K-5 IIs Low-Light AF test video

RonHendriks1966 · 10-27-2012, 09:30 AM

Originally posted by Winder

For web use I can clean up 6400 just fine, but for a decent sized print it will definitely show. At 6400 I am converting to B&W and using the grain as part of the "mood" of the image.

But then again the expectations for an image made in real life with iso6400 is different then one made in a studio with everything under control. I have some very fine soccer images made with iso6400 printed on A3 paper that are absolutely fabulous. Offcourse there are camera's that can do that job better, but not for the same pricetag that K-5 is.

Winder · 10-27-2012, 09:42 AM

Originally posted by RonHendriks1966

But then again the expectations for an image made in real life with iso6400 is different then one made in a studio with everything under control. I have some very fine soccer images made with iso6400 printed on A3 paper that are absolutely fabulous. Offcourse there are camera's that can do that job better, but not for the same pricetag that K-5 is.

It is not just noise that is the issue. The color and tonal range fall off at ISO 6400 to the point where it is an issue. For a magazine or basic print article where print quality is low or mid-level it is not an issue. My post processing skills are decent, but by no means at the expert level. People with the time/skill in post processing could probably do a better job, but so far I have not been happy with color ISO 6400 prints at any real size. If i have a project where I have the money I will pay a person to do my post processing. I have a friend who works for a national magazine as a graphics designer who does my critical Photoshop work.... I don't even own Photoshop. Most of my work I can do in LR4 with basic post work.

Winder · 10-27-2012, 09:45 AM - **2 Likes**

Originally posted by Adam

This was performed in an area with at most 0EV (the Pentax spot meter doesn't go any lower). The K-5 couldn't focus at all on any part of the scene, while the K-5 IIs locked on every time. Note that we were focusing on the floor, not the bed.
Headphones may be needed

I hope you are able to test the AF in some difficult (different colors) lighting conditions. Find a local Jazz or Blues club (they usually have really crappy light) with a good band and some cold beer and shoot a few test shots..... After 6 or 7 beers you can also compare how well the SR works. All in the name of science.

westmill · 10-27-2012, 10:00 AM

Im about the same as Ron in this arena. 6400 ISO especialy if shot in raw is good for most things although I would only use it at a wedding in a dire emergency lol. Where poss I try keep my ceiling at 3200. I will say though, 6400 is easily as good as my 3200 setting on the Nikon D300. Poss even better ! I use just 1600 ISO as my ceiling with the nikon. Again 3200 ISO is easily as good as 1600 ISO on the Nikon. So its pretty impressive

This was shot at 6400 at F2.8. Ive done very little, but it would print nicely at A4. You can get sharper more detailed prints with care and or more light. it was dark indeed when I took this the other night

Last edited by westmill; 10-27-2012 at 10:09 AM.

Class A · 10-27-2012, 03:34 PM

Originally posted by ogl

ISO6400 and 1.4 (I know only DA55/1.4 and FA50/1.4 with such aperture) + 1/40 = using SR and unacceptable IQ. Sigma 30/1.4 - I don't think that it's good lens.

The Sigma 30/1.4 is a great lens. Have a look at the respective sample image thread and tell me with which other lens (sans the FA 31/1.8) you can get results like this.

There is also the excellent Rokinon 85/1.4, the Nokton, the f/1.2 lenses, etc.

Originally posted by ogl

Did the same with K200D + FA24*/2- from semi-dark corridor to dark room - f4.5, ISO200, 1.6 s.

Assuming an 18% exposure that's just an EV of 2.6. Not a challenge.

Originally posted by ogl

Let's have a look at K-5 and K200D at full SNR - it's very interesting:

Where are these figures from?
Are they based on pixel noise or image noise?
If they are based on pixel noise, they'd be inadequate, favouring the K200D with its lower resolution.

ogl · 10-28-2012, 02:08 AM

Originally posted by Class A

Where are these figures from?
Are they based on pixel noise or image noise?
If they are based on pixel noise, they'd be inadequate, favouring the K200D with its lower resolution.

DXO Mark - Full SNR for each camera. You can go to K-5 and K200D's page with measurement
and you will see...

ogl · 10-28-2012, 02:11 AM

Originally posted by Class A

Assuming an 18% exposure that's just an EV of 2.6. Not a challenge.

Adam's test - EV0.

If you want you can make the same test with your camera...It's not very big problem for old cameras too. Not fast. Sometimes very slow. But it could.

Class A · 10-28-2012, 04:16 AM

Originally posted by ogl

DXO Mark - Full SNR for each camera. You can go to K-5 and K200D's page with measurement
and you will see...

