Pages: 1 ... 3 4 [5] 6   Go Down

Author Topic: new DxOmark test results  (Read 17167 times)

BJL

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 5665
Column parallel analog-digital conversion in EXMOR ... And elsewhere
« Reply #80 on: March 27, 2012, 08:57:36 AM »

The single headline advantage of Sony's EXMOR tecnnology is its Column parallel analog-digital conversion, which I believe helps greatly because the majority of sensor read noise in other designs enters during the fast transportation of the analog signal along the edge of the sensor and then to tue ADC, not in the photosites themselves. So digitizing before that step helps a lot. This approach also sems to make high frame rates easier to achieve, so is good for hogh resolution video.

The good news is that this technique is not a Sony exclusive, even though Sony has so far made the most of it in the world of larger sensors for "dedicated stills cameras".The same method was used first by Samsung, but in a video camera sensor, and it seems that video sensors have been the main area of application so far.

So my guess is that Canon is working on something similar --- unless it already has a different and even better idea in development. By the way, it also seems quite possible that some other sensors from other designers are also using column-parallel ADC, without talking about it so much (perhaps because it would sound rather "me too"). Some candidates are the sensors for Nikon One cameras, the video oriented multi-aspect ratio sensors of the Panasonic GH1 and GH2, and the sensor of unstated origins in the new Olympus OM-D E-M5. Panasonic has said that the GH1 and GH2 sensors produce digital output (as with EXMOR) whereas its other 4/3" sensor produce an analog output signal for off-board ADCs (as with Canon) and that those sensors are significantly more expensive to make than its other 4/3 sensors, which is why they are not used in all recent micro four thirds bodies. So maybe Panasonic is part-way there, but the new approach has problems with low yields, pushing the price up. And maybe Nikon and its sensor partners are also about the EXMOR monopoly, or already have.

P. S. it is often stated as obvious fact that the D800 sensor is from Sony, sometimes supported by the fact that Nikon does not have sensor fabrication equipment. However, AFAIK, neither Nikon nor Sony has said this, and the possibility remains that Nikon has been involved in the design of this sensor (perhaps in partnership with Sony, or Aptina, or one of the other good sensor designers that have been working mostly with smaller sensors up till now, such as Toshiba).  Fabrication can be contracted out to one of several competent fabs, so Nikon's lack of a sensor fab. is irrelevant. For example, Aptina is a fabless sensor designer.

Added: talking of Aptina (likely sensor supplier for Nikon One), it is about to release an APS-C sized sensor for "DSLR/mirrorless", which does 14-bit ADC on the camera chip itself, and offers 10 fps, 1080p30 video etc.:
http://www.aptina.com/products/image_sensors/mt9h004/
I have no idea how good it is or if any major camera maker will adopt it, but the competition seems to be getting stronger, and the idea that a camera maker is always better off doing all its own sensor design in-house rather than making use of the technologies offered in a competitive market place of companies that specialize in sensor design makes less sense than ever to me. Frankly, the all in-house philosophy verges on "log cabin in the mountains off the grid survivalism".
« Last Edit: March 27, 2012, 09:14:06 AM by BJL »
Logged

eronald

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 5717
Re: Column parallel analog-digital conversion in EXMOR ... And elsewhere
« Reply #81 on: March 27, 2012, 10:46:07 AM »

Frankly, the all in-house philosophy verges on "log cabin in the mountains off the grid survivalism".

And that is the reason why every mobile phone now has an ARM chip in it, designed by a small team from an obscure Apple II clone maker in the UK (Acorn).

Edmund
Logged

BJL

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 5665
Re: Column parallel analog-digital conversion in EXMOR ... And elsewhere
« Reply #82 on: March 27, 2012, 11:12:26 AM »

And that is the reason why every mobile phone now has an ARM chip in it, designed by a small team from an obscure Apple II clone maker in the UK (Acorn).
Yes: not a lot of completely DIY phone processors are there? Not even from giants like Samsung (which strangely makes its own ARM chips, yet also buys some from TI.) Apple and Samsung play the game well, balancing in-house exclusives with outsourcing from companies that do some things very well and in high volume, thus sharing in economies of scale. Processors for mobile devices show the fine points of the strategy: the most successful smart phone makers, Apple and Samsung, license ARM processor core designs and graphics processor designs, but then integrate them into custom designs with some in-house chip design expertise.

