90D pushed images look clean (no banding)

Started 11 months ago | Discussions
keyframe77 New Member • Posts: 20
90D pushed images look clean (no banding)
11

I case someone was wondering what they can expect after pushing their long exposure images by 5 stops, I am happy to report that the images exhibit no banding.

30 seconds @ ISO 100 +5 stops from raw in LR Classic with no corrections/NR:

Normal exposure (reduced size, for brightness reference only):

The only thing of note in my quick test was a small streak of what looks like "hot" pixels. This is with long exposure NR on (dark frame subtraction). There was typically only one and it was in a different place each time and had a different orientation:

Here it is on one of the sweaters in the closet (not in the image above):

Here's a couple on different pictures of the closet wall:

So I guess if you're shooting stars at 30-sec exposures, you'll have a few additional streaks in your images. But they will be easy to find, since their direction/orientation will be different from rest of the stars.

 keyframe77's gear list:keyframe77's gear list
Canon EOS 90D
Sharlin
Sharlin Contributing Member • Posts: 739
Re: 90D pushed images look clean (no banding)

Those look interesting. Almost like streaks caused by cosmic rays or other ionizing radiation. Difficult to explain just as a quirk of electronics due to their ”organic” look.

 Sharlin's gear list:Sharlin's gear list
Canon EOS 80D Canon EF 85mm F1.8 USM Sigma 18-35mm F1.8 DC HSM Art Canon EF 50mm F1.4 USM Sigma 10-20mm F4-5.6 EX DC HSM +3 more
1Dx4me Veteran Member • Posts: 9,051
Re: 90D pushed images look clean (no banding)

cool tests!

-- hide signature --

You miss 100 percent of the shots you didn't take!!! "Wayne Gretzky"

Dave
Dave Veteran Member • Posts: 5,280
Re: 90D pushed images look clean (no banding)

Almost something from nothing.

 Dave's gear list:Dave's gear list
Canon EOS 80D Canon EF 70-300mm f/4-5.6L IS USM Canon EF 135mm F2L USM Canon EF 100mm F2.8L Macro IS USM Canon EF-S 24mm F2.8 STM +9 more
Sharlin
Sharlin Contributing Member • Posts: 739
Re: 90D pushed images look clean (no banding)

Investigating a bit, seems likely that those streaks are caused by stray electrons.

 Sharlin's gear list:Sharlin's gear list
Canon EOS 80D Canon EF 85mm F1.8 USM Sigma 18-35mm F1.8 DC HSM Art Canon EF 50mm F1.4 USM Sigma 10-20mm F4-5.6 EX DC HSM +3 more
Traveller Veteran Member • Posts: 6,065
That is Pretty Amazing! Nice Test, Valuable Info Provided.n/t

Traveller

OP keyframe77 New Member • Posts: 20
Re: 90D pushed images look clean (no banding)

Sharlin wrote:

Investigating a bit, seems likely that those streaks are caused by stray electrons.

Wow. Thank you for looking into this, Sharlin. I never saw this with my now ancient EOS 20D. Does this mean that newer sensors are much more sensitive, then, to be able to pick up these things? I guess shorter exposures should have fewer and longer exposures more of these.

 keyframe77's gear list:keyframe77's gear list
Canon EOS 90D
OP keyframe77 New Member • Posts: 20
Re: That is Pretty Amazing! Nice Test, Valuable Info Provided.n/t

Traveller wrote:

Traveller

You're welcome, Traveller. I had this question before I bought mine and no answer was to be found. Sorry no EXIF. My LR export was set to strip that out. Fixed for future tests.

 keyframe77's gear list:keyframe77's gear list
Canon EOS 90D
Sharlin
Sharlin Contributing Member • Posts: 739
Re: 90D pushed images look clean (no banding)

keyframe77 wrote:

Wow. Thank you for looking into this, Sharlin. I never saw this with my now ancient EOS 20D. Does this mean that newer sensors are much more sensitive, then, to be able to pick up these things? I guess shorter exposures should have fewer and longer exposures more of these.

Well, have you ever tried pushing an almost black 20D photo by five stops as has been done here? I don't think the results are pretty, and these kind of transients are likely difficult to distinguish from the noise.

 Sharlin's gear list:Sharlin's gear list
Canon EOS 80D Canon EF 85mm F1.8 USM Sigma 18-35mm F1.8 DC HSM Art Canon EF 50mm F1.4 USM Sigma 10-20mm F4-5.6 EX DC HSM +3 more
OP keyframe77 New Member • Posts: 20
Re: 90D pushed images look clean (no banding)

My guess was right. Just tested 15, 8, and 4 second exposures in the same closet and no streaks to be found. What's interesting, however, is that I also took six 30-second exposures to test the effects of long exposure noise reduction (see https://www.dpreview.com/forums/post/63116504) and none of those shots had any streaks either. So if you're getting stray electrons in your images, just wait a little

 keyframe77's gear list:keyframe77's gear list
Canon EOS 90D
OP keyframe77 New Member • Posts: 20
Re: 90D pushed images look clean (no banding)

Sharlin wrote:

keyframe77 wrote:

Wow. Thank you for looking into this, Sharlin. I never saw this with my now ancient EOS 20D. Does this mean that newer sensors are much more sensitive, then, to be able to pick up these things? I guess shorter exposures should have fewer and longer exposures more of these.

Well, have you ever tried pushing an almost black 20D photo by five stops as has been done here? I don't think the results are pretty, and these kind of transients are likely difficult to distinguish from the noise.

