Multi-shot aka pixel shift going mainstream

[Theodoros]

Wouldn't the mechanism to stabilize the larger sensor would be much more challenging?
Sure Pentax/Ricoh have the conceptual methods in practise (they have since the K20d), but my understanding is it wouldn't just be as simple as upsizing this?

But I would like this!

There shouldn't be any problem what so ever... The K1 is sensor is 2.35x the area of the K3 sensor... the 645Z one is only 1.65x the one of the K1...

[Bart_van_der_Wolf]

There shouldn't be any problem what so ever... The K1 is sensor is 2.35x the area of the K3 sensor... the 645Z one is only 1.65x the one of the K1...

So it has a much larger mass, and thus inertia, and it needs to move larger distances to accommodate for the longer focal lengths. Not exactly the same problem only larger, but a much larger challenge. The magnetic fields may interfere with the other electronics, and will require more power, and thus generate more temperature increase near the sensor.

I'm not saying its impossible, but it is a much harder nut to crack than what you make it sound like.

Cheers,
Bart

[alatreille]

Thanks Bart.

[eronald]

I'm not saying its impossible, but it is a much harder nut to crack than what you make it sound like.

Cheers,
Bart

Bart

Engineers solve difficult problems for a living.

Stabilised MF is coming, we can expect it from people who know how to do stab ie. Pentax/Ricoh, or Hassy who are now partnered with a gimbal maker.

Phase will do it if/when they have to - they won't like it any more than Hassy because they will have to redo the back/case design. Pentax doesn't care about the box, they iterate faster, and anyway they need to do a body redesign for larger sensors.

On the positive side, 6 (not 5) axis stabilisation would mean free multishot, AF with manual tech ultra-wides, mild shift, tilt, focus stacking and other goodies that pro product and interior photographers would love.

We should like this idea. Progress at last.

Edmund

[Theodoros]

 Pentax doesn't care about the box, they iterate faster, and anyway they need to do a body redesign for larger sensors.

Actually it is the other way around... Pentax body is much larger than is needed for larger sensors... It is as big, that it is nearly enough for 6x6 full frame sensor (if there ever was one...). It is so big that it is waisting space without a reason what so ever... So big that it is much tougher for their Engineers to design UWA lenses on it... A wrong (original -from film days) design IMO... They should have redesign the body when they decided to enter digital MF... It would be much better if they would have gone for a new (smaller) design and used adapters to retain the lens compatibility. 

[ErikKaffehr]

Hi Bart,

It would be entirely possible to pass 15 or 16 bit of data trough 14 bit wide channel using non linear coding. My understanding is it works perfectly well as long as shot noise dithers any skipped values.

It seems that Phase One really pulled that white rabbit out of that hat with the 100 MP back , even if I assume that much of the credit needs to go to Sony, as I presume that Sony is doing the ADC-stuff.

It would be hard to ignore the fact that the 100 MP sensor is much larger than a 24x36mm. I you take two sensors of similar quality and put equally good lenses in front the larger sensor will provide a much better image. That image quality may not be needed, but that is another issue.

Best regards
Erik

... but a camera and support system is more than a sensor at a low price. Besides, the Phase One system uses a 16-bit processing pipe-line which achieves something like more than 14.3 stops of engineering Dynamic Range at base ISO. That is unlikely to be achievable with the Pentax if they use less than internal 16-bit processing.

But the Pentax looks like a nice camera with interesting features. It will depend on the lens arsenal whether it gets major traction in the market, and how well their service organization is implemented if they want professionals to use the system.

Cheers,
Bart

[Theodoros]

Never the less, one can't deny that the O/P is right... Multishot does go commercial and even if its not as good as with a multishot MFDB at the moment, it will certainly improve with time to overcome the issues that are related with first offerings. That said, it shouldn't be difficult to achieve this, given that artefact related issues are eliminated by the process itself... IMO, the work/research that has to be done is more related with having advanced software that offers colour calibration methods up to the level that the dedicated software for MFDBs currently does....

