Kodak KAF8300 v's Sony ICX814 *Noise comparison*

[swag72]

How many times have I heard that the Kodak KAF8300 sensor is ....... shall we say ........... past it's best. I hear this so many times compared to the Sony chips. It's a noisy old thing and can't compare to the clean Sony sensors.... that type of thing. While I was hunting out some M42 stuff, I chanced upon the following.....

14x1800s Ha - With the Sony ICX814, 3nm Astrodon and Tak FSQ85 0.73x - Calibrated with darks, flats and bias.

Would you believe that I'd just taken the following

14x1800s Ha - With the Kodak KAF8300, 3nm Astrodon and Tak FSQ85 0.73x - Calibrated with darks, flats and bias.

So I set about a stretch and levels of both images to get the same 90ADU figure on the same pixel. I then cropped them both from the same area and then magnified the area to look at the noise. I won't say right now which one I think wins, as there's a little visual below that I've done to hopefully show my findings. I've also attached the PI graph of noise in the PI subframe selector script.

Discuss

 

[hughgilhespie]

Gosh,

You don't think that using actual - errr - facts is a bit risky? I mean, obviously dearly cherished opinions will always win out over boring facts won't they.

Regards, Hugh

[wxsatuser]

I will throw the first punch in this fight but it's a glancing blow.

I'm enjoying the output of the QSI 583 and it looks good to me but otoh have'nt got an 814 to compare.

[swag72]

2 minutes ago, wxsatuser said:

I will throw the first punch in this fight but it's a glancing blow.

I'm enjoying the output of the QSI 583 and it looks good to me but otoh have'nt got an 814 to compare.

As I had the means to compare, with the only physical change being the sensor (of course the sky conditions are different) I just couldn't resist looking at the differences!

@hughgilhespie - I understand that trying to defend anything with facts (as much as they are) is a risky business

[gnomus]

I sense that you enjoyed that a great deal Sara... 

[ChrisLX200]

I never felt my 16803 was noisy - despite 9e readout noise (gasp!)

ChrisH

[GTom]

Very interesting! What was the sensor temperature for the cams?

[StuartJPP]

If people just got on with imaging instead of reading forums about untruths, conjecture, innuendo and occasional ignorance we'd actually see more pretty pictures, which is mainly what I want to see.

I very much believe that 80% is down to the individual and the rest the kit...but you can't say that for fear of hurting people's feelings (after all they spent a fortune on the kit).

I also believe that proof is in the final image, not in the graphs, charts or spec sheets.

 

[steppenwolf]

Yup, I'm imaging with my noisy out of date 8300 equipped CCD camera right now. To be honest, it's doin' just fine right now - like always I love t'internet but there is bias everywhere and that probably includes me. My advice? Just enjoy what you have and strive to get the best from it.

Great comparison, Sara - gosh, based on your results, I think I'd go for the *******

[swag72]

I'm not trying to make anything out of this.... just posting a couple of image from two sensors, same detail, same amount of imaging time and different sensor (one that everyone raves about ... Sony ICX814 and one that many say has had its day KAF8300).... I know which one I prefer and I hope it will redress the balance a little and people will be able to make up their minds based on unbiased evidence..... I spent too much time thinking that the KAF8300 was useless when in fact in my opinion, in noise at least it woops the Sony out of the water according to this basic comparison

Meanwhile people should look at REAL life scenarios instead of spouting theory with little else to base facts on

There's enough people out there spouting theory....... lets look at some real life results.

[swag72]

2 hours ago, GTom said:

Very interesting! What was the sensor temperature for the cams?

Temperature was -10 ... I always use this

[ollypenrice]

Give me a sec before I respond. As a user of the bronze age 11 meg Kodak chip it takes me a while to get my noise-proof ear plugs out. Oof, there we go. I can hear you now. My chip is so noisy that if I had any neighbours they'd complain. I know this chip is a noise disaster because I've read about it on Cloudy Nights.

Reality check. Tom and I calibrate our 'notorious' 11 meg chip data by using bad pixel maps and  master bias made nearly 2 years ago plus a hot pixel filter in AstroArt. 5 min, 10 min, 15min, 30 min subs all get the same files. Why don't we update our calibration files? Well, you see, we - ah - don't find there is anything to clean up once we have a stack. The oldest camera has a couple of dead columns but it has thousands of hours (quite literally) under its belt and these can be fixed. As for other forms of noise after calibration - well, we don't find any.

If you're not convinced I'll send you a stack.

Now sensitivity might be a different matter. I've seen some fast action here from Sony chips. Since I don't have one I can't comment from experience but, living in the country, I'd find a small plot hard to live on...

