Home · Register · Join Upload & Sell

Moderated by: Fred Miranda
Username  

  New fredmiranda.com Mobile Site
  New Feature: SMS Notification alert
  New Feature: Buy & Sell Watchlist
  

FM Forums | Leica & Alternative Gear | Join Upload & Sell

1       2       3       4              end
  

Leica Triple Resolution marketing vs reality. Another look.

  
 
highdesertmesa
Offline
• • • •
Upload & Sell: On
p.5 #1 · p.5 #1 · Leica Triple Resolution marketing vs reality. Another look.


Fred Miranda wrote:
Maybe it was a smart move from Leica's perspective. If the smaller DNG files had introduced artifacts, someone would have noticed and criticized the feature. As it stands, I believe it's a good alternative to editing in post-processing. I learned a lot from this thread. The take away is that downsampling of image files (effectively averaging out some of the noise) can improve the SNR and consequently contribute to a perceived improvement in DR. (as the reduction in noise makes subtle details more discernible)


Yes, since they didn't provide multiple options for the smaller DNG resample methods, then the method they chose makes the most sense.



Apr 08, 2024 at 04:19 PM
Fred Miranda
Offline
Admin
Upload & Sell: On
p.5 #2 · p.5 #2 · Leica Triple Resolution marketing vs reality. Another look.


highdesertmesa wrote:
Yes, since they didn't provide multiple options for the smaller DNG resample methods, then the method they chose makes the most sense.


I wonder how many Leica M11 shooters use the 18-megapixel option as their default resolution choice, opting for higher resolutions only when cropping or when capturing highly detailed images.



Apr 14, 2024 at 11:30 AM
highdesertmesa
Offline
• • • •
Upload & Sell: On
p.5 #3 · p.5 #3 · Leica Triple Resolution marketing vs reality. Another look.


Fred Miranda wrote:
I wonder how many Leica M11 shooters use the 18-megapixel option as their default resolution choice, opting for higher resolutions only when cropping or when capturing highly detailed images.


My guess would be zero. But S-DNG would be my preference if I did want to use the smaller DNGs. On the M11M, the S-DNGs look more like output from a native lower res sensor than the M-DNGs do. The S-DNGs seem more purposeful while the M-DNGs leaving me wanting.



Apr 14, 2024 at 12:06 PM
Fred Miranda
Offline
Admin
Upload & Sell: On
p.5 #4 · p.5 #4 · Leica Triple Resolution marketing vs reality. Another look.


highdesertmesa wrote:
My guess would be zero. But S-DNG would be my preference if I did want to use the smaller DNGs. On the M11M, the S-DNGs look more like output from a native lower res sensor than the M-DNGs do. The S-DNGs seem more purposeful while the M-DNGs leaving me wanting.


From previous threads, I think @zhangyue mainly shoots @18MP.



Apr 20, 2024 at 09:59 PM
thrice
Offline
• • • •
Upload & Sell: Off
p.5 #5 · p.5 #5 · Leica Triple Resolution marketing vs reality. Another look.


zhangyue wrote:
I donít see anything prevent binning in hardware level. Itís the cost vs benefit make it no go comparing to digital/software way because modern processes are more powerful and cheaper.



The fact that the sensor produces a digital output at a pixel level is not preventing them combining the analogue output of multiple pixels?

Again (I feel like a bit of a broken record) there is no way to do hardware binning on a CMOS sensor as they exist today due to architectural limitations.
If they redesigned them to allow pixel binning they'd likely sacrifice the lower read noise at native resolution and end up with something a lot like a CCD sensor.



Apr 24, 2024 at 07:54 PM
zhangyue
Offline
• • • • •
Upload & Sell: On
p.5 #6 · p.5 #6 · Leica Triple Resolution marketing vs reality. Another look.


thrice wrote:
The fact that the sensor produces a digital output at a pixel level is not preventing them combining the analogue output of multiple pixels?

