Shocked by A7ii vs A6300 noise comparison

[Stflbn]

When I first started messing with my a6300 files in Sony IDC (Now LR and C1 support it) I was shocked by how pleasing the noise looked compared to my old a6000 and thought to myself that it looked as good as my full frame A7ii files.

So this morning I shot the same thing with the A6300 and A7ii. Just looking at the grain / noise in each photo... and I remain shocked by how good the A6300 noise looks at higher ISO's. I don't know what voodoo Sony did to it, but it looks fantastic compared to a Full Frame sensor and if anything "I think" the a6300 noise may be less clumpy than the A7ii's.



A7ii with Green tag, A6300 with Red:







--
Sony A7ii A7s A6000
28 FE f2 | Sony 55 FE 1.8 | 70-200G FE F4 | 24-240 FE | Sony 90mm FE Macro
http://501concepts.com

 

[OFremen]

That's really interesting.

I remember reading a thread earlier where someone was saying that the A7's were only a stop or two better at low light than the A6000...so if the A6300 has better low light capabilities than the A6000 due to that new sensor then what you're seeing here makes sense !

 

[JimKasson]

When I first started messing with my a6300 files in Sony IDC (Now LR and C1 support it) I was shocked by how pleasing the noise looked compared to my old a6000 and thought to myself that it looked as good as my full frame A7ii files.

So this morning I shot the same thing with the A6300 and A7ii. Just looking at the grain / noise in each photo... and I remain shocked by how good the A6300 noise looks at higher ISO's. I don't know what voodoo Sony did to it, but it looks fantastic compared to a Full Frame sensor and if anything "I think" the a6300 noise may be less clumpy than the A7ii's.

The "voodoo" in the a6300 is conversion gain switching, with (probably) Aptina technology, a la the a7S and a7RII.



Details here: http://blog.kasson.com/?p=13810

Really nerdy and technical details here: http://blog.kasson.com/?p=13870

Jim

--
http://blog.kasson.com

 

[Stflbn]

Someone questioned that the a6300 could have lesser detail... which is also a no at iso 20,000 (big jump in noticeable detail from iso12,800 to 20,000 on both btw)







--
Sony A7ii A7s A6000
28 FE f2 | Sony 55 FE 1.8 | 70-200G FE F4 | 24-240 FE | Sony 90mm FE Macro

http://501concepts.com

 

[Seeky]

So, the DR of A6300 is better from ISO 400 onwards. But what about noise compared to A7II?

 

[JimKasson]

So, the DR of A6300 is better from ISO 400 onwards. But what about noise compared to A7II?

Engineering dynamic range is fullscale/read noise. So EDR and read noise are two sides of the same coin. If the EDR is higher, the RN is lower, if fullscale is the same. And, when the a7II is in compressed mode, fullscale is unarguably the same. When the a7II is in uncompressed mode, you could argue that fullscale is different, but it turns out that, whichever number you pick for fullscale with the a7II, RN sales with that, so that EDR turns out to be the same for the a7II whether it's in 14-bit uncompressed mode or 13-bit compressed mode.

It's explained here:

Sony a7II uncompressed read noise and EDR - the last word

Yesterday, I posted some dark-field histograms from the a7II at various ISO settings with the new uncompressed raw mode enabled. Today I'll show you what the read noise from those images looks like, and how it compares to the CRAW lossily compressed dark field images. I looked at a 200x200...

blog.kasson.com

Jim

 

[exdeejjjaaaa]

So, the DR of A6300 is better from ISO 400 onwards.

A7, A7II are using the sensor with the design that is several years old... but even then S/N __above deep__ shadows for FF will be better.

 

[Seeky]

So, the DR of A6300 is better from ISO 400 onwards. But what about noise compared to A7II?

Engineering dynamic range is fullscale/read noise. So EDR and read noise are two sides of the same coin. If the EDR is higher, the RN is lower, if fullscale is the same. And, when the a7II is in compressed mode, fullscale is unarguably the same. When the a7II is in uncompressed mode, you could argue that fullscale is different, but it turns out that, whichever number you pick for fullscale with the a7II, RN sales with that, so that EDR turns out to be the same for the a7II whether it's in 14-bit uncompressed mode or 13-bit compressed mode.

