Rumours that Panasonic is slowing/postponing Micro Four Thirds gear

[RAH]

It effectively shoots a quick video and stacks the images, yes. There’s no reason why the quality would be worse than advertised (6k photo stack being 18mp). It’s not 20, but it’s not worse than if you fired off a burst with an 18mp sensor and honestly, I doubt most people would notice the gap between that and 20 at anything approaching normal picture sizes

OK, thanks for the info. I wasn't worried about the MP size (a focus-stacked image using an E-M5.3 - i.e. 20MP - is somewhat cropped too). I thought maybe the image quality from a derived video wouldn't be as good somehow. But I guess not. That's fine then.

 

[11GTCS]

OK, thanks for the info. I wasn't worried about the MP size (a focus-stacked image using an E-M5.3 - i.e. 20MP - is somewhat cropped too). I thought maybe the image quality from a derived video wouldn't be as good somehow. But I guess not. That's fine then.

To my understanding the only reason a still frame from a video file would be worse than the resolution it is capable of (18mp for 6k, 8mp for 4K) is if it is a situation where electronic shutter would be inappropriate for one reason or another. There’s also the possibility of reduced bit depth, but if it’s a modern camera, even if the shot is only 10 or 12 bit (vs 16 for a still frame), I’d again doubt most people would get a result they were unhappy with. I don’t have a 6k photo mode camera, and when I did macro I just bracketed in camera. The stacking I did in post.

 

[RAH]

To my understanding the only reason a still frame from a video file would be worse than the resolution it is capable of (18mp for 6k, 8mp for 4K) is if it is a situation where electronic shutter would be inappropriate for one reason or another. There’s also the possibility of reduced bit depth, but if it’s a modern camera, even if the shot is only 10 or 12 bit (vs 16 for a still frame), I’d again doubt most people would get a result they were unhappy with. I don’t have a 6k photo mode camera, and when I did macro I just bracketed in camera. The stacking I did in post.

I guess the thing that has me confused is that in regular still mode, most digital cmaeras have a burst mode that is significantly slower than their video (which is I guess 30 frames/second, or something like that, maybe higher; whatever - faster). So why is that?

If it were as simple as the camera just firing off a video and making stills, you would expect burst mode to be super-fast and essentially the same as video, even on a DSLR with the flapping mirror. But often it is a relatively lousy 7 or 8 per second (yes, on some mirrorless it is much faster). But they are usually different. So I kind of assumed that still bursts are slower than video because the quality of video frames (i.e. when used as stills) was probably lower than individual images in still mode.

I could obviously be incorrect, but it does seem that there must be some reason that they even have a burst still mode (and have it as a bragging right when it is faster than competitors) on cameras that have strong video. Now that I'm thinking about it, maybe it's because 6K and 8K are NEW, so burst mode is a thing of the past that will go away once 6K and 8K become common on cameras. Hmmm, sounds about right, I think. This is interesting.

 

[D7k1]

FPS determines how viewers perceive video in relationship to the motion of the video and how the video is to be used. Shooting 18 mpx @ 30 FPS captures a lot of action that 10,12 or even 20 fps can't.

 

[DanS]

If it were as simple as the camera just firing off a video and making stills, you would expect burst mode to be super-fast and essentially the same as video, even on a DSLR with the flapping mirror. But often it is a relatively lousy 7 or 8 per second (yes, on some mirrorless it is much faster). But they are usually different. So I kind of assumed that still bursts are slower than video because the quality of video frames (i.e. when used as stills) was probably lower than individual images in still mode.

its a problem with multiple facets.

One aspect is how fast can you write to the sd cad. The fastest sd cards on the market will only guarantee sustained write speeds of 90 MB/s.

full res raw stills are out, because at 30 fps we are talking ~700 MB/s
full res & guallity jpeg stills are out, because at 30 fps we are talking ~300 MB/s

another aspect if how much processing power does the camera have and how much are consumers willing to pay for. On top of encoding the images and writing them to disk, the camera has to be able to adjust focus, exposure, the preview display, etc etc.

another aspect is how much power is needed to drive all that processing.

another is how much cooling is needed to keep everything from overheating.

