Home · Register · Join Upload & Sell

Moderated by: Fred Miranda

New fredmiranda.com Mobile Site

New Feature: SMS Notification alert

New Feature: Buy & Sell Watchlist

FM Forums | Canon Forum | Join Upload & Sell

3 4 end

R5's ISO-Invariance - a happy discovery for shooting Owls at dawn and dusk

Jeff
Offline
• • • • •
Upload & Sell: On

p.2 #1 · p.2 #1 · R5's ISO-Invariance - a happy discovery for shooting Owls at dawn and dusk

RoamingScott wrote:
In case you don't know, you should typically either shoot at either base ISO or ISO 400 on the R5 to maximize recovery via the dual gain properties of the sensor. This applies to every dual gain sensor, but the 2nd optimized ISO value (not base) will vary by body.

If I did this with my R5 I'd end up with pattern noise (term used loosely, i.e. horizontal banding) that would ruin the image. I believe the R5 is best shot around ISO 1600 for night images.

Nov 22, 2023 at 06:18 AM

Rudy Pohl
Offline
• •
Upload & Sell: Off

p.2 #2 · p.2 #2 · R5's ISO-Invariance - a happy discovery for shooting Owls at dawn and dusk

Nov 22, 2023 at 06:46 AM

alundeb
Offline
• • • • •
Upload & Sell: Off

p.2 #3 · p.2 #3 · R5's ISO-Invariance - a happy discovery for shooting Owls at dawn and dusk

Thanks for the examples! The color cast in the recovered ISO 1600 image concerns me.

Nov 22, 2023 at 06:49 AM

Rudy Pohl
Offline
• •
Upload & Sell: Off

p.2 #4 · p.2 #4 · R5's ISO-Invariance - a happy discovery for shooting Owls at dawn and dusk

Here's a closeup comparison. Please note the fine highlight feather details in the face. The underexposed-recovered image can by finessed in Photoshop into an half-decent acceptable image while the ISO 25,600 cannot.

Nov 22, 2023 at 07:01 AM

Rudy Pohl
Offline
• •
Upload & Sell: Off

p.2 #5 · p.2 #5 · R5's ISO-Invariance - a happy discovery for shooting Owls at dawn and dusk

alundeb wrote:
Thanks for the examples! The color cast in the recovered ISO 1600 image concerns me.

There is no ISO 1600 image, only an ISO 1000 image and an ISO 25,600 image.

Nov 22, 2023 at 07:03 AM

alundeb
Offline
• • • • •
Upload & Sell: Off

p.2 #6 · p.2 #6 · R5's ISO-Invariance - a happy discovery for shooting Owls at dawn and dusk

Rudy Pohl wrote:
There is no ISO 1600 image, only an ISO 1000 image and an ISO 25,600 image.
Sorry for the typo !
I see what you mean, the shutter speeds are different so the ISO 25600 image has less exposure than the ISO 1000 image. Hence the recovered ISO 1000 image has less noise.

Nov 22, 2023 at 07:06 AM

Rudy Pohl
Offline
• •
Upload & Sell: Off

p.2 #7 · p.2 #7 · R5's ISO-Invariance - a happy discovery for shooting Owls at dawn and dusk

alundeb wrote:
Sorry for the typo !
I see what you mean, the shutter speeds are different so the ISO 25600 image has less exposure than the ISO 1000 image. Hence the recovered ISO 1000 image has less noise.

Yes, that's correct which is why I did not post the ISO 25,600 because I didn't do a properly controlled comparison. However, I was since asked to post it so I have done so.

Nevertheless, it appears from the posts here and over at DP Review where I also posted this thread several people have experimented with the Canon R5 and R6 and have found that they get the best results in low-light shooting when limiting the ISO to 1600-3200 and letting the image be underexposed, and then recovering the lacking exposure in post.

EDIT:
A final thought: If we were take the underexposed image of ISO 1000 and multiple that ISO 1000 value 4.5 times we would end up with an ISO 24,000 despite being 1/400 secs. So I think the bottom line is that the R5 doesn't like these high ISO values regardless what shutter speeds are.

Rudy

Nov 22, 2023 at 07:14 AM

jedibrain
Offline
• • •
Upload & Sell: Off

p.2 #8 · p.2 #8 · R5's ISO-Invariance - a happy discovery for shooting Owls at dawn and dusk

exdeejjjaaaa wrote:
R5 is NOT dual gain sensor... it just Canon's tricks with forced NR applied @ nominal ISO values <= 800... compare MS vs ES where Canon's can't hide the fact

https://i.postimg.cc/BQCKhhMM/R5-NOT-DUAL-GAIN.png

usual classic Canon's see-saw which Canon makes look like a proper Sony dual gain sensor using NR

here is a real dual gain = MS or ES - same sh$t ... no NR detected

https://www.photonstophotos.net/Charts/PDR.htm#Sony%20ILCE-7RM3,Sony%20ILCE-7RM3(ES)

https://i.postimg.cc/d0hZVJvx/Sony-real-deal.png">

ISO invariance is investigated on the input inferred read noise chart, not on the DR chart. DR HAS to decrease as ISO increases because ISO is a multiplicative gain. If you receive 1 photon to the sensor, and at base ISO put one electron in the read-well, you can fit 65k photons in the well. If you are at ISO 1000 and put 1000 electrons in for every photon you receive, now you can only fit 6.5k photons in the read-well. (made up numbers for ISO and gain, for example,. But as ISO goes up, the gain applied also goes up)

