K3 vs K5 high ISO noise.

derekkite · 01-10-2014, 10:39 PM

But it isn't simply a matter of double the light. It is the characteristic of the sensor that matters for IQ. If I understand correctly, film wasn't linear either when it came to shutter speed, at one point there was a drop off. The digital sensors deal far better with longer exposures up to the point that you get heat noise or amp glow. The K3 seems quite resistant to both.

For exposure purposes we start with the assumption that it is linear, or close enough to it. For IQ, the relationships are not linear, especially with iso. I think that it isn't linear with aperture/shutter speed. I know that if I have the choice, ie. a subject that does not move, a longer exposure with a smaller aperture will give me more color and better contrast than a shorter exposure and wider aperture. I'm not talking about depth of field. If the circumstances dictate a faster shutter speed to eliminate shake or subject movement blur, I will get better IQ if I have a faster lens and lower iso for a given exposure length. And I'm better to use a tripod or develop the handholding skills so that I can get longer exposures without shake if possible. There are sweet spots for any given situation and if I get it right I can get a great shot. If I get it wrong it is noisy.

I'm talking about situations near the edge of the capabilities of the body and lens combination. For some reason I seem to take lots of shots near the edge, and I'm interested in finding the combination of settings that can get the best possible shot. That combination is different with the K3 vs the K5.

We have the luxury of discussing iso 3200 or iso 6400 shots. The range of conditions where a properly exposed shot will be good is very wide.

**Rondec** · 01-11-2014, 07:02 AM

I just took a photo of my daughter last night at iso 6400 with the DA *55. I would describe light as normal "living room light" for my home. I wouldn't normally shoot with less light than this.

And a crop of the above.

I guess I would say that there is noise in the dark areas. I didn't try to clean it up, but it does clean up pretty easily. I think that getting a sharp photo with decent exposure is paramount when you are shooting high iso. Back when I shot with a K7, it was true even at iso 800, but certainly when you get up to iso 6400 on APS-C, any under exposure is going to dramatically introduce noise.

kirkmahr · 01-11-2014, 08:26 AM

She is an angel! Beautiful photo.

normhead · 01-11-2014, 09:42 AM

Originally posted by Class A

Not only are you comparing with different noise reduction applied (in-camera), you also show the K-3 image at higher magnification.

You appear to have cropped out the same sized pixel area for both cameras, which shows pixels at the same size, whereas the K-3 pixels should be smaller by a factor of 2/3 to provide a fair comparison.

There is no question that a smaller pixel (from the K-3) will look worse than a larger pixel (from the K-5) when both are shown at the same size. The question is whether the K-3's larger amount of pixels can compensate the higher per-pixel-noise, accordingly.

Yes, please turn of all in-camera noise reduction.

Please also leave the EXIF data intact. In one comparison you show two crops, one from the K-5 and one from the K-3, but there is no difference in magnification this time. How did you scale down the K-3 image? If you leave the EXIF data intact, we can make sure that no errors happen (things like that happen, as your first post demonstrates).

Making fair comparisons isn't as easy as it sounds. The results from Tjompen1968 look like what I'd expect and are in line with what DxOMark results.

Here's why I did things the way I did.

Essentially, I've decided the 16K images are more than enough for what I do. What I want is more magnification from the K-3 and more ability to crop, for BiFs and moving animals.

So essentially, most of the time I only want 16mb of my 24 mb k-3, image, so the test id designed is to find out what I need to know.

So this is the test to find out what I need to know. Now if you have different tests, that you need, for the stuff you do, I might be able to help you out I might not. But please don't expect me to spend hours of my time to do tests you've devised for someone else's needs.

If for your needs, you need a different things done a different way... hop to it.

I post my stuff, not because I'm out to do the tests everyone wants, I do the tests that test the things I want to know, and post results so that others with a similar mind set don't have to repeat my work.

Thanks for understanding.

If I ran every test I'd like to do the way I'd want to do them, I'd be here for years. I have to focus on what's important. This test was exactly what I wanted. IN terms of testing with NR off, if I'm a little unclear as to why turning NR off would make the noise better. I'm curious, but what I'm hoping for is that it doesn't make any difference. With the K-3, I shoot one card for jpeg and one card for RAW, so I'd like those NR settings on for the jpegs, and off for the raw. So that's why I'm going to test the whole NR thing. I'm still hoping NR doesn't affect RAW to any noticeable degree. The jpeg card goes to my iPad, for image review in the field., and will eventually be my flu card slot.

**Rondec** · 01-11-2014, 10:15 AM

Originally posted by kirkmahr

She is an angel! Beautiful photo.