It's explained here:

http://blog.kasson.com/?p=12272

Jim

 

[JimKasson]

So, the DR of A6300 is better from ISO 400 onwards. But what about noise compared to A7II?

Engineering dynamic range is fullscale/read noise. So EDR and read noise are two sides of the same coin. If the EDR is higher, the RN is lower, if fullscale is the same. And, when the a7II is in compressed mode, fullscale is unarguably the same. When the a7II is in uncompressed mode, you could argue that fullscale is different, but it turns out that, whichever number you pick for fullscale with the a7II, RN sales with that, so that EDR turns out to be the same for the a7II whether it's in 14-bit uncompressed mode or 13-bit compressed mode.

It's explained here:

http://blog.kasson.com/?p=12272

Jim

So, can we classify the sensor in the a6300 as the first to beat FF sensors from around the same era?

I wouldn't say they were from around the same era; the a7II sensor was a carryover from the a7. And the a7 still has better EDR at loe ISO.

Jim

 

[JimKasson]

So, the DR of A6300 is better from ISO 400 onwards.

A7, A7II are using the sensor with the design that is several years old... but even then S/N __above deep__ shadows for FF will be better.

A bit better. The a7II FWC is a little over 60000 electrons as measured by me:

Modeling the a7II FWC and read noise for many ISOs - the last word

Now that I've taught DCRAW and my Matlab analysis program to be on the same page with the a7II files, I set about to find the read noise (RN) and the full well capacity (FWC) of the camera across a range of ISOs. I told the modeling program to consider all ISOs from 100 to

blog.kasson.com

and the 6300 about 45000 electrons, as measured by Bill Claff with data from me.

So, in regions where read noise and PRNU are not considerations, the signal to noise ratio of the a7II will be about log2(sqrt(61500/45000)) = 0.23 stops better.

Another way of looking at it is that the exposure would have to be log2(61500/45000) = 0.45 stops more for the a6300 to have the same SNR as the a7II.

Jim

 

[bclaff]

So, the DR of A6300 is better from ISO 400 onwards.

A7, A7II are using the sensor with the design that is several years old... but even then S/N __above deep__ shadows for FF will be better.

A bit better. The a7II FWC is a little over 60000 electrons as measured by me:

http://blog.kasson.com/?p=8169

and the 6300 about 45000 electrons, as measured by Bill Claff with data from me.

So, in regions where read noise and PRNU are not considerations, the signal to noise ratio of the a7II will be about log2(sqrt(61500/45000)) = 0.23 stops better.

Another way of looking at it is that the exposure would have to be log2(61500/45000) = 0.45 stops more for the a6300 to have the same SNR as the a7II.

My numbers are 59,267e- and 48,023e- for 0.15 stops and 0.30 stops respectively.
(Your value could be more accurate. I provide mine for an apples to apples computation.)

Regards,

 

[Jefftan]

The discussion is getting complicated

So does it mean that at base ISO (or up to ISO 400), A7II still has lower noise and more dynamic range but after ISO 400 the table turn and A6300 better?

If this is so than even if A6300 still lag full frame in low ISO , it is a very significant breakthrough for APS-C

A6300 definitely the most advance APS-C sensor now

 

[vett93]

When I first started messing with my a6300 files in Sony IDC (Now LR and C1 support it) I was shocked by how pleasing the noise looked compared to my old a6000 and thought to myself that it looked as good as my full frame A7ii files.

So this morning I shot the same thing with the A6300 and A7ii. Just looking at the grain / noise in each photo... and I remain shocked by how good the A6300 noise looks at higher ISO's. I don't know what voodoo Sony did to it, but it looks fantastic compared to a Full Frame sensor and if anything "I think" the a6300 noise may be less clumpy than the A7ii's.

A7ii with Green tag, A6300 with Red:



--
Sony A7ii A7s A6000
28 FE f2 | Sony 55 FE 1.8 | 70-200G FE F4 | 24-240 FE | Sony 90mm FE Macro
http://501concepts.com

To me, the images from A7II look cleaner from 1000 to 6400. Anyone else see that too or I need to calibrate my eyes?

