far with the Speed Booster. See the links I posted, Alister Chapman posted a profile that might work for you. I'm testing and learning the settings to create exactly what I want from scratch.

@Leang- 120 & 240fps should have the same aliasing behavior ( http://provideocoalition.com/awilt/story/high_speed_and_low_light_with_the_nex-fs700/ ). The tree branch in the example has aliasing internally- not over exposed. This can be helped by adding grain and/or using Neat Video. The Edge aliasing in the glasses should be helped using Cinegamma 2 and turning off or down "Detail" (sharpening). 60fps should have no additional aliasing vs. 24fps. In some cases shooting at 60fps and using After Effects Pixel Motion interpolation or Twixtor to create 120, 240fps or more could work well.

You can completely remove the aliasing in the glasses case by using a tracked effect over the glasses. In Premiere Pro that's a Track Matte Key. Once the area is tracked with a sufficient mask shape, you can apply a Gaussian blur of 14 (horizontal) and the aliasing will be gone. Add in some noise grain and the final result looks good.   Thus, testing with Detail turned off along with Cinegamma 2 would be a good start.

Picture Profile 1 appears to be the same as "Off"- modify it first.   You can copy from other profiles, reset back to default (original setting for that slot), as well as save them all to removable flash and reload them flash.   For your glasses example- try a gaussian blur of 4 (horizontal only) and add film/noise grain. Can't completely remove, but after this quick test unless you look for it, aliasing is not distracting. Same should work for the tree example. Trying out picture profile settings should be helpful. Use the on-screen histogram to check live exposure. Peaking works great for checking focus (as does the focus zoom button on the handle ("expanded focus")).

See the charts here: http://provideocoalition.com/awilt/story/high_speed_and_low_light_with_the_nex-fs700/ The FS700 has more vertical than horizontal sampling resolution. More vertical resolution makes sense since this is a 422 capable camera (where 422 has twice the vertical color resolution vs. 420). The FS700 was rated at 800 lines horizontal and 1000+ lines vertical using a trumpet chart. Horizontal TV lines looks to be around 1000 when using a chart with straight vertical lines. See the charts at the same link above showing what happens to resolution, detail and aliasing when using slow motion: even at 120 & 240fps, there will be some aliasing.   The FS700 has a Detail setting in the picture profiles which you can change. It allows separate horizontal and vertical sharpening. You might try setting up one picture profile with Detail set to OFF, and another set to Default, and another set to turn sharpening off on horizontal detail only. Then shoot a few clips and compare.   Also try Cinegamma2, which is supposed to prevent superwhites (you'll select this as a gamma setting in a picture profile). Also note you can copy picture profiles, so can have the same one in multiple places (say 1-3), where you only change a setting or two for each when testing. See: http://www.xdcam-user.com/2012/07/what-to-do-if-your-highlights-look-clipped-fs700-in-particular-but-applies-to-many-cameras/ http://www.xdcam-user.com/forum3/viewtopic.php?f=47&t=1069

provideocoalition site is slow, but appears to be working.   Why not use/create a custom picture profile? Minimally worth experimenting to see what you can come up with after reading the manual and the provideocoalition page.

re: 6D & bolex please post numbers & links and I will update list. GH2 & GH2 trumpet chart, both rated by reviewer as 700 lines before significant aliasing (looks like detail holds up well past 1000 though) http://provideocoalition.com/awilt/story/first_look_panasonic_lumix_dmc-gh3_dslm_hybrid_still_video_camera/P2

Try adjusting the picture profiles to prevent aliasing on bright edges (I've seen this too): e.g. use the Cinegammas as well as turning detail to -7 (or experiment with the Detail settings). http://provideocoalition.com/awilt/story/review_sony_nex-fs700_super35_lss_avchd_camcorder/P2 http://blog.abelcine.com/2012/09/18/sony-fs700-scene-files-from-abelcine/   Since aliasing is a form of high-frequency noise, perhaps try Neat Video to remove. A super-resolution program could remove this kind of aliasing- perhaps also check out the Dark Energy plugin for After Effects.   Here's a way to remove aliasing but will soften the final image: (blur if necessary) and downsample the image until aliasing is gone, then upsample using bicubic or better (such as Lanczos). A final sharpen can help.

FS100: GH2: From these charts, the FS100 has higher peak resolution but the GH2 has more area under the curve and thus better overall accutance. 5D2: 5D3: Similarly, the 5D3 has more contrast at higher frequencies (peak resolution), however it has lower area under the curve. Thus the 5D2 has better overall accutance straight from the camera. Sharpened 5D3 footage looks more detailed than 5D2 footage (which can't be sharpened much). Thus, overall peak accutance depends on how much the image can be sharpened in post. Sharpening increases microcontrast (but not peak sampled resolution) and will thus affect these types of tests. In summary, the FS100 provides higher peak resolution, and the GH2 provides better accutance. Out of the camera, the 5D2 provides better accutance and after post sharpening the 5D3 provides similar accutance and higher peak resolution. It might be possible to post-sharpen FS100 footage to bring accutance up to GH2 levels.

