Digital twin lens reflex.

Wolfgang Weisselberg wrote:
Bowser wrote:

The GH3, in single shot AF mode, is *very* quick, and easily much
quicker than the 5D II or 6D models I've shot.

'easily much quicker'? As in how many milliseconds? How about
you use a lens a step or 2 above kit lens or known slow-to-focus
focus lens (I hear tell the 85mm f/1.2 belongs in this class)
and do actually *measure* AF performance? And in the same
environment, too ...

The tech to put the same fast focusing optics on digital sensors is
still evolving. It can be done but it, being permanently attached,
hurts the image quality. DSLRs have the advantage that everything moves
out of the way for the photo.

I don't think embedded PDAF is the future. It's a stop-gap at best
until CDAF catches up. I believe it can, and will eventually.

It cannot. CDAF starts way too often into the wrong
direction. If it doesn't, it's embedded PDAF.

Not necessarily. There's often a chromatic difference between the
edges of front focus blur and back focus blur which could be
exploited by CDAF.

If you speed up the lens movement (and you must, for CDAF),
PDAF speeds will improve likewise. If you speed up sensor
reading and/or increase low light sensitivity, the PDAF
sensors will improve just as well.

Not necessarily, because CDAF "sensors" are selected from the very
large array of image pixel sensors, and the speed problem of finding
them in the first place is unique to CDAF.

One huge advantage of CDAF is focusing accuracy. There's no need to AF
calibration like there can be with PDAF since you're focusing right on
the sensor.

So add a 'calibrate PDAF'-function that pre-focuses with PDAF
(just for speed reasons), fine-focuses with CDAF (preferably
on a flat, orthogonal target), measures PDAF, checks again with
CDAF that the focus hasn't moved (else back to fine-focussing)
and thus gets the currently correct offsets for PDAF. Would be
a second with 1 sensor.

And for zooms which need different calibrations for each focal length?
And lenses with aperture related focus drift which need calibration
for each aperture? It's the nature of PDAF. however well calibrated,
to be an approximation based on simplifying assumptions. As megapixels
increase and lens quality improves yesterday's appropriate simplifying
assumptions become today's oversimplifications.

Not to mention the problems of curved planes of focus...

And on a tripod against a wall (or a newspaper on the wall) it'd
find all offsets within a minute or two --- with individually
microadjusting for this lens and focal length for every single
PDAF sensor. You'd need to do this only once or twice ...

PDAF can be as accurate as CDAF, just mucho faster.

When you don't care about the inherent residual inaccuracies. If you have an
old "nifty fifty" with spherical aberration do the best calibration
you can at f1.4, and then tell me how well it does at f5.6 with that
calibration. And did I mention curvature of the focus plane? :-)

--
Chris Malcolm

On 4/19/13 10:13 PM, philo wrote:
On 04/19/2013 02:44 PM, James Silverton wrote:
I enjoyed playing with a friend's much prized Rolleiflex film camera. Do
digital versions exist for cameras of this class? I was not able to
search one out.


No, there would be no need to do so.

I am convinced, we could get a huge potential out of a crossover of that
camera form with modern digital electronics and a really good electronic
retina-grade viewfinder behind flaps.
The typical Rolleiflex angle of view has advantages, moreover you could
have e.g a tele and a wide lens at the same time and combine both views
electronically.
The human eye works that way: high resolution fovea and wide view for
the rest.
--
One computer and three operating systems, not the other way round.
One wife and many hotels, not the other way round ! ;-)

Chris Malcolm wrote:
Wolfgang Weisselberg wrote:
Bowser wrote:

The GH3, in single shot AF mode, is *very* quick, and easily much
quicker than the 5D II or 6D models I've shot.

'easily much quicker'? As in how many milliseconds? How about
you use a lens a step or 2 above kit lens or known slow-to-focus
focus lens (I hear tell the 85mm f/1.2 belongs in this class)
and do actually *measure* AF performance? And in the same
environment, too ...

Well?

The tech to put the same fast focusing optics on digital sensors is
still evolving. It can be done but it, being permanently attached,
hurts the image quality. DSLRs have the advantage that everything moves
out of the way for the photo.

I don't think embedded PDAF is the future. It's a stop-gap at best
until CDAF catches up. I believe it can, and will eventually.

It cannot. CDAF starts way too often into the wrong
direction. If it doesn't, it's embedded PDAF.

Not necessarily. There's often a chromatic difference between the
edges of front focus blur and back focus blur which could be
exploited by CDAF.

Too weak, to hard to read, ... otherwise they'd be already
implementing it. And "often" is really fun with EVIL cameras.


If you speed up the lens movement (and you must, for CDAF),
PDAF speeds will improve likewise. If you speed up sensor
reading and/or increase low light sensitivity, the PDAF
sensors will improve just as well.

Not necessarily, because CDAF "sensors" are selected from the very
large array of image pixel sensors,

which means the individual pixels are small, and you *have*
to read the sensor reasonably often for EVFs to work. Which
means many pixels per second ... more noise, more heat.

and the speed problem of finding
them in the first place is unique to CDAF.

It comes down to reading pixels. PDAF also needs to read
pixels, just not of the main sensor, but of dedicated sensors.


