Bokeh Noise

Marcus

I've always admired the beautiful bokeh shots that get uploaded to Pixie but have never quite managed to replicate the effect myself...

In the below shot you can see terrible noise in each circle and I'm wondering what causes this? Is it the lens quality (15-85mm Canon EFS on 60D)?



By comparison, the below shot from MorningTea is relatively noiseless! This is what I'm after...

DaireQuinlan

you probably need to blow it out a little more. That radial pattern would get drowned out to some extent. Is your lens perfectly clean ? Do you have a filter on it ? Could just be the lens I guess, zoom = tons of lens elements so there's lots of room for aberrations to creep in.

5uspect

Are you using a Diffractive Optics lens by any chance?

See here:

http://www.the-digital-picture.com/Reviews/Canon-EF-70-300mm-f-4.5-5.6-DO-IS-USM-Lens-Review.aspx

Marcus

DaireQuinlan wrote: »

you probably need to blow it out a little more. That radial pattern would get drowned out to some extent. Is your lens perfectly clean ? Do you have a filter on it ? Could just be the lens I guess, zoom = tons of lens elements so there's lots of room for aberrations to creep in.

The lens has a relatively clean UV filter on the front, there could be dust of the back or on the sensor...

5uspect wrote: »

Are you using a Diffractive Optics lens by any chance?

No I'm using this lens: http://www.the-digital-picture.com/Reviews/Canon-EF-S-15-85mm-f-3.5-5.6-IS-USM-Lens-Review.aspx

I'm going to borrow an 85mm 1.2 and repeat to see what sort of difference the lens makes.

I also wonder if the light emission from the LED style Christmas lights have anything to do with it...

DaireQuinlan

I was looking quickly through a few shots on flickr, such as this one:

http://www.flickr.com/photos/shandilee/5196894990/lightbox/

Seems to be relatively common, from a quick sampling I found it on primes, wide zooms and long zooms alike. So I'd say maybe your LED theory might have some water, or your lens is dirty :-D

.Longshanks.

Full frame vrs a crop?

kelly1

I doubt it's Noise, Marcus, just an imperfection in the lens optics/design.

Marcus

Thanks all... I'll post the result I get with the MUCH more expensive 85mm 1.2 lens for comparison!

McVitae

This is almost defintely NOT noise caused by the sensor since the artefacts are circular in nature. if you look closely you can see a slight banding pattern of concentric rings within each circle. There is no meaningful way that a sensor could achieve this. The LED idea is an interesting one as LEDs are very imperfect light sources but this is usually in terms of imperfections in their colour spectrum and not in terms of thier emission with respect to thier viewing angle which might cause this. I bet my money on lens imperfections or the filter on the front

Hugh_C

McVitae wrote: »

This is almost defintely NOT noise caused by the sensor since the artefacts are circular in nature. if you look closely you can see a slight banding pattern of concentric rings within each circle

May not necessarily be the sensor, but could be dithering in the A/D converter. But probably optics.

Whatever, I kinda like it that perfect bokeh example Marcus posted is sorta dull and boring. Bring me gritty bokeh any day

kelly1

Hugh_C wrote: »

May not necessarily be the sensor, but could be dithering in the A/D converter.

Quite an imagination you have!

DaireQuinlan

Hugh_C wrote: »

dithering in the A/D converter.

16... no... 17... no... wait... what about 16. 16 looks about right. 16 it is so. Unlesssss, 15 ? no, way too low. Maybe I'll just play it safe and go with 16. 17 is tempting though. Ohhhh I don't know... Look. I'll do all the neighbouring pixels first and get back to this one. Right. Next pixel. Looks like a... 12 ! That's a big difference though. Hmm. Isn't the AA filter meant to smooth this crap out ? Let's say 13 and move on. Unless of course...

.
.
.
.
.

dacogawa

This is a weird one, I put your image into my video editing software & noticed something, on the bokeh circle at the bottom (2nd from the right of centre) there is one of the tiny circles that is very different from the rest of them in the image, it's the only one I can find at the edge of a bokeh circle but it has different shadowing.

So is this something to do with light hitting 1 fungus spore somewhere ? as the dots look very like spores (image below)



Any chance of taking a few more test images ?

dacogawa

Just googled it & found this link, I do think it's spores now, hope it's not sorry Marcus

http://blog-mo-mukha-mo.tabulas.com/2010/12/29/dirty-bokeh/

Marcus

I took the shot again last night, this time with the UV filter removed. After examining several shots I can see that the small circle imperfection (top right below) is likely sensor dirt as it appears on all images in exactly the same spot.



