Dear X5R owners,

I would like to find out can X5R be used for regular compressed H264 recordings in same way and with the same quality as X5 when RAW is not of importance. Until recently I hought that X5R without SSD performs exactly as X5, so I hoped to use X5R on some occasions just with MicroSD card to speed up my workflow when RAW is not of importance... After watching this video from Cinema5D I got confused - as they show that these two cameras, when both set at default D-Log ISO100, look significantly different - X5R PROXY (saved on MicroSD card) recording seems to be of totally flat and almost "plaster" look without detail, while X5 recording looks a bit oversharpened. Please look at their video at YT

So, I kindly ask you to share few seconds of compressed H264 rexording from MicroSD card, preferably without even inserting an SSD. It would be helpful if that recording is made in manual mode without any profiles. Did anyone of forum members get opportunity to personaly compare X5R SD recording to X5 recording? Is there really such difference as in this Cinema5D test?


Thank you for your support.

Looking flat out of the card is actually a good thing. It is to preserve dynamic range when transferring to Rec709, the broadcast color space standard for broadcasting, which is smaller than the camera color space.
Think of dynamic range (DR) as an accordion where each crease represents one stop of DR.
The accordion coming out of the camera is much more extended than what you could fit in the Rec709 color space box.
You have two choices. You can chop off both ends, which is what the X3 does or you can squeeze the accordion to fit it in without removing creases (dynamic range stops.)
Because all your DR is preserved, you can, in post bring back the flat footage to its initial glory.
LOG at its core is a curve applied to its squeezed accordion so you don't have to do it yourself. It is a bit more complicated than that but in essence that's it.
The X3 does not have LOG and bakes the picture in camera. This is why it is more critical to expose an X3 like camera correctly because you don't have much leeway to color correct your footage.
Beyond this, I don't know how they exposed the footage in the video.To me it does not mean much. I would watch as many films as possible on youtube or vimeo and such with the different camera and see if you like what you see. It is very possible to get great looking footage with all these cameras . You have to learn how to use them.
Philip Bloom an award winning DP uses the X5, not the X5R, and his footage is incredible.
I hope that helps

@Sky Lens, I think what the original poster is asking is this: the H264 recording of the X5 and the proxy recording on the SD card of the X5R should be identical. Looking at the image he posted, this is clearly not the case! This needs to be clarified by someone, ideally by another third party reviewer and/or DJI themselves. I need to know the answer to this question too. I cancelled an order for the X5R since I found out about this open question: why does the H264 proxy recording of the X5R look different/of a much lower quality then the H264 recording of the X5?

tholtmark@me.co Posted at 2016-7-17 17:01
@Sky Lens, I think what the original poster is asking is this: the H264 recording of the X5 and the  ...

If I did, my apologies.
I am not sure what they call proxy for the X5R. You have two choices, H264 encoded and MPEG4. I am talking about what is recorded on the MicroSD card not the SSD. The H264 recording is at a much higher bitrate on the X5R than on the X5 and while I never shot with the X5, I own the X5R and I can tell you the H264 footage is very good.
I am very skeptical of tests like that. I don't know why the picture look like that when blown at 400%. Did they use the MPEG4 setting in error? Or is that what they mean by proxies? Essentially the same sensor records to the SSD and the SD. Again I have the X5R and I record in both raw and H264. I am happy with both formats.
I will do a test

Hmm I wonder what firmware the x5r is running? the x5r proxy should be the same as the x5.

Sky Lens Posted at 2016-7-17 21:55
If I did, my apologies.
I am not sure what they call proxy for the X5R. You have two choices, H264 ...

Hello Sky Lens

I am trying to learn evry day and you posted some new information that I would like to get confirmed :

First
"The H264 recording is at a much higher bitrate on the X5R than on the X5"
Can you tell what bitrate the X5R records H264 at ? ( I have always thought it was the same on the two cameras.)

Second
"The X3 does not have LOG"
I have always been thinking that I was recording in D-Log with my X3. Is LOG something different maybe - something I have missed ?

Thank you in advance

nilsblix@gmail. Posted at 2016-7-24 17:33
Hello Sky Lens

I am trying to learn evry day and you posted some new information that I would like ...

