Peter Loughran “Having collected my GH4 body on Thursday 8th May (too impatient to wait another day for delivery so drove to collect it as soon as I heard it was at my supplier …my thanks to Paul and Martin at 3D Broadcast I set about running some tests to see exactly what had changed compared to my GH3.
One of the things I really wanted to clarify was exactly what the various crop factors really are as there have been many misleading figures and guesstimates bandied about online, suggesting figures for the ’4K’ formats of around 2.3 (Panasonics own literature and manual only mentions that “The angle of view of motion pictures in 4K is narrower than those of motion pictures in other sizes.” without actually specifying the precise figures).
I knew from previous experience with my GH3, and now confirmed as being exactly the same in the GH4 (including checking the crop factor information in the EXIF data), that the 16:9 FHD and HD (1080p/i and 720p) video formats are read from the full width of the sensor (4608 pixels) but read from a cropped sensor height of 2592 pixels to give the 16:9 frame format from the (4608×3456 pixel) 4:3 format sensor.
As the 4:3 sensor height must be cropped vertically to produce the 3:2 stills format and the 16:9 stills/video format there is obviously a reduction in the frame diagonal (which equals the required image circle) and hence a corresponding increase in the crop factor. There is also a 1:1 (square) stills image format available which maintains the full sensor height but crops off the sides.
Note that the older GH2 had an oversized ‘Multi Format’ sensor which allowed the horizontal pixel readout width to be increased as the vertical pixel readout height was reduced and thus give a constant diagonal and constant crop factor of 2.0
The GH3 and GH4 do not have that constant crop factor so ONLY THE FULL SENSOR 4:3 FORMAT HAS A CROP FACTOR OF 2.0 …all other formats crop into the sensor and thus have higher crop factors.
For the sake of clarity, here are the crop factors (relative to a Full Frame 35mm Still Format 3:2 36x24mm = diagonal of 43.27mm)) which apply to both the GH3 and GH4 when using the full sensor based crops:
4:3 (stills and video [VGA] using full sensor width and full sensor height) = 2.0
3:2 (stills using full sensor width and cropped sensor height) = 2.08
16:9 (stills and video [HD, FHD] using full sensor width and cropped sensor height) = 2.18
1:1 (stills using cropped sensor width and full sensor height) = 2.36
Figures are as calculated from sensor dimensions and format aspect ratios, and confirmed by examining EXIF data (using ExifTool 9.54) which shows a rounded ‘Scale Factor To 35mm Equivalent’ figure and also a more accurate (particularly when using longer lenses) ’35mm equivalent focal length’ from which the exact crop factor can be calculated.
For example: using the Panasonic 100-300mm lens at 300mm the EXIF data shows a rounded ‘Scale Factor’ of 2.0 (when using 4:3 image format) and 2.2 (when using 16:9 image format).
The ’35mm equivalent focal length’ figures shown in EXIF are 600.0mm and 653.0mm so dividing those by 300mm (actual lens figure) gives the more accurate crop factor figures of 2.0 and 2.18 respectively.
Unfortunately, examination of the GH4s EXIF data for 4K and C4K video shows that it does not include the ‘Scale Factor To 35mm Equivalent’ or ’35mm equivalent focal length’ information so I have had to resort to the ‘old school’ method of actually shooting a known scale in order to compare the actual captured images in both the full frame width stills/video and the cropped frame width video images produced in 4K and C4K.
My test subject was a 120cm builders spirit level which has a very clear mm scale along its edge. I set up the camera (locked down on a very substantial tripod) and by careful adjustment of distance and zoom (using the Panasonic 12-35mm f2.8 lens), shooting still frames and examining them enlarged on the cameras display to see exactly what was being captured, was able to frame the mm scale such that the horizontal frame width was capturing exactly 1,000mm.
(I was able to adjust the setup to an accuracy of less than 1mm on the scale so the subsequent calculations should be accurate to approximately 0.1%).
Subsequent video samples, including 4K and C4K crops, could then be compared and ‘measured’ by reading off how much of the 1,000mm scale could be seen in the image (the video clips produced were imported into my video editing program and full resolution frame-grabs exported for critical examination of exactly how much of the mm scale was captured).
Knowing those figures, and the aspect ratios of the different formats, then allowed me to calculate the crop factors of the 4K and C4K videos as compared to the known crop factor when using the full sensor width and thus the overall crop factors relative to the Full Frame 35mm Still Format (3:2 36x24mm = diagonal of 43.27mm) (Note: that’s a stills image format, not ’35mm’ or ‘Super 35mm’ movie formats).
Note: the 16:9 FHD/HD videos show exactly the same horizontal field of view as the still images so they are certainly downsampled from the full 4608 pixel sensor width.
