IDAS LPS-P3 Light Pollution Suppression Filter

[vlaiv]

2 minutes ago, Alexis said:

Thanks for the welcome -- but if you look at the filter specs (and I just looked at it through a pocket spectrometer as well -- that matches) you wouldn't expect a drop of signal of a factor 2.

 

Again, that depends.

If you take for example 600-700nm range to be red part of the spectrum as is with interference filters, then you would expect a drop to about 50% (or factor of x2) if observed spectrum is uniform:

I outlined 600-700nm range. Black curve is filter response curve. Area under the curve is roughly equal to area above the curve - this means that in 600-700nm range filter cuts uniform spectra roughly to half of its original strength.

You can of course have also other extremes - signal can be filtered by factor of x100 (to 1% of its original value) - and also filtered by factor of x1.05 - to 95% of its original value - but that would mean that signal itself is not very uniform.

Here are example of such cases:

Here we have example of "orange" signal - that is all concentrated in 620-630nm part. Such signal would be completely obliterated by this filter.

On the other hand - if that signal was in range 680-690nm like this:

It would only loose about 5% of its value.

This all holds true if we use very sharp cut off filters for colors - between 600-700nm for red.

Above data was taken with OSC camera. Such camera does not have sharp cut off filters - it QE graph looks like this:

Red is sensitive all the way down to 470nm (although small sensitivity) and even a bit below 430nm. Bulk of sensitivity of red goes down to about 570nm.

It is also worth noting that part between 600-650nm is about 20-25% more sensitive than part between 650-700nm.

D3 Filter blocks more sensitive part - meaning that more signal is cut by filter.

Above measured x2.5 stronger signal measured without filter - fits nicely to this if you observe red from 565nm to 700nm and its distribution vs filter cut off.

[Alexis]

Agreed, and it does explain something important for a photosensor that is less important for me when I use it visually (I'd be hard pressed to see anything above 610nm with night vision) -- but you're right that even I see less improvement for yellowish objects like the M81 arms than some other stuff, even visually.

But you also saw a lot less signal for green (1.9x), which for these objects is quite important, and that's more puzzling.

The red I expect to want to see most here is the H-alpha in M82, but that shouldn't suffer too much (and it takes quite a bit of stacking to pull it out of the noise anyway).

Not doubting either the images nor the data that comes out of them (if you just try to adjust levels on the two FITs it's quite obvious to the eye as well that your data is correct) -- just trying my best to understand the causes.
 

Edited June 2, 2021 by Alexis

[vlaiv]

4 minutes ago, Alexis said:

Agreed, and it does explain something for a photosensor that is less important for me when I use it visually (I'd be hard pressed to see anything above 610nm with night vision)

But you also saw a lot less signal for green.
 

Same logic holds - compare green channel in IMX571 response curve graph and filter response graph. You will get that it cuts down light in similar fashion  - but not as drastic as with red. It cuts it down to say 55% or so (factor of x1.9).

If you are solely interested in visual - best thing to do would be to actually try that filter on your scope. With imaging we are concerned with improving SNR per exposure (or for total image). For visual you want different thing - you want threshold of JND - just noticeable difference.

You want to make nebulosity be at least 10% than LP (actually some sources quote visual JND to be around 7% but let's go with 10%).

So it does not matter how much darker target gets - if sky gets more darkened in comparison to unfiltered version. It's a bit like using higher magnification. If you double the magnification - everything will get darker by factor of x4 in light intensity (light is spread over x4 larger surface). Cutting light to 50% is not that big of a deal.

Actual benefit for you will also depend on spectrum of your light pollution. How much darker does background sky get with filter?

Say you want to observe sqm22 target in sqm18 skies.

10% is 0.1 as ratio and that translates into 2.5 magnitudes of difference. Now, you have 4 magnitudes of difference between target and sky. Will this filter help you or not?

Say that your LP is such that this filter cuts it down by factor of x5 while it cuts down green part of spectrum of your night vision by factor of x2. This means that you improve contrast ratio by factor of 2.5. It no longer needs to be 0.1 - it can now be 0.04 when it is unfiltered as filter will raise that to 0.1

Again - we can translate 0.04 into magnitudes - it is ~3.5mags of difference - you still won't be able to see sqm22 target in sqm18 skies - but you'll be able to see sqm22 target in sqm18.6 skies for example (just barely detectable).

In any case - filter acted as if sky was improved by roughly 1 magnitude in darkness.

 

[Alexis]

Quote

If you are solely interested in visual - best thing to do would be to actually try that filter on your scope.

I don't only try, I use it regularly (the advantage of having a filter slide is that it's farily easy to just try -- you have little to lose). It replaced an Omega Optical GCE and is a lot better.

See above -- effectiveness varies with the objects. Granted it also depends a lot on the sources of light pollution -- it's spectacularly ineffective against metal halide/mercury vapour (we can "test" that at one of my darkish sites at less than an hour's drive: on Saturday the neighbouring village has a football match; luckily we stay up a lot longer than the duration of the match.) but quite good against the ubiquitous sodium street lighting which is really everywhere in Belgium.

Which is why I was slightly suprised at the lack of effectiveness in photography. Although in our skies it might still help more than in this example -- when they use a fast scope and expose 120s, most people here don't get an image that is as neutral as the posted examples, but a nice HP sodium glow.