Some comments.
First, the X5 and X5R .mov files are at an identical max bit rate and having shot using both, I can tell you they are not different. The example shown is from a respected review site (Cinema5D; reviewer Sebastia Wöber) but the writer appears to have a strong bias against the X5 but thinks very favourably about the X5R.

The X5 is capable of excellent footage but you need to be aware of how to avoid some situations. Sudden camera moves like quick pans and tilts show up the bit rate limitations through the grouping of frames in compression. This is discussed elsewhere and I will not elaborate here.

As with most cases, an expert user of an X5 can achieve superior results to an inexpert user of a X5R or even something superior.

Secondly, there is certainly a log option in the X3 which works the same as it does in the X5 but the imaging chip in the X5 is superior in many ways so its a different comparison. Under good lighting conditions, honestly, you can get really good results from an X3 and many happy customers can testify to that. Interchangeable lenses, control of aperture and shutter speed are pluses on the X5 platform. With the X5R, you can really toy around with the grading and there is less jerkiness with fast manoeuvres but you have to have both time, expertise and lots of additional software and better hardware to exploit its abilities.

Mike-the-cat Posted at 2016-7-25 03:56
Some comments.
First, the X5 and X5R .mov files are at an identical max bit rate and having shot u ...

You are totally right Mike-the-cat. I am a X5 user and I simply love that camera.

Sky Lens Posted at 2016-7-16 18:21
Looking flat out of the card is actually a good thing. It is to preserve dynamic range when transfer ...

In theory, yes. But in practice there are other factors at work. With the Inspire 1 the main obstacles to making any form of Log gamma work are the extremely high compression, the S/N ratio of the chip, and the 8-bit recording.

Working backwards, in order to make grading truly versatile, at least 10-bit recording is required. Yes, you are cramming more DR into the recording by using Log gamma, but with an 8-bit recording you are doing this at the expense of tonal range. Which can cause banding issues in areas of subtle graduation, such as blue skies.

With the Inspire 1 this is further compounded by the extremely high compression, which also introduces banding and macro blocking of its own. The comparatively low S/N ratio of the chip further compounds things, as does it's limited usable DR right from the beginning.

It is far better with the Inspire 1 to obtain as much tonal range as you can, rather than DR. This means more care with shooting scenarios and your lighting angles.

SimonMW Posted at 2016-7-26 18:22
In theory, yes. But in practice there are other factors at work. With the Inspire 1 the main obsta ...

Nice, technically rich reply. Thanks!

Another point about the Cinema 5D posting is that the X5 video was shot in Log. As Simon said, this really isn't great with 8 bit color recording and works under some circumstances. For most users, using None, None gives visually more pleasing results if you don't want to take the trouble to grade every shot.

SimonMW Posted at 2016-7-26 03:22
In theory, yes. But in practice there are other factors at work. With the Inspire 1 the main obsta ...

Banding is due to bit depth (which is tonal range).
LOG can be extremely effective at 8 bit. The Canon C-300 shoots LOG on a 8 bit depth.
You are not cramming more DR into the recording. Dynamic Range is a function of the camera. It is set. LOG preserves the recorded dynamic range to fit into a Rec709 color space. Just like an accordion.

SimonMW Posted at 2016-7-26 03:22
In theory, yes. But in practice there are other factors at work. With the Inspire 1 the main obsta ...

It is far better with the Inspire 1 to obtain as much tonal range as you can, rather than DR

How do you obtain more tonal range?

Sky Lens Posted at 2016-7-27 05:19
It is far better with the Inspire 1 to obtain as much tonal range as you can, rather than DR

How  ...

Shoot with LUT and color/contrast/sharpness settings that bake footage to look as close to final grade as possible so in color grading you'll only make minor adjustments. That way there won't be a need for drastic changes to your footage in post (which is where H264 compression imperfections are amplified - these artifacts are specific to this camera due to such low bandwidth limitation).
Cinelike LUT is best of all at the moment. It packs tons of shadow details at the same time protects highlights better than any other profile. Add a bit extra saturation and set sharpness around -1 (this seems to be most natural looking setting with little to no artifacts) and your footage at this stage requires minimal adjustments while also looking great and not lacking DR.