The manufactures have to take all the things above into account, plus all the standard stuff of interpreting the market trends, etc.

you can by a blackmagic ursa mini pro today that shots 8k 120fps, buts it's $10k, and the size of an old school camcorder from the 80's. Not to mention once you have it fully kitted out so that its actually usable, and you have a machine that can process the footage you will have dropped another $10-$15k, and you won't want want to pick it up.

 

[BDR-529]

To my understanding the only reason a still frame from a video file would be worse than the resolution it is capable of (18mp for 6k, 8mp for 4K) is if it is a situation where electronic shutter would be inappropriate for one reason or another.

Well No. And Yes. It depends on what kind of "video file" we are talking about here.

In standard MP4 file camera might record only, say, every 12th frame as an I frame which is essentially a stand-alone jpg image. And even this I frame might have less dynamic range and higher compression rate than a real still jpg from the same camera.

Everything between I frames is made of P and B frames which record only difference from previous frame or in case of B frame use both previous and next frame as reference to what this frame should look like. Also the resulting frame is a far cry from normal jpg quality when seen as a still image.

Some higher end cameras like GH5 can record in All-I mode where each frame is a standalone jpg image but even at 400Mbps a single image is only about 1,9MB, also roughly half of best quality still-jpg

Only when using external recorders like Atomos Ninja, it's possible to record RAW video which is more or less a stream of unprocessed RAW images and there you could really take any given frame and process it into higher quality still image.

 

[archaeopteryx]

I'm now as confused as you are about the focus stacking

All that's documented in the manuals is 4k and 6k stacking from post-focus, so I wouldn't expect more than that. However, since Panasonic doesn't update manuals after initial release I've been unable to preclude the possibility of stacking from brackets as a later addition.

It’s not 20, but it’s not worse than if you fired off a burst with an 18mp sensor

Actually, that's not guaranteed. The 6k bit rate on the G9 is 150 Mb/s (200 Mb/s from the GH5). So, depending on the amount of similarity between frames, the number of compressed bits may be lower in 4k or 6k post-focus relative the amount of unique information per frame. (The BGH1 has 400 MB/s but its manual has zero hits for post focus or focus stacking and it's in the video line, so I presume it lacks support.) This isn't a simple thing to quantify and most photographers aren't going to write the image analysis code to work out the details but there are a couple key things in practice.

The first is it's unclear if post-focus .mp4 gets converted to jpeg frames on camera with the inclusion SOOC jpeg processing components like i.Dynamic or i.Resolution but it seems they're most likely not involved. Off camera, these definitely don't apply. So there is a tradeoff between video fps for bracket acquisition speed and tone curve control which, in my experience, primarily affects exposure management for highlights. It can also be significant to noise. Panasonic doesn't document anything about how video mode settings affect post-focus but, at least on the G9, there's some evidence v-log is helpful to relaxing tone curve constraints.

The second is raw is an option in stills-based brackets but, so far as I know, raw video in post-focus isn't supported in m43 as its an S1 specific feature. This has tone curve and other processing implications somewhat beyond SOOC considerations. There are also differences in bit depth (8- or 10-bit video versus 12-bit raw) and (usually) sampling that aren't well documented but do have pixel peeping effects.

Since the value of capturing 18 MP 6k, 16 MP still, or 20 MP still frames rather than 8.3 MP 4k frames is presumably in ensuring the larger frames pixel peep well enough to be worthwhile it's perhaps worth noting that 6k needs 234 Mb/s to have the same bits per pixel as 100 Mb/s 4k. So far as I know, no one's ever done a comparison of stacks from 4k and 6k post-focus, well controlled or otherwise, but the mathematics don't put me in a hurry to get a 6k capable body.

Does that mean that it uses successive video frames from a video to stack into a still image?