Good point, Sharlin.

 keyframe77's gear list:keyframe77's gear list
Canon EOS 90D
Jim B (MSP) Forum Pro • Posts: 10,391
Re: 90D pushed images look clean (no banding)

Thanks - a good test and better results.

I think the 90D has some other secrets to share.

-- hide signature --

Jim
"It's all about the light"

 Jim B (MSP)'s gear list:Jim B (MSP)'s gear list
Canon EOS 90D
bclaff Forum Pro • Posts: 10,121
Re: 90D pushed images look clean (no banding)

Your result isn't too surprising to me.
Fixed Pattern Noise (FPN) testing at PhotonsToPhotos doesn't show any significant FPN in the Sensor Heatmaps (compared here to the 20D):

Best viewed at "Original Size"

-- hide signature --

Bill ( Your trusted source for independent sensor data at PhotonsToPhotos )

fredlord
fredlord Senior Member • Posts: 2,925
Thank you.

This is quite good compared to past sensor problems.

-- hide signature --

Fred Lord

 fredlord's gear list:fredlord's gear list
Canon EOS 5DS R Canon EOS 5D Mark IV Canon EOS 90D Canon EF 16-35mm F4L IS USM Canon TS-E 17mm f/4L +9 more
John Sheehy Forum Pro • Posts: 23,541
Re: 90D pushed images look clean (no banding)

Sharlin wrote:

Those look interesting. Almost like streaks caused by cosmic rays or other ionizing radiation. Difficult to explain just as a quirk of electronics due to their ”organic” look.

Well, if the noise impulses only occurred at the pixel level with no correlation to other pixels, even read noise would have very weak (1/f) low-frequency noise, and no one-dimensional noise (banding). There is really not much difference between simple, synthetic Gaussian noise lacking in correlation, vs photon noise, for samples where the average number of photons goes above 12. In fact, code which synthesizes Poisson noise usually uses Gaussian noise for any samples with populations of photos greater than 12. I once re-wrote such an algorithm such that it actually calculated a Poisson value for any source population, and it became very, very slow with high populations, and didn't look very different.

John Sheehy Forum Pro • Posts: 23,541
Re: 90D pushed images look clean (no banding)

Dave wrote:

Almost something from nothing.

Well, even with the poor quantum efficiency in current color cameras due to CFA filters, the photon noise that results from exposure is very fine and random, and cleans up very well with filtering or by reducing display size with a good resampling method. It is the added noises from reading the photosites after the initial capture with its photon noise, which is where most of the chunkier noise and banding exists, which survive filtering or display size reduction.

It's more like "not dumping a lot of ugly crap" on the relatively weak photon capture, than it is getting something from nothing. "Nothing", photons-wise, isn't approached until ISOs get into the billions, for an APS-C sensor.

Sharlin
Sharlin Contributing Member • Posts: 739
Re: 90D pushed images look clean (no banding)

John Sheehy wrote:

Well, if the noise impulses only occurred at the pixel level with no correlation to other pixels, even read noise would have very weak (1/f) low-frequency noise, and no one-dimensional noise (banding). There is really not much difference between simple, synthetic Gaussian noise lacking in correlation, vs photon noise, for samples where the average number of photons goes above 12. In fact, code which synthesizes Poisson noise usually uses Gaussian noise for any samples with populations of photos greater than 12. I once re-wrote such an algorithm such that it actually calculated a Poisson value for any source population, and it became very, very slow with high populations, and didn't look very different.

I’m not sure what your point is; the streaks I referred to are obviously highly spatially correlated but not in the way FPN or banding would be. They are randomly oriented and squiggly instead of straight, which is what I meant by ”organic-looking”.

 Sharlin's gear list:Sharlin's gear list
Canon EOS 80D Canon EF 85mm F1.8 USM Sigma 18-35mm F1.8 DC HSM Art Canon EF 50mm F1.4 USM Sigma 10-20mm F4-5.6 EX DC HSM +3 more
John Sheehy Forum Pro • Posts: 23,541
Re: 90D pushed images look clean (no banding)

Sharlin wrote:

John Sheehy wrote:

Well, if the noise impulses only occurred at the pixel level with no correlation to other pixels, even read noise would have very weak (1/f) low-frequency noise, and no one-dimensional noise (banding). There is really not much difference between simple, synthetic Gaussian noise lacking in correlation, vs photon noise, for samples where the average number of photons goes above 12. In fact, code which synthesizes Poisson noise usually uses Gaussian noise for any samples with populations of photos greater than 12. I once re-wrote such an algorithm such that it actually calculated a Poisson value for any source population, and it became very, very slow with high populations, and didn't look very different.

I’m not sure what your point is; the streaks I referred to are obviously highly spatially correlated but not in the way FPN or banding would be. They are randomly oriented and squiggly instead of straight, which is what I meant by ”organic-looking”.

I didn't realize at the time that you were talking specifically about the streaks; they are not what one would normally call "noise".

A bit more work would be required to figure out exactly what they are.

Sharlin
Sharlin Contributing Member • Posts: 739
Re: 90D pushed images look clean (no banding)

John Sheehy wrote:

I didn't realize at the time that you were talking specifically about the streaks; they are not what one would normally call "noise".

That's why I didn't use the word "noise" anywhere But yeah, I could have made more clear what I was referring to.

 Sharlin's gear list:Sharlin's gear list
Canon EOS 80D Canon EF 85mm F1.8 USM Sigma 18-35mm F1.8 DC HSM Art Canon EF 50mm F1.4 USM Sigma 10-20mm F4-5.6 EX DC HSM +3 more
Keyboard shortcuts:
FForum MMy threads