Historical trivia: ARM processor designs got their first big push into the mobile device market many years ago when adopted for the first touch-screen "pad" --- the ill-fated Apple Newton MessagePad. The movie of the iPad's success story could be called "The Revenge of the Newton".
Logged

ejmartin

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 575
Re: new DxOmark test results
« Reply #83 on: March 27, 2012, 12:00:07 PM »

I have yet to see an image (as in a photograph...with light and a real subject, you know what I mean...) that demonstrates this allegedly superior base iso DR of the Sony...

How about this one:

http://forums.dpreview.com/forums/read.asp?forum=1034&message=36903045
Logged
emil

ondebanks

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 860
Re: new DxOmark test results
« Reply #84 on: March 27, 2012, 12:58:07 PM »

2012 vs 2004  ;D
http://dl.dropbox.com/u/17452116/DynamicRange_5DMIIIvsAptus22.jpg

Very good. Now repeat it with both cameras set to ISO 400...or heck, even ISO 400 on the Canon and ISO 100 on the Aptus.

This will illustrate
(1) the problem with MFD: shooting above base ISO is merely underexposure without any compensating improvements in electronic noise;
(2) the problem with Canons: they only come into their best above base ISO. 

Ray
Logged

yaya

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1202
    • http://yayapro.com
Re: new DxOmark test results
« Reply #85 on: March 27, 2012, 01:11:16 PM »

How about this one:

http://forums.dpreview.com/forums/read.asp?forum=1034&message=36903045

???

All I can see is 2 reasonably similar, soft looking images where 1 (I think the 6400 iso) has less shadow noise than the other and the other one holds a bit more highlight detail (because it was so much under-exposed, obviously)

No offence but personally I wouldn't use them as an example for anything...

itsskin

  • Jr. Member
  • **
  • Offline Offline
  • Posts: 83
Re: new DxOmark test results
« Reply #86 on: March 27, 2012, 03:43:26 PM »

Very good. Now repeat it with both cameras set to ISO 400...or heck, even ISO 400 on the Canon and ISO 100 on the Aptus.

This will illustrate
(1) the problem with MFD: shooting above base ISO is merely underexposure without any compensating improvements in electronic noise;
(2) the problem with Canons: they only come into their best above base ISO. 

Ray

They are great compliments to each other. That's why I have them both.  ;D
Logged

Aku Ankka

  • Newbie
  • *
  • Offline Offline
  • Posts: 26
Re: new DxOmark test results
« Reply #87 on: March 27, 2012, 05:00:45 PM »

???

All I can see is 2 reasonably similar, soft looking images where 1 (I think the 6400 iso) has less shadow noise than the other and the other one holds a bit more highlight detail (because it was so much under-exposed, obviously)

Both images were exposed the same way - 1/30s, f/2. Also, "Under-exposure" is a subjective term and as such not useful.

ISO is not an exposure parameter - it only tells the camera how much to boost the signal. By applying the high ISO 6400 boost the camera loses the highlight details that would not be lost if a more proper ISO were used instead. The dynamic range which has been compressed into this (ISO 100 out of camera) image is a good sample. An inferior camera would have to sacrifice one or the other end.

Logged

ejmartin

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 575
Re: new DxOmark test results
« Reply #88 on: March 27, 2012, 05:09:59 PM »

???

All I can see is 2 reasonably similar, soft looking images where 1 (I think the 6400 iso) has less shadow noise than the other and the other one holds a bit more highlight detail (because it was so much under-exposed, obviously)

No offence but personally I wouldn't use them as an example for anything...

