Sony RX100: Getting the best video out of it...

[Bruno]

I just bought a Sony DSC-RX100. After a couple years shooting my personal projects on a Canon 7D, I needed something more compact that I could carry with me at all times, and the Sony DSC-RX100 looked like it could be the one. The great reviews, the ability to shoot 1080p at 50fps with a fast Zeiss IS lens, full manual control in video mode, focus peaking and a 16mm sized sensor convinced me to go for it!

 

My footage shot on the 7D improved quite a lot during the time I've been using it, knowing a camera's strong points and especially its limitations is very important to getting good images out of it, so the first thing I did with the Sony DSC-RX100 was to shoot some tests to help me decide which settings I'll be using when shooting video with it.

 

Like most people around here I learn so much from online reviews and discussion forums, and those have been a great help deciding my gear purchases, so I'm sharing what I learned from my tests as a way to return the favor and give something back to the community. There's been some good reviews of the Sony DSC-RX100 online, and some useful info spread around the internet, but I think this post will cover a lot of useful information for whoever's interested in this camera for video, and much of this info will also apply for any other similar camera.

 

I started by turning off all the automatic picture improvement options, as they usually degrade the quality of the image and make it less gradable, then I set the codec to AVCHD at 28mbits and 50p (PS).

 

SHOOTING MODE

For video shooting I'd recommend setting the top wheel to video mode and then selecting video-M for manual video shooting

The RX100 does have a dedicated Movie Recording button, and can shoot video on any Stills mode, but you might get aspect ratio and exposure changes once you hit the Record button in these modes. In video-M mode you'll get what you see on screen.

 

RECORD SETTINGS

 

The Sony RX100 can shoot movies in two different formats, MP4 and AVCHD. All MP4 options are below 1080p resolution though, so I won't get into those. In AVCHD mode however, we get 3 different 1080 options:

 

50i 24M (FX) (50i @ 24Mbps, Blu-Ray AVCHD disc compliant)

50i 17M (FH) (50i @ 17Mbps, DVD AVCHD disc compliant)

50p 28M (PS) (50p @ 28Mbps, Progressive Scan)

 

So it seems like we get 50i at 17Mbps and 24Mbps, and we get 50p at 28Mbps, but not really…

The 50i mode is actually capturing 25p images out of the sensor and encoding them as 50i footage, this means that we do end up with interlaced footage, but since it was captured progressively, de-interlacing it will produce a clean 25p image!

 

So if we're looking for the best possible video out of the RX100, we should use 50i 24M for 25fps video and 50p 28M for 50fps video.

In theory, shooting 50i 24M gives us the best bitrate per frame in this camera, almost twice as much as shooting 50p 28M.

Shooting 50fps however would have neighboring frames changing less than when shooting 25fps, helping the encoder do a better job, but still the per frame bitrate is lower, and here's a comparison that shows is. If you look at the darker areas in the back where the window is, you'll see that the 50i version is slightly cleaner.

 

 

Here's an example of something in motion shot at 50i and 50p, the 50i frame was de-interlaced and as you can see there's no interlacing artifacts at all.

 

 

Considering all of the above, I think it's safe to say that the 50i 24M mode, which is in fact 25p @ 24Mbps, will give you best video quality out of this camera.

 

CREATIVE STYLES

Creative Styles is the RX100's designation for Color Profiles.

My first test was to choose the flattest Creative Style the camera had to offer, so I shot some footage of all the different styles.

 

 

After looking at all these different images, I decided to go with the Portrait Creative Style, as it seems to be the flattest of them all.

 

EXPLORING THE PORTRAIT CREATIVE STYLE

Each Creative Style has settings for Contrast, Saturation and Sharpness that can be set from -3 to 3, so I shot some more footage using the Portrait Creative Style in a number of different settings.

 

 

Using the Portrait Creative Style at the minimum settings (Contrast: -3, Saturation: -3, Sharpness: -3) definitely (and obviously) seems to be the flattest style in this camera, but I had to check how well it graded and how it compares to using the default values (Contrast: 0, Saturation: 0, Sharpness: 0).

