Panasonic seems to be announcing something "BIG" on December 15

[wolf33d]

5 hours ago, Jonesy Jones said:

@Cliff Totten @jonpais @Andrew Reid

Not sure if you guys saw or read my incredibly long and boring post in the middle path thread, but I am pretty certain this new camera will either have the same sensor or a new iteration of the sensor that is in the current GH5. My logic is as follows:

• Luke has shared several images from his production with the new "camera" (or whatever it is). The resolution of the larger image is 6656 x 2496. Currently the GH5 has the ALL-I anamorphic mode that shoots 3328 x 2496. So, with a 2x anamorphic lens, which Luke is shooting with, the images will be exactly 6656 x 2496. He also shared 2 smaller images that were 3328 x 1248, which is obviously an exact 50% decrease from the larger and would make complete sense that he just descaled by 50% for those. There is a 4th image that is exactly 6656 x 2496/1.73333 = 3840 x 1440, which I think is Luke either just experimenting with the horizontal resolution of UHD, or the vertical resolution of Youtube's 1440 res. 
• Could the new proposed Starvis sensor have a 3328 x 2496 anamorphic mode within it's max 4168 x 2824? Sure. But does that seem a bit contrived? It does to me. Unless there is something special about 3328 x 2496. Which by the way is also an EXACT 1/3 decrease from the current GH5 open gate of 4992 x 3744. (I'm assuming that 1/3 decrease helps with ETC and IBIS.)
• Therefore, because these shared images are exact simple ratios of the anamorphic mode of the current GH5, I'm thinking that either A) this new camera has the exact same sensor as the current one, or B) it is a version 2 of the same sensor, with improved specs. We've actually seen this before with BM's URSA and Micro Cinema Cams. The version 2 of those same sensors gave us more fps and supposedly better IQ. This could be a version 2 of the current GH5 sensor, or a recalibration with improved specs. 
• Luke's title of "The Middle Path" has had me thinking. I've been focused on the word "Middle", however I think "Path" is just as significant. Luke is very familiar with Red, and even discussed Red in that exact same thread. Red is well known for their upgrade "Paths", and the more I think about it, the more this new thing seems like it could be an upgrade program with the current GH5. I know there is no precedence for this from Panasonic, but I am thinking there could be an upgrade path for current GH5 owners (since it just came out pretty recently) to whatever this new camera is. I think that upgrade will cost $500, or $2500 for the new camera itself. And this also comes from simple math based on Luke's multiple uses of 1524.8% times the camera in relation to using the $38,120 Arri anamorphic lens. 
• Bottomline, I am thinking that there is a new camera based on an improved version of the current sensor. It will cost $2500 or be a $500 upgrade for current GH5 owners. It will have improved video features. Better low light, audio, and auto focus would be my top guesses. But I think more DR and better color too (Luke's images are just superb). I don't think any of those specs will be "omg this is the last camera I'll ever need to buy" amazing, but they'll at least be better than they currently are. Maybe a better monitor? Maybe improved focus assist stuff? Maybe 10 bit 60 fps or 72 fps? Maybe more? Again, I think we should manage expectations here, but it's not inconceivable that all or most of those hypotheticals are at least improved to some degree for this new video centric GH5. 

Sounds good, good idea about the upgrade. But I do not see how can they upgrade existing GH5 with better AF? Software update for AF ok but that is nothing. I do not see them upgrading it to a DPAF then ... too big change for an upgrade of customer body. Do you think a paid upgrade could bring hardware change,,,? 

Also Luke said "best V LOG ever" so it could indeed be a paid upgrade. 

 

[Rinad Amir]

id be pist if i had to pay another 100euro for extra Dr , just saying

[Cliff Totten]

Wolf33d. - You could be right that this "S" model uses the same Sony IMX272 sensor that the current GH5 does today.

However, the idea of Panasonic using the same. Sony style A7r/s marketing model clone goes out the window.

G9 - Fast photo centric model

GH5 - Great video/photo hybrid model

GH5-S - Heavily optimized, for video only.

