Has the analogue processing stage of digital image capture changed in recent cameras ??

[Luca Mancini]

So I was reading the book "Manual of Photography" and there's a small part where they talk about the analog processing stage:

The analogue signals from the sensor are f i rst processed in an analogue pre-processor, which performs various func-tions, depending upon the sensor. These may include correlated double sampling (CDS) to suppress reset noise and to reduce f i xed pattern noise, automatic gain control, and generation of a reference black level for the reproduced images. The reference black is obtained from shaded pixels on the sensor, which produce a signal as a result of ther-mally generated dark current. This generated signal is then subtracted from the signals from the pixels in the sensing area of the array. Black level generation may be performed for each colour separately. Additionally, white balance may be performed on the analogue signal by using different gain settings for different colour channels, although this is most commonly performed on the digitized signal. The analogue front end may also control the operation of the image sensor by generating timing pulses. The signal is then digitized in the analogue-to-digital converter (ADC).

Considering the book was published in 2011, I'm wondering if all these still hold true or are even more jobs performed at the analogue stage or vice versa in recent cameras ??

 

[JimKasson]

So I was reading the book "Manual of Photography" and there's a small part where they talk about the analog processing stage:

The analogue signals from the sensor are f i rst processed in an analogue pre-processor, which performs various func-tions, depending upon the sensor. These may include correlated double sampling (CDS) to suppress reset noise and to reduce f i xed pattern noise, automatic gain control, and generation of a reference black level for the reproduced images. The reference black is obtained from shaded pixels on the sensor, which produce a signal as a result of ther-mally generated dark current. This generated signal is then subtracted from the signals from the pixels in the sensing area of the array. Black level generation may be performed for each colour separately. Additionally, white balance may be performed on the analogue signal by using different gain settings for different colour channels, although this is most commonly performed on the digitized signal. The analogue front end may also control the operation of the image sensor by generating timing pulses. The signal is then digitized in the analogue-to-digital converter (ADC).

Considering the book was published in 2011, I'm wondering if all these still hold true or are even more jobs performed at the analogue stage or vice versa in recent cameras ??

I'm assuming you're talking about raw filles. These days, black level subtraction is done in post. Dual conversion gain is now common; it was rare or nonexistent in 2011. White balance does not typically affect raw values, and is usually performed in post, using metadata from the camera for guidance. I don't know of still cameras that use ACG, except for auto-ISO.

The analog processing usually goes something like this: conversion gain, buffering, multiplexing, programmable gain amplifier, column ADC. CDS is likely used as well.

Eric Fossum is the best person here for more detail, I think.

 

[Detail Man]

... assuming you're talking about raw files. These days, black level subtraction is done in post. Dual conversion gain is now common; it was rare or nonexistent in 2011. White balance does not typically affect raw values, and is usually performed in post, using metadata from the camera for guidance. I don't know of still cameras that use ACG, except for auto-ISO.

The analog processing usually goes something like this: conversion gain, buffering, multiplexing, programmable gain amplifier, column ADC. CDS is likely used as well.

Eric Fossum is the best person here for more detail, I think.

Regarding further internal image-processing topologies, (the venerable) Iliah Borg wrote:

Apply WB; run 1st stage NR and sharpening; perform demosaicking; convert from camera colour to XYZ; apply 2nd stage of sharpening on Y; convert to LCH; apply tone curve to L; apply saturation to C; apply "look" to H; apply 2nd stage of NR over L and C; smooth H; apply convert to output space (like sRGB or AdobeRGB).

.

It is possible to sharpen raw data, and it is possible to select demosaicking algorithm and demosaicking parameters based on sharpening setting. Pre-sharpening is used sometimes in camera, but only for out of camera JPEGs, and not based on sharpening settings. I never saw any variations in in-camera demosaicking algorithms based on sharpening settings.

 

[Eric Fossum]

So I was reading the book "Manual of Photography" and there's a small part where they talk about the analog processing stage:

The analogue signals from the sensor are f i rst processed in an analogue pre-processor, which performs various func-tions, depending upon the sensor. These may include correlated double sampling (CDS) to suppress reset noise and to reduce f i xed pattern noise, automatic gain control, and generation of a reference black level for the reproduced images. The reference black is obtained from shaded pixels on the sensor, which produce a signal as a result of ther-mally generated dark current. This generated signal is then subtracted from the signals from the pixels in the sensing area of the array. Black level generation may be performed for each colour separately. Additionally, white balance may be performed on the analogue signal by using different gain settings for different colour channels, although this is most commonly performed on the digitized signal. The analogue front end may also control the operation of the image sensor by generating timing pulses. The signal is then digitized in the analogue-to-digital converter (ADC).