Then you aren't paying attention.  First of all, the discussion is about DR, not resolution so comments about 'softness' are irrelevant.  Second, shadow noise is the same since electronic read noise is the same for the two ISO being used and exposure is the same between the two shots.  Third, there is no 'underexposed' here; both shots have the same exposure; the difference is that the ISO 100 capture has enough DR to reveal highlight details that are blown in the ISO 6400 shot, while capturing the rest of the scene with the same level of noise as the ISO 6400 shot.

The point is that the camera had enough DR that the photographer could expose for the highlights at ISO 100, and still have enough latitude to bring up the shadows by six stops in post.

I suspect you have some fundamental misconceptions about how digital capture works.
Logged
emil

ondebanks

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 860
Re: new DxOmark test results
« Reply #89 on: March 27, 2012, 05:14:02 PM »

They are great compliments to each other. That's why I have them both.  ;D


Exactly; me too.

I think what's interesing though is that the Sony Exmor sensors are increasingly able to cover both bases in one camera. This is what all cameras/sensors should aspire to - well their users do anyway! I expect Canon to catch up on the low-ISO end sooner or later. But MFD will never catch up on the high-ISO end, if it stays stuck with CCD sensors.

Ray
Logged

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 10042
    • Echophoto
Re: new DxOmark test results
« Reply #90 on: March 27, 2012, 05:22:27 PM »

Hi,

Now that we have seen a decent demo of DR on an APS-C sensor, it would be interesting to see a similar demo on an MF sensor. I have not seen that many.

Here is one on the Leica S2 and the Nikon D3X based on images courtesy of Lloyd Chambers: http://echophoto.dnsalias.net/ekr/index.php/photoarticles/38-observations-on-leica-s2-raw-images

Best regards
Erik

Then you aren't paying attention.  First of all, the discussion is about DR, not resolution so comments about 'softness' are irrelevant.  Second, shadow noise is the same since electronic read noise is the same for the two ISO being used and exposure is the same between the two shots.  Third, there is no 'underexposed' here; both shots have the same exposure; the difference is that the ISO 100 capture has enough DR to reveal highlight details that are blown in the ISO 6400 shot, while capturing the rest of the scene with the same level of noise as the ISO 6400 shot.

The point is that the camera had enough DR that the photographer could expose for the highlights at ISO 100, and still have enough latitude to bring up the shadows by six stops in post.

I suspect you have some fundamental misconceptions about how digital capture works.

Nick-T

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 462
Re: new DxOmark test results
« Reply #91 on: March 27, 2012, 05:22:44 PM »

Then you aren't paying attention.

SNIP

I suspect you have some fundamental misconceptions about how digital capture works.

That's a hall-of-fame comedy comment.

On the subject of digital capture Yaya is one of the most knowledgable people I know  [apart from the Hasselblad guys :)]

Nick-T

yaya

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1202
    • http://yayapro.com
Re: new DxOmark test results
« Reply #92 on: March 27, 2012, 05:41:30 PM »

I suspect you have some fundamental misconceptions about how digital capture works.

Thanks for the complement...Appreciated! And BTW shadow noise in these two examples is NOT the same....

ejmartin

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 575
Re: new DxOmark test results
« Reply #93 on: March 27, 2012, 06:20:39 PM »

Thanks for the complement...Appreciated! And BTW shadow noise in these two examples is NOT the same....

Yeah, that was uncalled for.  I was however responding to the agressive tone of your reply.

Yes, there is more pattern noise in the ISO 100 shot.  You have pushed the shadows another 5+ stops to expose it; so now we're talking about an 11+ stop push of the original capture?  Please.  BTW the difference is simply the pattern noise; since it's coherent across the frame, very easy to correct in post.



On the subject of digital capture Yaya is one of the most knowledgable people I know  [apart from the Hasselblad guys :)]

Nick-T

Being a sales/product manager (albeit a very proficient one) does not necessarily translate into technical expertise about digital electronics.
Logged
emil

Nick-T

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 462
Re: new DxOmark test results
« Reply #94 on: March 27, 2012, 06:24:42 PM »


Being a sales/product manager (albeit a very proficient one) does not necessarily translate into technical expertise about digital electronics.

