The mysteries of green.

[ollypenrice]

Macavity, in another thread, linked to this article. http://nogreenstars.blogspot.co.uk/2013/03/why-are-there-no-green-stars.html It's excellent.

It explains why there are no green stars. (Quick explanation; the Sun peaks in energy output in the green part of the spectrum but the peak in terms of photon radiance, which is what our eyes perceive, is away from green down towards the reds. Also the eye's colour sensitivity derives from three types of receptor and the response of that for green overlaps heavily with that for red.)

So that got me thinking about the one shot colour Bayer Matrix and the question, 'Why does it have twice as many green pixels as red or blue?' When we shoot using a monochrome camera and RGB filters we don't shoot twice as much green as we do blue or red. This article http://en.wikipedia.org/wiki/Bayer_filter gives explanations. It seems the purpose of the double set of green filters is to allow them to serve as luminance. Bayer (whose christian name is Bryce - for completeness!) described the green pixels as 'luminance sensitive elements' and red and blue as 'chrominance sensitive elements.' But...

This is predicated on the fact that, in daylight, the human eye is most sensitive to green.

Now when we are taking astrophotos we are not working in daylight and we are not collecting light from the sun which is the source of our daylight spectrum. So do we want one shot colour cameras to have twice as many green pixels? I would have thought not. I would have thought that the 'luminance' function of green pixels would collapse on most astronomical targets. Many imagers use Ha (which is red) as partial luminance on emission nebulae, so a better matrix would be RRGB in these cases. For reflection nebulae I'd have thought RGB ideal.

So is this why tests show monochrome RGB-filtered cameras to be faster than OSC cameras? The OSCs have the wrong filter distribution.

And it also shows the clear advantage of using a luminance filter. The luminance filter doesn't have to guess which colour is the best to serve as luminance on this particular target, it gets the lot and the peak will be somewhere in that data. It doesn't matter where.

Olly

[Knight of Clear Skies]

There is an alternative to the bayer filter, which is mentioned in the wiki article above. It uses a matrix incorporating panchromatic cells for better low-light performance at the expense of reduced colour information. Unfortunately I don't believe there are any in production, but perhaps one day - it would be more suitable for astronomy than the traditional bayer.

This is one proposal for an RGBW filter, the white pixels are sensitive to all visible wavelengths.

[ollypenrice]

Yes, Earl has mentioned this in the past and it would be a step in the right direction. However, I think any matrix optimized for daytime is likely to be compromised for astronomy.

Olly

[Zakalwe]

I would assume that the astro CCD sensors are "general purpose" items. The manufacturers (Sony, Kodak, Micron etc) will develop the sensors to suit a wide range of applications, such as machine vision, microscopy, video work and so on. The vast majority of the taget market for these sensors will probably require something that approximates human vision.

The astro-CCD manufacturers take these general-purpose sensors and re-purpose them into their products. As the market for amateur astro gear is small, the likelihood of a sensor manufacturer doing huge amounts of research, development, testing, wafer fabrication for such a small market is uneconomical. 

[Knight of Clear Skies]

However, I think any matrix optimized for daytime is likely to be compromised for astronomy.

Yes, I think you're right, due to how human vision works. As I understand it, mono + filterwheel is definitely the best solution while OSC is cheaper and more convenient. With OSC if an imaging run is cut short a colour image can still be produced and seeing is consistent across all three channels, putting less of a burden on processing skills. This can potentially claw back some of the time lost by being a less efficient light-gathering solution. In dodgy UK weather I do wonder if there is a strong case for OSC over LRGB. Of course, narrowband imaging on mono opens up a lot more possibilities.

The astro-CCD manufacturers take these general-purpose sensors and re-purpose them into their products. As the market for amateur astro gear is small, the likelihood of a sensor manufacturer doing huge amounts of research, development, testing, wafer fabrication for such a small market is uneconomical. 

Yes, it's a similar story with telescopes and camera lenses - telescope design is more suitable for astro but lenses benefit from scale of production. There is some great kit out there for us to play with but the design boundaries probably aren't being pushed in some areas as it's a small market.

[Stargazer33]

Would it not work to have the RGB matrix in a triangle? So on the top line you would have RGBRGBRGB etc and then on the next line - shifted by half a pixel - BRGBRGBRG!

There would be an uneven edge to the matrix but that could be outside of the used area of the matrix.

[ollypenrice]

I realize that we have no hope of seeing alternatvive astro matrices, alas. It just struck me as interesting to follow the whole subject through. (Imagine an HaRGB matrix!)

Given that there is a surfeit of green from a Bayer matrix I'd question the idea that the colour balance will be right straight from the camera, though it's easy to adjust.

Olly

[Alien 13]

The problem is that our perception of colour is not linear white is not a 33% split so even if a camera was produced that could capture RGB in equal amounts our brain wouldnt like it and would tweek it back to white.

Alan

[themos]

Fuji innovates again

[ollypenrice]

Fuji innovates again

Still too much green, though!

Olly

[Knight of Clear Skies]

Yeah, some white lego bricks would be good.

[Zakalwe]

Still too much green, though!

Olly

Again, it's being developed to replicate human vision in daylight to suit it's target market.

What you are looking for is a specialist sensor for a very narrow market. I guess it's unlikely to be found.

[ollypenrice]

Again, it's being developed to replicate human vision in daylight to suit it's target market.

What you are looking for is a specialist sensor for a very narrow market. I guess it's unlikely to be found.