OK, mystery solved.

The "Full SNR" measurements are provided in terms of pixels (called "Screen" by DxOMark). This favours sensors with less MP.

The correct way to compare image quality (as opposed to individual pixel quality) is to look at the whole image (called "Print" by DxOMark).

Since the K-5 has ~1.6 times more pixels, you need to add ~2 dB (= 3 * log2(1.6)) to all K-5 "Full SNR" figures. When you do this, you'll see that the K-5 is ahead at all SNR levels.

Just compare the "Full SNR" figures at 18% with the figures in the regular regular "SNR 18%" chart. You'll see that the figures correspond to the (irrelevant) "Screen" figures and that the K-5 is ahead in the "Print" figures.

Class A · 10-28-2012, 04:19 AM

Originally posted by ogl

Adam's test - EV0.

If you want you can make the same test with your camera...It's not very big problem for old cameras too. Not fast. Sometimes very slow. But it could.

I'm not sure what you are saying.

The video is proof that the K-5 II locks focus in light levels where the K-5 struggles.

Also, we know that the K-5 has a low-light AF accuracy problem. From all I've heard this also has been addressed in the K-30 and K-5 II.

ogl · 10-28-2012, 06:36 AM

Originally posted by Class A

Just compare the "Full SNR" figures at 18% with the figures in the regular.

It's wrong way to compare SNR18%. We can't see the noise level in shadows and deep shadows.

ogl · 10-28-2012, 06:39 AM

Originally posted by Class A

I'm not sure what you are saying.

The video is proof that the K-5 II locks focus in light levels where the K-5 struggles.

Also, we know that the K-5 has a low-light AF accuracy problem. From all I've heard this also has been addressed in the K-30 and K-5 II.

K-5 has accurancy problem in any light condition.

Tom S. · 10-28-2012, 07:11 AM

Originally posted by ogl

K-5 has accurancy problem in any light condition.

What are you basing that statement on? Do you own a K-5? Or is this another one of your posts trying to stir up crap?

Class A · 10-28-2012, 07:23 AM

Originally posted by ogl

It's wrong way to compare SNR18%.

No, it is absolutely the correct way.

You need to compare full images. Higher MP cameras have noisier individual pixels but in combination the higher number of pixels average out to the same image noise (everything else being the same).

Are you really questioning DxOMark's approach to scale every sensor to 8MP in order to normalize the results?

I'm surprised they don't do it for the "Full SNR" figures, but surely there is a reason why the "Print" view is the default as opposed to the "Screen" view.

ogl · 10-28-2012, 08:21 AM

Originally posted by Tom S.

What are you basing that statement on? Do you own a K-5? Or is this another one of your posts trying to stir up crap?

Yes. Of course. Be more polite, please... Sold K-5 1 month ago.

ogl · 10-28-2012, 08:26 AM

Originally posted by Class A

No, it is absolutely the correct way.

You need to compare full images. Higher MP cameras have noisier individual pixels but in combination the higher number of pixels average out to the same image noise (everything else being the same).

Are you really questioning DxOMark's approach to scale every sensor to 8MP in order to normalize the results?

I'm surprised they don't do it for the "Full SNR" figures, but surely there is a reason why the "Print" view is the default as opposed to the "Screen" view.

You can think in this way...

I won't argue.

I've heard that they did it for every camera - full SNR too - to scale every sensor to 8MP. Didn't you hear that K-5 is noisy camera in shadows?

10-27-2012, 09:45 AM - 2 Likes	#33
Winder Veteran Member Join Date: Apr 2010 Location: Tennessee Posts: 6,617	Originally posted by Adam This was performed in an area with at most 0EV (the Pentax spot meter doesn't go any lower). The K-5 couldn't focus at all on any part of the scene, while the K-5 IIs locked on every time. Note that we were focusing on the floor, not the bed. Headphones may be needed I hope you are able to test the AF in some difficult (different colors) lighting conditions. Find a local Jazz or Blues club (they usually have really crappy light) with a good band and some cold beer and shoot a few test shots..... After 6 or 7 beers you can also compare how well the SR works. All in the name of science. Last edited by Winder; 10-27-2012 at 10:11 AM.

