Sony FDR-AX53 Camcorder

[VidThreeNorth]

[2020-10-14 21:45 Changed the name of this topic from "Guess the Mystery Camera", since I have identified it.]

A Fun Mystery: Can you Guess the Camera, Lens and recording settings?

I was recently testing a camera's video capabilities and you might find this fun and educational. These are samples from test files taken on two occasions. All files are maximum detail JPEGs ("C1" compression). The first test clip was "Full HD" and the second test clip was "Ultra HD". There was so much detail that I had to resize some images below "Full HD" in order to make them small enough to upload.

If you want to post your guesses, post them as comments as usual. I will post the actual details after a week (assuming I remember) .

"01-Cap-C00018-18h05m06s107-C1.jpg"
The first file was a "FullHD". The "01" image is the direct capture.

"02-Pro-C00018-18h05m06s107b-Rsz1600-C1.jpg"
The "02" file is my standard process. Unfortunately it had too much detail to be uploaded so I resized it down to 1600.


"03-Cap-C00005-15h17m11s469-Rsz1904-C1.jpg"
The "03" image was recorded in a 4K file, so the capture could not be uploaded in its full size. "03"was resized down to 1904 wide before compression. I could have resized to to 1980, but I decide to make it the same size as the "05" file.

"04-Crop-C00005-15h17m11s469-Detail01-C1.jpg"
The "04" file is a detail crop taken from "03".

"05-Pro-C00005-15h17m11s469b-Rsz1904-C1.jpg"
Like the "02" file above, when I applied my standard processing the "03" file, the resulting file became too large to upload as a 1920 wide image, and so I had to resize it down to 1904 wide so it would upload. As I explained above, this is why I resized the original capture file down to 1904 wide -- to make it the same size as this one.

[2020-09-26 20:49]
I actually have a wider selection of clips, but I did not want to upload a lot of samples in this game, but I decided to add a sample from a more "standard" clip.

"06-Cap-20h12m09s880-rsz1900-C1.jpg"
Again, the captured frame had too much detail when reduced to 1920 wide, so I further reduced it to 1900 wide. No adjustments have been made -- not even straightening the tilt.

"07-Crop-20h12m09s880-rsz1920-C1.jpg"
This detail crop is a "worst case" scenario because of the quick flowing river. It is one of the hardest things to capture in a video frame. People who do a lot of video work will know that the river actually looks fine when viewed at speed. In fact, the whole image looks really good when viewed at speed -- except for the tilt.

[2021-11-28 2:14]
More coverage of the FDR-AX53 at:
"Sony FDR-AX53 (SV-E10?) and Shure VP83F with Dead-Cat v. Wind"
"Sony FDR-AX53 (SV-E10?) and Shure VP83F with Dead-Cat v. Wind"

 

[VidThreeNorth]

Sony FDR-AX53

Aside from having a bit of fun, I started this topic this way for a reason, and it was not so much about this camcorder itself as the question of what bitrates do and do not do in the context of modern video codecs.

When I ordered the Sony FDR-AX53 camcorder, I was not sure if I would keep it. This is rare for me. Usually, I know with fair certainty if I want something or not before I buy. I have rarely used gone for a refund, even when I had a right to one. The AX53 was not something I thought I would want right now. I did have an intention of buying a new camcorder, but most likely early next year, and more likely not the AX53. So this an unexpected purchase in a few ways.

Why This Now?

For this year, I planned to buy a Panasonic Lumix G95 (yes, the G95) and later in the year, a Sony a6400, and some accessories, including maybe two or three lenses. Instead, I bought the Lumix G85 at a cost saving, and a number of accessories, including a couple of cheap used lenses (not the lenses I had planned), and the Sony a6400 became a problem due to short supply.

And then a deal came up for this camcorder, and it really solved some problems. But I didn't originally want this camcorder. The biggest problem with the FDR-AX53 camcorder is that Sony intended this product to be a "family recorder". That is to say it has some fancy first cabin hardware features but the firmware unnecessarily limits it to amateur functionality. It is meant to be used like an old-school "Kodak Brownie"-like product. It is meant for little more than "point - and - shoot". Some people would simply class this as a "pro-sumer" product. That is debatable.