 

 

On the top left you have a frame shot using the Portrait Creative Style, with all the settings set to 0, on top right you have a frame shot using the Portrait Creative Style in its flattest settings (Contrast: -3, Saturation: -3, Sharpness: -3).

On the bottom right frame I added some sharpness to the flat image, which responded quite well, and on the bottom left frame I added not only sharpness but also increased the Saturation and Contrast in order to match the top left frame (Contrast: 0, Saturation: 0, Sharpness: 0).

The result is an image that matches in color saturation and contrast, but with a much nicer detail and less compression artifacts. It looks sharper and cleaner overall, which made me decide to use this Creative Style and these settings from now on.

 

SHOOTING BLACK & WHITE

The following test is something I've been wanting to do for a while, regardless of the camera. The thinking behind this test was:

 

"If the camera is compressing B&W footage instead of color footage, maybe it can do a much better job at it since it doesn't have all the color information to process, so even using the same bit rate could give us better results."

 

Of course I don't know the details on the cameras' inner workings, but assuming the B&W Creative Style is applied BEFORE the footage is compressed to AVCHD, then this should work. Maybe.

 

 

So I shot some footage using the B&W Creative Style in its flattest settings (Contrast: -3, Saturation: -3, Sharpness: -3), which you can see on top left, and then some more footage using my new favorite Portrait Creative Style, also in its flattest settings (Contrast: -3, Saturation: -3, Sharpness: -3).

 

On the middle left frame, I increased the sharpness and the contrast on the image to make it less flat, and on the middle right frame I did the same, and also desaturated it. As you can see both images are different, since the B&W Creative Style's color conversion is not merely desaturating the image to create a B&W version, it's using a more clever process that also looks better, but anyway, the point here is to test the image compression and figure out which one gives cleaner results, so on the last test frames I increased the exposure by 2 stops to find out how well the images handled it.

On the bottom left frame you can see how much cleaner the image shot with the B&W Creative Style is, compared with the one shot using the Portrait Creative Style, it's actually beautifully clean and overexposing it by 2 stops didn't show any ugly artifacts at all.

 

So my conclusion on this one is, if you're shooting for black and white, and you're sure that's the look you'll want (since it's kind of hard to color B&W footage if you change your mind afterwards), then using the B&W Creative Style will give you far superior results!

 

DYNAMIC RANGE OPTIMIZER

The Dynamic Range Optimizer works when writing to compressed formats, such as JPG, MP4 or AVCHD. It has no effect when shooting RAW.

Its purpose is to capture more detail in the areas that are more prone to get lost when using compressed formats, such as dark shadows. It works in the darker areas of the image, making them brighter and producing a flatter image, which makes it easier on the image compression to achieve better results.

Here's a test scene shot using all the DRO levels available. There's also an Auto Mode, but I suspect it wouldn't give predictable results when shooting manual video.

 

 

The result is quite clear on every mode. Personally I think 5 is too much and might be actually degrading the image more than it helps, but lower settings definitely look not just useable but very useful in achieving a flat and clean image. I'd say using the DRO in its modes 2 and 3 would definitely help achieving a better flat image. I'll probably leave it at 2 all the time and increase it to 3 in situations with more contrast.

 

5DtoRGB

I've used 5DtoRGB on Canon footage since the early beta versions, and I honestly don't understand how come it's not used by everyone. 5DtoRGB features one of the best YCbCr to RGB compression out there, and it's free!!!

(the Pro version with batch capabilities costs $50 though)

 

5DtoRGB does a great job improving aliasing and compression artifacts and transcoding to 10-bit Prores (can also transcode do DPX image sequences and DNxHD files), or at least it did with Canon DSLR footage, so I thought I'd try it with the RX100.

 

 

The top frame is from the original AVCHD file and the bottom frame is from the Prores transcoded file out of 5DtoRGB.

