Sony HDR-CX405 and Their Earlier Budget Camcorders

[VidThreeNorth]

Looking at my earlier Topic about the Sony CX405 and its version of XAVC-S files, I think this is the topic I should have started in its place. I do not expect to post much. The only post I have in mind for this topic is this video. Ironically, I have already posted material and information in the earlier topic that should have gone here. If I find other things to add later, then this is where I will accumulate them.

I prepared this video with this topic in mind. I recorded the camera files on my Sony HDR-CX240B in AVCHD in 1080i format. According to the information I got (I believe from Sony) the camera renders frames at 29.95 fps and then stores them in two fields, each having half the screen data. The camcorder also supports 1080p. The main reason I do not use 1080p is that it uses more storage and I have never felt that the extra smoothness added significantly to the viewing experience. In the case of low light video, there is an added theoretical advantage that the longer time between frames can be used for longer exposures for even lower light, but if things are moving, and in a video, you expect things to be moving, that would result in blurring, so I do not count that as a substantial advantage.

As far as I know, if I use the AVCHD settings in the later CX-405, then there is effectively no difference between the output of that camera and this older CX240B.

All exposure is manual. Theatrical lighting is highly unpredictable, but in general, you can count on an experienced professional lighting team to keep it even enough overall.

The first part of this video (the "Yate Dance and Acrobatic company" -- the girls in the green dresses) was under exposed probably about 1 stop. The only "adjustment" I made in the whole video was to lighten this part. Looking at the results, I think I lightened it a bit too much compared to the aerial and skating parts, but on its own it is acceptable. Aside from the brightness though, on close examination, noise will probably show up more in this section, and should not be held against the camera.

The skating and aerial acrobatics are all a single file. After finishing the first song, the second song began with only the very short break that is in the video. The camcorder had done a very good job. It rarely lost focus and then, only briefly. The decreasing aperture when the zoom is farther out has not resulted in substantial noise. With proper exposure the colours generally turned out well.

Because YouTube does not allow higher bitrates than you see in this clip, effectively a camcorder or rig costing thousands of dollars more will probably not do a significantly better job.

As I noted above, I put this video together specifically to show the output of the Camcorder. This is not my normal output style. Especially for skating, and really in most of my videos, I use stabilization in post. Mostly I record with the camcorder on a monopod and with the camcorder's built in stabilization On. So in effect it is "triple stabilized". The last stabilization in post, however, does smooth out errors caused by the camcorder's stabilization. Sometimes when you pan the camcorder holds the image still until it reaches a limit and then it jumps. In post, most programs will find those jumps and smooth them out. Sometimes I get caught with the zoom wrong and I need make a quick change in the focal length. Editing programs will often smooth out those zoom errors as well.

However, stabilization in post always looses some quality. How much will depend largely on how much correction is done. But you can expect the result to be somewhat less sharp and less detailed. So this is one of the few times I have posted a video that can be used to show the capability of the camcorder (subject to YouTube recoding) fairly directly.

"20180831 CNE Highlights Yate Dance & Quatro Aerial & Ice V2"
published Sep 25, 2018 15:12

[2018-08-25 Replaced Video]
Sorry for changing the video so quickly but I decided to re-correct the exposure for the first part. The first part is a substantially different presentation many respects so they are never really going to match, but this version comes much closer. If you saw the first version, there is not that helpful to see the second.

 

[VidThreeNorth]

Sony CX405 Focusing

The Sony CX405 has a general auto-focus, face detection and manual focus.

Manual Focus:

So far, I have avoided manual focus because there is no focus aid. If you are in an auto-focus mode and then go straight to manual focus, then the last focus distance is kept when you enter manual focus. Then the focus screen is shown and the joystick is used to select between "nearer" and "farther". Each click of the joystick jumps to the next distance till it gets to a limit. It appears that the jumps depend on the current focal length because they are not always the same. Counting the clicks, it took 10 clicks from infinity down to 0.5m on one test. I think the number of clicks is probably the same at all focal lengths and the distances vary.