Again (I feel like a bit of a broken record) there is no way to do hardware binning on a CMOS sensor as they exist today due to architectural limitations.
If they redesigned them to allow pixel binning they'd likely sacrifice the lower read noise at native resolution and end up with something a lot like a CCD sensor.



CMOS dominate now is mainly cost and speed. CCD pixel binning happen in CHARGE domain before converting to Voltage and further convert to digital. In theory, CMOS should be able to do the same to do binning in voltage domain (summing amplifier) before converting to digital. However I believe first the cost is prohibitive make it not attractive. (At least this is one of reason consumer industry move away from CCD to CMOS. Even CCD is slow read out for whole sensor, CCD actually is still holding advantage on read noise over CMOS) 2nd, I am not sure it bring much benefit over doing binning in digital domain.

Digital domain binning is much preferred these days as it is scalable and can utilize modern process power and very flexible (can be downsize to any resolution) We are utilize the feature daily. for example: 100M file at ISO25600 can be very clean if we display them on pad or phone.



Apr 26, 2024 at 02:00 AM
thrice
Offline
• • • •
Upload & Sell: Off
p.5 #7 · p.5 #7 · Leica Triple Resolution marketing vs reality. Another look.


You literally just repeated what I said. Pixel binning on CMOS occurs after readout, with read noise already introduced. It is the same as averaging pixels in software. To do hardware binning requires binning before readout - analogue.

I'm not sure what you don't understand, but it seems perhaps we're making the same point but you don't realise it?

To do binning before readout and amplification on CMOS wouldn't be possible due to the architecture of CMOS. It would require a fundamentally different design very similar to CCD. If it were feasible and better then it would be the more mature and dominant process. CMOS refinement has allowed it to surpass CCD for almost all purposes.

Can you explain to me the difference between CMOS and CCD other than the read (A2D)/amp differences in the architecture?

zhangyue wrote:
CMOS dominate now is mainly cost and speed. CCD pixel binning happen in CHARGE domain before converting to Voltage and further convert to digital. In theory, CMOS should be able to do the same to do binning in voltage domain (summing amplifier) before converting to digital. However I believe first the cost is prohibitive make it not attractive. (At least this is one of reason consumer industry move away from CCD to CMOS. Even CCD is slow read out for whole sensor, CCD actually is still holding advantage on read noise over CMOS) 2nd, I am not sure it bring
...Show more



Apr 26, 2024 at 02:42 AM
zhangyue
Offline
• • • • •
Upload & Sell: On
p.5 #8 · p.5 #8 · Leica Triple Resolution marketing vs reality. Another look.


OK Let start from beginning

thrice wrote:
The fact that the sensor produces a digital output at a pixel level is not preventing them combining the analogue output of multiple pixels?


Can you elaborate this?

thrice wrote:
there is no way to do hardware binning on a CMOS sensor as they exist today due to architectural limitations.


What you mean no way? This is the whole point I am trying to make here. We are talking about hardware binning vs software binning. CCD combine X pixel charge and covert to "1" voltage and follow ADC. CMOS can combine X pixel Voltage to "1" voltage and follow ADC. IF the term of hardware binning mean binning in analog domain. I don't see "no way" or "impossible".

If you tell me that CMOS sensor means every single pixel has its own dedicate ADC build in, and output is in digital domain. OK, I will buy this argument.

thrice wrote:
To do binning before readout and amplification on CMOS wouldn't be possible due to the architecture of CMOS. It would require a fundamentally different design very similar to CCD. If it were feasible and better then it would be the more mature and dominant process. CMOS refinement has allowed it to surpass CCD for almost all purposes.

Why? and require what fundamental design?is that means it is possible now?

thrice wrote:
Can you explain to me the difference between CMOS and CCD other than the read (A2D)/amp differences in the architecture?


Which part you want to know? fundamental operation of CCD and CMOS is so straight forward. it is so similar to memory design or display panel operation. I don't see any challenge here. I'd like to understand which part make you think it is impossible to do binning at hardware level for CMOS?