With the same pixel count sensor, there is no room to improve the pixel well (photosite) saturation ability. These pixel wells will simply fill up and clip at the same amount as they do  today. The only way to increase dynamic range is to lower the density of sensor and get those photosite wells larger to hold more photons. This kills it's photo worthiness completely but could allow for significant dynamic range improvements.

I strongly suspect that they will buy Sony again for this "S" camera. Plus, this Starvis sensor seems to promise ALL the improved video specs that a GH5-S could ask for. (Minus 6k readout and 5k ovesampling) An 11mp video only sensor with 4.63 micron photosites will almost make you forget you are shooting MFT.  

This new quad Bayer pixel layout is freaking me out. How do you expose a cluster of 4 pixels with 2 different speeds and create ONE pixel value from that? Two pixels see into shadows deeper while the other two protect highlights better.

That is some crazy whichcraft!

Maybe this is why Luke mentioned "best V-Log ever"

If this crazy idea DOES lead to impressive dynamic range, then yes,...that WOULD make for great VLog!

CT

[Cliff Totten]

Does anybody know how to read these  Starvis IMX294 sensitivity numbers?

 

Sensitivity (F5.6)  Typ.   1900 mV  1/30s    accumulation (HCG)

Saturation signal  Min.    970 mV    Tj = 60 °C (LCG)

 

Can anybody pull a signal to noise ratio out of this, like "60db" or anything like that?  How do these numbers stack up against other known performing sensor numbers?

 

[IronFilm]

16 hours ago, Shirozina said:

It would be ironic if this new GH5 was a more 'stills' oriented camera with a higher MP count and reduced video capability.

GH5S = "S" for stills! ;-)

1 hour ago, Cliff Totten said:

G9 - Fast photo centric model

GH5 - Great video/photo hybrid model

GH5-S - Heavily optimized, for video only.

That three way split is probably the only way we'll see NDs put into a GH series body?! :-/

[wolf33d]

1 hour ago, Cliff Totten said:

Wolf33d. - You could be right that this "S" model uses the same Sony IMX272 sensor that the current GH5 does today.

However, the idea of Panasonic using the same. Sony style A7r/s marketing model clone goes out the window.

G9 - Fast photo centric model

GH5 - Great video/photo hybrid model

GH5-S - Heavily optimized, for video only.

With the same pixel count sensor, there is no room to improve the pixel well (photosite) saturation ability. These pixel wells will simply fill up and clip at the same amount as they do  today. The only way to increase dynamic range is to lower the density of sensor and get those photosite wells larger to hold more photons. This kills it's photo worthiness completely but could allow for significant dynamic range improvements.

I strongly suspect that they will buy Sony again for this "S" camera. Plus, this Starvis sensor seems to promise ALL the improved video specs that a GH5-S could ask for. (Minus 6k readout and 5k ovesampling) An 11mp video only sensor with 4.63 micron photosites will almost make you forget you are shooting MFT.  

This new quad Bayer pixel layout is freaking me out. How do you expose a cluster of 4 pixels with 2 different speeds and create ONE pixel value from that? Two pixels see into shadows deeper while the other two protect highlights better.

That is some crazy whichcraft!

Maybe this is why Luke mentioned "best V-Log ever"

If this crazy idea DOES lead to impressive dynamic range, then yes,...that WOULD make for great VLog!

CT

I was quoting someone

i do not know if they will bring same GH5 sensor. In fact do not hope... 

i hope for a completely new model with lower mpx count and PDAF

[DBounce]

Apparently this is now pushed back until the 8th of January. That is right around the time of CES 2018.

[majoraxis]

That makes sense!  It's going to be an interesting NAB this year... oops - too soon to starting think about what Blackmagic and others can do the latest Sony sensors... 1" global shutter ... back on topic.  Looking forward to whatever new Panasonic GH announces early next year.

[IronFilm]

16 hours ago, DBounce said:

Apparently this is now pushed back until the 8th of January. That is right around the time of CES 2018.

Makes a LOT more sense. So that it doesn't harm the Christmas sales. 

