All tests based on pre-production model
Here we come to some areas which aren’t perfect on the GH4, but which nevertheless are both a step forward from the GH3. I like what Panasonic have started here with 96fps slow-mo. This is the first consumer camera to deviate from the video standard of 60p to give us a genuinely useful creative tool which sets the ball rolling for improvements in future models.
I love slow-mo for adding more drama to a shot and a graceful feel to otherwise jarring movement.
This feature will in particular find a lot of use in low-budget action sports productions where the price of a Sony FS700 would normally prohibit any kind of slow-mo beyond the 1080/60p already offered by the GH3.
The GH4 goes right up to 96fps which can be conformed to a regular frame rate (like 24p) in-camera.
96fps
The Panasonic GH4 has a Variable Frame Rate feature in the menus which is available when 1080p at 100Mbit is selected in the codec menu.
Between 24fps and 60fps image quality is very good, the same as you’d get from the normal speed 1080p video modes. The 100Mbit/s data rate is enough to prevent any mud or blockiness ruining motion blur or fine detail. If you’d prefer to add slow-mo in post from regular speed footage, it is perhaps best to choose the 200Mbit/s codec and 1080/60p though. That bitrate is very impressive given the frame rate. AVCHD 2.0 on Sony cameras for example only goes to 28Mbit/s for 60p and the GH3 went to 50Mbit/s.
At 96fps the sensor mode changes to allow for the faster scan. It appears to line-skip or pixel bin more aggressively than at 60fps to gain speed.
The result is a softer image and one that has some issues with moire and aliasing.
This requires that you select your shots wisely. If you do it can yield nicely detailed results.
Shots that tend to work well are at the longer end, and quite close-in. Faces work well, head and shoulder framing, sports and the usual fast moving subjects which slow-mo suits so much. What doesn’t work so well is wide angle stuff, landscapes, architecture – but then these are still anyway so why shoot them in slow-motion?
I think for most applications people will find 96fps on the GH4 a lot of fun.
When you want more quality you will have to rent an Epic. The FS700 is also better but that is north of $7000.
It’s worth noting that ex-tele crop mode is not available above 60fps.
Rolling shutter
CMOS sensors expose video line by line with a very quick vertical scan which we call ‘rolling shutter’. The side effect is that the top of the frame is captured earlier in time than the bottom.
The end result we call ‘jello’ or ‘skew’ and it’s most noticeable when you have straight vertical lines running top to bottom through the shot and moving quickly left or right. Jello is also an issue when doing violent handheld camera work, crash zooms or whip pans.
[vimeo]http://vimeo.com/91881977[/vimeo]To solve this problem 100% is not easy, you need a global shutter or a mechanical shutter. High end cinema cameras like the Alexa and Epic just use a faster rolling shutter, as global shutter circuitry takes up room otherwise used by capturing light. This is why the Blackmagic Production Camera has a lower sensitivity than the Cinema Camera and why the Sony F55 has a lower native ISO than the F5.
Of course DSLR stills can use a mechanical shutter but this isn’t possible for video on a consumer camera. Motion picture film cameras use a mechanical rotary shutter and that’s exactly what is featured in the Sony F65 after the company moved from a CCD with global shutter in the F35. Such an expensive, large and power hungry device you will never likely to see on a consumer camera. Additionally a very fast sensor reading a huge amount of data would create too much heat for the small body to dissipate without a fan and would require much larger batteries, so sensors tend to be clocked slower in small consumer cameras.
Panasonic have made great efforts to speed up the sensor in the GH4 so in slow-mo at 96p and standard 1080/60/50p the GH4 has less rolling shutter than the GH3 thanks to a 50% faster sensor readout.
In 4K mode the sensor has to readout 4x the amount of data than in 1080p and the 50% increase in readout speed isn’t enough to compensate.
The result is that in 4K the GH4 has slightly worse rolling shutter than the GH3 in 1080p.
[vimeo]http://vimeo.com/91802787[/vimeo]In my personal opinion the severity of the GH4’s 4K rolling shutter skew is not as bad as what we’re seeing from the Sony A7S, but is close to the Blackmagic Cinema Camera 2.5K.
These tests are extreme – you would rarely use these angles or movements on a real shoot. The purpose of ‘a test’ is to exaggerate the rolling shutter skew in order to analyse it, otherwise you’d be hard pressed to tell any difference.
In electronic shutter mode for stills, the GH4 can silently shoot 16MP raw shots which is great – but of course these also use a rolling shutter so are prone to the related artefacts like skew.
Coming next
I’ll be shooting a comparison between the 5D Mark III raw and the GH4.
Can 8bit 4K from a 2.2x crop sensor area really match the cinematic look of 14bit raw from a full frame sensor?
And what is the right ‘special sauce’ we can add to the GH4 to get them to match?
We know we can take the 4K image and oversample for 4:4:4 colour sampling and 10bit luma. This makes for VERY good HD.
We know we can use Speed Booster to give the GH4 a Super 35mm size sensor (1.0x crop over the cinema standard and 1.5x crop over the 5D Mark III) to get the rendering of lenses a bit closer to full frame.
I want to see how this looks in reality.