I don't want to be rude but just want to match the tone of your reply here. I have a feeling you don't have knowledge depth here. If you want technical discuss, Id happy to do so.



Apr 26, 2024 at 03:24 AM
thrice
Offline
• • • •
Upload & Sell: Off
p.5 #9 · p.5 #9 · Leica Triple Resolution marketing vs reality. Another look.


zhangyue wrote:

If you tell me that CMOS sensor means every single pixel has its own dedicate ADC build in, and output is in digital domain. OK, I will buy this argument.


This is the fundamental difference between CMOS sensors and CCD sensors. This is the only point I have been trying to make. Not necessarily each pixel but certainly each column on modern sensors.
Edit: probably more importantly, per pixel amplifiers, rather than column amps on CCD.

zhangyue wrote:

I'd like to understand which part make you think it is impossible to do binning at hardware level for CMOS?


On a CCD sensor the analogue charge is summed for adjacent pixels then sent to the amplifier and ADC, that is what is considered hardware binning as all 4 pixels (on monochrome sensor in the most simplistic example) behave physically like one. Binning occurs in the charge domain.

On a CMOS sensor the pixel has its own amplifier and on modern BSI sensors each column its own ADC. It is possible to have a CMOS sensor without the ADC per-column but that leads to issues like amp glow (generic term caused by heat generating components on the periphery of the sensor) and additional noise. If you move the amplifier to no longer be at a pixel level then congratulations you've just made a CCD.

If the sensor does not have per-column ADC you can bin in the voltage domain but that is not considered hardware binning and includes read noise.
Sony Exmor R sensors like the IMX455 have CDS (correlated double sampler - reads twice and averages read noise) at the charge level to reduce analogue noise at a column level which also makes any kind of hardware (or voltage domain software) binning impossible.
They also have per-column ADC and a second digital domain CDS to further reduce noise. For these reasons any Exmor or Exmor R or Exmor RS based sensor cannot do voltage domain binning.

The design is so intrinsically locked to column (and in future per pixel) level ADC I doubt any kind of hardware binning is in our future for a CMOS sensor that performs with any kind of competitive noise level.

zhangyue wrote:

I don't want to be rude but just want to match the tone of your reply here. I have a feeling you don't have knowledge depth here. If you want technical discuss, Id happy to do so.


Don't feel rude, I welcome the response and look forward to more.

Edited on Apr 26, 2024 at 06:10 PM · View previous versions



Apr 26, 2024 at 04:46 AM
 


Search in Used Dept. 

LBJ2
Offline
• • • • •
Upload & Sell: On
p.5 #10 · p.5 #10 · Leica Triple Resolution marketing vs reality. Another look.


FWIW, I think many have already seen these results, I reviewed what some have been recommending to me. ReidReviews SL3: Studio Tests comparison @ ISO 800-6400 noise results between SL3 DNG-L vs SL3-DNG-M and SL3-DNG-L resized -M vs DNG-M.

The ReidReviews results @ ISO 800-6400 while a very different comparison process/method, seem to align with Mathphotographer's cleaner DNG-M results. Which only encourages me, to test again for myself.



Apr 26, 2024 at 11:56 AM
zhangyue
Offline
• • • • •
Upload & Sell: On
p.5 #11 · p.5 #11 · Leica Triple Resolution marketing vs reality. Another look.


The argument is really can CMOS do hardware pixel binning? Not if Sony CMOS sensor can do hardware binning AS IS?

Use this logic: M9 and S006 CCD sensor can't do pixel binning doesn't mean CCD can't do binning.

CCD and CMOS ultimately both covert light to code. CCD shift charge through shift register and CMOS shift voltage through Row and Column driver. binning can happen at charge level or voltage level before ADC.

thrice wrote:
If the sensor does not have per-column ADC you can bin in the voltage domain but that is not considered hardware binning and includes read noise.


I think this is only one initiate whole arguments. In my mind, any binning happens before ADC consider hardware binning. That is the point of all last couple of replies I was trying to make.

thrice wrote:
The design is so intrinsically locked to column (and in future per pixel) level ADC I doubt any kind of hardware binning is in our future for a CMOS sensor that performs with any kind of competitive noise level.


