New Nikon sensor tech

Ironheart

Digging into the specs a bit more, and comparing the latest 5-series to all previous Nikon camera releases, I don't think we've ever seen a two stop max ISO increase. Granted, the published ISO numbers are marketing numbers, but there is a consistency to them. Also this is the first time ever that we've seen a three stop improvement in extended ISO. That's the D5 compared to the D4s, looking at the D7200 compared to the D500 we see a one stop bump in max ISO, but a four stop ISO in extended. (Personally I think the max ISO on the D7200 is inflated it should also be a 2/3 max/extended increase).

Nonetheless, we see two obvious things. One, Nikon clearly has breakthrough sensor tech here, probably BSI like they used on the J5. I'll bet it's a Nikon design, and BSI would account for the higher low-light performance. The extra-extra bump on extended ISO likely comes through better noise reduction algorithms in expeed 5. Two, Nikon has two sensors, developed side by side, exact same specs and tech, the only difference is size. DX vs. FX. And one stop of light, exactly predicted by the difference in size. This is sort of a rare occurrence, like identical twins, and it will (hopefully) allow us to put the DX vs. FX arguments to bed once-and-for-all. )

I wanted to start a new thread to speculate on these twins, and track the info we find as folks eventually get these in their hands for testing and tear down. I want to focus specifically on the sensors and related technology like expeed 5. Thoughts?

WestEndFoto

I think that this is an excellent idea.

PB_PM

Each time a new DX body comes out people say the same thing, and then the real test come out and the FX always has at least a stop on the DX body. The D500 still has physically smaller pixels, it's just a matter of physics.

The D3 and D300 both had 12MP sensors, and the D3 was over a stop better than the D300. It's just the nature of the tech.

haroldp

BSI matters less as sensors get bigger because the fraction of surface real estate taken by the traces decreases.

Ironheart

Each time a new DX body comes out people say the same thing, and then the real test come out and the FX always has at least a stop on the DX body. The D500 still has physically smaller pixels, it's just a matter of physics.

The D3 and D300 both had 12MP sensors, and the D3 was over a stop better than the D300. It's just the nature of the tech.

Hmmm. I said that the FX has a one stop advantage. It isn't due to the size of the pixels though, as gapless micro lenses capture 99% of the light regardless. It's due to the overal physical size of the sensor. It's due to this really simple branch of physics called geometry :-)

heartyfisher

I seriously doubt that there is much real improvement to the sensor.. maybe 10% if we are lucky. The reason is the physics of the sensor tech. for many years the efficiency of energy conversion of light energy to electrical was about 10-15% (see solar panels) then around 2006-2009 there was a break through and there was steady improvement going from 20% through to 50% that translated to the huge ISO improvement over that time .. its slowed down since then. However, In addition to the raw energy conversion improvements there are other things like the microlense coverage improvements and electrical conductivity and efficiency and material science. I doubt BSI will help much with the relative large pixels of the D5/D500.

so in combination with that slight improvemt of sensor material (say 7% ) with the electrical and microlense design improvements (say another 7%) and larger pixels, I think raw sensor output between D500 and D7200 we would be lucky if we get 10-15% improvements.
Or about half to 2/3 EV improvement. That still very very good and I feel I may have over estimated it. we will see from the DXO measurements.

Where the big improvements seem to be, if I read right from the "confidential" info., would be in the Expeed processing power and new software algorithms. I have mentioned before that some 2-4 years ago I saw some amazing papers from various universities that use groups of 3,7 9, 13 etc pixels to deduce colour and improve noise and reduce moire. It makes sense that these have gone into expeed5..

Ironheart

Uh, someone correct me if I'm wrong, but a one stop improvement, which is what we've seen every generation, give or take, is a 100% improvement right? Each stop is double or half the amount of light so to go from ISO 12,800 to 25,600 is a 100% increase, not just numerically but in light sensitivity.

What we've been told (if the specs pan out in testing) is an unprecedented 300% increase in max ISO, from D4s at max ISO 25,600 to the D5 at max ISO 1,204,000. We'll have to wait and see...

WestEndFoto

Correct Ironheart.

Ironheart

BSI matters less as sensors get bigger because the fraction of surface real estate taken by the traces decreases.

Somebody better tell Sony, they went ahead and made a full frame BSI sensor anyway. Now if the Sony A7R II just had a better processor, a mirror and true 14-bit RAW... You'd have a D5 :-)
Kidding aside, I agree with you, but it is fun to speculate on what they've done to pump it up.

PB_PM

As far as I know Sony released a firmware update for the A7R II late last year that added uncompressed 14bit RAW. For Sony the problem isn't the processor, it's the software used for the AD conversion process, it's simply inferior to the one that Nikon's developed.

heartyfisher

Some people are suggesting that the new D500 sensor is based off the Sony Stacked Sensor configuration. if that is true then its possible that the raw ISO readout has indeed a 1 EV improvement..!!

From the images captured off the back of the camera (on the main blog) 512K iso is fairly clean ! should be fun having another go at bat photography again :-)

spraynpray

HAH! Hearty - everything looks good on the back of the camera!

