Analysis of RAW samples at Fred Miranda show weak DR

[ahsanford]

Re: DRpocalypse 2017

However running with this faster readout rate causing the noise difference, and just playing devil's advocate here, how do we explain the 5D4 killing it in DR at base ISO? 30 MP x 7 fps requires an even faster readout than the 6D2. Also a large jump from the 22 MP x 6 fps of the 5D3 and we all know the huge DR jump we saw there. I still find it odd that the 6D2 would not see a similar benefit and I can't imagine Canon would intentionally hold back their latest sensor tech.

Can anyone explain that to me?

A fair question.

A quick before/after glance at the three on-chip upgrades has shown a semi-consistent base ISO DR bump for cameras that received very different throughput increases. (Someone please proof that for errors, I banged that out in a hurry.)

So if the data is representative of a real production model -- and I'm not there yet -- I would not blame throughput for the DR result. I'd ask if the 6D2 was somehow left out of the on-chip ADC party, which seems unthinkable (hence my 'not being there yet').

- A

 

[x-vision]

Re: Bill Claff Chimes In

However running with this faster readout rate causing the noise difference, and just playing devil's advocate here, how do we explain the 5D4 killing it in DR at base ISO?

...

Can anyone explain that to me?

By the look of things, the 6DII does not have on-sensor ADCs.

This will be easily confirmed when someones does a teardown of the camera (e.g. LensRentals).
The external ADC chips, likely a pair of them on the 6DII, are easily identifiable.

 

[x-vision]

Re: DRpocalypse 2017

I'd ask if the 6D2 was somehow left out of the on-chip ADC party, which seems unthinkable (hence my 'not being there yet').

See my previous post above .

 

[x-vision]

Here are the four external ADCs on the 5DS (source: LensRentals):

 

[BillB]

Re: Bill Claff Chimes In

However running with this faster readout rate causing the noise difference, and just playing devil's advocate here, how do we explain the 5D4 killing it in DR at base ISO?

...

Can anyone explain that to me?

By the look of things, the 6DII does not have on-sensor ADCs.

This will be easily confirmed when someones does a teardown of the camera (e.g. LensRentals).
The external ADC chips, likely a pair of them on the 6DII, are easily identifiable.

If the data are correct, off sensor ADC might be an explanation. Won't have to wait much longer to find out.

 

[Aglet]

Re: Bill Claff Chimes In

It doesn't make sense to me either. 5D IV has a faster readout as you pointed out. Even 80D surpasses 6D in terms of DR. And Canon itself admitted that 6D II is made with the same tech as its latest sensors. So unless there are some shenanigans going on, either market preserving of fault in measuring, this is simply mind boggling. I'm patiently waiting for more reviews with retail cameras and more importantly, using updated and various software.

other than maybe saving on-sensor ADC for the 6d3...
it's still possible that demo 6d2 was made with previous tech just to sort out features and things not specific to the sensor. The one shred of hope.

 

[LonelyBoy]

Re: DRpocalypse 2017

So if the data is representative of a real production model -- and I'm not there yet -- I would not blame throughput for the DR result. I'd ask if the 6D2 was somehow left out of the on-chip ADC party, which seems unthinkable (hence my 'not being there yet').

- A

Not only that, but haven't people here been saying Canon confirmed the 6D2 got their modern sensor design? I didn't read those quotes myself, but I don't follow as closely as some. And like you said, it seems unthinkable that the T7i got the love and the 6D2 didn't. It makes me wonder if there are engineering mules out there that were testing other aspects of the body or electronics and happened to have old sensors in them, and those are the source of the leaked raws. Either way, I will be fascinated to see the outcome.

 

[CanonCams]

I went through 9 pages of this topic and nobody talks about high-ISO. If Fred Miranda is right, 6DII should be the best low light camera in the world. ...

That's a good point but a bit of hyperbole there.
Below the PDR chart at PhotonsToPhotos there is a sortable table.
My Low Light ISO is analygous to but on a different scale than DxOMark Sports.
It looks like the 6D Mark II low light performance is just a bit better than the 6D.
Remember to look at the stops rather than just the ISO value.

As per that chart, the MK II is the worst on the list for DR

 

[ahsanford]

I keep hearing increased throughput can adversely impact (was it?) read noise. The A9 somewhat underwhelmed recently versus the A7 line, and some folks attributed that to the face-melting 484 MP/s it was moving.