5DtoRGB automatically changed the Decoding Matrix setting to ITU-R BT.709, so I assume that's the one to use with the RX100 (Canon DSLRs like the 550D, 60D or 7D used the ITU-R BT.601 Decoding Matrix, the 5Dmk3 however used the ITU-R BT.709).

Looking at it like this there's not much of a difference, so I went looking in the channels.

 

 

The Red and Green channels looked quite clean in both versions, but looking closely at the Blue channel you can see how 5DtoRGB makes a pretty good job at smoothing out some of the compression blockiness, but mainly smoothing out the aliased lines you get on sharper edges.

Using 5DtoRGB won't do any miracles, but when shooting to 8 bit compressed codecs, every little bit helps, and using it along with a flat Creative Style will definitely help you getting cleaner and better images.

 

SHUTTER ANGLE / SHUTTER SPEED

The Sony DSC-RX100 has the annoying feature of only shooting 50fps (or 60fps on NTSC markets).

On one hand it's great to be able to shoot 50fps at 1080p, but on the other hand, shooting 25fps at the same bit rate would probably produce better results with less compression.

One of the advantages of this could be that you'd always have the extra frames in case you needed the slow motion effect, but unfortunately that's not quite the case, since the ideal shutter speed for 25fps real time playback is different than the ideal shutter speed for 25fps slow motion playback.

If you're planning on shooting for 25fps real time playback, then you should set your shutter to 1/50, but if you intend to shoot for slow motion playback at 25fps, then you should set your shutter speed to 1/100.

 

 

Using a shutter speed of 1/100 for real time 25fps playback will not give you enough motion blur, and the motion playback will not be as smooth as it should. Also, playing back footage shot at 1/50 shutter speed at 25fps slow motion will have too much motion blur, making its motion look rather fuzzy.

 

 

Here's a sample file you can download yourself.

This was shot at AVCHD, 1080 50fps with a shutter speed of 1/50, meant to be used on a 25fps timeline, playing at real time:

https://dl.dropbox.com/u/8569573/rx100review/RX100videoSample1.mov

 

 

That's it for now, I really hope it helps some people out. I have some videos I can share later on if you're interested, and I also might update this review with tests of the different Steady Shot modes once I get to them.

Keep in kind that these are only my findings and personal opinions, it would be great to hear from people with different opinions, or about settings you think would give better results.

 

Enjoy!

[richg101]

love your b+w test results.  WOW, the b+w is significantly cleaner in the shadows, much better gradation there.  is it due to lack of colour information, or due to the portrait profile being that flatter meaning it has more information to fillinto the 24mbs bandwidth - resulting in the compression hitting the darks harder than in b+w mode?  whatever, it looks a lot cleaner and makes me a lot more likely to use this profile more regularly.

 

cheers

[Bruno]

love your b+w test results.  WOW, the b+w is significantly cleaner in the shadows, much better gradation there.  is it due to lack of colour information, or due to the portrait profile being that flatter meaning it has more information to fillinto the 24mbs bandwidth - resulting in the compression hitting the darks harder than in b+w mode?

 

Thanks!

I'd say it's definitely due to the lack of color information, which allows for more detail to be stored using the same bit rate. The Portrait profile actually looks better than other color profiles at the same bit rate and is still the best option for color shooting in my opinion, but for black and white, definitely go with B&W (with all settings set to -3, for the flattest black and white image), the results are indeed amazing.

 

Also, keep in mind that all those tests are done without using 5DtoRGB transcoding, which could potentially make that B&W footage (and all the other footage) look and grade even better!

[see ya]

Thorough test but do you have a native source file from the camera to offer as well, so many tests online fail to offer it and therefore they're even more subjective for the reader based on the testers setup, workflow and tools. No one can coroborate or compare results.

Not familar with it but does it shoot other than AVCHD. How did you extract your images for your test frames?

[Bruno]

It also shoots MP4 files, but not at its highest bit rate, and not progressive, so if you wan the best quality it needs to be AVCHD.
For these tests I rewrapped the AVCHD files as Quicktime movies using ClipWrap, but for proper work I'll be using 5DtoRGB and transcoding to Prores.