Of course it doesn't. Job title aside Yair DOES have technical expertise, I know I have spoken with him.

 I suspect if you knew him you wouldn't be so quick to insult him, especially if you knew about that incident on the metro in Spain, and the special forces thing. (But I digress)

Nick-T

yaya

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1202
    • http://yayapro.com
Re: new DxOmark test results
« Reply #95 on: March 27, 2012, 07:06:46 PM »

ejmartin,

(perhaps it is time for you to step out of the shadows and tell us who you are and what you do? It seems silly to talk to a nickname...)

There is a huge difference between expertise and experience...although with experience (sometimes) comes expertise. I'm not sure if the opposite is true

Anyway this is the medium format & large sensor section and we are wasting bandwidth and people's time arguing about some snapshot images taken with a thumbnail size sensor.

ejmartin

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 575
Re: new DxOmark test results
« Reply #96 on: March 27, 2012, 07:22:33 PM »

ejmartin,

(perhaps it is time for you to step out of the shadows and tell us who you are and what you do? It seems silly to talk to a nickname...)


Who am I?  Emil Martinec; I am a physics professor at the University of Chicago.  I don't think my identity is a secret here.

Quote

Anyway this is the medium format & large sensor section and we are wasting bandwidth and people's time arguing about some snapshot images taken with a thumbnail size sensor.


The topic of discussion was the D800; the D7000 that took the linked images is essentially a cropped version of the D800 sensor, approximately the same size pixels, the D800 simply scales up the sensor size with the same (or slightly better) pixel performance.  It is therefore reasonable to infer the properties of the D800 from those of its little brother.
Logged
emil

bclaff

  • Jr. Member
  • **
  • Offline Offline
  • Posts: 77
    • Photons to Photos
Re: new DxOmark test results
« Reply #97 on: March 27, 2012, 07:29:53 PM »

Emil,
Quote
the D7000 that took the linked images is essentially a cropped version of the D800 sensor
Yeah, for the purposes of this discussion. D7000 4.75 microns, D800 4.85 microns
Photosite read noise very close:
http://home.comcast.net/~NikonD70/Charts/RN_ADU.htm#D7000_14,D800_14

Regards,
Bill
Logged
Visit me at Photons to Photos

david distefano

  • Full Member
  • ***
  • Offline Offline
  • Posts: 127
Re: new DxOmark test results
« Reply #98 on: March 27, 2012, 08:53:01 PM »

since i started this a few days ago it has morphed into math and physics. i have learned quite a bit. the d800, even in their wildest dreams, will not bring about the death of mf digital, but do you think prices will drop for mf digital backs in the future?
Logged

BJL

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 5665
Re: new DxOmark test results
« Reply #99 on: March 27, 2012, 10:12:33 PM »

... do you think prices will drop for mf digital backs in the future?
If anything, I expect the opposite: MF gear will move a bit further up the scale of lower volume and higher price. As 35mm format becomes accepted as "good enough" for some part of what previously required MF, and so some photographers choose 35mm where in years past they would have chosen MF, two things happen:
1. Reduced demand for MF makes the economies of scale worse, so prices go up even for gear with equal unit production costs.
2. MF makers shift their emphasis towards the higher quality part of their product range, to be clearly above what 35mm can match, maybe with things as simple as higher sensor resolution, which also pushes demand for lenses of ever higher resolution, ever more precisely aligned bodies and AF systems, and so on, so the gear becomes more expensive to make too.

We have already seen signs of this, with the prices of lenses and bodies often significantly higher now than in the film era.

P. S. of course any talk of the D800 or even D800E killing off DMF is either wild hyperbole or siliness, if only for only trivial reason: 80>36. That is, along with all the other much debated differences, resolution clearly does also contribute to a good number of camera purchasing choices, as shown by the higher prices of the higher resolution backs.
Logged
Pages: 1 ... 3 4 [5] 6   Go Up