So, here's my knuckleheaded please-educate-me question: does the resulting read noise that results from the max throughput of the rig affect every still the sensor takes, even if it's not shooting at high fps? Or is the sensor globally locked in to a certain level that affects all stills of the same ISO identically?

I'm completely spitballing if the RAW file everyone has been studying might have been from high FPS or quick bracketed trio that might not reflect (most) general use. I'm assuming that's a dead end, but I thought I'd ask.

- A

 

[ahsanford]

Re: Bill Claff Chimes In

If the data are correct, off sensor ADC might be an explanation. Won't have to wait much longer to find out.

Why would Canon not put on-chip on the 6D2?

Before I can even type "Why would the 6D2 not get something the 80D/M5/M6 has received?", there have been some instances of crop / cheaper cameras getting some niceties that the 6D line didn't -- 1/8000s shutters, 1/200 flash sync, higher fps, etc. but we could associate that with the added difficulty/cost to pull that level of performance off on a FF rig vs. a crop one.

e.g. In those three metrics above, you need a larger, more robust shutter, mirror box, etc., so it's a painful cost/weight/complexity delta that Canon feels is worthy for the 1D/5D lines but not to the 6D. Makes perfect sense to me.

But would the same logic apply to sensors?, i.e. due to smaller size and higher production volumes, you can offer on-chip ADC in an 80D but not for a 6D2?

I look to the far brighter people here than I to answer that.

- A

 

[bclaff]

I keep hearing increased throughput can adversely impact (was it?) read noise. The A9 somewhat underwhelmed recently versus the A7 line, and some folks attributed that to the face-melting 484 MP/s it was moving.

So, here's my knuckleheaded please-educate-me question: does the resulting read noise that results from the max throughput of the rig affect every still the sensor takes, even if it's not shooting at high fps? Or is the sensor globally locked in to a certain level that affects all stills of the same ISO identically?

I'm completely spitballing if the RAW file everyone has been studying might have been from high FPS or quick bracketed trio that might not reflect (most) general use. I'm assuming that's a dead end, but I thought I'd ask.

- A

It's a reasonable conjecture; but wrong.
The read-out rate isn't even a function of the ISO setting.
On some cameras it's affected by bit depth (Nikon 12-bit versus 14-bit), shutter mode (mechanical versus electronic), etc.
In the case of the 6D Mark II there's no reason to think it varies at all.

 

[x-vision]

Re: DRpocalypse 2017

Not only that, but haven't people here been saying Canon confirmed the 6D2 got their modern sensor design?

From Canon CPN Europe:

• Low light shooting has been further improved thanks to the EOS 6D Mark II using similar sensor technology as found on the award-winning EOS 5D Mark IV and EOS-1D X Mark II DSLRs.

So, a similar technology, not the same technology 8).

 

[bclaff]

Re: DRpocalypse 2017

However running with this faster readout rate causing the noise difference, and just playing devil's advocate here, how do we explain the 5D4 killing it in DR at base ISO? 30 MP x 7 fps requires an even faster readout than the 6D2. Also a large jump from the 22 MP x 6 fps of the 5D3 and we all know the huge DR jump we saw there. I still find it odd that the 6D2 would not see a similar benefit and I can't imagine Canon would intentionally hold back their latest sensor tech.

Can anyone explain that to me?

A fair question.

A quick before/after glance at the three on-chip upgrades has shown a semi-consistent base ISO DR bump for cameras that received very different throughput increases. (Someone please proof that for errors, I banged that out in a hurry.)

So if the data is representative of a real production model -- and I'm not there yet -- I would not blame throughput for the DR result. I'd ask if the 6D2 was somehow left out of the on-chip ADC party, which seems unthinkable (hence my 'not being there yet').

- A

Nice table.
Note that in the other three cases there was a clear change in Analog to Digital Converter (ADC) strategy.
Off-chip have a roll-off at low ISO settings and on-chip are more straight.

Still begs the question why the 6D Mark II seems to have off-chip ADC
The attached chart shows only 4 cameras rather than 6 for clarity.

 

[CanonCams]

Re: DRpocalypse 2017

Not only that, but haven't people here been saying Canon confirmed the 6D2 got their modern sensor design?

From Canon CPN Europe:

• Low light shooting has been further improved thanks to the EOS 6D Mark II using similar sensor technology as found on the award-winning EOS 5D Mark IV and EOS-1D X Mark II DSLRs.

