[biz-engineer]

16 bit already demoised DNG, 16 bit tiff.

Falk's statement led me to rethink about it.

DNG is supposed to undergo classic bayer demosaicing by existing pp software, if so, I don't see how you could use a DNG that is already demosaiced as it wouldn't comply with mainstream workflows. However, Pentax specific pixel shift algorithm with / without MC, are already available in the K1, so you can exports into a TIFF but.... unfortunately, it was already stated that this TIFF is 8bits deep. Ricoh should implement a 16bits TIFF via firmware update, or upsampled JPEG = 144Mpixels large option. Anyway, you could argue about the file size, but the pixel shifted raw are already as large as 144Mpixels files, IMO it's not an argument. Ricoh have to rethink about their approach of pixel shift for the workflow to be seamless for the end users.

---------- Post added 26-07-16 at 11:43 ----------

What I'm after the most is automatic in-camera focus stacking in RAW, more than anything. Love macro, hate macro rails.

Yes. I guess the issue for Ricoh is if they implementing some automated multishot technique in their cameras, the possibilities are unlimited (as soon as you are on tripod), but the amount of work for them is also unlimited. For example, you could think of automating bracketing to increase DR without image quality tradeoff (the camera could use the electronic shutter to take a first image and decide for a range of exposures and take two additional shots to combine them into one). Currently, you could ad a few automation sequences on top of current camera functions to make it a beast that no other brand can rival... but that's up to Ricoh to decide to implement that.

With the current K1, some of the things I'd like to do aren't so convenient because there are a few gaps between modes that aren't really fitting together. For instance, in camera HDR does have the auto-align feature, but composite/interval shooting doesn't have that option. Electronics shutter is used for pixel shift but not available as an option for image stacking or bracketing etc...

That said, implementing multishot techniques with a 36Mpixels sensor is a bit of an overkill. I'd rather see more multishot technique automation on apsc cameras.

---

Last edited by biz-engineer; 07-26-2016 at 02:55 AM.

[HopelessTogger]

Falk's statement led me to rethink about it.

DNG is supposed to undergo classic bayer demosaicing by existing pp software, if so, I don't see how you could use a DNG that is already demosaiced as it wouldn't comply with mainstream workflows. However, Pentax specific pixel shift algorithm with / without MC, are already available in the K1, so you can exports into a TIFF but.... unfortunately, it was already stated that this TIFF is 8bits deep. Ricoh should implement a 16bits TIFF via firmware update, or upsampled JPEG = 144Mpixels large option. Anyway, you could argue about the file size, but the pixel shifted raw are already as large as 144Mpixels files, IMO it's not an argument. Ricoh have to rethink about their approach of pixel shift for the workflow to be seamless for the end users.

---------- Post added 26-07-16 at 11:43 ----------


Yes. I guess the issue for Ricoh is if they implementing some automated multishot technique in their cameras, the possibilities are unlimited (as soon as you are on tripod), but the amount of work for them is also unlimited. For example, you could think of automating bracketing to increase DR without image quality tradeoff (the camera could use the electronic shutter to take a first image and decide for a range of exposures and take two additional shots to combine them into one). Currently, you could ad a few automation sequences on top of current camera functions to make it a beast that no other brand can rival... but that's up to Ricoh to decide to implement that.

With the current K1, some of the things I'd like to do aren't so convenient because there are a few gaps between modes that aren't really fitting together. For instance, in camera HDR does have the auto-align feature, but composite/interval shooting doesn't have that option. Electronics shutter is used for pixel shift but not available as an option for image stacking or bracketing etc...

That said, implementing multishot techniques with a 36Mpixels sensor is a bit of an overkill. I'd rather see more multishot technique automation on apsc cameras.

Focus bracketing with pixel shift. Imagine the file size!

[Simen1]

A decade from now the price of memory capacity, speed and computational power should be sufficient to introduce such a function, even with a large pixel inflation.

In the shorter term, I hope the next high end APS-C and FF cameras will shoot the pixel shift sequence much faster. Four frames in 1/3 second should be within reach. That should reduce motion blur problems significantly regardless of how good or bad the software handles the motion blur compensation function.