It might be intended for someone to use to record simple 4K family videos, with almost everything automated, to watch on their home 4K Sony TV set. The add-on features are simply video "novelties", like short bursts of slow-motion, or toy-like finishes in their "Picture Effects" but no ability to adjust "contrast" or "saturation" in a craftsman-like way. This was something I expected, first because I had bought and used a couple of Sony consumer Full-HD camcorders before, and I researched its manual. User adjustment of exposure compensation limited to +/- 1 stop? Rediculous!

But I knew a few work-arounds that could give me some pretty good results, and then there was the Sony "BOSS" stabilization system. Dated though it is, this was a huge draw. In effect, the lens and sensor are contained in a "sub-system" which is mounted in a 2 axis set of gimbals. And the sensor itself is a 1/2.5" back-illumanted device similar to a GoPro (excellent capabilities), and some other things. . . .

There were alternatives, such as the newer Canon HF-G50 which I would have preferred, but the price decided it. I did not get a wonderful price, but it was good enough to make it worthwhile.


Why The Mystery?

As for why I introduced this as a "mystery" camera, it was to show something about this camcorder. Specifically, I have previously written that most of the industry had dis-served the photographic community by ignoring the "50 mbps" video bitrate in H.264 video and jumping past it in the pursuit of UHD video. Sony did not do what I consider the ideal case (supporting a 1440P format), but at least they addressed it by providing "Full HD" 50 mbps and "UHD" 60 mbps formats. Yes, the FDR-AX53 captures video in both of these relatively unique formats, as well as supporting a few other formats such as 100 mbps UHD (24, 25 and 30 fps).

The question is whether these are good usable formats and thus, viable alternatives to high bitrate UHD video. There is no doubt that if one can afford to use the higher bitrates, then those might as well be used, but is there really a use-case for the 50-60 mbps formats? That was the point.

The Frame Captures:

Starting from the last pair, pictures "06" and "07", these are a pretty good example of the quality available at UHD in 100 mbps, as Sony wants you to see it. The settings are mainly the standard "automatic" settings. The only thing non-standard is that I have set the exposure EV = -1, because I do not like Sony's standard exposure in general. I find that Sony's default setting is, in my opinion over exposed. But other than that, the colours are the default settings. The degree of artifacting in the river is actually less than I expected. In that sense the image is very high quality.

The lens is 4.4mm - 88mm (35mm equivalent 26.8mm - 536mm) F2.0 - F3.8 and this was at around 450 - 500mm. I had zoomed out to the limit, but I think I had pulled it back a bit before this clip.

The image stabilizing was "standard", so only the two axis gimbals were in use. This shows the unfortunate limit that there is no "z-axis" ("roll") stabilizing in this mode. It is unclear at this point if there is any z-axis stabilizing in the advanced modes, which add digital stabilizing to the mechanical gimbal based stabilizing. The stabilizing did a good job in this case and the clip is very smooth.
[2020-1008 21:53 clarified "z-axis"]

I should add right now that although I have actually made a number of test clips with the camcorder, few were made using these "default" settings, so that when I decided to upload a "standard sample" I did not have much to chose from. All the other "standard" clips were assigned to specific uses, and I cannot upload them at this time (some are assigned to projects which might not be made public even later), and the rest are tests of various non-standard settings. That is why I ended up with a "tilted" image. It was not a deliberate choice, to demonstrate the limits of the stabilization, though I am actually glad that it occurred.

Anyway, this is a nice example of the AX53 image capability. The lens is generally sharp and without significantly bad flaw, and there is enough (and in my opinion, a bit too much) contrast and arguably good saturation. It's very nice.