Thanks. She's a pretty willing model when I need one and I'd rather take photos of my kids than of walls or things like that.

normhead · 01-11-2014, 11:03 AM

Originally posted by Rondec

Thanks. She's a pretty willing model when I need one and I'd rather take photos of my kids than of walls or things like that.

But hey, if your family lives 400 km away, a wall isn't so bad...

Class A · 01-11-2014, 02:20 PM

Originally posted by normhead

So essentially, most of the time I only want 16mb of my 24 mb k-3, image, so the test id designed is to find out what I need to know.

So this is the test to find out what I need to know.

You can of course test anything and in a manner that you see fit, but if you keep posting elsewhere in the forum that according to your observations the K-3 is a bit noisier than the K-5 then this is very misleading for everyone unless you include the fact that you are only using 16MP from the K-3.

If you include the latter fact, no one should be objecting you. If you don't then you are inviting objections and are misinforming those who do not suspect any special circumstances.

Originally posted by normhead

..., if I'm a little unclear as to why turning NR off would make the noise better.

The main thing is to create equal conditions for the comparison.

If you keep in-camera noise reduction on, you are not only comparing sensor performance but also the respective NR algorithms and the strengths chosen by the engineers.

Also, by leaving NR to good post-processing software you can indeed get better results compared to the in-camera processing.

audiobomber · 01-11-2014, 05:40 PM

Originally posted by Class A

You can of course test anything and in a manner that you see fit, but if you keep posting elsewhere in the forum that according to your observations the K-3 is a bit noisier than the K-5 then this is very misleading for everyone unless you include the fact that you are only using 16MP from the K-3.

If you include the latter fact, no one should be objecting you. If you don't then you are inviting objections and are misinforming those who do not suspect any special circumstances.

The main thing is to create equal conditions for the comparison.

If you keep in-camera noise reduction on, you are not only comparing sensor performance but also the respective NR algorithms and the strengths chosen by the engineers.

Also, by leaving NR to good post-processing software you can indeed get better results compared to the in-camera processing.

What he said. This thread should be titled "K3 vs K5 high ISO noise - uneven playing field." You're comparing the K-3 to something the K-5 can't do (16mp crop from 24mp),

john5100 · 01-11-2014, 06:51 PM

Here's one with the K3 that I screwed up and it's still okay.

ISO 6400 pushed to more than ISO 12,800 as I underexposed big time. I would say that it's closer to ISO 15,000. I have another show tonight where I'll try and make up for it.

crussellsprout · 01-11-2014, 07:06 PM

Nice shot, John. I also think the K3 can perform quite well up to 6400 or even 12800 ISO when necessary, though of course it's even better when you can get by with lower.

Here are a couple tests, first at ISO 6400 and second one at 12800.

Del K3 ISO 6400 | Flickr - Photo Sharing!

K3 High-ISO test | Flickr - Photo Sharing!

**Sailor** · 01-11-2014, 07:09 PM

Originally posted by Rondec

I just took a photo of my daughter last night at iso 6400 with the DA *55. . . . . . . . .

[

What a wonderful young lady - you've gotta be one proud DAD.

Jer

**Sailor** · 01-11-2014, 07:17 PM

This thread has turned my whole world upside down

. Until now, I had the absolute conviction that RAW files were ejected from the camera totally unmolested (other than an initial exposure) allowing me carte blanche with regard to post processing modifications without having to fiddle with JPEG-type camera settings. Now I'm finding that there are in-camera controls for NR for RAW files, which leaves me wondering what other adjustable stuff is there that I didn't think I had to consider

.

Jer

audiobomber · 01-11-2014, 08:28 PM

Originally posted by Sailor

This thread has turned my whole world upside down

. Until now, I had the absolute conviction that RAW files were ejected from the camera totally unmolested (other than an initial exposure) allowing me carte blanche with regard to post processing modifications without having to fiddle with JPEG-type camera settings. Now I'm finding that there are in-camera controls for NR for RAW files, which leaves me wondering what other adjustable stuff is there that I didn't think I had to consider

.

Jer

Which camera settings are imported depends on the raw program. DCU5 shows the raw view with all camera settings. Lightroom imports white balance, noise reduction and sharpness. Some raw programs ignore the camera settings. In all cases though, all raw settings can be manipulated without loss, whether they were read by the raw program, or not.

normhead · 01-11-2014, 09:46 PM

Quote:

You can of course test anything and in a manner that you see fit, but if you keep posting elsewhere in the forum that according to your observations the K-3 is a bit noisier than the K-5 then this is very misleading for everyone unless you include the fact that you are only using 16MP from the K-3.