[Bart_van_der_Wolf]

Bart

Engineers solve difficult problems for a living.

Hi Edmund,

That's only true if they get a budget to spend that time/effort. R&D and Manufacturing often take strange shortcuts from a quality point of view, usually because they are forced to do so by monetary or time constraints. To give an idea, a 14-bit processing pipeline is a few dollars cheaper than a higher quality 16-bit pipeline, yet 14-bit is the norm, even in equipment that costs thousands to produce.

Then the issues that are created have to solved from someone else's budget, e.g. firmware and test runs to reduce noise and calibration issues, designing heatsinks and ventilation (where perhaps multiple read-outs of the CMOS could have been averaged if enough precision was used to resolve the noise).

Stabilised MF is coming, we can expect it from people who know how to do stab ie. Pentax/Ricoh, or Hassy who are now partnered with a gimbal maker.

Phase will do it if/when they have to - they won't like it any more than Hassy because they will have to redo the back/case design.

Sure, it's a matter of time.

Pentax doesn't care about the box, they iterate faster, and anyway they need to do a body redesign for larger sensors.

I don't know, perhaps they can reuse a lot of the old film 645 body concepts.

On the positive side, 6 (not 5) axis stabilisation would mean free multishot, AF with manual tech ultra-wides, mild shift, tilt, focus stacking and other goodies that pro product and interior photographers would love.

We should like this idea. Progress at last.

No objection to that, it's just not as simple in practice as some want us to believe. Besides, as long as our subject matter is not stationary, subject movement remains problematic for multi-shot.

Cheers,
Bart

[ErikKaffehr]

Hi,

I don't agree with pipeline costs. Analogue digital conversion is now on the sensor and it is highly probably it is of ramp type converter. Very simple construction and I am pretty sure the precision can be increased.

Processors generally come in multiples of 8-bit sizes, so there are 8, 16, 32 and 64 bit processors. But I don't think todays processors can handle all image processing so they are with high probability combined with some special processing hardware and that can be 14 bits wide.

Until now, sensors were not good enough to generate more than 14 bits worth of data, but it seems that the new sensor on the IQ3-100 can do that. So Phase One adopted so they can use 16 bit data.

Certainly, going to 16 bits is associated with costs as it requires a generation of ASICs, but the cost is not in the extra two bits but in the redesign.

Or, it could be that that the data channel is actually sixteen bit wide, but only 14 bits are used as it is enough to carry all information. When Sony added uncompressed option the new file format is actually 16 bits, but just using 14 bits.

Best regards
Erik

Hi Edmund,

 To give an idea, a 14-bit processing pipeline is a few dollars cheaper than a higher quality 16-bit pipeline, yet 14-bit is the norm, even in equipment that costs thousands to produce.

[eronald]

Bart,
 
   A +1.7x MF surface sensor eg. Sony 33x44mm  = 1.68*24x36  is made with exactly the same wafer as the 24x36mm, so it's total weight will be *less* than 1.7x because the side connections will scale as square root of surface. I do agree that a magnetic suspension and voice coils are going to need to be heftier, but eg. Pentax has just upscaled its stabiliser to full-frame and there is no reason to think that going up another step won't now be a routine exercise. In fact one wonders whether Sony isn't going to integrate some of the necessary sensors and electronics in the sensor family - after all they now have several cameras themselves that are stabilised, others in the Olympus brand which they part-own, I believe, and sensor customers are adopting the tech too.

 In-body stabilisation really has become a critical feature like AF and somebody is going to make the move to bring it to MF. BTW, many years ago I tried to put down a tripod to take some twilight shots on the Champs Elysés in Paris. It was hopeless, the handheld images were much much better - traffic vibration was simply too strong. I know that some members of this forum dispense with AF, but many do love it, and will choose a 35mm camera with AF instead of an MF camera with manual focus for many jobs. And so they doubtless feel about stabiliisaton,and multishot. Frankly I cannot understand why multishot and even filter wheels are not standard on museum-grade repro setups.
 