10-28-2012, 06:36 AM	#40
ogl Banned Join Date: Dec 2007 Location: Sankt Peterburg Photos: Gallery \| Albums Posts: 8,382	Originally posted by Class A Just compare the "Full SNR" figures at 18% with the figures in the regular. It's wrong way to compare SNR18%. We can't see the noise level in shadows and deep shadows. Last edited by ogl; 10-28-2012 at 08:22 AM.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Low light AF/AL speed test	Hey Elwood	Pentax K-5 & K-5 II	4	02-07-2012 09:05 PM
Video with low light on K-r-need help.	LightMeter	Pentax K-r	4	11-15-2011 03:46 PM
Pentax K-5 low light AF torture test	zaiko	Pentax K-5 & K-5 II	15	07-08-2011 08:30 AM
K5 low light video on youtube	FunkyMonk	Pentax K-5 & K-5 II	16	02-04-2011 07:31 AM
Low-light video	dshack	Video Recording and Processing	9	04-13-2010 06:44 PM

10-27-2012, 09:30 AM	#31
RonHendriks1966 Banned Join Date: Jan 2009 Photos: Gallery Posts: 9,675	Originally posted by Winder For web use I can clean up 6400 just fine, but for a decent sized print it will definitely show. At 6400 I am converting to B&W and using the grain as part of the "mood" of the image. But then again the expectations for an image made in real life with iso6400 is different then one made in a studio with everything under control. I have some very fine soccer images made with iso6400 printed on A3 paper that are absolutely fabulous. Offcourse there are camera's that can do that job better, but not for the same pricetag that K-5 is.

10-27-2012, 09:42 AM	#32
Winder Veteran Member Join Date: Apr 2010 Location: Tennessee Posts: 6,617	Originally posted by RonHendriks1966 But then again the expectations for an image made in real life with iso6400 is different then one made in a studio with everything under control. I have some very fine soccer images made with iso6400 printed on A3 paper that are absolutely fabulous. Offcourse there are camera's that can do that job better, but not for the same pricetag that K-5 is. It is not just noise that is the issue. The color and tonal range fall off at ISO 6400 to the point where it is an issue. For a magazine or basic print article where print quality is low or mid-level it is not an issue. My post processing skills are decent, but by no means at the expert level. People with the time/skill in post processing could probably do a better job, but so far I have not been happy with color ISO 6400 prints at any real size. If i have a project where I have the money I will pay a person to do my post processing. I have a friend who works for a national magazine as a graphics designer who does my critical Photoshop work.... I don't even own Photoshop. Most of my work I can do in LR4 with basic post work.

10-27-2012, 10:00 AM	#34
westmill Veteran Member Join Date: Jan 2012 Location: Stoke on Trent Photos: Gallery \| Albums Posts: 1,146	Im about the same as Ron in this arena. 6400 ISO especialy if shot in raw is good for most things although I would only use it at a wedding in a dire emergency lol. Where poss I try keep my ceiling at 3200. I will say though, 6400 is easily as good as my 3200 setting on the Nikon D300. Poss even better ! I use just 1600 ISO as my ceiling with the nikon. Again 3200 ISO is easily as good as 1600 ISO on the Nikon. So its pretty impressive This was shot at 6400 at F2.8. Ive done very little, but it would print nicely at A4. You can get sharper more detailed prints with care and or more light. it was dark indeed when I took this the other night Last edited by westmill; 10-27-2012 at 10:09 AM.

10-27-2012, 03:34 PM	#35
Class A Pentaxian Join Date: Aug 2008 Location: Wellington, New Zealand Posts: 11,251	Originally posted by ogl ISO6400 and 1.4 (I know only DA55/1.4 and FA50/1.4 with such aperture) + 1/40 = using SR and unacceptable IQ. Sigma 30/1.4 - I don't think that it's good lens. The Sigma 30/1.4 is a great lens. Have a look at the respective sample image thread and tell me with which other lens (sans the FA 31/1.8) you can get results like this. There is also the excellent Rokinon 85/1.4, the Nokton, the f/1.2 lenses, etc. Originally posted by ogl Did the same with K200D + FA24*/2- from semi-dark corridor to dark room - f4.5, ISO200, 1.6 s. Assuming an 18% exposure that's just an EV of 2.6. Not a challenge. Originally posted by ogl Let's have a look at K-5 and K200D at full SNR - it's very interesting: Where are these figures from? Are they based on pixel noise or image noise? If they are based on pixel noise, they'd be inadequate, favouring the K200D with its lower resolution.

10-28-2012, 02:08 AM	#36
ogl Banned Join Date: Dec 2007 Location: Sankt Peterburg Photos: Gallery \| Albums Posts: 8,382	Originally posted by Class A Where are these figures from? Are they based on pixel noise or image noise? If they are based on pixel noise, they'd be inadequate, favouring the K200D with its lower resolution. DXO Mark - Full SNR for each camera. You can go to K-5 and K200D's page with measurement and you will see...

