Moon Maid's sensor comparison

Started Feb 29, 2012 | Discussions
jenella
jenella Senior Member • Posts: 1,009
Re: Moon Maid's sensor comparison

Tony Beach wrote:

jenella wrote:

...just more speculation IMO despite all the fancy charts and graphs.

No one really gives a damn about your opinion.

Maybe not, but you cared enough to take the time to quote me...thank you!

 jenella's gear list:jenella's gear list
Fujifilm X100S Nikon D700 Nikon D7000 Nikon D4 Nikon AF-S Nikkor 14-24mm f/2.8G ED +8 more
OP Marianne Oelund Veteran Member • Posts: 7,766
Re: Color me curious

Luke Kaven wrote:

The more I think of it, the more I realize that the D3s had this peak in it. I think it's also reflected in the engineering DR measurements.

Yes, it shows there as a "flatter" curve through the high-ISO range.

Could you help me to understand what is going on there functionally?

Perhaps, but this does land squarely in the domain of "speculation." It appears that Nikon have found a way to achieve a gain boost of up to 1EV very early in the analog chain, perhaps even in the sensor itself. This allows read noise (in ADU) at ISO 12800, for example, to be where it would have been at ISO 6400.

Is this what you would call the decisive edge in low-light performance for the D4 design?

I'd say that's a fair statement (applies to D3s also).

I had held out some hope that the squeaky clean Exmor would perform on a par when seen at 12MP.

Not in this generation. Perhaps later, they may find a way to incorporate a design feature similar to the D3s/D4 to control high-ISO read noise better. Also keep in mind that the Exmor analog gain tops out around ISO 1000, and sensitivities above that are achieved only by numerical scaling.

Mikael Risedal
Mikael Risedal Veteran Member • Posts: 4,623
Re: Where speculation ends. Who are you geronimo?

You reminds me of a member who tried to explain to Eric Fossum how a sensor works here at dpreview
Who are you?

geronimo wrote:

ukat123 wrote:

Surely someone who calls himself Geronimo has a huge ego, but a small pee pee

OK. Lets' get back on track and discuss this great analysis from Marianne now.

Aww, obviously I have touched a nerve with the cult.

rhlpetrus Forum Pro • Posts: 25,860
Re: You said you don't like it ... (and ISO100?)

Kaj E wrote:

rhlpetrus wrote:

Thanks a lot, very instructive and quite in line what we had expected. One way to check the "stops" difference is to plot, say, ISO6400 for D4 and various ISOs for D3s and check which one matches the 6400 from D4. That would be the "ISO stops" advantage of the D4 at ISO6400 over D3s in SNR. What do you think?

Renato, the point is that the difference in stops is different at different EVs.
--

Yes and no. If curves are within 1/3 EV of each other along a wide enough range, you may consider it a tie. Anyway, when people estimate "such and such stops above/below", it's just that, a rough estimate looking at an image in overall terms, usually looking for noise in some midtone smooth area.
--
Renato.
http://www.flickr.com/photos/rhlpedrosa/
OnExposure member
http://www.onexposure.net/

Good shooting and good luck
(after Ed Murrow)

 rhlpetrus's gear list:rhlpetrus's gear list
Leica D-Lux (Typ 109) Nikon D7000 Nikon 1 V1 Nikon AF-S DX Nikkor 12-24mm f/4G ED-IF Nikon AF Nikkor 35mm f/2D +4 more
rhlpetrus Forum Pro • Posts: 25,860
Re: Anything under -3EV is pretty dark

Iliah Borg wrote:

If you take Zones, and set the saturation to Zone VIII on a print it gives 5 stops from saturation to loss of detail on a print.

More or less my estimate, +2.5 at saturation, -3 for loss of detail.
--
Renato.
http://www.flickr.com/photos/rhlpedrosa/
OnExposure member
http://www.onexposure.net/

Good shooting and good luck
(after Ed Murrow)

 rhlpetrus's gear list:rhlpetrus's gear list
Leica D-Lux (Typ 109) Nikon D7000 Nikon 1 V1 Nikon AF-S DX Nikkor 12-24mm f/4G ED-IF Nikon AF Nikkor 35mm f/2D +4 more
geronimo Regular Member • Posts: 304
Re: Insatiable

Marianne Oelund wrote:

geronimo wrote:

I didn't ask about a vast array of data, I asked about your vast array of sample images. So, you only used a few images to extract your vast array of data. I'm assuming these were from non-production cameras, using non-production firmware?

