The GetDPI Photography Forum

Great to see you here. Join our insightful photographic forum today and start tapping into a huge wealth of photographic knowledge. Completing our simple registration process will allow you to gain access to exclusive content, add your own topics and posts, share your work and connect with other members through your own private inbox! And don’t forget to say hi!

Why is (small) MF color/IQ better than FF?

Paratom

Well-known member
Maybe I need to work harder on the profiles, but how to do this? And we read it so often that its much easier o work with MF-files in post processing. Is this a myth? (I dont think it is)
 

MGrayson

Subscriber and Workshop Member
Maybe I need to work harder on the profiles, but how to do this? And we read it so often that its much easier o work with MF-files in post processing. Is this a myth? (I dont think it is)
I don't think the solution is to come up with your own profiles (that IS hard), but to be fearless about changing the individual color response, (saturation, hue, brightness). Play aggressively with all parameters (for instance, moving vibrance and saturation in opposite directions), not to produce outlandish results, but to understand the space of possibilities. Somewhere, you find the one you were looking for. The malleability of MF files, at least as they come OOC, lets you explore this space with plenty of room. You don't have to first undo the manufacturers own decisions, doing damage to the data in the process.

Or you can accept that Phase One made all the correct decisions for YOU and just go with that :D Actually, I agree that Phase One makes excellent choices, and if they made a system I could carry around all day, I'd buy it in a second. But I object to the notion that ANY one system is better than any other.

Matt
 
Last edited:

MGrayson

Subscriber and Workshop Member
While the SL images are fine the images from the S seem to offer (sometimes) better skin tone and its easier to work with them. If we think that pixel density is not far away, then it seems its not just size of sensor or size of pixel, but other things. And Leica doesnt offer for either camera native software.

I am not starting this discussion to fight FF vs MF, I started it because I really wonder why cant they put FF sensors with MF-sensor-quality in FF bodies for those people, who are fine with a little less resolution but want the color and tonality of MF?
The SL pixels are EXACTLY the same size as the S. It is, indeed, a crop of the sensor. The lenses make a difference. In M language, the SL lenses have a Summicron look and the S lenses more Summilux (well, about midway between the two). The test I did was SL, S006, S007, and X1D. They could all produce the same portrait with some messing around, but the lenses (SL 24-90, S120, and X1D 90, all shot at f/4) had different OOC looks.

I do not have an SL-S adapter, or I could take that variable out of the equation.

(This is, of course, my opinion - not fact.)

Matt
 

ErikKaffehr

Well-known member
While the SL images are fine the images from the S seem to offer (sometimes) better skin tone and its easier to work with them. If we think that pixel density is not far away, then it seems its not just size of sensor or size of pixel, but other things. And Leica doesnt offer for either camera native software.

I am not starting this discussion to fight FF vs MF, I started it because I really wonder why cant they put FF sensors with MF-sensor-quality in FF bodies for those people, who are fine with a little less resolution but want the color and tonality of MF?
Hi,

It would help if you supplied a little more information:

  • What gear are you comparing?
  • What raw converter you are using?
  • Which profiles are you using?
  • Which RGB are you evaluating in?

I shared this image shot with Phase One P45+, Sony Alpha 900 and Sony A7rII a while ago:
Tricolore2.jpg

  • Would be interesting if you see a difference?
  • Which one you prefer?

All these images were processed using my own DCP profiles in Lightroom.

Best regards
Erik
 

dougpeterson

Workshop Member
  • Would be interesting if you see a difference?
  • Which one you prefer?

All these images were processed using my own DCP profiles in Lightroom.
I think that generally when the question is subjectively pleasing color this test misses the mark in at least four ways:
- A contrived lab scene/lighting rather than a creative/interesting scene
- Using non native raw processor (LR) instead of the one where the hardware/software are coordinated (C1+P1)
- Creating lowest-common-denominator computer-generated profile that seek to standardize the response of the color
- Ignoring that different cameras influence the behavior of the photographer

In short: rather than have each look its individual best, your test seeks to make them all look the same – you more-or-less succeeded.

It's like trying to decide whom you prefer of Bach, Queen, Tchaikovsky, and John Lee Hooker by playing their music through a heartless MIDI synthesizer, exactly-as-written, in the same room. It ignores that each artist incorporated the instruments of their era, improvised from the written music, and expected to perform in different venues. It ignores all the things that make music music rather than sound.

I totally get the technical/science-oriented approach of reducing all the variables. But at some point you abstract so far away from the original question that the results are meaningless.

Much less scientific, but much more practically useful, in my opinion, is to simply to take a few hours (minimum) with each camera, making images the way you would if you owned that camera. The way you handle the camera may be different, where you stand in the scene may be different, the software you use may be different, the way you work up the files may be different. Even the way you print or present the work may be different. With so many confounding variables you certainly won't be scientifically exploring the percentage of correlation ascribed to a specific variable, but you will be answering the questions "which camera better helps me execute my vision?" and "which camera do I enjoy using the most?" which is, for 99.9% of photographers, the salient questions.
 