[Cinegain]

Typical. Pre-order promotion (free battery grip) for the G9 ends 24th of this month. Do think it's time to get rid of some camera gear though. D5300 has served me well, but a liveview approach for stills on this is the worst and the video capabilities... so-so, it's just no mirrorless camera with all the innovation and features. Do love me some APS-C/S35 goodness, but... apparently you can't have it all. The original E-M1... has been a good sport too, but is starting to feel dated and most of the more dedicated stills shooting happens with the G80 these days. That one itself has been great too, but it's probably a good idea to ditch all three and get myself in on that G9 promo and make that one the new dedicated stills cam. And then the GH5... what is the GH5s going to do? Will it be the ultimate video focused camera to compliment the G9? Or will the GH5 hold its own? Hum. Anyways, one thing's for sure: Panasonic is on top of the game and they're coming after our wallets!

[DBounce]

I would love to see this tech make it’s way into the GH5s... 8k/4K Organic Sensor

[Kisaha]

Fail. That will hurt their GH5 sales for sure, no way I am getting a GH5, if I was in the market for one, knowing that that will be an announcement on a month exactly. This gives time to other competitors to make - or brake - their own announcements just after Christmass. I guess 2018 it will be the decisive year. Everyone seem to have a competitive mirrorless on the market. Panasonic and Sony are leading the way so far, but this is the last change for Nikon, and one of the last chances for Olympus, then Fuji, and to a lesser degree, Canon.

@Cinegain sorry to bring that up, but that APS-C camera exists, and I do have most. It is called Samsung NX1, and except 4K/60p has everything the G9 has and more. Including the battery grip was always less expensive than the G9. oh, and also has this little screen on top that a lot of people made fun of 3 years ago!

[Cinegain]

You're conveniently forgetting the 'issues' it has and shares with the Fujifilm X-T2 (which I'd probably pick over it, if I'd pick any APS-C mirrorless camera currently out there at all), though : no sensor stabilization, no forward facing screen, limited use mount. I did consider it when looking for somewhere to go from the GH4 (of course then came the generation with sensor stabilization for 4K video starting with the GX80/GX85) and Sony and Canon EOS-M were definitely out of the consideration window and they're still nowhere near my line of sight. That's why I'd say that the best mirrorless APS-C camera probably is the X-T2. Can't see Sony cleaning up their act and Canon is too stubborn to get with the times, I'll give 'em that the basics are alright, like color and C-AF, but you can't be that one trick pony when you see what else is out there. Nikon has signs that it's aiming for the high-end fullframe mirrorless market and of course there's nothing new to be expected from Samsung, so... that pretty much concludes the APS-C mirrorless scene. Though I won't lie, a videocentric X-T2 from Fujifilm with sensor stabilization, a touchscreen, headphone jack on the body and preferably as well a frontfacing screen and silky smooth C-AF, would be rather tough to resist... yet, Panasonic has come a long way and for those like me, with a vast native lens line-up, lenses which partly are fairly compact and light (e.g. Laowa 7.5mm f/2 MFT & Leica 15mm f/1.7), they're most interesting if you're on-the-go a lot. And they just cover it all. Like, why can't anyone do a camera body like that, with features like that, for APS-C? Luckily I usually am able to somehow manage the downsides of the smaller sensor size, I'd rather pay the penalty there in terms of missing out on a slight flexibility advantage of a bigger sensor, but then having a camera that's more or less a textbook example of how to make something that's a joy to use (like having sensor stabilization, a front-facing screen, a mount with a vast native selection, incl. very compact lenses and electronic adapters for EF lenses, et cetera).

Of course weighing the pros and cons of each system turns out differently for different people. In the GH4 days the NX1 might've been the ultimate pick... and I almost had one 2 years ago, if it weren't for both me and the seller misinterpreting a promotion that was listed on their site (the NX1 + 16-50mm f/2-2.8 S lens was listed for 1899,- EUR and their deal was that if you had any old camera to trade-in, you'd get a 600,- cashback bonus (though I was told to keep whichever old worthless camera I had for trade-in), so we both concluded the set was going to set me back 1299,- EUR and I went ahead and send 'em my details to make up the invoice. Then the news came that a colleague mentioned the listed price already included the trade-in cashback, so it was going to take 1899,-. At this time it was already well-known Samsung had pulled the plug out of the NX line-up, so that was a little too much to get into a system that wasn't going anywhere and I backed out).