As @11GTCS said, the basic answer is yes. The more complex answer is video and still photography have different assumptions about the most desirable handling of rolling shutter artifacts. Stacking from post-focus is somewhere in the middle due to the translation from video frames to a still image. Since there's usually an effort to minimize motion during (auto)focus bracketing it tends not to matter but my experience is the effects can be significant in macro situations like flowers with some wind motion (and become more pronounced with camera vibration at photomacrographic magnifications).

(a focus-stacked image using an E-M5.3 - i.e. 20MP - is somewhat cropped too)

In practice it's nearly always the case alignment of stacked images results in cropping. The extent of the crop is controlled primarily by the bracketing method and lenses used but there is also some software contribution. (It's easier just to plan for the crop than to arrange near-ideal telecentricity or rear standard bracketing.)

Now that I'm thinking about it, maybe it's because 6K and 8K are NEW, so burst mode is a thing of the past that will go away once 6K and 8K become common on cameras.

Panasonic cut SH burst from the G95 but I suspect that's just feature crippling. Sony's flyers indicate the lack of 30-60 fps still burst isn't a sensor limitation so it's presumably due to some combination of processing engine and card write speed limitations in the direction @DanS mentioned.

All of the Panasonic 4k cameras can use 4K to focus stack.

Definitely not the G7 and I think probably not the GX8 or GX80. All the later ones (including the GX800), yes.

@archaeopteryx says that he

Wrong gender there.

 

[RAH]

In practice it's nearly always the case alignment of stacked images results in cropping. The extent of the crop is controlled primarily by the bracketing method and lenses used but there is also some software contribution. (It's easier just to plan for the crop than to arrange near-ideal telecentricity or rear standard bracketing.)

On an E-M5.3 (probably on an E-M1 too), when you use focus stacking, it shows you the crop frame that will be used when the final image is produced. Yes, I agree that is an alignment thing. It's nice that it shows you what you'll get so you can adjust your framing to allow for that.

Wrong gender there.

Oops! Sorry!

 

[BDR-529]

Definitely not the G7 and I think probably not the GX8 or GX80. All the later ones (including the GX800), yes.

I found an article in Camerajabber where they claim that Focus Stacking works on all panny bodies which support 4k Photo and therefore also Post Focus-features. I believe that this feature was added to older models in FW update.

"Both the Panasonic G80 / G85 and G7 offer the company’s 4K Photo functionality, which allows you to extract a still frame from a short burst of 4K video. They also employ Panasonic’s Focus Stacking and Post Focus features, which allow you to take still frames from that short burst of 4K video".

This sounds logical since panny uses DFD AF which means that they don't rely on limited number of fixed PDAF-points but can analyze the actual image instead and detect which areas are in focus on any given frame.

This Post Focus-feature is a very clever marketing gimmick that actually does what it promises as long as user is shooting something static. In Post Focus-mode camera is basically shooting a short 4/6k video clip but focusing every frame at different points in the scene.

After Post Focus "image" (which is in reality a short mp4 video clip) is taken, user can view it in the screen and touch any point that that should be in focus and camera instantly "focuses" there. In reality it just browses through all frames in the video clip untill it finds one that was focused on requested area and presents it on the screen.

 

[Hypilein]

I can confirm that the GX8 has no Post-Focus and shoots 4k.

 

[archaeopteryx]

I found an article in Camerajabber where they claim that Focus Stacking works on all panny bodies which support 4k Photo and therefore also Post Focus-features.

Camerajabber is incorrect. While post-focus was added to the G7 in a firmware update, in camera focus stacking was not. All you get on the G7 is the ability to go through the bracket and choose frames to export. Since this a ridiculously tedious for setting up a stack the usual workflow is to copy off the .mp4, shred it to frames with ffmpeg, trim the bracket, and stack that. While I don't have a GX8 to confirm, I think the GX8 is probably the same.

This sounds logical since panny uses DFD AF which means that they don't rely on limited number of fixed PDAF-points but can analyze the actual image instead and detect which areas are in focus on any given frame.

Which would be great if Panasonic actually got it right, but they didn't. So, particularly in photomacrography, depth reconstruction errors can occur during stacking due to focus steps being incorrectly skipped in the post-focus process. Some of this is probably also due to limitations to AF motor precision and accuracy in making steps.