Focus distance is displayed on screen, but only for about 1 sec. I see no reason for the time limit. The focus distance should be kept on the screen.

Auto Focus:

Auto focus might be affected by the "Scene" but there is no information about that, so I will assume that it is not.

In general, it appears that the auto-focus favors the nearest object it finds (see below).

Face Detect:

General Comments:

I have tried face detect -- sometimes accidentally, and it works. I have never seen it select more than 1 face, so I will assume that this is a limit. If the faces is moving it is tracked, and if there are multiple faces, it will sometimes jump to another face. Sometimes if there are many faces, it might not lock onto one.


The Frames:

I have uploaded two frames from different clips, but at the same event -- The 2018 Toronto Polish Street Festival.

"01-Upld-Mod-17h34m45s073-C1.jpg"
click here: Sony XAVC-S in the HDR-CX405

Look back at "01-Upld-Mod-17h34m45s073-C1.jpg". The whole story of this frame starts a couple of clips earlier. I was walking down the street and came to a Polka Band on the west side. I started recording it, and had reasonably good focus on the band leader. But no "face detect" came up, so I did not know it was On. Then a spectator walked closer to the band stand and turned around. The "face detect" box came on and locked onto the spectator, putting the band leader out of focus. I stopped recording and turned off the "face detect" and started a new clip. After recording the band a while, I turned and saw the people dancing. So I started to record it, resulting in the clip that the frame was taken from. During the clip, another spectator walked in front of me, and held up a phone on a selfie stick to record himself, a lady with him (probably his wife), and then other stuff. The guy's head is the blue cap, and his companion is in front of him.

So where is it focused? It is actually very hard to tell. The lens was set fairly wide and the lighting good enough to set a small F-stop. If you look down the street in the sunny surfaces, depth of field is carried far down. But in the shadows, things go fuzzy. This is the effect of sharpening. About the only way to tell is to estimate objects of roughly the same distance and look for things in the shadows. Those objects will go "soft" more progressively.

I think it is focused on the guy's hat or the back of his companion's head. The hairs on the back of the guy's head are very detailed. If the focus was in the middle of the dancers, or further, I would expect his head hairs to be more noticeably out of focus. There is also the back of a blonde girl's head to their left which seems to confirm this with fairly good sharpness and detail. This is a "hard call" -- which is good, but I do think that the actual focus distance was on the "nearest" object (the hat), rather than the dancers in the middle of the screen.


"01a-2018-10-09-08h34m44s742-C1.jpg"
[Note: Faces have been altered.]

The second picture was part of the first clip I recorded at this event and I think is short to mid telephoto. I would guess around 150mm equivalent (35mm camera). I did not know that "face detect" was on, but it must have been since I do not recall turning it on later. Yet I did not notice a box around any heads. Ideally, I would have preferred focusing down the street, but after looking at various objects, I am lead to believe that it is actually focused on the "Road Closed" sign or the yellow traffic block fencing.

Between these two frames, along with past experience with the Sony CX240 predecessor, I think I can conclude that the auto-focus favours the nearest objects, and if "face detect" is on, then it will tend to pick the nearest face as well. Either way, Sony is relying on depth of field to produce a generally acceptable image. Overall, it works acceptably enough for many situations. Ironically, it is not really what I would prefer. I would prefer the auto-focus to give preference to the middle of the screen, and "face detect" as well to prefer faces near the middle of the screen. But without some kind of focus aid, I am still reluctant to try manual focus.

I can add that the Canon HF R40 and R70 autofocus seems to work similarly. On the Canons however, there is a fairly good manual focus aid. The screen is magnified to a degree that is settable by the user. I am not sure how far it magnifies. It might be 4X. I have not tried manual focus on the Canon R70 yet, but I do think it will turn out usable.
[2018-10-09 18:54
I should also mention that the Canon R40 and R70 have touch screen point focus and separate touch screen spot exposure selection, but no way to combine the two functions. I have tried the functions and they appear to work well, but I have not used them in a real recording yet -- which is why I forgot to mention it.]