It'd also be interesing to see if in _this_ example the SNR would still be worse if the exposure for one light frame was longer (at least with this sensor).

Edited June 3, 2021 by Alexis

[vlaiv]

1 hour ago, Alexis said:

I don't only try, I use it regularly (the advantage of having a filter slide is that it's farily easy to just try -- you have little to lose). It replaced an Omega Optical GCE and is a lot better.

See above -- effectiveness varies with the objects. Granted it also depends a lot on the sources of light pollution -- it's spectacularly ineffective against metal halide/mercury vapour (we can "test" that at one of my darkish sites at less than an hour's drive: on Saturday the neighbouring village has a football match; luckily we stay up a lot longer than the duration of the match.) but quite good against the ubiquitous sodium street lighting which is really everywhere in Belgium.

Which is why I was slightly suprised at the lack of effectiveness in photography. Although in our skies it might still help more than in this example -- when they use a fast scope and expose 120s, most people here don't get an image that is as neutral as the posted examples, but a nice HP sodium glow.

It'd also be interesing to see if in _this_ example the SNR would still be worse if the exposure for one light frame was longer (at least with this sensor).

Yes, it has a lot to do with type of light pollution.

I'm using IDAS P2 for imaging. We also have quite a bit of yellow type HPS installed all over the city. Here P2 does a good job.

Here is aerial photo of my city (I roughly marked my current location with arrow):

SNR with/without filter will not depend on exposure length used.

Only difference exposure length makes is with respect to read noise - other than that, there is no difference as both signal, and LP noise grow with exposure time in same manner for filtered / unfiltered scenario (math is really simple - there is signal and signal noise called shot noise and there is light pollution signal / background and its associated noise, and a bit of thermal signal and its noise - each signal grows linearly with time and each noise component is square root of that signal - filter just reduces particular signal strength by some percent - except for thermal signal - which is very small anyway compared to rest).

[Alexis]

Quote

SNR with/without filter will not depend on exposure length used.

That depends on the source of the noise and what dominates. If sensor-induced noise dominates more when you filter because you suppressed more signal, you might see better performance with longer exposures. Of course if there is readout noise in the sensor you still need enough exposures to get rid of that as well so for a given total integration you need to keep the amount of subs high enough as well, so it's not a simple subject. Given it's not simple I like experiments -- the proof of the pudding etc.

We're totally in agreement about shot noise in the picture, of course.

Edited June 3, 2021 by Alexis

[vlaiv]

8 minutes ago, Alexis said:

If sensor-induced noise dominates more when you filter because you suppressed more signal, you might see better performance with longer exposures.

What sort of sensor induced noise are you talking about?

From your response - you seem to distinguish between read noise and this other kind. Are you referring to dark current noise? (that is usually very small with cooled astro cameras - often lower or equal to read noise in exposure times normally used).

[Alexis]

Quote

that is usually very small with cooled astro cameras

Agreed. I haven't been paying enough attention to what sensor was used here.

Edit: yeah, for a cooled ASI2600 colour I doubt it's that relevant.

Here, though, readout noise might be relevant, no, since we're just looking at one sub? Or is it low enough for it to be irrelevant given the signal in both the filtered and unfiltered image? [I'm sorry if the question sounds stupid, I don't know the particulars of the different sensors that well, and you're much better at this than I am 😉 .]

Edited June 3, 2021 by Alexis

[vlaiv]

18 minutes ago, Alexis said:

Agreed. I haven't been paying enough attention to what sensor was used here.

Edit: yeah, for a cooled ASI2600 colour I doubt it's that relevant.

Here, though, readout noise might be relevant, no, since we're just looking at one sub? Or is it low enough for it to be irrelevant given the signal in both the filtered and unfiltered image? [I'm sorry if the question sounds stupid, I don't know the particulars of the different sensors that well, and you're much better at this than I am 😉 .]

Well, we have stats above. Those are in ADU units, and if I'm not mistaken gain was set to around 0.25 e/ADU (from fits header).

Yep, it's about 0.25 (to round things up).

Measured background values with filter:

R: 613.77
G1: 2222.72
G2: 2222.75
B: 1290.32

Even if we take lowest one - 600 and multiply that with e/ADU to get number of electrons - ~125e.

So LP alone had signal of 125e or about 11.8e of noise (square root of signal).

ASI2600MC Pro has about 1.5e of read noise at gain 100 - which was used (marked is read noise and e/ADU value that we saw in fits header):

If LP noise is ~11 - that is at least x6-7 larger than read noise. Since noise adds like linearly independent vectors (square root of sum of squares) - 1.5e of read noise is going to make about 1% of difference to overall noise at most.

sqrt(1.5^2 + 11^2) = sqrt(2.25+121) = sqrt(123.25) = 11.1

It really makes minimal difference over already present noise - and this is for filtered red channel - which had by far the least amount of light pollution present (all others had more LP and hence higher LP noise - so read noise was even more inconsequential).

In general - read noise is really not issue with "normal" exposure lengths (few minutes or more) in light polluted areas - since LP noise easily swamps read noise.

 

 

[Alexis]

See? It's faster for me to ask you ;-). Thanks for the illustration.

[Lee_P]

Thanks @vlaiv, you're a legend! From visual inspections I thought that the D3 was having minimal impact one way or another, but it's so much better having numbers to compare. 

[uhb1966]

I just wanted to say THANKS A LOT @Lee_P For doing this awesome work!