10-28-2012, 02:11 AM	#37
ogl Banned Join Date: Dec 2007 Location: Sankt Peterburg Photos: Gallery \| Albums Posts: 8,382	Originally posted by Class A Assuming an 18% exposure that's just an EV of 2.6. Not a challenge. Adam's test - EV0. If you want you can make the same test with your camera...It's not very big problem for old cameras too. Not fast. Sometimes very slow. But it could.

10-28-2012, 04:16 AM	#38
Class A Pentaxian Join Date: Aug 2008 Location: Wellington, New Zealand Posts: 11,251	Originally posted by ogl DXO Mark - Full SNR for each camera. You can go to K-5 and K200D's page with measurement and you will see... OK, mystery solved. The "Full SNR" measurements are provided in terms of pixels (called "Screen" by DxOMark). This favours sensors with less MP. The correct way to compare image quality (as opposed to individual pixel quality) is to look at the whole image (called "Print" by DxOMark). Since the K-5 has ~1.6 times more pixels, you need to add ~2 dB (= 3 * log2(1.6)) to all K-5 "Full SNR" figures. When you do this, you'll see that the K-5 is ahead at all SNR levels. Just compare the "Full SNR" figures at 18% with the figures in the regular regular "SNR 18%" chart. You'll see that the figures correspond to the (irrelevant) "Screen" figures and that the K-5 is ahead in the "Print" figures.

10-28-2012, 04:19 AM	#39
Class A Pentaxian Join Date: Aug 2008 Location: Wellington, New Zealand Posts: 11,251	Originally posted by ogl Adam's test - EV0. If you want you can make the same test with your camera...It's not very big problem for old cameras too. Not fast. Sometimes very slow. But it could. I'm not sure what you are saying. The video is proof that the K-5 II locks focus in light levels where the K-5 struggles. Also, we know that the K-5 has a low-light AF accuracy problem. From all I've heard this also has been addressed in the K-30 and K-5 II.

10-28-2012, 06:39 AM	#41
ogl Banned Join Date: Dec 2007 Location: Sankt Peterburg Photos: Gallery \| Albums Posts: 8,382	Originally posted by Class A I'm not sure what you are saying. The video is proof that the K-5 II locks focus in light levels where the K-5 struggles. Also, we know that the K-5 has a low-light AF accuracy problem. From all I've heard this also has been addressed in the K-30 and K-5 II. K-5 has accurancy problem in any light condition.

10-28-2012, 07:11 AM	#42
Tom S. Veteran Member Join Date: May 2008 Location: S.E. Michigan Photos: Albums Posts: 4,317	Originally posted by ogl K-5 has accurancy problem in any light condition. What are you basing that statement on? Do you own a K-5? Or is this another one of your posts trying to stir up crap?

10-28-2012, 07:23 AM	#43
Class A Pentaxian Join Date: Aug 2008 Location: Wellington, New Zealand Posts: 11,251	Originally posted by ogl It's wrong way to compare SNR18%. No, it is absolutely the correct way. You need to compare full images. Higher MP cameras have noisier individual pixels but in combination the higher number of pixels average out to the same image noise (everything else being the same). Are you really questioning DxOMark's approach to scale every sensor to 8MP in order to normalize the results? I'm surprised they don't do it for the "Full SNR" figures, but surely there is a reason why the "Print" view is the default as opposed to the "Screen" view.

10-28-2012, 08:21 AM	#44
ogl Banned Join Date: Dec 2007 Location: Sankt Peterburg Photos: Gallery \| Albums Posts: 8,382	Originally posted by Tom S. What are you basing that statement on? Do you own a K-5? Or is this another one of your posts trying to stir up crap? Yes. Of course. Be more polite, please... Sold K-5 1 month ago.

10-28-2012, 08:26 AM	#45
ogl Banned Join Date: Dec 2007 Location: Sankt Peterburg Photos: Gallery \| Albums Posts: 8,382	Originally posted by Class A No, it is absolutely the correct way. You need to compare full images. Higher MP cameras have noisier individual pixels but in combination the higher number of pixels average out to the same image noise (everything else being the same). Are you really questioning DxOMark's approach to scale every sensor to 8MP in order to normalize the results? I'm surprised they don't do it for the "Full SNR" figures, but surely there is a reason why the "Print" view is the default as opposed to the "Screen" view. You can think in this way... I won't argue. I've heard that they did it for every camera - full SNR too - to scale every sensor to 8MP. Didn't you hear that K-5 is noisy camera in shadows?