The claim about DFD points relative to PDAF is somewhat disingenuous for four reasons.

1. Panasonic omits focusing on the margin of the 4k crop so, in addition to alignment cropping, compositions have to be planned with the potential for out of focus border cropping. It's not a big deal but does sometimes throw people who are expecting to be able to use the full 8.3 MP of 4k as opposed to usually getting 7.something MP. Maybe 6.something if there's more "foreground" than usual to crop out.
2. So far as I know, the only other focus bracketing implementation which operates like post-focus is Fuji's multi-focus. As best I could tell from Fuji's manuals and firmware notes, last I checked that was limited to certain X-A bodies. Which don't have PDAF.
3. All other bodies I know of with autofocus bracketing implement it as a series of steps of size determined a priori by a step size parameter. It follows the steps do not use autofocus, instead relying on closure of the autofocus control loop through sensing of the focus group(s) positions. This at least partially avoids post-focus's problems with not always laying down an entirely gap-free set of frames along the optical axis.
4. Even if other bodies did include AF focus confirmation in autofocus bracketing, the (D)SLRs I know with autofocus bracketing are the Nikon D810+. So limitations in (D)SLR PDAF point grids would become relevant only if Nikon makes an improbably large firmware feature change. Much the same is true for all mirrorless bodies. If Fuji ever does bubble up multi-focus as a feature it's likely the bodies which would get have full sensor PDAF coverage (since that's the X-T3 generation and later) and that the autofocus bracket would run in hybrid AF (since, so far as I know, falling back to only PDAF is an Olympus-specific thing).

Depending on the lens' autofocus speed and amount of focus breathing there's also potentially issues with variable frame keystoning due to the focus elements moving during post-focus exposures. This is mainly an issue with the Panasonic-Leica 45 f/2.8 macro's notoriously slow AF and it's nearly always avoided by trimming off the unneeded starting and ending frames in the bracket. So it's only rarely a functional issue, and usually only when you forget to trim and then have to go back, trim, and restack. However, for most macro compositions, the number of frames skipped is typically small and so the time savings from dropping those frames at 30 fps is also small.

What Panasonic is missing is a really simple 30 fps 4k autofocus bracketing mode which pulls to the closest AF point and then continues with a priori steps until the farthest AF point is reached. This would be an easy thing to code, minimizes issues around post-focus step size variation, and maintains post-focus's automation of the parts of focus bracketing which are time consuming to set up. But, five years on, they still haven't implemented this minor enhancement.

In reality it just browses through all frames in the video clip until it finds one that was focused on requested area and presents it on the screen.

Minor detail: judging by the long load time, I'm pretty sure what happens are the frames are indexed by depth when you open the post-focus video. After that it's quick to access some of the frames on camera but if you want all of the frames in a macro situation you're going to be shredding the video file.

 

[BDR-529]

I can confirm that the GX8 has no Post-Focus and shoots 4k.

I did check this from the local Panasonic website (Europe/PAL models) and they too claim that Post Focus has been available for both GX8 and G7 since FW update 2.0 which was released on 25th of November 2015

"[Post Focus] is now available.
New function which enables the user to select in-focus area and pull out favorite shots, after shooting."
https://av.jpn.support.panasonic.com/support/global/cs/dsc/download/fts/dl/gx8.html

This feature was also not there when camera came out of the box and it might not be included in the original printed manual if there ever was one.

They don't indeed mention anything about in-camera Focus Stacking, though and based on original reviews this feature was only available on PC software which means that the whole Post Focus mp4 clip has to be uploaded to PC for processing.

I have never found any use for Focus Stacking or Post Focus but it would be interesting to test them out one of these days when I don't have anything better to do. HDR on the other hand is a very usefull feature but that has been around since the dawn of digital photography.

 

[BDR-529]

What Panasonic is missing is a really simple 30 fps 4k autofocus bracketing mode which pulls to the closest AF point and then continues with a priori steps until the farthest AF point is reached. This would be an easy thing to code, minimizes issues around post-focus step size variation, and maintains post-focus's automation of the parts of focus bracketing which are time consuming to set up. But, five years on, they still haven't implemented this minor enhancement.