I am not disagreeing this. The point is really can CMOS do hardware binning not does it make financial sense to do hardware binning. Industry has already made this decision long time ago that pretty much all binning will be in digital domain.



Apr 26, 2024 at 05:44 PM
RustyBug
Offline
• • • • • • •
Upload & Sell: On
p.5 #12 · p.5 #12 · Leica Triple Resolution marketing vs reality. Another look.


zhangyue wrote:
Industry has already made this decision long time ago that pretty much all binning will be in digital domain.


Any chance that Medium Format is different ... you know, those super expensive Phase models, etc.



Apr 26, 2024 at 06:23 PM
thrice
Offline
• • • •
Upload & Sell: Off
p.5 #13 · p.5 #13 · Leica Triple Resolution marketing vs reality. Another look.


zhangyue wrote:
I think this is only one initiate whole arguments. In my mind, any binning happens before ADC consider hardware binning. That is the point of all last couple of replies I was trying to make.


If you're asserting that voltage domain binning is hardware binning then that's fine, you're entitled to that opinion, but that is not the definition accepted by industry and includes read noise which is the primary source of noise in modern CMOS sensors unless you're shooting long exposures at very high gain.

This means voltage domain binning is far less effective than charge domain (hardware) binning.

No CMOS sensor can do charge domain binning.

Okay let's move on.



Apr 28, 2024 at 02:46 AM
thrice
Offline
• • • •
Upload & Sell: Off
p.5 #14 · p.5 #14 · Leica Triple Resolution marketing vs reality. Another look.


Plenty of CCD cameras still exist, especially in astronomy, machine vision and scientific research. These often allow hardware binning.

Phase One IQ160, IQ180, IQ260 and IQ280 all allow hardware (charge domain) binning on their CCD sensors. They call it Sensor+. I found a post by Doug Peterson where he says it is some proprietary trick not downsampling but I think he means true hardware binning as he says it's wired into the sensor and if you read the Teledyne Dalsa white papers the sensors have binning modes.

Funny the comment about the M9 earlier, the KAF-18500 (M9) and KAF-10500 (M8) developed specifically for Leica have no hardware binning (the sensor doesn't have multiple binning modes). However, the Kodak KAF8300 CCD used in the Olympus E500 doesn't offer binning in the Olympus but the sensor is used widely in astrophotography and has multiple hardware (charge domain) binning modes.

RustyBug wrote:
Any chance that Medium Format is different ... you know, those super expensive Phase models, etc.




Apr 28, 2024 at 03:03 AM
RustyBug
Offline
• • • • • • •
Upload & Sell: On
p.5 #15 · p.5 #15 · Leica Triple Resolution marketing vs reality. Another look.


thrice wrote:
Plenty of CCD cameras still exist, especially in astronomy, machine vision and scientific research. These often allow hardware binning.

Phase One IQ160, IQ180, IQ260 and IQ280 all allow hardware (charge domain) binning on their CCD sensors. They call it Sensor+. I found a post by Doug Peterson where he says it is some proprietary trick not downsampling but I think he means true hardware binning as he says it's wired into the sensor and if you read the Teledyne Dalsa white papers the sensors have binning modes.

Funny the comment about the M9 earlier, the KAF-18500 (M9) and KAF-10500 (M8) developed specifically
...Show more

Interesting.

Reminds me of my Kodak SLR/C. It had some unique kind of program shift via ISO / exposure times. Makes me wonder if it was related. That camera ... "acted" like CCD, yet all research / literature into it that I could find, suggested it was actually CMOS, not CCD. It was super clean at base ISO 160, 320 was okay, 640 tolerable (only) in good light. A noise monster, after that.

Pretty much a one trick pony at base ISO ... and even though it went down to an ISO of 6, I never really got a good grip on using the lower ISO ranges. They had a restricted / limited exposure time associated with them. Kind of a pre-LENR thing I guess. If I was a betting man, I'd say it was CCD (even if it wasn't). IIRC, it may have been a sensor similar to one of the earlier Leica cameras (I forget).