This is about the sensors, not processing controlled by the firmware. Are you suggesting that the production sensors will be different?

I didn't make comments on sensors - I made comments on your use of non-production cameras using non-production firmware.

If none of that mattered, then you wouldn't feel the need to continue your experiments, as you mention below, using production cameras.

I don't know what other threads you may have posted images in, and I won't be spending time researching you or any myriad of other posts you may have made prior to this one, but it is in this thread that you are making scientific claims, and as such, it should be in this thread that you also post links to the data and data sources that you are using in asserting this thread's claims.

Enough time has been spent on this already. It's offered as a "heads up" for those who are interested in such things, and will soon be replaced with data from production cameras. Toward that end, my effort will not be wasted, since it's simply a matter of inserting new measurements into an existing spreadsheet.

Well it's such a shame to hear you cannot spend 1 second to paste a link to your supporting one or two sample images from non-production cameras that you used to create Estimation Charts.

rhlpetrus Forum Pro • Posts: 25,860
Re: Moon Maid's sensor comparison

perhort wrote:

These estimates suggest that the D800's low light-performance lies between those of the D3 and the D3s, and that it is about 0,5 to 1 stop behind the D3s' (shadows at iso 6400). This is what one would expect if the D800 is more or less a D7000 sensor writ large. What is more surprising is that the D4 is practically identical to the D3s indicating no low light improvement whatsoever. Or perhaps not so surprising, given that Nikon has stated that it will take a more balanced approach towards noise reduction. But then this should perhaps show up in the performance at the low iso end? Perhaps the investigation underestimates the D4's performance in this regard? It would also be interesting to know some more about the method and the data used.

Here's why D4 so close to D3s, they are already very close to the ideal sensor:
http://forums.dpreview.com/forums/read.asp?forum=1021&message=40764201
--
Renato.
http://www.flickr.com/photos/rhlpedrosa/
OnExposure member
http://www.onexposure.net/

Good shooting and good luck
(after Ed Murrow)

 rhlpetrus's gear list:rhlpetrus's gear list
Leica D-Lux (Typ 109) Nikon D7000 Nikon 1 V1 Nikon AF-S DX Nikkor 12-24mm f/4G ED-IF Nikon AF Nikkor 35mm f/2D +4 more
Mikael Risedal
Mikael Risedal Veteran Member • Posts: 4,623
Re: Raul

As I said before - You reminds me of a member who tried to explain to Eric Fossum how a sensor works here at dpreview
Who are you Geronimo?

geronimo wrote:

Raul wrote:

you are in a minor league; don't go playin' with the pros.

You cultists think too highly of yourselves.

Jon Rty Veteran Member • Posts: 3,838
Re: Insatiable

Marianne measured the performance of the sensor. Do you think Nikon will change the sensor?

It's a shame that you do not value science, but that's your prerogative. I know you may feel threatened by tests as like these as you don't understand them, but take comfort in the thought that others do.

geronimo wrote:

I didn't make comments on sensors - I made comments on your use of non-production cameras using non-production firmware.

If none of that mattered, then you wouldn't feel the need to continue your experiments, as you mention below, using production cameras.

Well it's such a shame to hear you cannot spend 1 second to paste a link to your supporting one or two sample images from non-production cameras that you used to create Estimation Charts.

Great Bustard Forum Pro • Posts: 42,943
Just want to jump in...

Marianne Oelund wrote:

noirdesir wrote:

What might be instructive would be a curve for a perfect sensor (ie, 100% QE, zero read noise). Just to show the physical limits of sensor technology.

Just draw a parallel line to the D4 graph in the EV +1 to +3 range, 2.2dB above the D4, then extend it to the left. That's a 100% QE sensor with no read noise.

D4: Only 2.2dB from ideal (for Bayer sensors)!

...to express my appreciation for the excellent work -- thanks!