I'm not actually looking for one because I'm very much wedded to LRGB but astronomy's only hope would be a large market coinciding with our needs, maybe in medicine or industry - or the military.

Olly

[Zakalwe]

I'm not actually looking for one because I'm very much wedded to LRGB but astronomy's only hope would be a large market coinciding with our needs, maybe in medicine or industry - or the military.

Olly

Good point Olly.

I guess, as with most things, it's a compromise that OSC users will have to live with.

having said that, as technologies advance, the increased sensitivity of chips will go some way to make up for the shortfalls?

[cfpendock]

 I'm very much wedded to LRGB

Olly

I know that it is not strictly to do with this subject, but I seem to recall you saying that for some targets you took RGB unbinned.  If that was the case, what would be the benefit in using luminance as well?

Chris

[lensman57]

I would assume that the astro CCD sensors are "general purpose" items. The manufacturers (Sony, Kodak, Micron etc) will develop the sensors to suit a wide range of applications, such as machine vision, microscopy, video work and so on. The vast majority of the taget market for these sensors will probably require something that approximates human vision.

The astro-CCD manufacturers take these general-purpose sensors and re-purpose them into their products. As the market for amateur astro gear is small, the likelihood of a sensor manufacturer doing huge amounts of research, development, testing, wafer fabrication for such a small market is uneconomical. 

Quite right, I would class them as industrial sensors. We are just not worth the investment for the returns or the cameras will have to be even more rediculously priced. I wonder if Fairchild Semiconductor would like to get in the amatteur market.

A.G

[ollypenrice]

I know that it is not strictly to do with this subject, but I seem to recall you saying that for some targets you took RGB unbinned.  If that was the case, what would be the benefit in using luminance as well?

Chris

With the  present rigs I never bin the RGB. My current pixel scales are 1.8 and 3.5 arcsecs per pixel and that is already coarse enough, I reckon. Also, in my processing, I try to use no luminance on stars so I need nice tight ones in RGB.

The advantage of luminance is very simple; it captures red and green and blue simultaneously and is, therefore, three times faster than RGB. In fact my own testing suggests it's nearer to four times faster. (I guess because the filters don't provide total overlap.) So in an LRGB shoot the L part is 3 times more efficient than shooting through filters, whether OSC or RGB. This gives a speed advantage of around 6 to 4.

Another thought occurs to me. The RGB filters on an OSC are presumably not very sophisticated. I'd guess they are absorption filters. Now the Baader LRGB interferometric filterset is supposed to be more efficient and is also supposed to combat light pollution. I've heard it said, though I've no experience of this whatever, that RGB beats OSC in light pollution. I stress that I don't make this claim, I'm just repeating it, but this might be the explanation if the claim is true.

Olly

[Gina]

The market is not even all astro imagers as a large proportion of us use LRGB with a mono camera. (Or NB with mono camera.)

[D4N]

Since colour is simply how we perceive different wavelengths of we will always have the issue of translating what a sensor detects into something we can visualise.

It's my understanding that the RGGB setup is designed to output an image as close as possible to what we would actually see regardless of whether that actually records an accurate representation of what wavelengths of light arrived at the sensor.

TSED70Q, iOptron Smart EQ pro, ASI-120MM, Finepix S5 pro.

[themos]

I wish someone would make a sensor with big pixels for the long exposures interspersed with small pixels for short-exposure guiding.

[ollypenrice]

Since colour is simply how we perceive different wavelengths of we will always have the issue of translating what a sensor detects into something we can visualise.

It's my understanding that the RGGB setup is designed to output an image as close as possible to what we would actually see regardless of whether that actually records an accurate representation of what wavelengths of light arrived at the sensor.

TSED70Q, iOptron Smart EQ pro, ASI-120MM, Finepix S5 pro.

Yes, but that's 'actually see in daylight.' Of course we look at the picture in daylight, so perhaps that's your point. Complicated!   

I wish someone would make a sensor with big pixels for the long exposures interspersed with small pixels for short-exposure guiding.

Why do you want small pixels for guiding? I guide with large pixels binned 2x2. It's the centroid calculation that matters, I'd have thought, so resolution isn't the key parameter. (I'm not talking about gudining at propfessional focal lengths here. I know nowt about that.)

Olly

[D4N]

I think the problem there would be reading from the small pixels while the large ones are still exposing.

The fuji super ccd has small and large pixels interspersed hexagonally but they are designed to be read at the same time to increase the dynamic range.

I'm fine with my OAG though

TSED70Q, iOptron Smart EQ pro, ASI-120MM, Finepix S5 pro.

[Earl]

The image taken off Iankings site of he sbig setup is very luminance heavy, id really like to see soem images done with these, there are some but not many.

A custom debayer will need to be applied, and the colour data is going to be weaker than the luminance, i feel.

Which leads me to think is this a OSC camera or  Mono camera that also does colour.

Edit still a green lean to it colour wise.

[Earl]

Why does the combined RGB not equal luminance? is it a sensetivity issue? (can you debayer a single green pixel rather than using 2?)

[ollypenrice]

Why does the combined RGB not equal luminance? is it a sensetivity issue? (can you debayer a single green pixel rather than using 2?)

I don't know. Maybe if you look at the three colour passbands they don't fully overlap, so there is some loss of transmission. When I look at the SN of 3 Hrs RGB and an hour's L I never feel the RGB is quite as good. I usually weight it at about 28% or so.

Olly