Panny did rewrite DFD system from scratch for S5 and desription of updated 6k (18MP image) Post Focus system sounds just like the one you described.

Unfortunately the only MFT body which has this new DFD system is BGH1 which is so video-centric that panny apparently eliminated pratically all still image features from the SW.

Better wait for GH6 then but arrival of that model starts to sound like a play by Samuel Beckett.

 

[D7k1]

Using Focus Bracketing with RAW is what I do.

Quite pleased with how the G9 handles these functions, but stacking in PS with focus bracketing Raw is the way to go IMHO.

 

[BDR-529]

If we jump back to GH6 for a second, there are now two new rumours which sound silly alone but make a lot of sense together.

Last year it was taken for granted that GH6 will have 8k video but a week ago at least one major camera site claimed in their 2021 predictions that this amount of pixels in a tiny m4/3 sensor is just silly and GH6 is going to have 6k video instead.

This would be perfect for hybrid camera because 6k video requires 19MP in 16:9 form factor which also means that 4:3 sensor needs 25MP. This sensor would maximize pixel size and low light performance while also having resolution that matches FF sensors where 24MP has become a de facto standard.

The only problem is that no 25MP MFT sensor is known to exist even on a drawing board. Enter the second new rumour which says that GH6 will use 33MP (8k) Super 35 sensor instead. This is not as stupid as it sounds. 35 super is very much alive in professional cinematography and there are plenty of sensors which are using the very latest technology.

How would this work then? Take Blackmagic who is using 16:9 "Super 35mm" sensor which has an image area of 23mm*13mm. Compare this with m4/3 image area of 17,3mm*13mm. MFT camera could also use the full height of the Super 35 sensor and just leave some pixels on the left and right edges unused.

Result would be exactly the one needed: cropping 17,3mm center of 23,1mm wide 33MP sensor would give GH6 a brand new 25MP "m4/3" sensor with all the latest bells and whistles like dual-ISO, BSI, pixel binning and so forth.

 

[speedy]

If we jump back to GH6 for a second, there are now two new rumours which sound silly alone but make a lot of sense together.

Last year it was taken for granted that GH6 will have 8k video but a week ago at least one major camera site claimed in their 2021 predictions that this amount of pixels in a tiny m4/3 sensor is just silly and GH6 is going to have 6k video instead.

This would be perfect for hybrid camera because 6k video requires 19MP in 16:9 form factor which also means that 4:3 sensor needs 25MP. This sensor would maximize pixel size and low light performance while also having resolution that matches FF sensors where 24MP has become a de facto standard.

The only problem is that no 25MP MFT sensor is known to exist even on a drawing board. Enter the second new rumour which says that GH6 will use 33MP (8k) Super 35 sensor instead. This is not as stupid as it sounds. 35 super is very much alive in professional cinematography and there are plenty of sensors which are using the very latest technology.

How would this work then? Take Blackmagic who is using 16:9 "Super 35mm" sensor which has image area of 23mm*13mm. Remember that m4/3 image area is 17,3mm*13mm. MFT camera could also use the full height of Super 35 sensor and leave some pixels on the left and right edges unused.

Result would be exactly the one needed: cropping 17,3mm wide center of 23,1mm 33MP sensor would give GH6 a brand new 25MP "m4/3" sensor with all the latest bells and whistles like dual-ISO, BSI, pixel binning and so forth.

That sounds very plausible to me. Does Sony make that particular sensor?

 

[BDR-529]

That sounds very plausible to me. Does Sony make that particular sensor?

This is the interesting part. Panasonic and Fuji signed a deal in 2013 for development of Organic/CMOS sensor that would replace the silicon diode used in current sensors. This is the Holy Grail of sensor technology.

In February 2018 Panasonic announced "industry's first" [email protected] OPF sensor that has also global shutter. They even demonstrated a prototype camera that used this sensor in 2019 when it was promised to be ready for 2020 Tokyo Olympics.