[Cliff Totten]

4 hours ago, DBounce said:

I would love to see this tech make it’s way into the GH5s... 8k/4K Organic Sensor

I have been told that these organic sensers require ALLOT of power to drive. They supposedly also get very hot and will probably only show up in a large, fan cooled camcorder body.

Who knows for sure? Maybe they have been able to conquer the heat and power problems and are now able to get them into a small passive cooled body under low wattage???

CT

3 hours ago, Kisaha said:

Fail. That will hurt their GH5 sales for sure, no way I am getting a GH5, if I was in the market for one, knowing that that will be an announcement on a month exactly. This gives time to other competitors to make - or brake - their own announcements just after Christmass. I guess 2018 it will be the decisive year. Everyone seem to have a competitive mirrorless on the market. Panasonic and Sony are leading the way so far, but this is the last change for Nikon, and one of the last chances for Olympus, then Fuji, and to a lesser degree, Canon.

@Cinegain sorry to bring that up, but that APS-C camera exists, and I do have most. It is called Samsung NX1, and except 4K/60p has everything the G9 has and more. Including the battery grip was always less expensive than the G9. oh, and also has this little screen on top that a lot of people made fun of 3 years ago!

It's my guess that the GH5 and GH5-S will not match other in video features.

The GH5 offers a BEAUTIFUL 6k readout and 5k to 4k (Sony style) full pixel readout and scaling engine. The GH5 offers enormous resolving power in 4k.

The GH5-S will be better in low light and could offer 13+ stops of dynamic range...but at the cost of that georgous resolutuon the GH5 has.

I think they will both produce two different "types" of images. I will own BOTH for different reasons!  ;-)

 

 

[sam rides a mtb]

Thinking more and more that its going to be the EVA sensor that they put in this new GHS35 camera

[Cliff Totten]

Andrew, do you want to take a shot at trying to explain how this IMX294 quad bayer circuit works?

We have a cluster of 4 photosites, 2 have one exposure time while the other 2 have a shorter exposure time. One set digs deeper into the shadows, while the other set protects highlights. Pixel "size" is not their trick...exposure "time" is their trick!

How do you average or bin those 4 values into one pixel ourtput high dynamic range number?

This is going to be a major selling point of this GH5-S. ( if its using a IMX294)

 

[seku]

I was wondering about that as well... I guess there are multiple solutions :

1. Two subpixel just expose for a longer time than the other, up to the max shutter speed defined, whereas the 2 other exposes way shorter. In that case, where the longer exposure blows out, you use the 2 short exposure subpixels to "recover" detail in the highlights. If this is the case, i kind of worry about motion blur artifacts, as we would be running different exposure times. As the sensor site talk about "Integration time"... i think this is what we will be getting.

2. I would kill for this : something similar to Magic Lantern  Dual ISO : Each pair of subpixels can have its own ISO set. No motion blur artifacts. and you "keep" the good signal (low iso for highlights, higher ISO for shadow detail). This only works if the sensor does not scale linearly with ISO amplification, like Canons. I do not know about how the Sony sensors behave.

3. Similarly, maybe there's an ND over a pair of subpixels? (ok, that is far-fetched)

[Cliff Totten]

The only thing with having 2 pixels with different ISO is that "gain" never increases dynamic range or signal to noise ratio....only light/photons do this. Dynamic range happens at the photosite level and measured by its pixel well saturation amount. Photosites that saturate (100% full) very quickly will have lower dynamic range. While photosites that can take in allot more photons without saturating will have more dynamic range. Signal "gain" doesnt improve the photosites saturation point. Dynamic range is determined before gain ever enters the picture. Remember, raw sensor data has no actual "gain" applied to it yet. Gain/ISO is a "post processing" tool to increase brightness of an already fully exposed sensor signal. In other words, "exposure" is completed and finished before signal gain is applied later down the road.