Now let's get to the images, which are the ones I wanted everyone to look at carefully:

The first frame capture (in files "01" and "02") was recorded in "Full HD" at 50 mbps. I used the "Soft Hi-Key" Picture Effect. This sets a reduced contrast, less sharpening, and less saturation, but at a high exposure. But by using "zebras" set to 100% I can use "manual exposure" and the display, to keep the exposure in a good range. In this case, I allowed some of the sky (visible through the upper branches of the trees) to clip, but I kept the bright plants near the middle of the frame well below clipping.

The sensor and these settings did a perfect job. If you look up the path and into the shadows, you will see detail in the wood fence and even into very dark shadows. Yes, there is noise in those levels, but getting as much detail as I did was very satisfying for me. It is very close to all the detail and dynamic range possible for the sensor. Understand what this implies. Even a proper "professional" level flat image profile including wider controls for contrast and exposure, probably could not do much better. It would take a better sensor.

Look at the second capture (files "03", "04" and "05"). This was a UHD file recorded at 60 mbps using the same "Soft Hi-Key" and manual exposure using "zebras". I kept the exposure down to preserve detail in the shadows. The wind blowing causing the flowers to wave around fairly violently might have reduced the sharpness a bit, but I do not thing it has much to do with the softness of the white petals. I think that is more a result of the limits of the lens, combined with the effect of the colour sub-sampling YUV420 artifacting. Look around at the detail file and you will see that there really is not much artifacting in this image at all compared to the 100 mbps file image (the ducks above). Noise in the shadows is very smooth, as is the out focus areas past the white flowers. Looking at the processed version, I could have added a touch more saturation, and some people might prefer a bit more contrast or brighter highlights, but this is a look that I like for it.

I am not say that there is "no" difference between "60 mbps" and "100 mbps" at UHD. There certainly is, and I have made a number of clips that show differences. But there is an attitude in the industry, mainly in the heads of people who have no real experience using the lower bitrates, as low as 60 mbps at UHD, that they are horribly inferior, whereas, in a lot of cases, unless you recorded them to deliberately show differences and look at them carefully side by side, you probably could not tell. Jumps from 60 to 80 mbps, or 80 to 100, or 100 to 120 are all generally negligible. My Yi-M1 with its 80 mbps files, in many cases, is probably about as good as the much more expensive Canon M6 MkII with its 120 mbps files.

Using the AX53:

So what are the use cases I am planning for the AX53? The main use is for the zoom. Trying to replicate the variable speed power zoom range from around 27mm to around 536mm equivalent in par - focus on my M43 camera or anything larger is expensive and cumbersome. Just buying the lenses needed to cover this range would cost me more than this camcorder.

Unfortunately, the main limit of this camera is that it is not weather sealed. As-is, it will be very strictly limited to indoors and sunny days. I can get a weatherproof outer cover which will serve well enough if I need to use it in some weather conditions. This has been done for many years now by many photographers. But I will probably never use it that way. That is why I have my G85.

Lately, many still photographers seem to be interested in trying to "double over" to video by buying hybrid cameras or simply using expensive re-purposed DSLRs. As a start point, that is not a bad idea. And it works. But eventually, some want to get into zoom, and I have suggested buying or renting a camcorder. I don't know if anyone took that advice. But this is why. There seem to be some really nice new camcorders available at what I consider an "affordable" price range (above and below $1,000 US), and for the most part, camcorders like these are being ignored. I am particularly attempting to show that 50 - 60 mbps H.264 can be a useful file format, but also, more broadly, I am showing that if you want to do better video, then adding a camcorder like this can be a good expansion of your system.

 

[VidThreeNorth]

More 60mbps v. 100 mbps

Continuing the exploration of the AX53 and H.264 bitrates, here are two more file captures. The captures are of similar scenes, but taken on different days with different lighting. The first capture was from a 60 mbps file and the second was from a 100 mbps files. All the uploads are highest JPEG detail compressed. Take a good look at them and ask yourself if you could tell which frame was from the 60 mbps file and which was from the 100 mbps file. I think that it would be impossible to say from these two frame captures. In fact, the original file sizes of the captured frames were very close:

Original Capture Sizes:

"01-60mbs-013-ScnAuto-20h32m43s402.png" was 14,516,106 bytes

"02-100mb-04-Standard-20h51m55s002.png" was 14,853,557 bytes

Is there artifacting? Yes, there is, but quantifying the amount of artifacting or the detail reduction in these frames would be difficult. If you cannot see the difference, what is the point?