[see ya]

Have you established whether the camera source is full luma at the high end like say a FS100 16-255. Does your clip wrap / 5dtorgb route ensure levels go 16-235 before you extract your image frames for evaluation? Otherwise all your sample frames will be skewed, more blown highlights and clipped color channels.

[Bruno]

ClipWrap doesn't do anything at all to the images, it just rewraps the .mts file into a .mov file, which I opened directly in Photoshop, that was my workflow for these tests. It's possibly that the color channels were clipped since Photoshop uses Quicktime, but for the sake of these tests it wouldn't change which Creative Style is the flattest or the cleanest to shoot with.

 

5DtoRGB was only used in the section where I wrote about it.
From what I understood, 5DtoRGB transcoding avoids or fixes the Quicktime issue with clipped color channels, I'll look more into that and post any relevant updates.

 

Meanwhile I added a small Quicktime link to the original post, it's been rewrapped from the original AVCHD .mts file, but not recompressed.

[see ya]

Thanks for the link to the rewrapped file but do you have a source file to ensure all metadata is there?

 

"but for the sake of these tests it wouldn't change which Creative Style is the flattest or the cleanest to shoot with."

 

You write a lot about 5DToRGB for a test on Creative Styles. :-) But it's clear to see in your AVCHD vs 5DToRGB Prores image that somethings not right with levels or color matrix in the conversion's you've done, the base of the toy is more orange in the ProRes and the prores has less contrast, so wrong color matrix and levels handling.

 

So apart from the comparison of Creative Styles like for like there's little to glean from this test, it's a given that if less saturation and contrast is dialled in then the image will be flatter.

 

Thing is by not handling the full luma levels correctly and converting with the wrong color matrix going from YCbCr to RGB then RGB channels get clipped, highlights blown, more contrast added artificially, all making artifacts more promentent that were probably already there in the source, the difference between what QT gives and 5DToRGB is subjective.

 

I see no point in using 5DToRGB for transcoding unless there's problems with editing speed. For quality a simple gentle denoise of the original source in a 32bit NLE workflow will do the job better.

[Chrad]

SHUTTER ANGLE / SHUTTER SPEED

The Sony DSC-RX100 has the annoying feature of only shooting 50fps (or 60fps on NTSC markets).

On one hand it's great to be able to shoot 50fps at 1080p, but on the other hand, shooting 25fps at the same bit rate would probably produce better results with less compression.

 

I believe it actually does shoot 25p when you choose the 50i option. Just drop it into a 25p timeline and you're good...no need to deinterlace.

[jgharding]

Try using the Dynamic Range Optimiser as well.

 

It just lifts the blacks up out of the most destructive area of the codec, it's pretty good.

[Bruno]

Thanks for the link to the rewrapped file but do you have a source file to ensure all metadata is there?

 

Sure, I'll post one later on today.

 

You write a lot about 5DToRGB for a test on Creative Styles.

 

I wrote about 5DtoRGB on the 5DtoRGB section, and wrote about Creative Styles on the Creative Styles section, different things, and in the Creative Styles section I did not use 5DtoRGB.

 

So apart from the comparison of Creative Styles like for like there's little to glean from this test, it's a given that if less saturation and contrast is dialled in then the image will be flatter

 

That was not the point of the test. What I tried to show is that there is an advantage to be had by shooting with a flat profile, since it makes it easier on the codec compression and gets you cleaner images, with less compression artifacts. Shooting with less saturation and contrast and then turning saturation and contrast up in post will give you better looking images than shooting with the saturation and contrast up to begin with.

Obviously I was not trying to prove that when you turn saturation down you get an image with less saturation. :)

 

 

Thing is by not handling the full luma levels correctly and converting with the wrong color matrix going from YCbCr to RGB then RGB channels get clipped, highlights blown, more contrast added artificially, all making artifacts more promentent that were probably already there in the source, the difference between what QT gives and 5DToRGB is subjective.