--
"Keep calm and take photos"
Photography enthusiast, from 12mm to 500mm

 

[Seeky]

When I first started messing with my a6300 files in Sony IDC (Now LR and C1 support it) I was shocked by how pleasing the noise looked compared to my old a6000 and thought to myself that it looked as good as my full frame A7ii files.

So this morning I shot the same thing with the A6300 and A7ii. Just looking at the grain / noise in each photo... and I remain shocked by how good the A6300 noise looks at higher ISO's. I don't know what voodoo Sony did to it, but it looks fantastic compared to a Full Frame sensor and if anything "I think" the a6300 noise may be less clumpy than the A7ii's.

A7ii with Green tag, A6300 with Red:



--
Sony A7ii A7s A6000
28 FE f2 | Sony 55 FE 1.8 | 70-200G FE F4 | 24-240 FE | Sony 90mm FE Macro
http://501concepts.com

To me, the images from A7II look cleaner from 1000 to 6400. Anyone else see that too or I need to calibrate my eyes?

--
"Keep calm and take photos"
Photography enthusiast, from 12mm to 500mm

Yes, I see that too; it struck me how biased I was in expecting the A6300 to rival the A7II, but the A7II shows less noise.

Aside from that, I'm not sure I would judge anything from this image since focus seems variable and most part in this crop is OOF.

--
Flickr: https://www.flickr.com/photos/keesw/

 

[HFLM]

The discussion is getting complicated

So does it mean that at base ISO (or up to ISO 400), A7II still has lower noise and more dynamic range but after ISO 400 the table turn and A6300 better?

If this is so than even if A6300 still lag full frame in low ISO , it is a very significant breakthrough for APS-C

A6300 definitely the most advance APS-C sensor now

I don't know whether it is more advanced, since I find the D7200 images, for example, to show less noise in the DPreview Raw comparison. Measurements will certainly shed more light on this. Additionally, why not comparing to recent FF (A7rii, A7s) and APSC sensors (D500 will certainly be a good candidate)? The A7ii is, in my and many other's opinion, not in the same ballpark as a D750, for example, and represents one of Sony's weaker sensors.

 

[JimKasson]

So, the DR of A6300 is better from ISO 400 onwards.

A7, A7II are using the sensor with the design that is several years old... but even then S/N __above deep__ shadows for FF will be better.

A bit better. The a7II FWC is a little over 60000 electrons as measured by me:

http://blog.kasson.com/?p=8169

and the 6300 about 45000 electrons, as measured by Bill Claff with data from me.

So, in regions where read noise and PRNU are not considerations, the signal to noise ratio of the a7II will be about log2(sqrt(61500/45000)) = 0.23 stops better.

Another way of looking at it is that the exposure would have to be log2(61500/45000) = 0.45 stops more for the a6300 to have the same SNR as the a7II.

My numbers are 59,267e- and 48,023e- for 0.15 stops and 0.30 stops respectively.
(Your value could be more accurate. I provide mine for an apples to apples computation.)

Thanks for chiming in, BIll. I just finished some FWC calculations for the a6300, using the 128 single shot shutter mode images that I use for PRNU and considering them as 64 pairs. I get different answers for the ISO 100 images and the ISO 400 images:


FWC from ISO 100 images, RGGB, in electrons


FWC from ISO 400 images, RGGB, in electrons

This is using the standard camera model, which doesn't know that the conversion gain (and thus, in fact, the FWC) changes at ISO 400.

There doesn't seem to be a consistent pattern to the two data sets.

This was kind of a quick and dirty FWC analysis (Inspired by you, I am trying to see if I can reduce the number of images required to characterize a camera), and, if there's sensor nonlinearity at these high mean signal levels the results could be wrong.

Jim

--
http://blog.kasson.com

 

[jchapell]

--
Sony A7ii A7s A6000
28 FE f2 | Sony 55 FE 1.8 | 70-200G FE F4 | 24-240 FE | Sony 90mm FE Macro
http://501concepts.com

This is very interesting to me - after I switched from the a6000 to the A7, I noticed the ability to push the images in post significantly more. I’ve tried to educate myself from jimkasson’s site over time, but I assume this is a combination of the better noise handling, dynamic range, as well as ISO invariance.

- Have you (or anyone) tested the iso invariance of the a6300 sensor?