Depending on what it recorded, it could be necessary to record nearly identical scenes and study a large number of frames carefully to definitively identify a 60 mbps file from a 100 mbps file.

 

[VidThreeNorth]

"Cinematone"

In the Comment #3 above, frame capture 2 from the 100 mbps file was actually from a pair of files made to compare the "Standard" image style with a new style called "Cinematone". This setting is new and overrides "Scene Selection" forcing it to "Auto" and "Picture Effect" forcing it to "Off". It is intended to emulate the look of a "film camera". In that sense, it is like Fuji's "film simulations". I was going to wait till I had a chance to learn more about Cinematone before posting any comparison files that I have, but I might not be getting that research done for a while, so I am uploading this much now.
[2020-10-17 00:47 corrected reference to Comment #3 rather than Comment #2]

Again, this is recorded at 100 mbps. I think that all three of the original video files were exposed at EV = -1. As I wrote above, I think that Sony's default exposure is over-exposed by that much. If I find out that I am wrong about the exposure compensation I used, I will post a correction later.
[2020-10-17 00:48 Clarified.]

The original frame capture was a bit smaller than the "Standard" profile frame capture. The difference is not enough to say conclusively that this is a general characteristic of Cinematone.

Likewise, the sample files that I am uploading are a bit smaller than the "standard" image equivalent files.

Original Captured Size:
"03-100mb-05-Cinematn-21h15m55s701.png" 14,636,121 bytes

From the files I have uploaded there is not much that can be told about Cinematone. It appears that Cinematone is about a stop darker. From other files I have, I think Cinematone is a bit lower saturation. Contrast and sharpening seem similar to "Standard".

I intend to make a small study to quantify the image differences, but it might not be until next year. I am a bit busy right now, and really, I have seen enough to know what Cinematone is good to be used for.

 

[VidThreeNorth]

Sony AX53 Gammas, Standard and Cinematone

I found that I needed to get some data earlier than I thought. I recorded the clips on October 11, 2020, but at the time, I was not in a hurry to get the analyzing done. In fact, I was not entirely happy with this set of clips and I thought I might re-record them. The flaws were not important to the tests. The Canon R70 "control" clip was slightly out of focus, and if anything, that might actually make the test better. On the other hand, because the Sony AX53 was pointed downward a bit more, I did not zoom in as tightly as I wanted. Later I realized that I could have aimed both cameras a bit higher and the result might have been better. But again, the higher amount of background in the Sony clips does not invalidate the results. So although I am annoyed when I look at the sample frames because they were not perfect, doing them over is not necessary.

This is a limited test of gammas for two of the three profiles that I will be using on the AX53.

Control:

"CanR70-010-Std-EV-Mns0-19h57m47s970-C1.jpg"

The "control" for this set was a file from my Canon Vixia R70. I have not been using this lately, but I used the R40 (the same image characteristics) for a few years while I was recording 3D. So it is a look that I am familiar with, and one that should be generally well known. Actually, it is not as well known as people might think, because Canon's cameras do not all look the same. Moreover, most photographers and cinematographers change the standard appearance of their output to their own taste. So just because a Canon camera was used, does not mean the output was Standard Canon.

You might notice that I missed the focus. That happened because I did not notice it. When doing these gamma and colour tests, I am not paying close attention to focus because in general it will not matter. Actually, a "bit" out of focus might even help the test (yeah, sure.... .

