[ds7711]

Just upgraded from Nikon D7000 to pentax K1. The main purpose is to do occasional astrophotography using the astrotracer. So, after getting the new K1, I tested the noise at different ISO settings: 1) different ISO, same exposure (increase shutter time); 2) same exposure time, different ISO, then use Adobe Camera to adjust the exposure to the same level (all exposure were adjusted to ISO 6400 equivalent). The following is my subjective results.

First, different ISO, same exposure.
The result shows there seems to be two critical points: ISO 800 and ISO 25600. The noise levels from ISO 100 to 800 are really close. After that, increasing ISO leads to increased noise (looks like linear). However, if ISO is more than 25600, there's a big increase in noise. When ISO is more than 51200, the noise and color become unacceptable. Following are the images (around 70% ~ 100% crop, didn't remember the exam numbers... Lens: Pentax FA 50mm, F1.7, shot at F4). Corresponding images were uploaded and sorted by ISO.


Secondly, if I shot at relatively low ISO (800 and higher) and then adjust to ISO 6400, the noise level seems to be a constant. The noise level is similar when shooting at ISO 12800 and then decrease to 6400. However, shooting at 25600 (decrease to 6400) results in some clipping (maybe artifacts from change in light conditions).
ISO: 100















ISO: 6400





ISO: 51200



Edit: images from imgur are not showing up. Trying to fix but failed, but images are accessible through the link.

It seems, setting ISO to 800 may yield best results given that it provides the best dynamic range. I'm still learning so the conclusion probably is not true. Let me know if anyone has some insights or suggestions.

Best,
Ming

Attached Images

View Picture EXIF

File Date

 

View Picture EXIF

File Date

 

View Picture EXIF

File Date

 

View Picture EXIF

File Date

 

View Picture EXIF

File Date

 

View Picture EXIF

File Date

 

View Picture EXIF

File Date

 

View Picture EXIF

File Date

 

View Picture EXIF

File Date

 

View Picture EXIF

File Date

 

---

Last edited by ds7711; 06-25-2016 at 07:42 PM. Reason: Images not showing up

[Na Horuk]

Cool test! It looks really good, I remember how some cameras had terrible colours and noise already above ISO 1600.
Can you please take one more test shot like that, though? Like the last one, ISO 51200, but with pixel-shift

[JinDesu]

So these are cropped to 1280x900? Or some other size, then downsized to 1280x900 (being the forum size or something)?

Because these look damn good up to 51,200 - and I'd use the 51,200 if I needed to.

[ds7711]

So these are cropped to 1280x900? Or some other size, then downsized to 1280x900 (being the forum size or something)?

Because these look damn good up to 51,200 - and I'd use the 51,200 if I needed to.

@JinDesu These were directly cropped to 1280 x 900 (to my best knowledge, crop in ACR, directly save to jpg). You may lose dynamic range while using high ISO, however.

---------- Post added 06-26-16 at 06:02 AM ----------

Cool test! It looks really good, I remember how some cameras had terrible colours and noise already above ISO 1600.
Can you please take one more test shot like that, though? Like the last one, ISO 51200, but with pixel-shift

I'm mainly concerned with high ISO performance in shooting stars, where pixel shift doesn't apply.

So, I probably won't do that test. But I'm curious to see your results. I bet it will be much better as it average data from 4 frames.

[jbinpg]

There is no averaging with pixel shift. The information from all 4 frames is integrated but there is no pixel averaging as I understand it.

[rawr]

Related to the above, while we wait for DxOMark to test the K-1 to give us a ISO number and score for the K-1, we can use Bill Claff's charts:

Photographic Dynamic Range versus ISO Setting

He gives the K-1 almost identical numerics to the D810, and his data [effectively] suggests that ISO 3170 is the point at which K-1 image quality starts to visibly degrade. That seems about right to me.

(ISO 1394 is his figure for the D7000, which sounds good, given the similarity of the D7000 and the K-5).

[biz-engineer]

Makes sense in the way that eyes are sensitive to 6 bits of tone gradation. Visible noise at ISO3200 instead of ISO1600 means that you can pull 5 stops of shadow into the 8bits space of JPEG and still have visually appealing tone gradation and colors definition.

[normhead]

Related to the above, while we wait for DxOMark to test the K-1 to give us a ISO number and score for the K-1, we can use Bill Claff's charts:

Photographic Dynamic Range versus ISO Setting

He gives the K-1 almost identical numerics to the D810, and his data [effectively] suggests that ISO 3170 is the point at which K-1 image quality starts to visibly degrade. That seems about right to me.

(ISO 1394 is his figure for the D7000, which sounds good, given the similarity of the D7000 and the K-5).

Very cool chart.

Nikon D810 , 11.83 PDR, Low Light ISO 3206, K-1
Pentax K-1 , 11.36 PDR Low light ISO, 3170

At least from this report it would appear that the reports that the K-1 was a half stop better than a D810 appear to be in question. The D810 appears to be about a half stop better DR, but both could comfortably shoot at 3200 ISO, if you can live withe the loss of DR.