JoelM

Well-known member
Also, if you had used the same lens for all three, that would give more credit to the "experiment".

Joel
 

ErikKaffehr

Well-known member
Hi Joel,

I would say that the experiment is pretty accurate.

If a significant difference is noted, it would be a good question if that would be attributed to the lens.

But, if the results are regarded similar, that would essentially indicate that using correct color profiles compensate for most differences in the imaging chain.

Best regards
Erik


Also, if you had used the same lens for all three, that would give more credit to the "experiment".

Joel
 

ErikKaffehr

Well-known member
Hi,

My take that the thread is about color. The intent of the posting is to give an objective illustration of the role of sensor technology vs camera profiles.

Best regards
Erik


Hi,

It would help if you supplied a little more information:

  • What gear are you comparing?
  • What raw converter you are using?
  • Which profiles are you using?
  • Which RGB are you evaluating in?

I shared this image shot with Phase One P45+, Sony Alpha 900 and Sony A7rII a while ago:
View attachment 139723

  • Would be interesting if you see a difference?
  • Which one you prefer?

All these images were processed using my own DCP profiles in Lightroom.

Best regards
Erik
 

ErikKaffehr

Well-known member
Hi,

If you see such a comparison relevant, here is a layered TIFF: http://echophoto.dnsalias.net/ekr/Articles/Temp/Tricolore/Stacked_with_samples.tif

The top layer contains samples from the fruits measured with a spectrometer and there is a layer for each camera.

Just a small comment. The only MFD system that really ever had 16 bits is the Phase One IQ3100MP and probably others sharing the same sensor.

According to Anders Torger, who picked apart Capture One's file format it used to be 14 bits with two bits added in raw conversion.

But, 14 bits have always been more than enough as MFD files never had more than 14 bits of relevant information. More like 12 bits of relevant data and 16 bits of noise.

Best regards
Erik

Erik,
Honestly I can not see that much difference here in this sample you posted.
 

dougpeterson

Workshop Member
Just a small comment. The only MFD system that really ever had 16 bits is the Phase One IQ3100MP and probably others sharing the same sensor.
In addition to the IQ3 100mp...

IQ3 100mp Trichromatic
IQ3 100mp Achromatic
IQ4 100mp Trichromatic
IQ4 150mp
IQ4 150mp Achromatic
 

ErikKaffehr

Well-known member
Hi Doug,

I guess that my statement: ' the Phase One IQ3100MP and probably others sharing the same sensor.' covers all the Phase One 100 MP sensors. I don't know how it is with the Hasselblad H6D100c, as I have never seen a raw file from that camera.

WRT the 150MP sensor, the incomplete spec I have seen from Sony has listed as 14 bits. But there was no mention of engineering DR, have you any figure for that. You don't need more than 14 bits to encode 14 bit of DR. So, what DR, per pixel, does the sensor have?

It would be possible to output 16 bit of data if the ramp rate on the ADCs was reduced to 1/4th. I don't recall the Sony specs, but my guess is that FPS would be around 0.8-1 FPS with full 16 bit readout. IS that correct?

I am not a Phase One dealer, but you perhaps can fill in the details:

  • What is default FPS on the IQ4150 series?
  • What is the real bit depth at that frame rate.
  • What is the DR per pixel, that is unnormalised, DR at default frame rate?
  • What is the DR per pixel, that is unnormalized, DR at slowest frame rate?

This things are not easy to find out, but I expect that a Phase One dealer should be able to give correct answers.

Best regards
Erik





In addition to the IQ3 100mp...

IQ3 100mp Trichromatic
IQ3 100mp Achromatic
IQ4 100mp Trichromatic

IQ4 150mp
IQ4 150mp Achromatic
 

dougpeterson

Workshop Member
Erik: the IQ4 150mp has several raw file modes. The IIQ-16 and IIQ-16 Extended are both 16-bit readout. There are also two modes for 14-bit (smaller file via compression only) and 14-bit Sensor+ (much smaller file via compression and lower resolution) both of which allow faster capture rates.

More info here: http://phaseoneiq4.com
 

ErikKaffehr

Well-known member
Another small demo...

Hi,

Below are some samples shot with my P45+ and my Sony A7rII.

Portrait examples.jpg

  • Top row is P45+,processing in Lightroom, with my own profiles. Last image is generated with C1-s tone curve
  • Center row is P45+ processed in Capture One. Note that Capture One makes the image 1 EV brighter by default
  • Bottom row is A7rII, processed in Lightroom with Adobe Standard and Lumarriver profiles

The idea with this posting is to show how much profiles, raw processors and sensors contribute.