Sky Lens Posted at 2016-7-26 22:19
It is far better with the Inspire 1 to obtain as much tonal range as you can, rather than DR

How  ...

ge0se already answered for me. You get maximum tonal range by using the default gamma, not the log gamma.

Look at it like this. With an 8-bit colour system like you have on the Inspire 1 recordings, you have 256 graduations of each colour. By using a log gamma you are trying to compress the full DR range of the sensor into a small bit-space by re-allocating and prioritising certain areas of the brightness curve. One particular area that this causes problems in is highlights and shadows, which are at a lot of risk of banding issues in 8-bit colourspace due to tonal compression, since the log gamma curve allocates less of the brightness curve to these areas. Such subtle graduation doesn't play well with the H264 compression either.

With Log gamma if your camera chip is capable of, say, 12-bit colour pipeline with a DR of, say 12 stops, and your recording system is 8-bit colour H264 with a recording signal DR of 5 stops, then you are trying to not only quantise 4096 graduations of colour down into 256, but you are also trying to compress via a log curve 12 stops of DR into a picture that can be displayed within 5-stops of brightness. Hence the reason why log gamma footage looks flat.

When you grade that log footage, you are effectively stretching much of that tonal range back out again so that the picture looks normal on most displays again. The banding may not be as noticeable on the ungraded Log gamma footage, but will become apparent when it is graded.

To effectively use log gamma you really need 10-bit recording ability so that you have 1024 graduations to play with compared to 256. Cameras like the A7S are 8-bit, but they are also very low noise, which gives them a distinct advantage, although even that camera is still at risk of tonal compression issues.

Some of this can be countered by properly exposing for log gamma, and taking account of the tonal compression in the highlights. This means using an exposure that "looks" wrong on the monitor and histogram, and will appear dark. Not many people have the confidence to do that or understand how the camera is remapping the different brightness areas to the recorded file. Although it still remains that the compression system on the Inspire and the S/N ratio makes using Log gamma troublesome, and it is much better to get the location recording as close as possible to the desired final result.

Sky Lens Posted at 2016-7-26 22:19
It is far better with the Inspire 1 to obtain as much tonal range as you can, rather than DR

How  ...

ge0se already answered for me. You get maximum tonal range by using the default gamma, not the log gamma.

Look at it like this. With an 8-bit colour system like you have on the Inspire 1 recordings, you have 256 graduations of each colour. By using a log gamma you are trying to compress the full DR range of the sensor into a small bit-space by re-allocating and prioritising certain areas of the brightness curve. One particular area that this causes problems in is highlights and shadows, which are at a lot of risk of banding issues in 8-bit colourspace due to tonal compression, since the log gamma curve allocates less of the brightness curve to these areas. Such subtle graduation doesn't play well with the H264 compression either.

With Log gamma if your camera chip is capable of, say, 12-bit colour pipeline with a DR of, say 12 stops, and your recording system is 8-bit colour H264 with a recording signal DR of 5 stops, then you are trying to not only quantise 4096 graduations of colour down into 256, but you are also trying to compress via a log curve 12 stops of DR into a picture that can be displayed within 5-stops of brightness. Hence the reason why log gamma footage looks flat.

When you grade that log footage, you are effectively stretching much of that tonal range back out again so that the picture looks normal on most displays again. The banding may not be as noticeable on the ungraded Log gamma footage, but will become apparent when it is graded.

To effectively use log gamma you really need 10-bit recording ability so that you have 1024 graduations to play with compared to 256. Cameras like the A7S are 8-bit, but they are also very low noise, which gives them a distinct advantage, although even that camera is still at risk of tonal compression issues.

Some of this can be countered by properly exposing for log gamma, and taking account of the tonal compression in the highlights. This means using an exposure that "looks" wrong on the monitor and histogram, and will appear dark. Not many people have the confidence to do that or understand how the camera is remapping the different brightness areas to the recorded file. Although it still remains that the compression system on the Inspire and the S/N ratio makes using Log gamma troublesome, and it is much better to get the location recording as close as possible to the desired final result.