I mention it because it makes me wonder what / if that was a binning approach they were implementing.



Apr 28, 2024 at 05:07 AM
thrice
Offline
• • • •
Upload & Sell: Off
p.5 #16 · p.5 #16 · Leica Triple Resolution marketing vs reality. Another look.


I remember the SLR/C being spoken about in threads like the fat pixel mfdb thread on getdpi.

Unfortunately I know nothing about it's sensor and Eastman Kodak don't seem to have published anything.

It does seem to be a CMOS sensor with true ISO/gain instead of being ISO invariant like the Sony sensors. Very interesting.

With the M9 and my phase one p45 (same pixel arch) CCDs performance was impeccable at base iso but fell apart higher. Oddly enough the M9 did better than the P45 at a pixel level with high ISO.

I still feel CCD at base iso has a purity and lack of noise that is hard to describe versus CMOS, but a lot of that is probably my memory and not borne out in comparison to the latest CMOS sensors.

RustyBug wrote:
Interesting.

Reminds me of my Kodak SLR/C. It had some unique kind of program shift via ISO / exposure times. Makes me wonder if it was related. That camera ... "acted" like CCD, yet all research / literature into it that I could find, suggested it was actually CMOS, not CCD. It was super clean at base ISO 160, 320 was okay, 640 tolerable (only) in good light. A noise monster, after that.

Pretty much a one trick pony at base ISO ... and even though it went down to an ISO of 6, I never really got a good grip on
...Show more



Apr 29, 2024 at 09:49 PM
RustyBug
Offline
• • • • • • •
Upload & Sell: On
p.5 #17 · p.5 #17 · Leica Triple Resolution marketing vs reality. Another look.


thrice wrote:
I still feel CCD at base iso has a purity and lack of noise that is hard to describe versus CMOS, but a lot of that is probably my memory and not borne out in comparison to the latest CMOS sensors.


I'm with ya ... but, given the availability of CCD sensors in contemporary cameras (of interest), not a lot of CCD options to use in practical terms. That, and the "gap" diff has narrowed, but my (historical) appreciation for CCD remains, even if I'm shooting CMOS mostly these days.



Apr 29, 2024 at 10:50 PM
thrice
Offline
• • • •
Upload & Sell: Off
p.5 #18 · p.5 #18 · Leica Triple Resolution marketing vs reality. Another look.


RustyBug wrote:
I'm with ya ... but, given the availability of CCD sensors in contemporary cameras (of interest), not a lot of CCD options to use in practical terms. That, and the "gap" diff has narrowed, but my (historical) appreciation for CCD remains, even if I'm shooting CMOS mostly these days.


It is true the CMOS typically has a higher noise floor. This is due to per-pixel amplifier noise, more amps = more noise. This is a random noise in the signal which cannot be subtracted out.

As CMOS technology advances it will become less and less of an issue. However, even with current state of the art like the IMX455, I believe a cooled CCD will have less noise than a cooled CMOS at low gain settings, but also typically lower quantum efficiency (function of pixel architecture and CFA). In this example there is no CFA and the whole optical train and exposure are identical:






http://www.atscope.com.au/BRO/tutorials/QHY600.html

The CCD sensor in question is the Kodak KAF-16803 (9 micron pixels introduced in 2006) and the CMOS is the Sony IMX455 used in all current ~60Mp full frame cameras.

In the interest of transparency, the above image has dark frames and bias frames subtracted and flat field calibration. I am actually not sure which one has less noise, as the IMX455 on the right has a lower black point and higher sensitivity (QE) so it is picking up more small stars which give the illusion of noise.



May 06, 2024 at 09:30 PM
1       2       3       4              end






FM Forums | Leica & Alternative Gear | Join Upload & Sell

1       2       3       4              end
    
 

You are not logged in. Login or Register

Username       Or Reset password



This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.