The sensors used here very different:

  • P45+ has a 2007 generation Kodak CCD sensor, it should be quite comparable to the sensor used in the Pentax 645D, Leica S (typ 006).
  • A7rII has a 2015 generation Sony CMOS sensor.

CFA (Color Filter Array) design may differ between vendors. Phase One offers the the IQ4100MP sensor with two different CFA designs.

Best regards
Erik




Ok,
I am a user of various different sensor sizes (dx/FF and small medium format).
One reason for me to use MF has allways been color, IMO better skin tones and better tonality.
But why does this have to be the case? Why shows (for example) the Leica S007 better color and tonality than the SL and the M10? Is it 16bit? And if so, why cant they make a FF sensor with 16bit?
The small MF options are still way less flexible to use and slower to use than FF. If there as a FF sensor wit the same IQ like a small MF, just with a little less resolution, this would be great IMO. Can we expect this in the near future?
 
Last edited:

Paratom

Well-known member
Hi Erik,
thanks a lot.
Besides what a converted is capable of it is important how to work with a piece of software. I dont think it is possible to really learn and to be good at 3 or 4 different converters. When C1 didnt support the Leica S anymore years ago I switched to LR. Occasionally I did run some comparisons, but for me it was not possible to come to a final conclusion which software works better for my cameras. (Leica S/M/SL and some Olympus and now x1d).
So for me LR is the starting point, even if other software has advantages. For my Hassy files I sometimes use Phoucs (because I think it has to be optimized for Phocus), but I can not say I find it miles ahead of LR.
If I hd a Phase back I would probably use C1 again. Just because I think Hardware and software and Profile from same company means it should be optimized to work good.
Well, in case of Leica or Nikon or Canon or Panasonic we dont really have this option, we need to use third party.

So for my purpose I do not only want a camera with a good sensor, I also want one where I do not have to spend much time experimenting with profiles and color adjustments, something where I just need to correct exposure (if necessary), WB, and maybe work a bit on shaddows and tones.
Most MF cameras I have owned (various Leica S and x1d) seem to give a pretty good starting point when loading into LR with the standard settings.
 

Geoff

Well-known member
I think that generally when the question is subjectively pleasing color this test misses the mark in at least four ways:
- A contrived lab scene/lighting rather than a creative/interesting scene
- Using non native raw processor (LR) instead of the one where the hardware/software are coordinated (C1+P1)
- Creating lowest-common-denominator computer-generated profile that seek to standardize the response of the color
- Ignoring that different cameras influence the behavior of the photographer

In short: rather than have each look its individual best, your test seeks to make them all look the same – you more-or-less succeeded.

It's like trying to decide whom you prefer of Bach, Queen, Tchaikovsky, and John Lee Hooker by playing their music through a heartless MIDI synthesizer, exactly-as-written, in the same room. It ignores that each artist incorporated the instruments of their era, improvised from the written music, and expected to perform in different venues. It ignores all the things that make music music rather than sound.

I totally get the technical/science-oriented approach of reducing all the variables. But at some point you abstract so far away from the original question that the results are meaningless.

Much less scientific, but much more practically useful, in my opinion, is to simply to take a few hours (minimum) with each camera, making images the way you would if you owned that camera. The way you handle the camera may be different, where you stand in the scene may be different, the software you use may be different, the way you work up the files may be different. Even the way you print or present the work may be different. With so many confounding variables you certainly won't be scientifically exploring the percentage of correlation ascribed to a specific variable, but you will be answering the questions "which camera better helps me execute my vision?" and "which camera do I enjoy using the most?" which is, for 99.9% of photographers, the salient questions.
Very good points.
 

ErikKaffehr

Well-known member
Hi Matt,

I am pretty sure that SL and the S(typ 007) use different sensor designs. To be more precise I would suggest that the S (typp 007) uses the sensor designed by CMOSIS for Leica, using ST Semiconductor technology, while i would guess the SL uses technology from TowerJazz, a company owned by Leica's technology partner Panasonic.

It is quite obvious if you check the image below:
chart (10).jpg


The bump you see on the SL and the Q indicates that the sensors have 'dual gain conversion', patented by Aptina as DR-Pix.

So S and SL use very different pixel designs.

Color depends not on sensor design though, it is much more about colour profile IR and UV filtering and CFA designs.

Most sensor makers use CFA filter compositions from FujiFilm. IR and UV filtering may come Corning Glass or Hoya, just as examples.

Best regards
Erik


The SL pixels are EXACTLY the same size as the S. It is, indeed, a crop of the sensor. The lenses make a difference. In M language, the SL lenses have a Summicron look and the S lenses more Summilux (well, about midway between the two). The test I did was SL, S006, S007, and X1D. They could all produce the same portrait with some messing around, but the lenses (SL 24-90, S120, and X1D 90, all shot at f/4) had different OOC looks.

I do not have an SL-S adapter, or I could take that variable out of the equation.

(This is, of course, my opinion - not fact.)

Matt
 
Top