Olly

[swag72]

1 minute ago, ollypenrice said:

......... I know this chip is a noise disaster because I've read about it on Cloudy Nights......

Exactly my point Olly and why I felt compelled to make this comparison...... There are many posts out there across many forums where people try to dismiss the KAF8300 (and you are using an even noisier chip!!) but it would seem without A SHRED of evidence beyond theoretical forum clap trap....So I hope that this thread and visual will help people like me....... Someone who was gullible enough to believe the naysayers and who couldn't easily find the facts behind the guff.....to make a more informed choice.

So I guess I am saying that after looking at the above.... a comparison that is as scientific as I can make it....... make your own decision about what an outdated dog the KAF8300 really is when compared to the 'clean and venerable' Sony chip.

I can do nothing more than offer a comparison of the two under circumstances that are as similar as almost anyone would make them

[andrew s]

Each to his own approach but if you want a quantitative answer you can do a Photon Transfer Curve for your CCD. Google it with narrowbandimaging.com

Regards Andrew

[ollypenrice]

1 minute ago, swag72 said:

Exactly my point Olly and why I felt compelled to make this comparison...... There are many posts out there across many forums where people try to dismiss the KAF8300 (and you are using an even noisier chip!!) but it would seem without A SHRED of evidence beyond theoretical forum clap trap....So I hope that this thread and visual will help people like me....... Someone who was gullible enough to believe the naysayers and who couldn't easily find the facts behind the guff.....to make a more informed choice.

So I guess I am saying that after looking at the above.... a comparison that is as scientific as I can make it....... make your own decision about what an outdated dog the KAF8300 really is when compared to the 'clean and venerable' Sony chip.

I can do nothing more than offer a comparison of the two under circumstances that are as similar as almost anyone would make them

In the end anyone in his or her right mind is going to look at the pictures that are published out there, look at the optics, the cameras and the exposure details, and make a sensible choice.

Olly

[andrew s]

9 minutes ago, ollypenrice said:

In the end anyone in his or her right mind is going to look at the pictures that are published out there, look at the optics, the cameras and the exposure details, and make a sensible choice.

Olly

Such faith in human rationality is touching Olly

Regards Andrew

[vlaiv]

Not sure if this proves that KAF8300 produces less noisy image than ICX814, well it obviously is so in this case, but there is really simple explanation for this:

Kodak has 5.2um pixel size, and Sony has 3.69um pixel size. Shots were taken with the same scope, meaning that resolution of ICX814 image is higher.

It would be interesting to compare images produced at the same conditions - target position, sky transparency, LP (though I doubt it has any serious impact since this is narrow band), aperture and resolution.

Otherwise, to actually compare these images, if shooting conditions were sufficiently similar - one should bin them to same resolution. For example binning of ICX814 to 5.2um would give SNR boost of x1.4

[johnrt]

Hee hee, what fun!

I have to agree with the comment made above, that in the right hands these cameras "so old there almost steam powered" with Kodak chips will be trotting out world class images for many a year to come.

If I was starting out today, I wouldn't be worrying about the graphs and statistics, but looking at the images these cameras can turn out and thinking if I can make one half as good, I;ll be more than happy.

Good point about the pixel size and resolution though.

I wonder where it will all stack up in 5 years time with the rise of CMOS?

 

Sony CCD chip owner, of Sevenoaks.

[swag72]

I don't have the ICX814 chip anymore to do any sort of further experimentation.... but as it stood, this was an interesting comparison for me. I guess I'd say that it's a real life example of the difference without getting tied up in the resolution argument (which I do understand).

I want images as large as I can get them and so in the real world binning wouldn't be an option for me.... so from this, would it be safe to say that without making a picture from the ICX814 1,4x smaller, you'd need 1.4x as much exposure as the KAF8300? Or is the relationship not linear in that way?

[vlaiv]

Very simple and approximate formula would be: 4 x exposure == bin 2x2 == stack 4 frames. So time has linear dependence on area but quadratic in side length. If pixel size ratio is x1.4 (5.2:3.69) you would need x2 exposure length to get same SNR if you don't consider read, dark and sky noise.

[steppenwolf]

Quote

There's enough people out there spouting theory....... lets look at some real life results.

+1

I've never made an equipment choice based on graphs and specifications alone - I want to SEE what can be achieved with a certain piece of kit and the Internet makes that simple to do provided you cross-reference your finds.