What is interesting is that the green light on the left of the image is a fluorescent street light, not the christmas tree LED lights and has a completely different, smoother look. The street light was approx 30M away while the tree lights were about 5M away.

McVitae wrote: »

I bet my money on lens imperfections or the filter on the front

Above is with the UV filter removed, so it's not the filter...

dacogawa wrote: »

So is this something to do with light hitting 1 fungus spore somewhere ? as the dots look very like spores (image below)

Any chance of taking a few more test images ?

Lens glass is perfectly clean and I can't see any evidence of a fungus...

DaireQuinlan

Looks like a classic airy disk. I'd also say that spot on the left must be on your lens, not your sensor, as the concentric pattern around it looks like a diffraction pattern. Doesn't sensor dust normally just look like a blob ?

-edit- Actually, just looking at my flickr contact stream,
www.flickr.com/photos/bertdesign/8283853732

Voigtlander Nokton f/1.5 . Not a cheap lens :-D

kelly1

I think it's due to cosmic background radiation. Any advance on that?

dacogawa

Could be stuff on the sensor, have you tried the self cleaning mode ?

http://www.canon-europe.com/Support/Consumer_Products/products/cameras/Digital_SLR/EOS_60D.aspx?faqtcmuri=tcm:13-821764&page=1&type=faq

Also just to make sure everything is fine with the lens too, I'd try the flashlight test

http://www.kenrockwell.com/tech/flashlight-test.htm

& I grabbed this from google, did it before myself on my old camera, it shows up everything

How to see sensor dust
If your camera has dust on its sensor, you can quickly spot it by doing the following:
Set your camera on Aperture Priority Mode.
Set your metering mode to Matrix/Evaluative Metering.
Set your camera ISO to the lowest number such as ISO 100 or 200.
Turn off Auto ISO.
Turn off autofocus and set your lens on manual focus.
Set your aperture to the largest number available for your lens by rotating the camera dial. For example, the minimum aperture on the Nikon 50mm f/1.4G is f/16, so if I were shooting with this lens, I would set my aperture to f/16.
If you are outside, point your camera up at the clear blue sky and take a picture. If you are indoors, find plain white paper, zoom in all the way so that the paper fits the whole frame, then make sure that the lens is completely out of focus and take a picture. If you are in front of a computer, open up a text editor such as Notepad, maximize it to the screen and then get as close to the monitor as possible so that only the white color is visible in the frame. Make sure that your focus is way off (completely out of focus) – that way only dust particles will be visible.
Zoom in on the image (rear camera LCD), scroll from left to right and top to bottom all over the image and see if you can find any dark spots.
If you cannot see any, your sensor is clean. If you see dark spots like in the above example, then your sensor has dust on it.

Hugh_C

DaireQuinlan wrote: »

16... no... 17... no... wait... what about 16. 16 looks about right. 16 it is so. Unlesssss, 15 ? no, way too low. Maybe I'll just play it safe and go with 16. 17 is tempting though. Ohhhh I don't know... Look. I'll do all the neighbouring pixels first and get back to this one. Right. Next pixel. Looks like a... 12 ! That's a big difference though. Hmm. Isn't the AA filter meant to smooth this crap out ? Let's say 13 and move on. Unless of course...

.
.
.
.
.

All I can say is ... it's f0cking dithering OK ? The source is no less than the august Engineers in UL.

Hugh_C

kelly1 wrote: »

I think it's due to cosmic background radiation. Any advance on that?

Pfft!

Neutrinos

sineadw

Hugh_C wrote: »

Pfft!

Neutrinos

...travelling faster than the light hitting the sensor, therefore causing abberations.

QED

DaireQuinlan

sineadw wrote: »

...travelling faster than the light hitting the sensor, therefore causing abberations.

QED

Also spelling errors. In a similar vein though, on a post-it note stuck to one of my monitors I have the word 'correlation'. The 'rr' is circled and marked with a red '2', and the 'l' is circled and marked 'only 1 l'

sineadw

DaireQuinlan wrote: »

Also spelling errors. In a similar vein though, on a post-it note stuck to one of my monitors I have the word 'correlation'. The 'rr' is circled and marked with a red '2', and the 'l' is circled and marked 'only 1 l'

Af ffs! I even looked it up, and *still* managed to misspell it..

Adrian.Sadlier

Hugh_C wrote: »

Pfft!

Neutrinos

Yes, and they've mutated!