 

I see no point in using 5DToRGB for transcoding unless there's problems with editing speed. For quality a simple gentle denoise of the original source in a 32bit NLE workflow will do the job better.

 

As I said in the original post, these are just some of the first tests I ran, there's a lot more to test. I didn't look into any color clipping issues yes, but it's definitely on my list.

My 5DtoRGB test dealt only with how it smooths out nasty compression aliasing in the color channels, which it does, and is very helpful when grading. I'll definitely compare it with alternatives such as denoising in 32bit as you suggested though.

[Bruno]

I believe it actually does shoot 25p when you choose the 50i option. Just drop it into a 25p timeline and you're good...no need to deinterlace.

 

Really? That is quite misleading. I confess I have no interest in shooting interlaced footage, so I stayed away from 50i, but I'll definitely give it a shot!

[Bruno]

Try using the Dynamic Range Optimiser as well.

It just lifts the blacks up out of the most destructive area of the codec, it's pretty good.

 

Sounds good, I'll add it to the list!

[Zach]

Really? That is quite misleading. I confess I have no interest in shooting interlaced footage, so I stayed away from 50i, but I'll definitely give it a shot!

I believe its actually 30p that it shoots. I have yet to try that though

[richg101]

I was told the 50i option is just a wrapper and if dropped into a 25p timeline it acts exactly the same as 25p.  On the nex5n the 50i option is no different to 25p option when dropped into your 25p timeline.  i imagine it is the same with the rx100

[see ya]

I was told the 50i option is just a wrapper and if dropped into a 25p timeline it acts exactly the same as 25p.  On the nex5n the 50i option is no different to 25p option when dropped into your 25p timeline.  i imagine it is the same with the rx100

 

Yes, numerous consumer range cameras from various manufacturers seem to do this, use an interlaced flagged container but store progressive frames, Canon did it with the HV series HDV mpeg2 camcorders causing confusion. I think it was another one of those artificially seperating the codec between consumer and professional ranges or some avoidance of the cost for licensing maybe.

[Bruno]

Yes, numerous consumer range cameras from various manufacturers seem to do this, use an interlaced flagged container but store progressive frames...

 

It seemed to me that the footage is actually interlaced, shot as progressive but interlaced before encoding.

The good thing is that de-interlacing it works perfectly well, leaving no weird artifacts as real interlaced footage would.

 

I've just updated the original post with a new "Record Settings" segment right at the beginning, I also uploaded 3 source camera files, please let us know what your thoughts are regarding color clipping and interlacing!

[zaz]

Damn Bruno, this is very awesome! Personally, I noticed a bit of banding in that highlight in the upper right hand corner (in the new record settings section). seems to decrease as you move from 17 to 28 mbps.

seriously man - this is killer. raising the bar, bru

[see ya]

Thanks for adding the MTS files. The clip wrapped sample you originally posted isn't interlaced. Again I think its a work around to avoid paying bigger license fees by delivering it as flagged interlaced where in reality it's two identical 25P streams, picking either one doesn't matter, so just need to tell the NLE to interpret as 25P then NLE picks alternate streams, one frame from one then next frame from other and so on so you still end up with 25p

If you're seeing interlacing lines in the clip wrap source recording profile I think your NLE or QT is trying to hard.

Haven't had chance to look at the MTS files yet.

re luma levels, its clear the source is full range in the top end, but looks like some sort of highlight roll off before hard clip, unless you were able to use zebras or highlight warning to judge exposure.

What creative style was used for the clip wrap file?

Could you provide a low light sample and in a more contrasty profile to see what happens with luma at low end?

[lmackreath]

I have had a play with numerous profiles on the RX100 since my purchase last year. To be honest I have not seen much difference using a flat profile compared to a standard profile when it comes to grading. The camera is great but even at 1.8 is not great in low light with the small sensor and grading then becomes near impossible.

 

Here is a couple of examples shot with the standard picture profile with everything set at 0 except for contrast which is at -1

 

https://vimeo.com/58391656

 

https://vimeo.com/58208672

 

https://vimeo.com/54039461