Look here for ISO invariance. You're looking for a point where the line goes flat. That means there is no more or less noise captured at the next ISO setting, which means the results will be identical if you push exposure in post vs. recording in the camera. R5 is close to invariance after ISO 400. R6 is not. The astrophotography guys I hang out with on other forums educated me on all of this.

https://www.photonstophotos.net/Charts/RN_e.htm

Brian

Edited on Nov 22, 2023 at 08:15 AM · View previous versions

Nov 22, 2023 at 08:13 AM

jedibrain
Offline
• • •
Upload & Sell: Off

p.2 #9 · p.2 #9 · R5's ISO-Invariance - a happy discovery for shooting Owls at dawn and dusk

I should add, the value of shooting at the lower ISO in this case is you retain the DR associated with that ISO.

Nov 22, 2023 at 08:14 AM

Jeff
Offline
• • • • •
Upload & Sell: On

p.2 #10 · p.2 #10 · R5's ISO-Invariance - a happy discovery for shooting Owls at dawn and dusk

jedibrain wrote:
Look here for ISO invariance. You're looking for a point where the line goes flat. That means there is no more or less noise captured at the next ISO setting, which means the results will be identical if you push exposure in post vs. recording in the camera. R5 is close to invariance after ISO 400. R6 is not. The astrophotography guys I hang out with on other forums educated me on all of this.

https://www.photonstophotos.net/Charts/RN_e.htm

Brian

Generally agree with everything you wrote, and probably somewhat semantic, but I'd consider the R5 nearing invariant at ISO 800, since noise reduction is being applied below that (sort of apples:oranges). If you really want to get picky, I think ISO 1600 is where most of us astro people shoot our R5s at (generally speaking), though could probably live with ISO 800 unless we wanted to be able to effectively preview the image on the LCD in the field. So many variables, especially when trying to compare different cameras.

The linked PhotographyLife article is pretty good, but mostly ignored the significance of dark current which can muddle the waters for 'stats' on older cameras (for example, my 5Ds is best shot at ISO 4000 to overcome pretty horrendous horizontal banding, compared to ISO 1600-3200). Not sure where on PhotonsToPhotos this would be revealed in a graph.

Nov 22, 2023 at 09:39 AM

Jeff
Offline
• • • • •
Upload & Sell: On

p.2 #11 · p.2 #11 · R5's ISO-Invariance - a happy discovery for shooting Owls at dawn and dusk

Jeff wrote:
Please post the ISO 25600 image for comparison, otherwise the main point of the thread is lost. I would be curious to see the difference (to see what you saw to prompt you to post).

Rudy Pohl wrote:
OK Jeff,

Here you go.
https://live.staticflickr.com/65535/53348542336_777e4bf33e_k.jpg

Thanks for that, very helpful even taking into consideration the differences in exposure.

Just out of curiosity, I assume you were using either ES or MS for both examples, yes? If not, I believe that using ES for (only) the higher-ISO image could produce at least some of the differences in noise characteristics.

Nov 22, 2023 at 09:43 AM

Jeff Nolten
Offline
• • • •
Upload & Sell: Off

p.2 #12 · p.2 #12 · R5's ISO-Invariance - a happy discovery for shooting Owls at dawn and dusk

Noob question: So does this mean I can get away with using my f4 zoom for opportunistic night sky shots rather than carry an extra f2.8 prime?

Nov 22, 2023 at 09:49 AM

Rudy Pohl
Offline
• •
Upload & Sell: Off

p.2 #13 · p.2 #13 · R5's ISO-Invariance - a happy discovery for shooting Owls at dawn and dusk

Jeff wrote:
Thanks for that, very helpful even taking into consideration the differences in exposure.

Just out of curiosity, I assume you were using either ES or MS for both examples, yes? If not, I believe that using ES for (only) the higher-ISO image could produce at least some of the differences in noise characteristics.

Hi Jeff, yes electronic shutter for both.
Rudy

Nov 22, 2023 at 10:00 AM

Scott Stoness
Offline
• • • • • •
Upload & Sell: On

p.2 #14 · p.2 #14 · R5's ISO-Invariance - a happy discovery for shooting Owls at dawn and dusk

A conclusion of shooting lower iso and boosting, suggests that Raw is not Raw?

In theory Raw, should permit achievement of the same results. Because Raw would record the original photons, not the amplified ones ?

Thus the difference would be whether you shot at longer (lower noise because more time to bring in photons but more risk of motion blur) or shorter shutter (higher noise) speeds?