WestEndFoto

Yes, the marketing department is on charge of programming how those JPEGs are processed.

manhattanboy

Some people are suggesting that the new D500 sensor is based off the Sony Stacked Sensor configuration.

Who is stating this?

haroldp

Why 'They' are.

heartyfisher

LOL ! the amorphous cloud of internet experts! of course. I myself am just wishfuly hoping that it may possibly remotely could be true...

Nikoniser

First Review is up with some of the high ISO pictures, he is saying its about 1 stop of improvement.

http://oleliodden.com/photo-gear/field-reviews/beta-test-report-nikon-d5/

Nikoniser

First Review is up with some of the high ISO pictures, he is saying its about 1 stop of improvement.

http://oleliodden.com/photo-gear/field-reviews/beta-test-report-nikon-d5/

Exact quote is

"Based on the jpg-files I have reviewed and edited, the image quality up to ISO 12 800 looks very clean. I would not have any problems including most of these ISO 12 800 images in my image bank or for high quality printing. With the D4S cameras this image quality was equal to approx. ISO 4000 – 5000. This is something like 1.5 stop improved ISO performance, which is very good."

Ton14

Sobering thought, making these photo's in full daylight f/7.1, 1/1000th + with a € 7000.- camera and 2 lenses € 12.999.- each, is not really the hardest field testing. Show me photo's from the darkest pub with very bad light please.

spraynpray

Yes, @Ton14, been there, done that with my D750 - that sorts out the good from the mediocre ok.

manhattanboy

Why 'They' are.

Finally found the source for the thought that the sensors are backside illuminated here:
This thread with pics
There is some talk of stacked sensor, but the consensus leans towards not stacked.

Paperman

Every time a new camera comes out, someone says/claims he sees 1 to 1.5 stop increase in high ISO performance ( unfortunately under conditions Ton14 has indicated - the light being so much the photographer had to push to 1/5000-1/8000 speeds to get most high ISO shots ). Then we all see that the improvement is more like 1/8th a stop once the tests are run.

D3s - ISO 3253 (2009)
D4 - ISO 2965 (2012)
D4s - ISO 3074 (2014)

Someone please factor in the Mp increase for me and we'll see what changed in 5-6 years.

Many seem to fall into the trap of looking only at the top pushed ISO value ( 52400-204800 ) assuming that relates to high ISO quality increase. Does it ?? :-?

WestEndFoto

Yeah, I am really quite skeptical of ISO claims.

It is easy to clean up noise, if you are willing to sacrifice resolution. You get to the point where a 36 megapixel image from a D800/810 is effectively 2 or 3 megapixels to keep the noise invisible. However, the colour starts to go off, and that is impossible to recover. Few consumers understand this which makes it easy for camera manufacturers to claim superior high ISO performance or 1 to 2 stops improvements in every generation. Or the gain may be real, but only in a narrow range, say ISO 6,400 - 12,800. Of course, the camera manufacturer will shoot their promotional shots at this ISO, glossing over the fact that the improvement at ISO 1,000 (where it really counts for me) is minor.

Of course, there are legitimate ways of reducing noise that have provided very useful gains over time, but they are incremental.

I think that the bottom line is to know the limitations of your equipment and shoot around those limitations. For me, that means packing a tripod around on my vacations.

donaldejose

The original poster is correct in pointing out that Nikon consistently sets "native" ISO ranges at what they feel is acceptable and makes the remainder "extended ranges marked High 1 and so forth without listing the ISO numbers. If Nikon's conventions and ratings are consistent generation to generation then Nikon has made more progress (more than a stop?) with the D5 and D500 generation than we usually see in one generation (less than a stop). The question is why? What accounts for this additional improvement in ISO? If the ratings are based on jpg images it could be both an improved sensor and improved software noise reduction. However, if the ratings are based on RAW files it would have to be just an improved sensor. We don't know how Nikon produces its ratings. If the two new sensors were BSI (back side illuminated) you would expect Nikon to mention this in their press releases and they do not so I think we must assume for now that they are not BSI sensors. I will watch this thread because I too am very interested in what real gain Nikon has achieved and how they have achieved it.

manhattanboy

That source I linked seems to think that the D5 and D500 sensors were significantly different designs. Both appear to be Sony/Renases (I spelled that wrong but forgive me) though, and that is what the booth at CES also claimed.

spraynpray

BSI is supposed to be a huge step improvement for Hi ISO from what I have read, so I am hoping for a big real-world increase in IQ at medium high ISO's. We will see though. It will only be a big improvement in really low light performance that will make me want one now.

Personally, I have become dispirited with any others reporting and on a topic like this, I would only trust a side-by-side comparison shoot with my D750 like I did for the D7100 with my D7000. These days most bloggers just quickly read around the net and regurgitate the info or worse still add their own distortion based on their own baggage. The 'best' way to shoot nightscapes as printed in magazines and on 99% of blogs is a case in point.

It must go: Click - chimp - "hmmm, that worked - hey I'm an expert! Must do a blog or a video with quirky behaviour for my fans". Yikes.