[savoche]

In general, to get lower high-ISO noise you have to go to a bigger sensor. Sure, newer technology helps, but a bigger sensor at the same tech level will still be better. At the end of the day there will always be shot noise, it's an inherent part of light. You can't break the laws of physics.

Of course, you can get amazing results with "computational photography", but that's a different ballgame altogether.

I have both the K3II and the K1MII, at ISO levels below 6400, IMO the K3II has less noise, above that they're fairly equal up to 12800, where the K1MII starts to pull away.

That is very different from my experience with a K-3 and K-1. I can't tell them apart up to ISO 400, but already at 800 the K-1 is somewhat less noisy. At 6400 it's really noticeable. I'd say the K-1 is only slightly worse at 25k than the K-3 is at 6400 when downsampled to the same size.

All that said, I won't hesitate to use the K-3 at ISO 25k if that's what it takes. It still cleans up decently. The main problem isn't noise, but rather lack of dynamic range. At these high ISO values there is, what, 3 stops of DR left? For extreme shooting it's probably better to stay at ISO 100 and underexpose to protect the highlights. These days the sensors are pretty darn close to ISO invariant. It does require more work when post processing, though.

Umm, sorry for veering off topic

[iheiramo]

K1ii will give you that reasonably noiceless ISO1600 to work with. Maybe even ISO3200.

[mikesbike]

Actually, the noise difference between the K-3 and the KP is substantial, according to both test labs and user reports. Differences can be seen between models in image comparisons from dpreview lab reports and from Imaging Resource's comparometer. Technology continues to advance, which is a good thing. The latest sensor design claims to offer greater improvements yet.

[Smoke665]

In general, to get lower high-ISO noise you have to go to a bigger sensor. Sure, newer technology helps, but a bigger sensor at the same tech level will still be better. At the end of the day there will always be shot noise, it's an inherent part of light. You can't break the laws of physics.

Of course, you can get amazing results with "computational photography", but that's a different ballgame altogether.



That is very different from my experience with a K-3 and K-1. I can't tell them apart up to ISO 400, but already at 800 the K-1 is somewhat less noisy. At 6400 it's really noticeable. I'd say the K-1 is only slightly worse at 25k than the K-3 is at 6400 when downsampled to the same size.

All that said, I won't hesitate to use the K-3 at ISO 25k if that's what it takes. It still cleans up decently. The main problem isn't noise, but rather lack of dynamic range. At these high ISO values there is, what, 3 stops of DR left? For extreme shooting it's probably better to stay at ISO 100 and underexpose to protect the highlights. These days the sensors are pretty darn close to ISO invariant. It does require more work when post processing, though.

Umm, sorry for veering off topic

"That is very different from my experience with a K-3 and K-1"

Possibly because of differences in the K3II and the K1MII?

"For extreme shooting it's probably better to stay at ISO 100 and underexpose to protect the highlights"

I would disagree with this as a blanket statement. There are two types of noise in a digital image, shot noise and read noise. Shot noise occurs when random photons are emitted by lightsources which gives rise to variations in the light hitting each part of the sensor. Read noise occurs as the analogue signal is amplified prior to the analog to digital convertor. When you increase ISO in camera you're amplifying shot noise but the signal to noise ratio is the same, because the read noise occurs after amplification. When you shoot at low ISO and boost the exposure post, you increase both shot noise and read noise.

High ISO doesn't create shot noise. If you have a dim scene, with little light hitting the sensor then you'll have noise in the shadows regardless of the ISO. Likewise a brightly lit scene will have more light hitting the sensor resulting little noise regardless of ISO setting. The biggest influence on noise then, is the amount of light falling on the sensor, assuming the ISO is set correctly to avoid over or under exposing the image.

A better way to manage noise is to allow as much light as possible to hit the sensor without over/underexposing the image and to use the highest ISO possible without over/underexposing the image.

[aslyfox]

I suggest in cases like this, consider renting equipment before buying

Information on Businesses that offer cameras and lenses for rent - Page 3 - PentaxForums.com

Discount Code from LensRental.com good for one use until December 31 - PentaxForums.com

[savoche]

Just to make it clear, I think of this as more of an academic discussion than something that will make a whole lot of practical difference.

Possibly because of differences in the K3II and the K1MII?

Possibly, but I'd expect the difference to be even more pronounced then. You may well be right, though.

I would disagree with this as a blanket statement.