That's inaccurate, the last four images are the full uncropped image, I'm comparing a 24 MP image to a 16 Mp image, and the 16 Mp image is better for noise, at both 1600 and 6400 ISO.

Not that you're going to apologize for once again misrepresenting the truth. That seems to be your MO. By the way, you're on my ignore list.

jsherman999 · 01-11-2014, 10:02 PM

Originally posted by normhead

OK try again....

K-5 1600

K-3 1600

K-5 6400

K-3

Ok, now there's a difference, using the K-3 at 6400 there is enough noise to produce sharpening artifacts. Also in the K-3 at 1600 but not as distinct. So this is what I was observing.. the K-3 is going to be more dependent on noise reduction software, or mean I have to use less sharpening than I'm accustomed to on the whole image. Much of this can be reduced by turning down the sharpening... but this is my fairly standard level of sharpening using the K-5.

.

This is a terrible comparison. The K-5 image obviously has some in-camera NR going on, which you can tell by the 'melted-plastic' look of the objects, and on top of that you show the K-3 image at different magnification.

???

.

01-11-2014, 08:28 PM	#43
audiobomber Veteran Member Join Date: May 2008 Location: Sudbury, Ontario Photos: Albums Posts: 6,806	Originally posted by Sailor This thread has turned my whole world upside down . Until now, I had the absolute conviction that RAW files were ejected from the camera totally unmolested (other than an initial exposure) allowing me carte blanche with regard to post processing modifications without having to fiddle with JPEG-type camera settings. Now I'm finding that there are in-camera controls for NR for RAW files, which leaves me wondering what other adjustable stuff is there that I didn't think I had to consider . Jer Which camera settings are imported depends on the raw program. DCU5 shows the raw view with all camera settings. Lightroom imports white balance, noise reduction and sharpness. Some raw programs ignore the camera settings. In all cases though, all raw settings can be manipulated without loss, whether they were read by the raw program, or not. Last edited by audiobomber; 01-12-2014 at 07:27 AM.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
New K3 vs K5ii high iso RAW tests	ihasa	Pentax K-3 & K-3 II	8	11-04-2013 09:15 AM
DR/Noise performance: pushing exposure vs high ISO	HSV	Photographic Technique	8	03-12-2013 06:51 PM
K-R VS X100 High Iso Noise Comparison	Ryan Cole	Pentax DSLR Discussion	20	02-01-2013 02:36 PM
K5 High ISO Noise Reduction Setting	daphnejohn	Pentax K-5 & K-5 II	2	11-06-2011 07:45 PM
High ISO Noise vs. Underexposure Correction	yoon395	Photographic Technique	7	08-27-2011 06:13 AM

01-10-2014, 10:39 PM	#31
derekkite Pentaxian Join Date: Sep 2011 Location: Nelson B.C. Posts: 3,782	But it isn't simply a matter of double the light. It is the characteristic of the sensor that matters for IQ. If I understand correctly, film wasn't linear either when it came to shutter speed, at one point there was a drop off. The digital sensors deal far better with longer exposures up to the point that you get heat noise or amp glow. The K3 seems quite resistant to both. For exposure purposes we start with the assumption that it is linear, or close enough to it. For IQ, the relationships are not linear, especially with iso. I think that it isn't linear with aperture/shutter speed. I know that if I have the choice, ie. a subject that does not move, a longer exposure with a smaller aperture will give me more color and better contrast than a shorter exposure and wider aperture. I'm not talking about depth of field. If the circumstances dictate a faster shutter speed to eliminate shake or subject movement blur, I will get better IQ if I have a faster lens and lower iso for a given exposure length. And I'm better to use a tripod or develop the handholding skills so that I can get longer exposures without shake if possible. There are sweet spots for any given situation and if I get it right I can get a great shot. If I get it wrong it is noisy. I'm talking about situations near the edge of the capabilities of the body and lens combination. For some reason I seem to take lots of shots near the edge, and I'm interested in finding the combination of settings that can get the best possible shot. That combination is different with the K3 vs the K5. We have the luxury of discussing iso 3200 or iso 6400 shots. The range of conditions where a properly exposed shot will be good is very wide.