MrSkelter Contributing Member • Posts: 705
Re: Sharpening the pencil

Marianne Oelund wrote:

MrSkelter wrote:

I'd like to see numbers, or graphs at a much greater scale. Everything is so clustered that, even if she has no agenda, the graphs make similar performance hard to separate clearly.

This presentation was chosen to make differences which are difficult to see in images, also difficult to see on the graphs. Yes, they're clustered, but in the same manner as the image quality from the cameras.

That seems like an odd choice. Normally analysis is supposed to make hard to discern differences easier to see.

I am planning on providing numbers, but only from production samples.

Great. That's good to know.

I'd also like to see the data-points plotted as curves. If that's not accurate points would be better. Straight-lines, as we have here, are meaningless and the crossover points are deceptive. Sensor performance doesn't 'tip' at arbitrary ISO/EV values.

It only makes sense to sharpen the pencil to a certain point. What's the value in knowing the intersect abscissa to better than 0.2EV uncertainty?

I'd counter that if it's not relevant then why link the points in a way that adds confusion and visual noise, instead of lessens it? As far as I can see curves would be most accurate, points would be correct if spartan. I'm just curious as to the choice - I don't see the benefit.

Horshack Veteran Member • Posts: 7,729
Re: Color me curious

Luke Kaven wrote:

Could you help me to understand what is going on there functionally?

Is this what you would call the decisive edge in low-light performance for the D4 design? I had held out some hope that the squeaky clean Exmor would perform on a par when seen at 12MP.

That High ISO "knee" has been a big mystery ever since Marianne posted her first D3s QE numbers long ago. Maybe a full analyze set of D4/D800 files will reveal more about it.

noirdesir Forum Pro • Posts: 13,586
Is this based on the green pixels or average of all pixels?

I had once asked Bobn whether DxO is based on the green channel only or an average and he was not sure at that moment.

Tony Beach Forum Pro • Posts: 11,961
Re: Moon Maid's sensor comparison

jenella wrote:

Tony Beach wrote:

jenella wrote:

...just more speculation IMO despite all the fancy charts and graphs.

No one really gives a damn about your opinion.

Maybe not,

Definitely not, and even more so since Marianne was kind enough to respond to you and instead of thanking her for her efforts or for taking some time to explain to you what she had done, you come back with yet another snarky post. You have clearly become nothing more than a rather lame troublemaker in this forum.

but you cared enough to take the time to quote me...thank you!

I'm glad that you are confirming that I captured the essence of your post. It reminded me of some parody of a moron dismissing "fancy booklearning."

OP Marianne Oelund Veteran Member • Posts: 7,766
Production unit not required

geronimo wrote:

I didn't make comments on sensors - I made comments on your use of non-production cameras using non-production firmware.

If none of that mattered, then you wouldn't feel the need to continue your experiments, as you mention below, using production cameras.

It doesn't matter whether it's a production unit - you've drawn an incorrect conclusion. It only matters that I will have it in my hands, to perform tests with, and take controlled test images at ISO settings where I do not yet have direct data. A pre-production unit would be just as suitable, but alas, I have no access to one. Do you know where they are being sold now?

Well it's such a shame to hear you cannot spend 1 second to paste a link to your supporting one or two sample images from non-production cameras that you used to create Estimation Charts.

Dealing with you is like dealing with a spoiled child. The real shame is that you continue to reply with insults, and cannot be bothered to do simple things for yourself. Keywords "D800 sample NEF" immediately turned up this:
http://forums.dpreview.com/forums/read.asp?forum=1021&message=40554548

perhort
perhort Regular Member • Posts: 134
Re: Moon Maid's sensor comparison

I see now that the D4 is about 0.5-1 stop better than the D3s in the shadows at iso 200. This seems logical given Nikon's statement that they would take a more balanced approach and given that they have lowered base iso to 100. So if the numbers will hold up, the D4 seems to embody a slight improvement over the D3s after all.

rhlpetrus wrote:

perhort wrote:

These estimates suggest that the D800's low light-performance lies between those of the D3 and the D3s, and that it is about 0,5 to 1 stop behind the D3s' (shadows at iso 6400). This is what one would expect if the D800 is more or less a D7000 sensor writ large. What is more surprising is that the D4 is practically identical to the D3s indicating no low light improvement whatsoever. Or perhaps not so surprising, given that Nikon has stated that it will take a more balanced approach towards noise reduction. But then this should perhaps show up in the performance at the low iso end? Perhaps the investigation underestimates the D4's performance in this regard? It would also be interesting to know some more about the method and the data used.

