RDF on the reason for the lack of Phase Detection

[SrMi]

Advertisement (gone after registration)

Red Dot Forum had another camera talk on YT, this time about Q3 Monochrom. They also explain the reason for the lack of PDAF (@44:42).

(Note that the video contains a transcript that you can peruse to find the most interesting parts to listen to)

 

[stbeyer]

I really like RDF, but I think they are speculating here. 

I am also speculating, but I genuinely believe that Leica just buy this sensor: https://www.sony-semicon.com/files/62/pdf/p-13_IMX455AQK_BQK_ALK_Flyer.pdf

Sony simply does not produce a monochrome variant with pdaf pixels. End of story. 

 

[SrMi]

19 minutes ago, stbeyer said:

I really like RDF, but I think they are speculating here. 

I am also speculating, but I genuinely believe that Leica just buy this sensor: https://www.sony-semicon.com/files/62/pdf/p-13_IMX455AQK_BQK_ALK_Flyer.pdf

Sony simply does not produce a monochrome variant with pdaf pixels. End of story. 

 

Both @jonoslack and RDF got the same information from Leica. This is not speculation.

Sony does not publish spec sheets for customers' sensor designs. The fact that there is no spec sheet with monochrome and PDAFs does not mean that Sony cannot make them. If you listen to the RDF explanation of how OSPDAF is implemented, you will understand that it is possible to implement OSPDAF using a monochrome sensor.

P.S.: X2D and GFX100 use the "same" sensor, but different OSPDAF implementations (number of rows and row locations).

 

Edited 17 hours ago by SrMi

[stbeyer]

 

1 hour ago, SrMi said:

If you listen to the RDF explanation of how OSPDAF is implemented, you will understand that it is possible to implement OSPDAF using a monochrome sensor.

I listened to it, but it does not sound coherent to me. It does not make sense to me that you can interpolate away a Bayer color array but not the PDAF pixels, of which you know the exact coordinates. 

It may be true that a custom sensor could have been sourced from Sony, but the fact that Leica uses the exact 3 configuration offered in the document I linked suggests to me that they are constrained to the standard models for some reason. Maybe cost for low production volumes, for technical contrasints on Sony's side. I don't know. But RDF's explanation does not make sense to me. 

Edited 16 hours ago by stbeyer

[SrMi]

40 minutes ago, stbeyer said:

I listened to it, but it does not sound coherent to me. It does not make sense to me that you can interpolate away a Bayer color array but not the PDAF pixels, of which you know the exact coordinates. 

Maybe @jonoslack's post is clearer:

I had some very detailed discussions about this with Leica - PDAF is fine with Monochrom sensors - it will work perfectly well (just the same). The problem is dealing with the missing information - colour sensors can deal with it in conjunction with the demosaicing. I suggested to them that they could manage it with the 'defect list' for dead/ malfunctioning pixels (which both colour and monochrom sensors have) .. but this list has limited memory, and cannot deal with the number of PDAF sensors . . . . . . of course that memory could be changed, but then we are looking at real money - and actually the focusing on the Q3M is fine anyway! (unless you after birds in flight)

40 minutes ago, stbeyer said:

But RDF's explanation does not make sense to me. 

It makes sense once you understand how OSPDAF is implemented and that the sensors in the specs are delivered with a differing microlens layer.

Edited 16 hours ago by SrMi

[jonoslack]

1 hour ago, stbeyer said:

 

I listened to it, but it does not sound coherent to me. It does not make sense to me that you can interpolate away a Bayer color array but not the PDAF pixels, of which you know the exact coordinates. 

It may be true that a custom sensor could have been sourced from Sony, but the fact that Leica uses the exact 3 configuration offered in the document I linked suggests to me that they are constrained to the standard models for some reason. Maybe cost for low production volumes, for technical contrasints on Sony's side. I don't know. But RDF's explanation does not make sense to me. 

Hi there I spent lots of time trying to get my head around this with Leica. 

With a colour sensor you can (with work) include the references of the PDAF Pixels in the demosaicing. Okay

A Monochrom sensor has no demosaicing - so you can't deal with them like that! you could include them in the 'defect' list (which monochrom sensors have) but the memory allocated to defect pixels is not enough to accommodate the 100s of PDAF Pixels . . . . which means that to do it this way you need to change memory allocations on the sensor and do the programming to match it - possible but very expensive. . . . . . . . and anyway the Q3M AF seems very good!