01-11-2014, 07:02 AM	#32
Rondec Loyal Site Supporter Join Date: Mar 2009 Location: Gladys, Virginia Photos: Gallery Posts: 27,662	I just took a photo of my daughter last night at iso 6400 with the DA *55. I would describe light as normal "living room light" for my home. I wouldn't normally shoot with less light than this. And a crop of the above. I guess I would say that there is noise in the dark areas. I didn't try to clean it up, but it does clean up pretty easily. I think that getting a sharp photo with decent exposure is paramount when you are shooting high iso. Back when I shot with a K7, it was true even at iso 800, but certainly when you get up to iso 6400 on APS-C, any under exposure is going to dramatically introduce noise.

01-11-2014, 08:26 AM	#33
kirkmahr Junior Member Join Date: Nov 2013 Location: Ham Lake, Minnesota Posts: 27	She is an angel! Beautiful photo.

01-11-2014, 09:42 AM	#34
normhead Pentaxian Join Date: Jun 2007 Location: Near Algonquin Park Photos: Gallery \| Albums Posts: 40,451 Original Poster	Originally posted by Class A Not only are you comparing with different noise reduction applied (in-camera), you also show the K-3 image at higher magnification. You appear to have cropped out the same sized pixel area for both cameras, which shows pixels at the same size, whereas the K-3 pixels should be smaller by a factor of 2/3 to provide a fair comparison. There is no question that a smaller pixel (from the K-3) will look worse than a larger pixel (from the K-5) when both are shown at the same size. The question is whether the K-3's larger amount of pixels can compensate the higher per-pixel-noise, accordingly. Yes, please turn of all in-camera noise reduction. Please also leave the EXIF data intact. In one comparison you show two crops, one from the K-5 and one from the K-3, but there is no difference in magnification this time. How did you scale down the K-3 image? If you leave the EXIF data intact, we can make sure that no errors happen (things like that happen, as your first post demonstrates). Making fair comparisons isn't as easy as it sounds. The results from Tjompen1968 look like what I'd expect and are in line with what DxOMark results. Here's why I did things the way I did. Essentially, I've decided the 16K images are more than enough for what I do. What I want is more magnification from the K-3 and more ability to crop, for BiFs and moving animals. So essentially, most of the time I only want 16mb of my 24 mb k-3, image, so the test id designed is to find out what I need to know. So this is the test to find out what I need to know. Now if you have different tests, that you need, for the stuff you do, I might be able to help you out I might not. But please don't expect me to spend hours of my time to do tests you've devised for someone else's needs. If for your needs, you need a different things done a different way... hop to it. I post my stuff, not because I'm out to do the tests everyone wants, I do the tests that test the things I want to know, and post results so that others with a similar mind set don't have to repeat my work. Thanks for understanding. If I ran every test I'd like to do the way I'd want to do them, I'd be here for years. I have to focus on what's important. This test was exactly what I wanted. IN terms of testing with NR off, if I'm a little unclear as to why turning NR off would make the noise better. I'm curious, but what I'm hoping for is that it doesn't make any difference. With the K-3, I shoot one card for jpeg and one card for RAW, so I'd like those NR settings on for the jpegs, and off for the raw. So that's why I'm going to test the whole NR thing. I'm still hoping NR doesn't affect RAW to any noticeable degree. The jpeg card goes to my iPad, for image review in the field., and will eventually be my flu card slot.

01-11-2014, 10:15 AM	#35
Rondec Loyal Site Supporter Join Date: Mar 2009 Location: Gladys, Virginia Photos: Gallery Posts: 27,662	Originally posted by kirkmahr She is an angel! Beautiful photo. Thanks. She's a pretty willing model when I need one and I'd rather take photos of my kids than of walls or things like that.

01-11-2014, 11:03 AM	#36
normhead Pentaxian Join Date: Jun 2007 Location: Near Algonquin Park Photos: Gallery \| Albums Posts: 40,451 Original Poster	Originally posted by Rondec Thanks. She's a pretty willing model when I need one and I'd rather take photos of my kids than of walls or things like that. But hey, if your family lives 400 km away, a wall isn't so bad...

01-11-2014, 02:20 PM	#37
Class A Pentaxian Join Date: Aug 2008 Location: Wellington, New Zealand Posts: 11,251	Originally posted by normhead So essentially, most of the time I only want 16mb of my 24 mb k-3, image, so the test id designed is to find out what I need to know. So this is the test to find out what I need to know. You can of course test anything and in a manner that you see fit, but if you keep posting elsewhere in the forum that according to your observations the K-3 is a bit noisier than the K-5 then this is very misleading for everyone unless you include the fact that you are only using 16MP from the K-3. If you include the latter fact, no one should be objecting you. If you don't then you are inviting objections and are misinforming those who do not suspect any special circumstances. Originally posted by normhead ..., if I'm a little unclear as to why turning NR off would make the noise better. The main thing is to create equal conditions for the comparison. If you keep in-camera noise reduction on, you are not only comparing sensor performance but also the respective NR algorithms and the strengths chosen by the engineers. Also, by leaving NR to good post-processing software you can indeed get better results compared to the in-camera processing.