Also, if I take my YiM1 as an alternate camera for either the Canon R70 or the Sony CX405, the YiM1 auto-focus favours farther objects instead of near objects, and it also has a good "touch screen" single point auto focus. So I can, to an extent grab the YiM1 and use it for situations where it is advantageous. I have not actually tried working this way, but it is a possibility.

 

[VidThreeNorth]

I have not really kept track of the lower line camcorders. I bought the Canon R40's with "enough" research, but the CX240 was bought with very minimal research. I bought the cheapest used camera that could modify to take a polarizer. Looking at the CX405 deficiency in focusing, I decided to look at the Sony product lineup more closely. I now know that the HDR-CX455, which is currently their "next up" product (the CX440 has been dropped in Canada) has a larger, and much higher resolution touch screen and has touch focus with object tracking. It costs about double the price of the CX405, (a bit more than my Canon R70, which can still be bought new, or the current Canon R80), but I would say is worth it. On the other hand, if Sony made one change to the CX405 in their manual focus interface (stop making the distance disappear 1 second after it appears) I would find it more generally adequate.

Anyway, it does look like I am going to continue using my Canon R70 too, despite its shortcomings, for its touch focus.

 

[VidThreeNorth]

When I planned my 2019 Toronto High Park Cherry Blossoms video, I originally estimated a final version 3 - 4 min. Full HD short subject. The recordings would be in two parts. I planned to start with my Yi-M1 recording in 4K, until the battery ran down (since I only have 1 battery for that camera), and then continue with the Sony CX405 in Full HD, XAVC-S.

The Yi-M1 had to do the "heavy lifting" since I expected that all the most important files would be in the first group. The "establish" and the "main story" were all in that group. The CX405 was mainly "elaboration" (aka B-roll) and a "closer". I did not really get a particular "closer", but it was not important.

In the back of my mind I was skeptical that this was going to work out because I felt that I would want to get close to matching colours. I do not feel that matching colours and looks is as important as some other photographers feel it is, but for this video I felt it should be very close, and the Sony CX405 looks very different to me. In the end, I felt that I wanted to keep the higher quality of the Yi-M1 UHD files intact, so I cut the length to 2 min., and did not use any of the CX405 files. That had the extra benefit of not having to match between the cameras.

One idea I had regarding matching, was to start with the Sony's "Soft High Key" profile. I have not done any previous testing of this profile yet but it sounds like it reduces contrast and saturation, which from a "post production" viewpoint sounds pretty good. Since this was the first time I used this setting, I only tried a few clips, so I could play with the output.

The first thing I found was that it definitely was "high key" and that for general outdoor use, one is generally going to get a lot of highlight clipping. Since that CX405 does not have the common exposure compensations ("EV = +/- N") I would need to use the full manual controls. I've been doing this quite a bit for "theatrical" videos using the normal profile but this was my first try using the "soft high key" profile.

I tried one clip using "4 clicks" exposure decrease. As far as I can tell, this was probably 1.3 - stops below the default auto exposure. Since the cherry blossoms are a very light pink to begin with, and the sun was not blocked, this did not turned out to be enough.

After adjusting the colours for the test frame, including an increase in green, it still look a bit cold to me. This might have been proper for the scene, but I don't think so. But for the corrections to work in this case, I think I needed to start from a slightly lower exposure. I think "5 - 6 clicks" below the default would probably have been better for this scene. But really, I need to get used to the look on the view screen. That is what I use when I used manual exposure with the normal profile.

I am pleased that the "shadow detail" is very well preserved in this setting. The bark on the trees on the left of the picture is very good. I could not use C1 compression because the resulting file was too big to upload, so I ended up using C2 instead. But the loss of detail at C2 is actually negligible, so I will not be uploading a "detail crop" version for it.


"02-2019-05-22-10h46m59s682-C1.jpg"
- "4 clicks" darker than profile's standard exposure
- Frame capture with no adjustments.

"02a-2019-05-22-10h46m59s682-SmartFix-C2.jpg"
The same frame after adjustments.