Edmund

Hi Edmund,

That's only true if they get a budget to spend that time/effort. R&D and Manufacturing often take strange shortcuts from a quality point of view, usually because they are forced to do so by monetary or time constraints. To give an idea, a 14-bit processing pipeline is a few dollars cheaper than a higher quality 16-bit pipeline, yet 14-bit is the norm, even in equipment that costs thousands to produce.

Then the issues that are created have to solved from someone else's budget, e.g. firmware and test runs to reduce noise and calibration issues, designing heatsinks and ventilation (where perhaps multiple read-outs of the CMOS could have been averaged if enough precision was used to resolve the noise).

Sure, it's a matter of time.

I don't know, perhaps they can reuse a lot of the old film 645 body concepts.

No objection to that, it's just not as simple in practice as some want us to believe. Besides, as long as our subject matter is not stationary, subject movement remains problematic for multi-shot.

Cheers,
Bart

[Bart_van_der_Wolf]

A +1.7x MF surface sensor eg. Sony 33x44mm  = 1.68*24x36  is made with exactly the same wafer as the 24x36mm, so it's total weight will be *less* than 1.7x because the side connections will scale as square root of surface.

I get that, but don't forget it is a 3D issue, surface x thickness. The volume, and as a result the mass and inertia, increases a lot.

I do agree that a magnetic suspension and voice coils are going to need to be heftier, but eg. Pentax has just upscaled its stabiliser to full-frame and there is no reason to think that going up another step won't now be a routine exercise.

The point is that it is not just the chip that is set in motion, it's an entire assembly that is suspended in a magnetic field. Other than that, I do not know for which mass it was designed, so it may or may not require a new design.

In fact one wonders whether Sony isn't going to integrate some of the necessary sensors and electronics in the sensor family - after all they now have several cameras themselves that are stabilised, others in the Olympus brand which they part-own, I believe, and sensor customers are adopting the tech too.

I don't know how their sensor division is setup. Maybe the silicon fabrication part is a separate unit.

In-body stabilisation really has become a critical feature like AF and somebody is going to make the move to bring it to MF. BTW, many years ago I tried to put down a tripod to take some twilight shots on the Champs Elysés in Paris. It was hopeless, the handheld images were much much better - traffic vibration was simply too strong.

Yes, traffic induced vibration is hell (try shooting from a bridge (we have a lot of those in the Netherlands) with traffic and trams/buses/trucks passing), but other sources of vibration can also be a nuisance, like wind motion on the higher levels of high buildings. Try shooting natural light interiors with swinging lamps hanging from the ceiling..., no fun. Or production lines in manufacturing plants with heavy equipment and indoor transportation. Yes, image stabilization is a bliss if it works well, and it's almost a must with multishot.

Frankly I cannot understand why multishot and even filter wheels are not standard on museum-grade repro setups.

Yes, filter wheels offer a lot of flexibility, also for multi-spectral analysis.

Cheers,
Bart

[eronald]

I get that, but don't forget it is a 3D issue, surface x thickness. The volume, and as a result the mass and inertia, increases a lot.

The point is that it is not just the chip that is set in motion, it's an entire assembly that is suspended in a magnetic field. Other than that, I do not know for which mass it was designed, so it may or may not require a new design.
 
Cheers,
Bart

I'm not sure it's a 3D issue, as noted the chip is cut from the same wafer thickness, and the number and thus surface of edge connections scale roughly at root of surface. I see no reason the cover glass and cleaning assembly should be any different per unit square. Of course Phase and Hassy probably would have to abandon the cover glass mounted on the case because of tilt issues -tilt would change the optical path. Which might be the real reason why they haven't done it already - goodbye easily detachable dustproof back, hello back-protection cover slide.

Yes, this is counterintuitive.

Think of it as using 70mm film rather than 35mm. It can be the same substrate, and there is not necessarily much more surface area lost to perforations.

Come on, as engineers the most wonderful moment is when we notice that a problem is actually reasonably straightforward.

Edmund