Perhaps this is just a weakness of whatever Raw converter (Adobe, DPP...) that you might be using ?

Not saying the above is true but just trying to understand.

Nov 22, 2023 at 10:11 AM

arbitrage
Offline
• • • • • • •
Upload & Sell: On

p.2 #15 · p.2 #15 · R5's ISO-Invariance - a happy discovery for shooting Owls at dawn and dusk

Your examples show what you are claiming but based on data this shouldn't be the case. At best, the low ISO shot pushed up to match the 25,600 exposure will be equal but often once you get into the 4+ stop pushes the proper ISO file sometimes will win out. You are seeing the opposite of that.

You can see controlled studio results of this on DPReview's R5 review in the Dynamic Range ISO Invariance test charts. Their charts use 6400 and compare lower ISO pushes to that. At the dual gain ISO of 400, pushed 4 stops to match the ISO 6400 shot the IQ and DR are identical. They don't offer the ability to test against a 25,600 shot so maybe Adobe is messing up the NR on the ISO 25,600 shot?? Not sure.

Nov 22, 2023 at 10:14 AM

alundeb
Offline
• • • • •
Upload & Sell: Off

p.2 #16 · p.2 #16 · R5's ISO-Invariance - a happy discovery for shooting Owls at dawn and dusk

I think we need to repeat that the ISO 1000 and ISO 25600 images was shot with different shutter speeds and different exposures. The ISO 25600 image we see is pushed 1.5 stops and corresponds to ISO 64000

Nov 22, 2023 at 10:17 AM

Jeff
Offline
• • • • •
Upload & Sell: On

p.2 #17 · p.2 #17 · R5's ISO-Invariance - a happy discovery for shooting Owls at dawn and dusk

Jeff Nolten wrote:
Noob question: So does this mean I can get away with using my f4 zoom for opportunistic night sky shots rather than carry an extra f2.8 prime?

Nope, sorry. Well, I mean, you can certainly use it, but...

The more light (relatively speaking) you put into the file, the less noise there will be. Further, without a star tracker, you quickly become limited by star trails before you approach 'adequate' (astro) exposure, if you care about such things.

Nov 22, 2023 at 10:48 AM

Jeff
Offline
• • • • •
Upload & Sell: On

p.2 #18 · p.2 #18 · R5's ISO-Invariance - a happy discovery for shooting Owls at dawn and dusk

alundeb wrote:
I think we need to repeat that the ISO 1000 and ISO 25600 images was shot with different shutter speeds and different exposures. The ISO 25600 image we see is pushed 1.5 stops and corresponds to ISO 64000

Which is almost sure the 'rub' here.

Nov 22, 2023 at 10:49 AM

jedibrain
Offline
• • •
Upload & Sell: Off

p.2 #19 · p.2 #19 · R5's ISO-Invariance - a happy discovery for shooting Owls at dawn and dusk

Jeff wrote:
Generally agree with everything you wrote, and probably somewhat semantic, but I'd consider the R5 nearing invariant at ISO 800, since noise reduction is being applied below that (sort of apples:oranges). If you really want to get picky, I think ISO 1600 is where most of us astro people shoot our R5s at (generally speaking), though could probably live with ISO 800 unless we wanted to be able to effectively preview the image on the LCD in the field. So many variables, especially when trying to compare different cameras.

The linked PhotographyLife article is pretty good, but mostly ignored the significance...Show more →

I did say 'close to invariance' at 400....

Also agree that this is not the only factor in choosing an ISO for AP work (or any other photography for that matter). I was shooting with a 1000D and M6mkII until just recently. I found shooting ISO 400 and pulling up in post gave slightly better results than shooting where the curve flattens out between 3200 and 6400 for that camera. Less hot pixels, similar noise in the shadows, less banding. I have a lot of 'signal' her due to the light pollution though, so may be a different story under darker skies.

I've really only gotten familiar with the DR and read noise charts on that site, so can't help with how to view banding related issues.

Brian

Nov 22, 2023 at 10:59 AM

Rudy Pohl
Offline
• •
Upload & Sell: Off

p.2 #20 · p.2 #20 · R5's ISO-Invariance - a happy discovery for shooting Owls at dawn and dusk

I posted this thread in my initial excitement of accidentally discovering the ability to recover in post an R5 image that was underexposed by 4.5 stops. This was big deal for someone whose been struggling with poor quality low-light Owl images for over a year since getting the R5.

Unfortunately I had not noticed that I had inadvertently in the field dropped my shutter speed from 1/1600 to 1/400 two quick shots resulting in a 4-stop drop in ISO. By the time I realized it this thread had taken on a life of its own and had become a confusing mess. I was hoping something useful might emerge from it all.

Anyways, I think it could time just let this one fade away into obscurity that it so well deserves.

Rudy

Nov 22, 2023 at 11:00 AM

1 2 3 4 end
FM Forums \| Canon Forum \| Join Upload & Sell 1 2 3 4 end

This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.