What this IMX294 sensor appears to be doing is "time" shifting each pair of pixels in the cluster. One set is exposed longer and takes in MORE light and saturates quickly but the other set has a shorter exposure time, taking in LESS light and protecting saturation. The "more light" value will help your shadows and the "less light" values protect the highlights. 

In the cluster....2 pixels will saturate easy while the other will saturate less! The question is....how in the Hell do you merge these different values into ONE high dynamic range pixel value?

My guess? Every value below middle grey comes from the slower exposed pixels while every value above middle grey comes from the faster exposed pixels??

I would also guess that these quad pixel clusters will produce a much "softer" 4k image over the current GH5's 5k oversampled image?? This is going to happen anytime you bin, sum or average chunks of pixels together.

So, if you set your camera for 1/30 second shutter speed, in your image, the darker pixels are drawn from the slower exposure photosites but the brighter pixels have maybe a 1/40 or 1/50 exposure time? 

Damn....thats a whicked concept!

 

[slonick81]

15 hours ago, Cliff Totten said:

The only thing with having 2 pixels with different ISO is that "gain" never increases dynamic range or signal to noise ratio....

But you'll get two ranges instead of one - for lighter and darker parts independently. These ranges may be narrower than one from doubled photosite, but in combination they can cover more range. Let's say 2xsized photosite has 10 stops of DR, 1xsized photosite has 9 stops of DR - proportional to area, but you can expose two of them with bracketing of 4 stops and get 13 stops with good range mapping (mid tones of one site cover problematic zones of another photosite - shadows or highlights).

15 hours ago, Cliff Totten said:

Signal "gain" doesnt improve the photosites saturation point. Dynamic range is determined before gain ever enters the picture. Remember, raw sensor data has no actual "gain" applied to it yet.

 

But you may want to amplify low signal befor ADC to get less digitising error. And anyway what you get in .raw file is not actual sensor data, there is still a lot of processing done.

15 hours ago, Cliff Totten said:

What this IMX294 sensor appears to be doing is "time" shifting each pair of pixels in the cluster. One set is exposed longer and takes in MORE light and saturates quickly but the other set has a shorter exposure time, taking in LESS light and protecting saturation. The "more light" value will help your shadows and the "less light" values protect the highlights. 

So, if you set your camera for 1/30 second shutter speed, in your image, the darker pixels are drawn from the slower exposure photosites but the brighter pixels have maybe a 1/40 or 1/50 exposure time? 

You can simply drain charge from brighter site through resistor so it'll store proportionally less electrons with same exposure time.

16 hours ago, Cliff Totten said:

Every value below middle grey comes from the slower exposed pixels while every value above middle grey comes from the faster exposed pixels??

Middle grey is different for dark and bright photosites. Some kind of smart blending in ovelapping area of dynamic range that evaluates non-linearity in shadows and highlights and noise amont in shadows.

16 hours ago, Cliff Totten said:

I would also guess that these quad pixel clusters will produce a much "softer" 4k image over the current GH5's 5k oversampled image??

Seems to. I guess it's still possible to get good resolution in the middle of DR, where both types of sites would be usable but shadows and highlights should suffer.

[maxotics]

1 hour ago, slonick81 said:

But you may want to amplify low signal befor ADC to get less digitising error. And anyway what you get in .raw file is not actual sensor data, there is still a lot of processing done.

If you're amplifying before the ADC wouldn't that circuitry distort the signal before digitising, just trading one source of error (analog) for another (digital)?  How would any specific pixel know whether to amplify or not?  I mean, that's the rub isn't it, no pixel can know beforehand how many photons will be sent its way.  Anyway, can you give more specifics about how the sensor works in that regard?  

When you say a "RAW file is not actual sensor data, there is still a lot of processing done", sure the camera manufacturers differ in how they program their ADC (where black levels might be set, say, which is data non-destructive assuming you won't need the full 14bits), but I don't know of any other "lot of processing" done.  Can you elaborate?   My experience is that RAW seems pretty close to actual sensor data.  Indeed, if you just output a non-debayered grayscale TIFF with dcraw (least amount of processing) it's unusable without a lot of complex processing.