01-11-2014, 05:40 PM	#38
audiobomber Veteran Member Join Date: May 2008 Location: Sudbury, Ontario Photos: Albums Posts: 6,806	Originally posted by Class A You can of course test anything and in a manner that you see fit, but if you keep posting elsewhere in the forum that according to your observations the K-3 is a bit noisier than the K-5 then this is very misleading for everyone unless you include the fact that you are only using 16MP from the K-3. If you include the latter fact, no one should be objecting you. If you don't then you are inviting objections and are misinforming those who do not suspect any special circumstances. The main thing is to create equal conditions for the comparison. If you keep in-camera noise reduction on, you are not only comparing sensor performance but also the respective NR algorithms and the strengths chosen by the engineers. Also, by leaving NR to good post-processing software you can indeed get better results compared to the in-camera processing. What he said. This thread should be titled "K3 vs K5 high ISO noise - uneven playing field." You're comparing the K-3 to something the K-5 can't do (16mp crop from 24mp),

01-11-2014, 06:51 PM	#39
john5100 Veteran Member Join Date: May 2008 Location: Seattle, WA Posts: 1,725	Here's one with the K3 that I screwed up and it's still okay. ISO 6400 pushed to more than ISO 12,800 as I underexposed big time. I would say that it's closer to ISO 15,000. I have another show tonight where I'll try and make up for it.

01-11-2014, 07:06 PM	#40
crussellsprout Forum Member Join Date: Jul 2012 Posts: 60	Nice shot, John. I also think the K3 can perform quite well up to 6400 or even 12800 ISO when necessary, though of course it's even better when you can get by with lower. Here are a couple tests, first at ISO 6400 and second one at 12800. Del K3 ISO 6400 \| Flickr - Photo Sharing! K3 High-ISO test \| Flickr - Photo Sharing!

01-11-2014, 07:09 PM	#41
Sailor Loyal Site Supporter Join Date: Nov 2006 Location: Coastal Texas Photos: Gallery Posts: 26,203	Originally posted by Rondec I just took a photo of my daughter last night at iso 6400 with the DA *55. . . . . . . . . [ What a wonderful young lady - you've gotta be one proud DAD. Jer

01-11-2014, 07:17 PM	#42
Sailor Loyal Site Supporter Join Date: Nov 2006 Location: Coastal Texas Photos: Gallery Posts: 26,203	This thread has turned my whole world upside down . Until now, I had the absolute conviction that RAW files were ejected from the camera totally unmolested (other than an initial exposure) allowing me carte blanche with regard to post processing modifications without having to fiddle with JPEG-type camera settings. Now I'm finding that there are in-camera controls for NR for RAW files, which leaves me wondering what other adjustable stuff is there that I didn't think I had to consider . Jer

01-11-2014, 09:46 PM	#44
normhead Pentaxian Join Date: Jun 2007 Location: Near Algonquin Park Photos: Gallery \| Albums Posts: 40,451 Original Poster	Quote: You can of course test anything and in a manner that you see fit, but if you keep posting elsewhere in the forum that according to your observations the K-3 is a bit noisier than the K-5 then this is very misleading for everyone unless you include the fact that you are only using 16MP from the K-3. That's inaccurate, the last four images are the full uncropped image, I'm comparing a 24 MP image to a 16 Mp image, and the 16 Mp image is better for noise, at both 1600 and 6400 ISO. Not that you're going to apologize for once again misrepresenting the truth. That seems to be your MO. By the way, you're on my ignore list.

01-11-2014, 10:02 PM	#45
jsherman999 Veteran Member Join Date: Dec 2007 Photos: Gallery Posts: 8,237	Originally posted by normhead OK try again.... K-5 1600 K-3 1600 K-5 6400 K-3 Ok, now there's a difference, using the K-3 at 6400 there is enough noise to produce sharpening artifacts. Also in the K-3 at 1600 but not as distinct. So this is what I was observing.. the K-3 is going to be more dependent on noise reduction software, or mean I have to use less sharpening than I'm accustomed to on the whole image. Much of this can be reduced by turning down the sharpening... but this is my fairly standard level of sharpening using the K-5. . This is a terrible comparison. The K-5 image obviously has some in-camera NR going on, which you can tell by the 'melted-plastic' look of the objects, and on top of that you show the K-3 image at different magnification. ??? .