Here's why D4 so close to D3s, they are already very close to the ideal sensor:
http://forums.dpreview.com/forums/read.asp?forum=1021&message=40764201
--
Renato.
http://www.flickr.com/photos/rhlpedrosa/
OnExposure member
http://www.onexposure.net/

Good shooting and good luck
(after Ed Murrow)

 perhort's gear list:perhort's gear list
Nikon D700 Nikon AF-S Nikkor 70-200mm f/2.8G ED VR Nikon AF Nikkor 24mm f/2.8D Nikon AF Nikkor 35mm f/2D Nikon AF Nikkor 85mm f/1.8D
OP Marianne Oelund Veteran Member • Posts: 7,766
Re: Sharpening the pencil

MrSkelter wrote:

That seems like an odd choice. Normally analysis is supposed to make hard to discern differences easier to see.

Most of the analyses I've performed over the last 30 years as an engineer, have been to uncover unformation which was previously unseen, not merely to amplify what is already apparent.

I'd counter that if it's not relevant then why link the points in a way that adds confusion and visual noise, instead of lessens it? As far as I can see curves would be most accurate, points would be correct if spartan. I'm just curious as to the choice - I don't see the benefit.

The connecting lines are a visual aid to lead the eye from point to point. Many of the points are difficult to locate by themselves, since they can be hidden behind others.

With the lines, I'm only getting complaints from you. Without the lines, I'm sure many others would have been complaining about the difficulty of reading the plots.

CriticalI Senior Member • Posts: 1,777
Ignore the nonsense

Marianne, don't get sucked in - he's just a sour little wannabe who should be ignored.

And thanks for your hard work - the graphs make it much easier to understand for which my thanks.

Marianne Oelund wrote:

geronimo wrote:

I didn't make comments on sensors - I made comments on your use of non-production cameras using non-production firmware.

If none of that mattered, then you wouldn't feel the need to continue your experiments, as you mention below, using production cameras.

It doesn't matter whether it's a production unit - you've drawn an incorrect conclusion. It only matters that I will have it in my hands, to perform tests with, and take controlled test images at ISO settings where I do not yet have direct data. A pre-production unit would be just as suitable, but alas, I have no access to one. Do you know where they are being sold now?

Well it's such a shame to hear you cannot spend 1 second to paste a link to your supporting one or two sample images from non-production cameras that you used to create Estimation Charts.

Dealing with you is like dealing with a spoiled child. The real shame is that you continue to reply with insults, and cannot be bothered to do simple things for yourself. Keywords "D800 sample NEF" immediately turned up this:
http://forums.dpreview.com/forums/read.asp?forum=1021&message=40554548

-- hide signature --

Regards,
Steve

OP Marianne Oelund Veteran Member • Posts: 7,766
Green channel for luminance noise

noirdesir wrote:

I had once asked Bobn whether DxO is based on the green channel only or an average and he was not sure at that moment.

The green channel is considered to be the luminance signal for these purposes, and as far as I'm aware, is used by everyone to assess luminance noise.

noirdesir Forum Pro • Posts: 13,586
Re: Green channel for luminance noise

Marianne Oelund wrote:

noirdesir wrote:

I had once asked Bobn whether DxO is based on the green channel only or an average and he was not sure at that moment.

The green channel is considered to be the luminance signal for these purposes, and as far as I'm aware, is used by everyone to assess luminance noise.

Thanks. Would you expect the relative performance of, eg, the blue to the green channel to differ noticeably between cameras? Particularly the blue channel noise would be much more important, in relative terms, under incandescent light. Iliah Borg had in the past suggested/reported that the D3 and D3s have a better blue channel performance than the D3x (in relative terms to their green channel performance).

Keyboard shortcuts:
FForum MMy threads