- Could you (or anyone) include a comparison showing an image that has been pushed several stops in post? It comes down to more than noise, a6000 images that have been pushed hard gain odd color casts as well as noise compared to the A7/A7II.

Thank you for the interesting discussion! As a wilderness/adventure guy, if the a6300 performed in line with the a7 I would jump ship to get back to my beloved 10-18mm and 12mm F2 that were so teeny tiny and excellent.



--
Jordan
St. Louis, MO

Jordan Chapell

Jordan Chapell Photography: Landscape, wilderness, and adventure photography from National parks, Colorado 14ers, alpine mountaineering, hiking, climbing, and backpacking.

www.jordanchapell.com

 

[JimKasson]

This is very interesting to me - after I switched from the a6000 to the A7, I noticed the ability to push the images in post significantly more. I’ve tried to educate myself from jimkasson’s site over time, but I assume this is a combination of the better noise handling, dynamic range, as well as ISO invariance.

- Have you (or anyone) tested the iso invariance of the a6300 sensor?

Yes. It is not ISo invariant, since, like the a7RII and the a7S, the conversion gains changes with ISO. In the a6300 the low conversion gains settings are ISO 320 and below, and the high ones are ISO 400 and above.

- Could you (or anyone) include a comparison showing an image that has been pushed several stops in post?

Here are some such images with the a6300:

Sony a6300 — pushing in post at low ISOs, single shot shutter - the last word

This is part of a long series of posts about the Sony a6300. The series starts here. So now I'm going to do a series like that of the preceding post, but with a big difference. I'm going to leave the exposure the same, vary the ISO in full stop intervals, and push the images

blog.kasson.com

Sony a6300 — pushing in post at higher ISOs, single shot shutter - the last word

This is part of a long series of posts about the Sony a6300. The series starts here. Now I'm going to do a test similar to the one in the preceding post, except that I'm going to start at ISO 12800. That's the first ISO where the camera does some big time spatial frequency for

blog.kasson.com

Sony a6300 — more pushing in post at higher ISOs, single shot shutter - the last word

This is part of a long series of posts about the Sony a6300. The series starts here. In the preceding post, I showed a series of images with identical exposures at decreasing ISO settings with pushes in post processing to make the brightness the same. I started at ISO 12800 and worked down to a

blog.kasson.com

Jim

 

[jchapell]

This is very interesting to me - after I switched from the a6000 to the A7, I noticed the ability to push the images in post significantly more. I’ve tried to educate myself from jimkasson’s site over time, but I assume this is a combination of the better noise handling, dynamic range, as well as ISO invariance.

- Have you (or anyone) tested the iso invariance of the a6300 sensor?

Yes. It is not ISo invariant, since, like the a7RII and the a7S, the conversion gains changes with ISO. In the a6300 the low conversion gains settings are ISO 320 and below, and the high ones are ISO 400 and above.

- Could you (or anyone) include a comparison showing an image that has been pushed several stops in post?

Here are some such images with the a6300:

http://blog.kasson.com/?p=13933

http://blog.kasson.com/?p=13956

http://blog.kasson.com/?p=13967

Jim

 

[JimKasson]

I just finished some FWC calculations for the a6300, using the 128 single shot shutter mode images that I use for PRNU and considering them as 64 pairs. I get different answers for the ISO 100 images and the ISO 400 images:


FWC from ISO 100 images, RGGB, in electrons


FWC from ISO 400 images, RGGB, in electrons

This is using the standard camera model, which doesn't know that the conversion gain (and thus, in fact, the FWC) changes at ISO 400.

There doesn't seem to be a consistent pattern to the two data sets.

This was kind of a quick and dirty FWC analysis (Inspired by you, I am trying to see if I can reduce the number of images required to characterize a camera), and, if there's sensor nonlinearity at these high mean signal levels the results could be wrong.

I was rewriting the Matlab Photon Transfer code to take multiple samples from a xy gradient image so I can do PDR with fewer shots, and I aimed the code at the same image files as for the above, with a 4x6 grid of 400x400 pixel (in each raw plane) samples, for a total of 24x64 = 1536 400x400 samples.

Here's what I got with that:


ISO 100, single shot shutter mode


ISO 400, single shot shutter mode.

Now i get essentially the same result for both ISO settings.

Jim

--
http://blog.kasson.com