Fast burst focus bracketing may potentially be one of the main new features as well and it can be used for much more then just macro. But its highly unlikely it can be combined with pixel shift.

[Nicolas06]

Fast burst focus bracketing may potentially be one of the main new features as well and it can be used for much more then just macro. But its highly unlikely it can be combined with pixel shift.

Burst rate typically linked to sensor read performance and shutter. With electronic shutter there still the sensor read performance. You'd need 12FPS at 36MP to take 4 pictures with pixel shift in 1/3third of a second. That's quite high expectations by today standards for FF cameras.

Even A7R-II with a BSI modern sensor has a mere 5FPS... Sure it has 42MP but 5FPS is far from 12FPS.

On next APSC flagship if we stay 24MP and go BSI, that might be possible but still a challenge.

[Simen1]

I wasn't thinking about 36 Mp, but 20-24.

Nikon D5 does 21 Mp at 14 fps
Canon 1DX markII does 18 Mp at 16 fps
Samsung NX1 does 28 Mp at 15 fps, and this camera is almost two years old

All these are optimized for longer series then four shots, so I think 12 fps for 1/3 of a second is quite doable on the next Pentax top FF and APS-C models. Four shots don't require a large buffer and it won't heat up the AD circuits as much as longer series.

[HopelessTogger]

I wasn't thinking about 36 Mp, but 20-24.

Nikon D5 does 21 Mp at 14 fps
Canon 1DX markII does 18 Mp at 16 fps
Samsung NX1 does 28 Mp at 15 fps, and this camera is almost two years old

All these are optimized for longer series then four shots, so I think 12 fps for 1/3 of a second is quite doable on the next Pentax top FF and APS-C models. Four shots don't require a large buffer and it won't heat up the AD circuits as much as longer series.

For faster buffer clearing, would having UHS-II support do the trick?

[Nicolas06]

For faster buffer clearing, would having UHS-II support do the trick?

It clear the buffer faster if the camera can benefit of such write speed. It will not change one bit the max FPS/burst rate or the time required for a pixel shot as there more than enough buffer for 4 images anyway. What the limiting factor is the sensor read speed.

D5 read speed: 294MP/s
Canon 1DX read speed: 288MP/s

Required read speed for 36MP at 12FPS (so pixel shift at 1/3s): 432MP/s ... 50% more than the best action camera out there.

Samsung NX1 is at 420MP/s, but Samsung did stop investing on cameras so this technology may never be really used in combination with pixel shift. Pentax doesn't make it own sensor and doesn't use Samsung sensors neither.

If Pentax provide a sport oriented FF it might do it, likely not at 36MP... And it could be argued that 36MP without pixel shift might better and more conveniant than 24MP wiht pixel shift. And in the case you can do PS at 24MP, most of the time, you may be able to do PS at 36MP with the other camera too...

---

Last edited by Nicolas06; 07-26-2016 at 11:22 PM.

[Simen1]

All I say is that 4 image pixel shift in 1/3s is technically feasible with current sensor technology and buffer size and might show up in the next future top models (a year or two from now). I didn't say it should be the sustained speed to the memory card anytime soon.

Then I did speculate further into the future: a decade from now. Maybe memory cards will be so fast that a ram buffer isn't needed anymore and that full sensor readout may be done at even faster rates, enabling pixel shifted focus stacking, even with higher pixel counts.

[falconeye]

DNG is supposed to undergo classic bayer demosaicing by existing pp software, if so, I don't see how you could use a DNG that is already demosaiced

I believe Nicolas already mentioned it:
There is a so-called "Linear DNG" subformat which is supposed to be demosaiced data.
E.g., it is output by the DNG converter if selecting to output downscaled DNGs.

The problem with linear DNG is that it is 3x the size of uncompressed RAW, i.e., about the size of pixel-shift RAWs. Moreover, I am not sure about available bit-depths for linear DNG. Eventually, the ability to detect motion will improve over time and a true RAW format with all four sub frames would most probably still remain the best option for archiving.

[Rondec]

I wasn't thinking about 36 Mp, but 20-24.