"black" Luma: 37.2
"black + 1" Luma: 83.1
"black + 2" Luma: 144.9
"white - 2" Luma: 207.7
"white - 1" Luma: 234.0
"white" Luma: 255.0 (Clipped)
Contrast: = 150.9


Sony AX53:
Standard:

The "Standard" profile seems to be similar to what I have seen on the Sony CX405B camcorder. I find that it is good for "theatrical" scenarios. I will use it to record figure skating shows, stage presentations, and general dramatic presentations.

It does generally need decreased exposure from Sony's standard. There are a couple of tricks that can be used with auto exposure. The first it "Spotlight" which I had on the CX405B, but I would add the EV= -1. That combination might be enough. If not, then full manual with zebras should work. I would set the zebras at 100% and try to keep "in-bounds".

"S-AX53-C0002-Std-EV-Mns0-20h33m51s125-C1.jpg"
EV = -0.0

"black" Luma: 37.4
"black + 1" Luma: 83.9
"black + 2" Luma: 141.5
"white - 2" Luma: 196.9
"white - 1" Luma: 240.2
"white" Luma: 255.0 (Clipped)

Contrast: ("white - 1": 240.2) - ("black + 1": 83.9)
= 156.3


"S-AX53-C0003-Std-EV-Mns1-20h55m12s378-C1.jpg"
EV = -1.0

"black" Luma: 17.3 (Clipped?)
"black + 1" Luma: 52.9
"black + 2" Luma: 93.4
"white - 2" Luma: 138.1
"white - 1" Luma: 181.4
"white" Luma: 228.2

Contrast: ("White - 1": 181.4) - ("Black + 1": 52.9)
= 128.5


Special note about Clipping the "Black"

I am still coming to terms with the problem of "what is Black?" and when is it "Clipped?" The problem is that in the early days, I was taught that 20 was "black" because that is how it would be displayed. Values below 20 were useful because they could be recovered in post when necessary. Apparently, later it bacame "16", and then fairly recently, "0" started to be used. So for now, I am not sure how to look at a value like 19. Is that "Black" or just very dark grey? So far in this particular test I ran into the problem twice. I am including the actual Red, Green, Blue values for the pixels I tested, and added some comments. When you see me comment that a Luma around 16 or 20 was "Clipped", this is what I saw.

"S-AX53-C0003-Std-EV-Mns1-20h55m12s378-C1.jpg"
EV = -1.0

"black" Luma: 17.3 (Clipped?)
Red, Green, Blue:
16, 16, 16 (Black?)
18, 18, 20
16, 16, 16 (Black?)
16, 18, 20 (Black?)
17, 17, 19
Totals: 83, 85, 91
Averages: 16.6, 17.0, 18.2
Luma: 17.3 (Clipped?)

* NOTE: This is the worst situation for me. If 16 is black, then this is not really clipped because it is an accurately read 16. If black is 20, then really, all the pixels are clipped since there are only a couple of pixels with any component as high as 20, and those pixels will actually have lumas below 20 in aggregate. If black is 0, then this is clearly not clipped because it is just "dark grey". Keep in mind that I am only checking the values of 5 pixels, and even counting probabilities of noise, there are likely to be pixels with components below 16.

"S-AX53-C0005-Cnmtn-EV-Mns1-22h06m28s505-C1.jpg"
- EV = -1.0

"black"
Red, Green, Blue:
19, 22, 20 (Clipped?)
19, 23, 22 (Clipped?)
18, 18, 20 (Clipped?)
19, 19, 22 (Clipped?)
19, 19, 22 (Clipped?)
Totals: 94, 101, 106
Averages: 18.8, 20.2, 21.2
Luma: 20.1 (Clipped?)

* Note: This is really not a problem since black as "20" is apparently simply "outdated" and nothing is near 16. But for now, I don't think it is bad for me to treat 20 as "black" when I am setting up for a recording. I am just being very "conservative".