mark- lol, wish I had the free time! I'm happy to have kicked this off, hopefully the community can take over and contribute to making it more accurate and complete. hmcindie- thanks for the provideocoalition link- found FS700 charts: http://provideocoalition.com/awilt/story/high_speed_and_low_light_with_the_nex-fs700/ http://provideocoalition.com/awilt/story/review_sony_nex-fs700_super35_lss_avchd_camcorder/P3 : he states 800 TVl/ph (horizontal resolution) and 1000+ vertical resolution. Another chart: http://provideocoalition.com/awilt/story/high_speed_and_low_light_with_the_nex-fs700/P2 . From both charts, it appears the FS700 has more vertical than horizontal sampling resolution. Since the straight line chart (vs. curved/trumpets) had better results, it would appear the de-Bayering and image processing may have something to do with the behavior. Note in the trumpet chart near the numbers, resolution is highest- extending past 1000 lines. Thus, it would be helpful to have both charts available when testing resolution. jgharding- I understand what you mean. I have an FS700 and it's hard to believe the strange trumpet chart results, but there they are. The FS700 does OK in the straight line charts, and it looks much better than the 5D3 for real imagery, so I'm happy with the upgrade. The secret to the ARRI is their color science- they've got highlight falloff and skin tone control nailed- that's why it's used so much in Hollywood. Thanks for the links guys- found this which is helpful in better understanding the FS700 camera picture profiles: http://provideocoalition.com/awilt/story/review_sony_nex-fs700_super35_lss_avchd_camcorder/P2

bruno & mark - good points, in a perfect world, however no test is ever perfect. Again, worst case the measured resolution will less than actual, so we know which way the trend can go. Looking at the lines of the ISO12233 chart, it's easy to see when aliasing starts whereas extinction of detail is less clear. In the case of the 5D3, I did a very crude estimate with the ISO12233 chart, and estimated 800+ lines (using a scaled still as a reference), whereas Jason's chart was full sized and shows at least 810 lines (which matches the 1620 bitmap frame reported by the ML team). I didn't see any clear problems with the other charts. Did you? Are there test results which contradict the ones in this thread (if so- link?)?

Hey mark- a chart test is pretty straightforward, right? Align and focus correctly, shoot. Other than having the chart too close, errors in set up result in reduced measured resolution. In my day job I work with cameras, pixels, and math. For example, we could do an automated FFT/DCT for a groups of photographed(video) lines at various resolutions and look at the frequency bins to get an empirical measurement. For such a test, we'd want to clean up via sharpening the source frames to remove as much low-frequency information as possible (improving accutance (a form of microcontrast) as much as possible). I suspect the slashcam.de test did something similar and didn't post-sharpen the 5D3 footage, resulting in an abnormally low high-frequency measurement. The charts I found looked pretty solid. Which one(s) do you feel weren't done by the book?

I realized that we could reasonably simulate test chart results in software. A camera system (Gaussian) blurs light before the sensor (anti-aliases) then decimates (A to D samples pixels) into discrete photosites. I downloaded the ISO12233 test chart and converted the PDF into a 1200 pixels/inch bitmap (Photoshop CS6). I then Gaussian blurred it 3 pixels and downsampled (bicubic smoother) to 1920x1080, then sharpened it to make the results more clear (to increase accutance): Note aliasing starting above 900 lines. I went back to the blurred 16K image and did the same process down to 1620x910 (+ sharpened): Note that aliasing starts to become apparent around 600 lines. Blurring and decimating is a common method to resample pixels. However, there are better resampling methods than bicubic such as the sinc-based Lanczos (Premiere Pro uses Lanczos). In the 1920 example, we see it is pretty clean all the way up to 1000 line pairs. Compare to an actual FS100 test chart shot: http://www.dvuser.co...fs100/chart.jpg Looks like a solid 1000+ TV lines (with two interesting diagonal aliasing bands). Now compare to the GH2 : Different chart; Barry Green rates this as around 800 lines perhaps could be rated higher, but not as high as the FS100 chart result. We could say 900 or so before detail extinction. http://www.dvxuser.com/V6/archive/index.php/t-232565.html Philip Bloom's test compared the hacked GH2 to the FS100's 24Mbps codec. It appeared to me that the FS100 low detail/resolution could be due to overcompression. If using an external recorder, the FS100 should be able to do much better in that test case. The 5D2 was rated at 440lp/ph (vertical resolution), so that would be 880 vertical pixels and if horizontal resolution is similar, we'd get 782lp/pw or 1564 horizontal resolution. Thus approximately 1564x880 pixel resolution. It's possible horizontal resolution is less- if anyone has a reference, please post it. http://provideocoalition.com/awilt/story/nab_2011_-_scce_charts/ The 5D3 providing around 1620x910 resolution (the bitmap size before rescaling to 1920x1080 per ML folks) or around 810 TV lines seems reasonable: https://vimeo.com/39517721 (890 line estimate was perhaps a bit too generous). Closeup of the 5D3 still scaled to 1920x1080 (represented full 1920x1080 resolution on the device): Closeup of 5D3 1080p video post-sharpened: We can see decent line pairs out past 800. 810 would be fair and matches the 1620x910 pre-scaled bitmap reported by the ML team. As always, if there are better chart or other real-world test results, please post them.

FS700 + Speed Booster + Canon 50mm F1.4 (T1.0) + 1 candle @ ISO 500. Recorded with 100Mbps 422 LGOP MPEG2 (Nanoflash). Candle: 24 inches from sensor; face 30 inches from sensor. Straight from camera (stock picture profile 4 (not using Cinegamma)). [attachment=397:J1.jpg]   Using one of the Cinegammas should help with the edges of the candle (aliasing and/or oversharpening, slightly visible on the right side of flame). The red/magenta color fringing may be due to the Speed Booster (don't have the regular Metabones adapter to compare). Not an issue for me, just interesting.