Nikon D5 does 21 Mp at 14 fps
Canon 1DX markII does 18 Mp at 16 fps
Samsung NX1 does 28 Mp at 15 fps, and this camera is almost two years old

All these are optimized for longer series then four shots, so I think 12 fps for 1/3 of a second is quite doable on the next Pentax top FF and APS-C models. Four shots don't require a large buffer and it won't heat up the AD circuits as much as longer series.

You can have these sorts of things, but it requires a little newer sensor and would be quite a bit more expensive. The sensor in the A7r II probably could get close. Certainly the sensors in the D610/D750/6D aren't capable of this sort of read out speed.

[HopelessTogger]

It clear the buffer faster if the camera can benefit of such write speed. It will not change one bit the max FPS/burst rate or the time required for a pixel shot as there more than enough buffer for 4 images anyway. What the limiting factor is the sensor read speed.

D5 read speed: 294MP/s
Canon 1DX read speed: 288MP/s

Required read speed for 36MP at 12FPS (so pixel shift at 1/3s): 432MP/s ... 50% more than the best action camera out there.

Samsung NX1 is at 420MP/s, but Samsung did stop investing on cameras so this technology may never be really used in combination with pixel shift. Pentax doesn't make it own sensor and doesn't use Samsung sensors neither.

If Pentax provide a sport oriented FF it might do it, likely not at 36MP... And it could be argued that 36MP without pixel shift might better and more conveniant than 24MP wiht pixel shift. And in the case you can do PS at 24MP, most of the time, you may be able to do PS at 36MP with the other camera too...

To be honest I can live with four frames per second (although six would be good and doable I'd have thought). It's the buffer clearing time that does my head in. If I could shoot say five frames per second without the buffer filling up and then taking what seems until Christmas to clear, that'd be good.

[D1N0]

You can have these sorts of things, but it requires a little newer sensor and would be quite a bit more expensive. The sensor in the A7r II probably could get close. Certainly the sensors in the D610/D750/6D aren't capable of this sort of read out speed.

The a7RII does 5fps (also in crop mode) I wouldn't expect any speed wonders from it. You can ofcourse shoot 4k video with it and just uses frames from that. You'll have 8mp 30fps.

[Nicolas06]

To be honest I can live with four frames per second (although six would be good and doable I'd have thought). It's the buffer clearing time that does my head in. If I could shoot say five frames per second without the buffer filling up and then taking what seems until Christmas to clear, that'd be good.

To substain 4.5 FPS of a K1 indefinitely with 45MB on average per picture, you need 180MB/s actual write speed. From what I have seen only the most expensive SD card on the market can do that; Count 65€ for a 32GB lexar for example while a classical sandisk that is already quite fast is 12€.

But apparently the K1 processing power isn't enough and once the buffer is full, game is over anyway. Honestly for what I shot, I don't care.

If optimal burst performance is key for you, maybe you want a sport camera instead?

[Simen1]

You can have these sorts of things, but it requires a little newer sensor and would be quite a bit more expensive. The sensor in the A7r II probably could get close. Certainly the sensors in the D610/D750/6D aren't capable of this sort of read out speed.

Yes, it would require a new sensor. Don't you think its plausible that the next FF and APS-C top models will get newer sensors? Its happened before and could happen again. The pricing of the top APS-C camera have usually been the same or shifted slightly down, compared to its predecessors launch prise, even if the newer model got a newer sensor. Why do you think a new sensor would be quite a bit more expensive? And why should Pentax be limited to re-use sensors that already sits in older cameras? Pentax may choose a sensor thats not implemented in any camera yet. Pentax could be the first too.

[HopelessTogger]

To substain 4.5 FPS of a K1 indefinitely with 45MB on average per picture, you need 180MB/s actual write speed. From what I have seen only the most expensive SD card on the market can do that; Count 65€ for a 32GB lexar for example while a classical sandisk that is already quite fast is 12€.

But apparently the K1 processing power isn't enough and once the buffer is full, game is over anyway. Honestly for what I shot, I don't care.

If optimal burst performance is key for you, maybe you want a sport camera instead?

I do want a sports camera, it's true. I'm just thinking of the next generation K-1 in a few years time. UHS-3 by then perhaps.