Unfortunately Panasonic sold their entire semiconductor unit to Nuvoton in November 2019 but in May 2020 Hitachi announced a broadcast 8k camera that uses Super 35 OPF sensor which sounds awful lot like panny technology.

" Using a CMOS image sensor with an organic photoconductive film (OPF) enables a higher saturation charge compared to silicon photodiodes, thus increasing dynamic range without the diminished image quality inherent with previous range expansion technologies.

The camera’s Super 35mm OPF CMOS sensor delivers 8K video with 7680x4320 resolution. The SK-UHD8060B is claimed to offer greater than 400% dynamic range, maximizing the effectiveness of its support for the Hybrid Log Gamma (HLG) HDR specification, and combines it with built-in signal-to-noise management to optimize visual performance."

Sharp was first to introduce 8k (33MP) broadcast camera in 2017 and it uses Super 35 sensor as well so this is already "old" technology in bizarrely expensive cine or TV cameras but more prosumer-oriented companies like Blackmagic have already introduced 12k (80MP !!) super 35 sensors in their flagship models.

Super 35 is not really a standard because each manufacturer uses somewhat different sized sensors but some 16:9 aspect ratio ones have also same height as M4/3 image area so it would be very easy to use one in MFT body by just leaving pixels unused on both edges.

 

[D7k1]

What ever the Gh6 is going to be, I think it might be between 2.5 and 3K US. It would take me a year to save up that much allowance, but such a camera has great allure for me (I do shot say 60% stills 40% video). I wonder about heat management and the size of the Gh6, but I still think 8K is going to be the play.

 

[L0n3Gr3yW0lf]

That's very, potentially, wonderful rumour and I do hope it's true, Super 35mm sensors are design for fast read data off the sensor as well as high bandwidth for recording quality. What that brings to photography is, possibly, increase (or more stable) performance at high speed shooting and/or highee quality recording even for Raw formats even in still (potentially 14-bit bursts, most mirrorless 35mm FF downgrade to 12-bit at high bursts, or less rolling shutter with electronic shutter, improved noise performance at high speeds, etc).
Right now E-M1 X and Mark III top at 30FPS (if I remember correctly) burst with electronic shutter while keeping the AF-C and AE continously updated, so an improvement would be raising that to 60 FPS for Pro Capture.
Also potential improvement would be improved HHHR or lower performance cost for HHHR that could be added at lower cost for cameras like E-m5 and/or Pen F.
If JIP can get a hold on that sensor tech.
For Panasonic side it would apply the same, just with the Panasonic set of features.

Another bug advantage to Micro Four Thirds would be if they can get the Super 35mm with Global Shutter, that would give a big boost against Sony A9 and Canon R5 with such high performance in hopefully future models like E-M1X Mark II and G9 Mark II... It would be cheaper to make such sensors for M4/3 sensor size then 35mm FF SIZE. GREAT FOR BOTH VIDEO AND STILLS ALIKE.

 

[BDR-529]

It would be cheaper to make such sensors for M4/3 sensor size then 35mm FF SIZE. GREAT FOR BOTH VIDEO AND STILLS ALIKE.

The main reason why I believe this rumour is not entirely without a merit is the fact that pretty much every single "Super 35" sensor seems to have somewhat different dimensions and they all are apparently coming from different manufacturers and using different technologies up to OPF.

This indicates that every Super 35 sensor is a very high-tech, very expensive and very, very low volume product in an industry where economies of scale is everything.

Currently Super 35 sensors are only used in real TV/Cine/broadcast-cameras which don't see consumer cameras as their direct competitors.

Manufacturers and their customers might be more than willing to sell one Super 35 sensor to Panasonic because GH6 alone would increase production volumes to entirely new level which causes prices to drop substantially - for every customer.

Unfortunately the most likely outcome is that GH6 - if we ever see one - would just use the 41MP Sony MFT sensor which has already been in production under their surveillance category since September. In the likely case that price tag of this 8k camera body alone matches the combined price of Panasonic S5 + 24-105mm f/4 Macro O.I.S, it will be very easy for me to decide which way to go.