Umm, you forgot some context:

The main problem isn't noise, but rather lack of dynamic range. At these high ISO values there is, what, 3 stops of DR left? For extreme shooting it's probably better to stay at ISO 100 and underexpose to protect the highlights. These days the sensors are pretty darn close to ISO invariant.

There are enough conditions there to not make it a blanket statement, I think

There are two types of noise in a digital image, shot noise and read noise. Shot noise occurs when random photons are emitted by lightsources which gives rise to variations in the light hitting each part of the sensor. Read noise occurs as the analogue signal is amplified prior to the analog to digital convertor. When you increase ISO in camera you're amplifying shot noise but the signal to noise ratio is the same, because the read noise occurs after amplification. When you shoot at low ISO and boost the exposure post, you increase both shot noise and read noise.

This is all true. It is also true that (many) modern sensor systems have very little read noise. With a K-1 there is very little difference in visible noise between an ISO 100 shot raised 5 stops in post and an ISO 3200 shot. I will not claim that the K-1 is fully ISO invariant, but I believe it is very close.

High ISO doesn't create shot noise. If you have a dim scene, with little light hitting the sensor then you'll have noise in the shadows regardless of the ISO. Likewise a brightly lit scene will have more light hitting the sensor resulting little noise regardless of ISO setting. The biggest influence on noise then, is the amount of light falling on the sensor, assuming the ISO is set correctly to avoid over or under exposing the image.

Yes, I agree. ISO has nothing to do with shot noise. The number of photons reaching you sensor has everything to do with shot noise. Increase the amount of light 100 times and noise only increases 10 times - resulting in an increased signal-to-noise ratio and thus less visible noise (although more noise in absolute terms).

A better way to manage noise is to allow as much light as possible to hit the sensor without over/underexposing the image and to use the highest ISO possible without over/underexposing the image.

Certainly, more light = less visible noise. But the higher the ISO the easier it gets to blow the highlights. This might force you to decrease your exposure to protect the highlights, resulting in a lower SNR. The point of shooting at a lower ISO is to retain better dynamic range. As I stated above, when you get to ISO 25k with the K-3 there is very little dynamic range left. Even the K-1 II will be down to something like 5 stops of DR.

So, if we accept as true that read noise is negligible and thus only need to worry about shot noise, we can get better results shooting at lower ISO because the dynamic range will be larger.

(There certainly are other drawbacks with this method, such as making chimping much harder and also the requirement of more post processing.)

Then again, if we have a system with noticeable read noise, such as most Canon sensors, I would agree with you all the way. And also if we shoot JPEG, of course.

[Smoke665]

Just to make it clear, I think of this as more of an academic discussion than something that will make a whole lot of practical difference.

Possibly, and I don't want to go off rail from the OP's post, but in thinking about this, it may well apply, as his first sentence was his "concern about noise".

These days the sensors are pretty darn close to ISO invariant. It does require more work when post processing, though.

You know this jogged my curiosity. I've always heard this, but never really investigated much, so I decided to do a little "unscientific" testing. I took four shots, manual exposures, all at 1/200, f/7.1, but varied only the ISO. From left to right/top to bottom 2 stops underexposed per camera meter +2 stops compensation in post, Meter centered no correction, 1 stop overexposed per meter -1 stop compensation in post, and finally 2 stops over exposed per camera meter, -2 stops compensation in post. Blinkies were just starting to show in the last one, but my experience with the K1MII is that you can push it to the point where they blink without blowing highlights. No other compensation or adjustments done post or in camera.





Same order here are the histograms of the files after exposure correction in post. Notice the slight addition of midtone data in the histogram of #2 where the exposure was correct.



And finally some extreme crops to compare the noise





After looking at these, I'm not sure that it makes much difference on noise under, on, or over, but it does appear to affect midtones.

Attached Images

     

---

Last edited by Smoke665; 04-11-2020 at 03:03 PM.

[savoche]

After looking at these, I'm not sure that it makes much difference on noise under, on, or over, but it does appear to affect midtones.

Thanks for taking the time and effort to do this. As expected it's hard to draw any hard conclusions - much thanks to a fantastic camera. For this to maybe make any practical difference at all I suspect we'd have to try this with an even more contrasty scene and at higher ISOs. The DR at ISO1600 is still around 10 stops on the K-1 II thanks to the accelerator chip, about 2-2.5 stops more than the K-3 (according to Photons to Photos).

In practical shooting, though, if I need to protect the highlights I prefer shooting in TAv mode with a stop or two of underexposure (which really isn't underexposure; it's the same exposure but at a lower gain/ISO).

Again, thanks for an interesting comparison!