Quote

Very simple and approximate formula would be: 4 x exposure == bin 2x2 == stack 4 frames. So time has linear dependence on area but quadratic in side length. If pixel size ratio is x1.4 (5.2:3.69) you would need x2 exposure length to get same SNR if you don't consider read, dark and sky noise.

Fascinating and I don't doubt that you are right but in the real world of imaging, I know that I would be much more interested in looking at relevant images than deciphering datasheets and carrying out the mathematics! There's nothing like other people's real life experiences but as usual, cross-reference each source to remove bias to get an overall picture (see what I did there with two imaging terms?).

[GTom]

55 minutes ago, vlaiv said:

Very simple and approximate formula would be: 4 x exposure == bin 2x2 == stack 4 frames. So time has linear dependence on area but quadratic in side length. If pixel size ratio is x1.4 (5.2:3.69) you would need x2 exposure length to get same SNR if you don't consider read, dark and sky noise.

Looking at the specs, I see only the read noise as a clear advantage of the sony sensor. Dark/thermal is around the same. Maybe the specs werent that accurate and sony is actually worse in thermal? 

Looking at the sample pics i was wondering how the icx694 (allright, bigger pixels than the 814) can be champ in limiting magnitude...

[swag72]

54 minutes ago, vlaiv said:

Very simple and approximate formula would be: 4 x exposure == bin 2x2 == stack 4 frames. So time has linear dependence on area but quadratic in side length. If pixel size ratio is x1.4 (5.2:3.69) you would need x2 exposure length to get same SNR if you don't consider read, dark and sky noise.

So in a real world scenario........ I would need to obtain double the amount of data from a ICX814 sensor to get the same amount of noise as a KAF8300? 

[RichLD]

I bought a second hand QHY9 purely based on the NB images of JP Metsavainio - I have to say it's been a marvellous (and relatively cheap in AP terms) workhorse and has given me many hours of pleasure, never letting me down. If I had taken too much notice of all the bad press these cameras used to get I would have been put off my purchase and would have probably never have been able to afford a mono camera...

The KAF8300 gives me very reasonable image quality on prints of up to 20"x16" - it's a great combo with the ED80 as well, and I got it with the filter wheel for a grand 

[vlaiv]

2 minutes ago, swag72 said:

So in a real world scenario........ I would need to obtain double the amount of data from a ICX814 sensor to get the same amount of noise as a KAF8300?

Not really, there are a lot of factors that you need to take into account to actually determine exposure time difference between two sensors. For example the approach that I use goes something like (I actually go other way around, given time budget I try to estimate SNR):

Fix expected SNR, and then based on following (and all of these are measurable quantities, but some of them are approximated based on previous knowledge, like LP levels, transparency and such): target position combined with sky extinction / transparency, LP levels, QE of each sensor, optical train losses, aperture of system, resolution of each sensor, read noise, dark noise, amount of light from target ( I usually go for the least bright parts when using this for estimating SNR, but for comparison of sensors you can fix this to some meaningful value of mag/arcsec squared), any hardware binning, any software binning, sub length and then stick that into spreadsheet and get total number of subs / total amount of exposure needed.

So if you do all of this for given two sensors on narrow band target in really dark skies and broadband galaxy kind of target in heavy LP skies you will end up with two totally different total exposure ratios. So its not as easy as saying, oh, you need x2 exposure on this sensor compared to the other as it depends on all written above.

33 minutes ago, steppenwolf said:

Fascinating and I don't doubt that you are right but in the real world of imaging, I know that I would be much more interested in looking at relevant images than deciphering datasheets and carrying out the mathematics! There's nothing like other people's real life experiences but as usual, cross-reference each source to remove bias to get an overall picture (see what I did there with two imaging terms?).

Yes, I know what you mean, I on the other hand appreciate the fact that there is a way to predict things prior to actually conducting the experiment

 

39 minutes ago, GTom said:

Looking at the specs, I see only the read noise as a clear advantage of the sony sensor. Dark/thermal is around the same. Maybe the specs werent that accurate and sony is actually worse in thermal? 

Looking at the sample pics i was wondering how the icx694 (allright, bigger pixels than the 814) can be champ in limiting magnitude...

I think that same approach can be taken when comparing icx694 to both 814 and kaf8300, so it depends on quite a few factors ...

As for limiting magnitude, if we are talking about star detection limiting magnitude, I would advocate a bit different approach: I think that way to go about that is to consider pixel size. Given scope aperture and focal length I would go for pixel size (or binning) that would give me best SNR on a single star. I would aim for pixel size that is just big enough to capture most of the light from a stellar source (depends on airy disk size / seeing psf) but not bigger in order not to capture too much sky light (as it introduces additional noise).