"Cinematone":

I have mixed feeling about what I have found in Cinematone. So far, I think this will do well for "direct to audience" general presentations. That is to say, if I will be distributing something I record without any processing at all, I think this is what I would like people to see. The output looks about 1/2 stop darker than "Standard", which should allow a good "automatic" exposure, with the EV = -1-stop compensation used where necessary. The "Spotlight" setting could also help this. The biggest disappointment here is that contrast has been raised. This makes achieving proper exposure more difficult, and almost ensures loss of detail at one end or the other (most likely at the black end). But at least the auto exposure is closer to what I prefer, so I think that directly from the camera, it should be better.


"S-AX53-C0004-Cnmtn-EV-Mns0-21h42m57s304-C1.jpg"
- EV = -0.0

"black" Luma: 32.3
"black + 1" Luma: 72.7
"black + 2" Luma: 134.8
"white - 2" Luma: 197.7
"white - 1" Luma: 239.7
"white" Luma: 254.8 (Clipped)

Contrast: ("White - 1": 239.7) - ("Black + 1": 72.7)
= 167.0


"S-AX53-C0005-Cnmtn-EV-Mns1-22h06m28s505-C1.jpg"
- EV = -1.0

"black" Luma: 20.1 (Clipped?)
"black + 1" Luma: 43.5
"black + 2" Luma: 82.2
"white - 2" Luma: 130.7
"white - 1" Luma: 181.3
"white" Luma: 222.1

Contrast: ("White - 1": 181.3) - ("Black + 1": 43.5)
= 137.8

[2020-10-25 20:24]

NOTE: The graph I posted on 2020-10-22 had errors. This is a corrected graph. Sorry about that!

"AX53-Gammas01a-C68.jpg"

A graph of all the gammas discussed in this post. Note that the slop of the Canon R70 "control" is about the same as the "Cinematone" gamma, except that the Sony "Cinematone" extends further due to its higher dynamic range. Also, these are my standard test settings and not what I would actually want to use. The best exposure would probably be about EV = -0.5 using the "Standard" profile.

 

[VidThreeNorth]

Sony AX53 Gammas: Soft High-Key:

For this test, I started with an uncorrected default automatic exposure, and made a second clip with EV= -1.0, which is how I started all my tests. However, I then made a third clip using manual exposure, the way I will normally use this camcorder. I set Zebras to 100% and try to keep the whites within bounds by adjusting until the zebras disappear in all areas that I want in-bounds. In general, that means that I might need to allow the sun to clip if it is present. I might allow some clouds, or "sparkles" to clip, but that would be decided on a "per case" basis.

If I grade in post, which I expect will be the most common way I will use this camcorder, this is my preferred profile. Contrast is usably low for a bit of grading. I have not really analyzed the colour saturation. It seems to be less saturated than Canon. I do not anticipate it will be a problem. Specific colour adjustment beyond overall temperature shifting would, again, be a "per case" decision.


"S-AX53-C0006-SftHKy-EV-Mns0-14h22m56s252-C1.jpg"
EV = -0.0

"black" Luma: 72.9
"black + 1" Luma: 126.5
"black + 2" Luma: 176.3
"white - 2" Luma: 248.1
"white - 1" Luma: 255.0 (Clipped)
"white" Luma: 255.0 (Clipped)

Contrast: ("White - 1": 255.0) - ("Black + 1": 126.5)
= 128.5


"S-AX53-C0007-SftHKy-EV-Mns1-14h50m28s830-C1.jpg"
EV = -1.0

"black" Luma: 48.7
"black + 1" Luma: 93.9
"black + 2" Luma: 137.6
"white - 2" Luma: 177.7
"white - 1" Luma: 226.4
"white" Luma: 255.0 (Clipped)

Contrast: ("White - 1": 226.4) - ("Black + 1": 93.9)
= 132.5


"S-AX53-C0008-SftHKy-Man-15h13m53s798-C1.jpg"
EV = Manual

"black" Luma: 36.7
"black + 1" Luma: 72.3
"black + 2" Luma: 106.5
"white - 2" Luma: 142.3
"white - 1" Luma: 173.9
"white" Luma: 214.0

Contrast: ("White - 1": 173.9) - ("Black + 1": 72.3)
= 101.6