I have used the (arbitrary) observed/implied captured mm dimensions for the different aspect ratio formats (16:9 4K/FHD/HD and 17.07:9 C4K) in order to simplify the calculations in this test. The same % crops could be applied to the actual sensor dimensions if the true sensor area framing sizes were required.
GH4 VIDEO CROP FACTOR MEASUREMENTS:
(Camera & lens locked off to show 1000mm scale as exactly filling the horizontal frame axis).
FHD 1920×1080 (16:9) & HD 1280×720 (16:9) Full Sensor Width = 1000mm = 100% of sensor width.
4K 3840×2160 (16:9) Cropped Sensor Width (showing 85mm to 918mm) = 833mm = 83.3% of sensor width
C4K 4096×2160 (17.07:9) Cropped Sensor Width (showing 57mm to 946mm) = 889mm = 88.9% of sensor width
(interestingly, the 4K and C4K pixel crops are not exactly in the horizontal centre of the sensor but fractionally displaced to the right; possibly to do with how the sensor readout electronics are routed?)
Crop Factor calculations:
FHD 16:9 Format (full sensor width) @ 1000×562.5mm = diagonal of 1147.35mm
4K (16:9) Format @ 833×468.56mm = diagonal of 955.74mm
Crop factor of 4K (16:9) as compared to FHD 16:9 (full sensor width) = 1147.35/955.74 = 1.20
C4K (17.07:9) Format @ 889×468.56mm = diagonal of 1004.92mm
Crop factor of C4k (17.07:9) as compared to FHD 16:9 (full sensor width) = 1147.35/1004.92 = 1.14
These crop factors ARE IN ADDITION TO the known (GH3 & GH4) 16:9 (full sensor width) crop factor of 2.18 as compared to Full Frame 35mm Still Format (3:2 36x24mm = diagonal of 43.27mm).
Crop factors as compared to Full Frame 35mm Still Format (3:2 36x24mm = diagonal of 43.27mm):
FHD 1920×1080 (16:9 Full Sensor Width) crop factor = 2.18 (as previously calculated and confirmed by EXIF data).
4K 3840×2160 (16:9 Cropped Sensor Width) crop factor = 2.18 x 1.2 = ~2.6 (test calculated: 2.616)
C4K 4096×2160 (17.07:9 Cropped Sensor Width) crop factor = 2.18 x 1.14 = ~2.5 (test calculated: 2.485)
These measured figures correspond exactly (allowing for slight measurement variables in this practical test) to my previously calculated figures which were based on the assumption that the GH4s 4K and C4K video modes are taking a 100% pixel-for-pixel crop from the sensor.
This is essentially the same method as is used in the ‘Extra Teleconverter’ modes, on the GH3 and GH4, when pixel-for-pixel crops are used to extract 1920×1080, 1280×720 and even 640×480 (VGA) video from the centre of the sensor …giving corresponding levels of magnification (hence the ‘Extra Teleconverter’ label) as compared to the same resolutions when downsampled from the full sensor area crops.
The calculated, and observed, 4K 16:9 crop factor of 1.2x (over and above the HD/FHD 16:9 full sensor width crop factor) ties in perfectly with Panasonics own stated ‘Extra Teleconverter’ crop factor of 2.4 for FHD 1920×1080.
The 4K 16:9 100% (pixel-for-pixel) crop has twice as many pixels in each axis (3840×2160) as the FHD 16:9 100% (pixel-for-pixel) crop of 1920×1080 pixels so the ‘Extra Teleconverter’ type crop factor is only half as much.
My observation so far is that the 4K and C4K videos are MUCH sharper, even when played back at FHD (1920×1080)(I don’t have a 4K display as yet), than is the cameras own FHD (at 50p at up to 100Mbps IBP and 200Mbps All-i) which is downsampled by a factor of 2.4 from the full sensor width pixel resolution (4608/1920 = 2.4).
The full resolution frame-grabs show very clearly that the FHD frames are not only softer but also contain much more aliasing and moire artifacts than the 4K (in this test the mm scale is effectively unreadable on the FHD video but every single mm can be easily read on the 4K samples …partly helped by the additional crop factor magnification, but even when viewing the full resolution frame-grabs at the same subject size the difference is very obvious).
I suspect that this is probably due to the somewhat awkward 2.4 factor downsampling of the (16:9) 4608×2592 sensor resolution to 1920×1080 and suggests that the in camera FHD video might be better quality if Panasonic had taken it from the same 3840×2160 area of the sensor as the 4K uses (thus giving an exact downsampling factor of 2x in each axis).
Of course, that would entail having a FHD crop factor of 2.6 (same as the 4K) so the ‘must have a bigger sensor & shallowest possible depth of field’ fanatics would be outraged…
As things stand it looks as if I’m going to want to shoot at 4K for the extra quality even when editing/displaying to FHD …if only the GH4s 4K was available at 50p! …now where’s my GH5!..?
For all your video production needs in Scotland, get in touch with Small Video Company Ltd