Considering the book was published in 2011, I'm wondering if all these still hold true or are even more jobs performed at the analogue stage or vice versa in recent cameras ??

I'm assuming you're talking about raw filles. These days, black level subtraction is done in post. Dual conversion gain is now common; it was rare or nonexistent in 2011. White balance does not typically affect raw values, and is usually performed in post, using metadata from the camera for guidance. I don't know of still cameras that use ACG, except for auto-ISO.

The analog processing usually goes something like this: conversion gain, buffering, multiplexing, programmable gain amplifier, column ADC. CDS is likely used as well.

Eric Fossum is the best person here for more detail, I think.

The general trend is to push towards all digital as soon as possible in the signal processing chain. What Jim says above is still true although CDS and CMS can be performed digitally (after the ADC). Architecturally, with 3D stacked BSI processes, column-parallel is going towards cluster-parallel or pixel-parallel. The photography-sensor (big die, big pixels) area is slower to adopt this because of cost and less compelling design considerations, compared to cell phone cameras.

For 20+ years, I have been thinking about going digital right at the single electron level (photons and electrons are already quantized so why not?) In the past 5+ years there has been increasing activity and technology development in this area with CIS-QIS and SPAD-QIS photon-counting image sensors.

 

[Eric Fossum]

Apply WB; run 1st stage NR and sharpening; perform demosaicking; convert from camera colour to XYZ; apply 2nd stage of sharpening on Y; convert to LCH; apply tone curve to L; apply saturation to C; apply "look" to H; apply 2nd stage of NR over L and C; smooth H; apply convert to output space (like sRGB or AdobeRGB).

.

It is possible to sharpen raw data, and it is possible to select demosaicking algorithm and demosaicking parameters based on sharpening setting. Pre-sharpening is used sometimes in camera, but only for out of camera JPEGs, and not based on sharpening settings. I never saw any variations in in-camera demosaicking algorithms based on sharpening settings.

I think all or almost all of these processes are done in the digital domain these days.

 

[Iliah Borg]

These days, black level subtraction is done in post.

For lossless formats.

 

[Detail Man]

Regarding further internal image-processing topologies, (the venerable) Iliah Borg wrote:

Apply WB; run 1st stage NR and sharpening; perform demosaicking; convert from camera colour to XYZ; apply 2nd stage of sharpening on Y; convert to LCH; apply tone curve to L; apply saturation to C; apply "look" to H; apply 2nd stage of NR over L and C; smooth H; apply convert to output space (like sRGB or AdobeRGB).

.

It is possible to sharpen raw data, and it is possible to select demosaicking algorithm and demosaicking parameters based on sharpening setting. Pre-sharpening is used sometimes in camera, but only for out of camera JPEGs, and not based on sharpening settings. I never saw any variations in in-camera demosaicking algorithms based on sharpening settings.

I think all or almost all of these processes are done in the digital domain these days.

Indeed. Hard to imagine stuffing analog circuit implementations of such (generalized) signal processing operations into such physical micro-spaces within which designs are developed - not to mention limiting challenges surrounding use of discrete test-points in development.

 

[JimKasson]

These days, black level subtraction is done in post.

For lossless formats.

And some lossy ones, like Sony's craw.

 

[Iliah Borg]

These days, black level subtraction is done in post.

For lossless formats.

And some lossy ones, like Sony's craw.

I was meaning to say that preserving black level "in raw" is one of the characteristics of lossless formats.

With cRAW and many other pseudo-raw formats, IMHO it depends on how we define black level; in other words, is a sum of constant bias and black level (the latter can be partially or even fully subtracted beforehand - leaving us with just bias, or some form of sandcastle) is black level?

 

[jiberlin]

So I was reading the book "Manual of Photography" and there's a small part where they talk about the analog processing stage:

...

Considering the book was published in 2011, I'm wondering if all these still hold true or are even more jobs performed at the analogue stage or vice versa in recent cameras ??

To my knowledge, dual-gain sensors is the only change in analog processing in the past. And possible the remove of analog gain switching caused by better ADCs.

 

[jogging]

I have a question about stacking. For the 3D stacked BSI sensor, the circuit for logic processing is placed below a BSI image sensor die. The circuit will generate heat and increase the temperature of BSI image sensor. Will this increase dark current?

 

[Eric Fossum]

I have a question about stacking. For the 3D stacked BSI sensor, the circuit for logic processing is placed below a BSI image sensor die. The circuit will generate heat and increase the temperature of BSI image sensor. Will this increase dark current?

Yes, and yes. But this is a matter of degree.

Generally the power dissipation is kept low with exactly this in mind, and does limit what one can perform in a 3-D stacked architecture unless there is cooling, passive or active, included in the package.