One huge advantage of CDAF is focusing accuracy. There's no need to AF
calibration like there can be with PDAF since you're focusing right on
the sensor.

So add a 'calibrate PDAF'-function that pre-focuses with PDAF
(just for speed reasons), fine-focuses with CDAF (preferably
on a flat, orthogonal target), measures PDAF, checks again with
CDAF that the focus hasn't moved (else back to fine-focussing)
and thus gets the currently correct offsets for PDAF. Would be
a second with 1 sensor.

And for zooms which need different calibrations for each focal length?

You take the short end and the long end and interpolate.
Works very well if you have reasonable glass.

And lenses with aperture related focus drift which need calibration
for each aperture?

CDAF does NOT close the aperture to focus. So CDAF has the
same problem.

It's the nature of PDAF. however well calibrated,
to be an approximation based on simplifying assumptions. As megapixels
increase and lens quality improves yesterday's appropriate simplifying
assumptions become today's oversimplifications.

PDAF is *faster* and can be /reliably/, /repeatably/ *as
accurate* as CDAF.

Not to mention the problems of curved planes of focus...

Which affect CDAF just the same.


And on a tripod against a wall (or a newspaper on the wall) it'd
find all offsets within a minute or two --- with individually
microadjusting for this lens and focal length for every single
PDAF sensor. You'd need to do this only once or twice ...

PDAF can be as accurate as CDAF, just mucho faster.

When you don't care about the inherent residual inaccuracies.

Which have been *proven* to be controlable so that you can't
measure the differences between PDAF, CDAF and manual
focussing.

If you have an
old "nifty fifty" with spherical aberration do the best calibration
you can at f1.4, and then tell me how well it does at f5.6 with that
calibration. And did I mention curvature of the focus plane? :-)

Did I mention that these problems affect CDAF the identical
way they affect PDAF? Like 'focussing only on one point' and
'CDAF using wide open aperture'?

-Wolfgang

Laszlo Lebrun wrote:

The typical Rolleiflex angle of view has advantages, moreover you could
have e.g a tele and a wide lens at the same time and combine both views
electronically.
The human eye works that way: high resolution fovea and wide view for
the rest.

Have you ever noticed that photographs are supposed to be sharp
even outside a tiny view angle (like the one of the fovea)?

-Wolfgang

In article ,
says...

On 4/19/13 10:13 PM, philo wrote:
On 04/19/2013 02:44 PM, James Silverton wrote:
I enjoyed playing with a friend's much prized Rolleiflex film camera. Do
digital versions exist for cameras of this class? I was not able to
search one out.


No, there would be no need to do so.

I am convinced, we could get a huge potential out of a crossover of that
camera form with modern digital electronics and a really good electronic
retina-grade viewfinder behind flaps.
The typical Rolleiflex angle of view has advantages, moreover you could
have e.g a tele and a wide lens at the same time and combine both views
electronically.
The human eye works that way: high resolution fovea and wide view for
the rest.

I'd like to see some sample shots where that technique was used to good
effect before I could be convinced that such a camera would be more than
a niche product.

Many of the latest generation of DSLR allow waist-level shooting you
know.

On Wed, 8 May 2013 09:13:02 -0400, "J. Clarke"
wrote:

In article ,
says...

On 4/19/13 10:13 PM, philo wrote:
On 04/19/2013 02:44 PM, James Silverton wrote:
I enjoyed playing with a friend's much prized Rolleiflex film camera. Do
digital versions exist for cameras of this class? I was not able to
search one out.


No, there would be no need to do so.

I am convinced, we could get a huge potential out of a crossover of that
camera form with modern digital electronics and a really good electronic
retina-grade viewfinder behind flaps.
The typical Rolleiflex angle of view has advantages, moreover you could
have e.g a tele and a wide lens at the same time and combine both views
electronically.
The human eye works that way: high resolution fovea and wide view for
the rest.

I'd like to see some sample shots where that technique was used to good
effect before I could be convinced that such a camera would be more than
a niche product.

Many of the latest generation of DSLR allow waist-level shooting you
know.


I used to do that with my old Topcon.
--

Regards,

Eric Stevens

On 5/8/2013 6:12 PM, Eric Stevens wrote:
On Wed, 8 May 2013 09:13:02 -0400, "J. Clarke"
wrote:

In article ,
says...

On 4/19/13 10:13 PM, philo wrote:
On 04/19/2013 02:44 PM, James Silverton wrote:
I enjoyed playing with a friend's much prized Rolleiflex film camera. Do
digital versions exist for cameras of this class? I was not able to
search one out.


No, there would be no need to do so.

I am convinced, we could get a huge potential out of a crossover of that
camera form with modern digital electronics and a really good electronic
retina-grade viewfinder behind flaps.
The typical Rolleiflex angle of view has advantages, moreover you could
have e.g a tele and a wide lens at the same time and combine both views
electronically.
The human eye works that way: high resolution fovea and wide view for
the rest.

I'd like to see some sample shots where that technique was used to good
effect before I could be convinced that such a camera would be more than
a niche product.

Many of the latest generation of DSLR allow waist-level shooting you
know.


I used to do that with my old Topcon.

the problem is that the higher end models do not have an articulated lCD
that allows waist level viewing.

--
PeterN