So, a similar technology, not the same technology 8).

Similar meaning worse than 6D?

 

[Luds34]

Re: Bill Claff Chimes In

However running with this faster readout rate causing the noise difference, and just playing devil's advocate here, how do we explain the 5D4 killing it in DR at base ISO?

...

Can anyone explain that to me?

By the look of things, the 6DII does not have on-sensor ADCs.

This will be easily confirmed when someones does a teardown of the camera (e.g. LensRentals).
The external ADC chips, likely a pair of them on the 6DII, are easily identifiable.

This data definitely points to that. It is the most logical explanation given exactly where the DR is falling.

However the decision to stick with off die ADCs on this camera doesn't make much sense to me from a strategy, company perspective. We'll know soon enough by I'm going to remain optimistic.

 

[Khalai]

Re: Bill Claff Chimes In

However running with this faster readout rate causing the noise difference, and just playing devil's advocate here, how do we explain the 5D4 killing it in DR at base ISO?

...

Can anyone explain that to me?

By the look of things, the 6DII does not have on-sensor ADCs.

This will be easily confirmed when someones does a teardown of the camera (e.g. LensRentals).
The external ADC chips, likely a pair of them on the 6DII, are easily identifiable.

This data definitely points to that. It is the most logical explanation given exactly where the DR is falling.

However the decision to stick with off die ADCs on this camera doesn't make much sense to me from a strategy, company perspective. We'll know soon enough by I'm going to remain optimistic.

If it's truly the reason for there preliminary tests, then I'm not hesitating to call this major design flaw from Canon. But same as you - I'm still remaining optimistic until there are several reviews from various sites and more importantly, from retail cameras and even more importantly, from real scene photos.

 

[Mikehit]

I seem to recall a similar pre-release discussion regards the (I think) 50D. People talked about how the 50D was significantly worse than the 40D regards pixel noise, but the sensor technology, the type of noise, and the processing added to make a better imaging machine overall. I wonder if the same thing is happening here?

 

[BillB]

Re: DRpocalypse 2017

However running with this faster readout rate causing the noise difference, and just playing devil's advocate here, how do we explain the 5D4 killing it in DR at base ISO? 30 MP x 7 fps requires an even faster readout than the 6D2. Also a large jump from the 22 MP x 6 fps of the 5D3 and we all know the huge DR jump we saw there. I still find it odd that the 6D2 would not see a similar benefit and I can't imagine Canon would intentionally hold back their latest sensor tech.

Can anyone explain that to me?

A fair question.

A quick before/after glance at the three on-chip upgrades has shown a semi-consistent base ISO DR bump for cameras that received very different throughput increases. (Someone please proof that for errors, I banged that out in a hurry.)

So if the data is representative of a real production model -- and I'm not there yet -- I would not blame throughput for the DR result. I'd ask if the 6D2 was somehow left out of the on-chip ADC party, which seems unthinkable (hence my 'not being there yet').

- A

Nice table.
Note that in the other three cases there was a clear change in Analog to Digital Converter (ADC) strategy.
Off-chip have a roll-off at low ISO settings and on-chip are more straight.

Still begs the question why the 6D Mark II seems to have off-chip ADC
The attached chart shows only 4 cameras rather than 6 for clarity.

The data looks like it came from a sensor without onboard ADC. It seems very likely that the 6DII sensor has onboard ADC. Soon we will have an answer to this Canondrum.

 

[Khalai]

Re: DRpocalypse 2017

The data looks like it came from a sensor without onboard ADC. It seems very likely that the 6DII sensor has onboard ADC. Soon we will have an answer to this Canondrum.

;D

*giggles*

 

[LonelyBoy]

Re: DRpocalypse 2017

Not only that, but haven't people here been saying Canon confirmed the 6D2 got their modern sensor design?

From Canon CPN Europe:

• Low light shooting has been further improved thanks to the EOS 6D Mark II using similar sensor technology as found on the award-winning EOS 5D Mark IV and EOS-1D X Mark II DSLRs.

So, a similar technology, not the same technology 8).

Yeah, but "similar" has meaning. It's not going to mean "well they're both digital imaging sensors", it means "they're broadly the same technology, but tuned differently and they obviously have different resolution".

At least, it would be an interesting false advertising case if it didn't.