You have just acquired a brand new monochrome camera and want now to purchase RGB filters?

You already have a set of SOH 8nm narrow-band filters, but Astrobobby told you 3nm is better than 8nm?

OK, this page could be useful. 

Before we start

No matter if we are dealing with Red-Green-Blue filters or narrow-band filters, there are few things we need to know before ordering a new set of filters. 

For technical and practical reasons described below, it is better to see your filters as a set, more than an heterogeneous group of individual filters. A good combination of filters will help you to save precious time during and after your shooting session. 

Even more, your set of filters must be adapted to your camera, and maybe even your focuser.

• Really? 

Yes. And now let's try to see why

What is a filter for astrophotography?

In astrophotography as in your coffee machine, a filter is a device that rejects what we don't want, and passes what we do want. Intuitively, we recognize that the best coffee filter is not necessarily the one with the smallest pore size. You coffee will take much longer, and won't be any better. On the other hand, having a bigger porosity could let some coffee ground pass through and result into a really unpleasant experience. As a compromise, coffee lovers prefer a pore size between 10 and 20 microns, allowing water and good aromas to pass through. 

In astrophotography, a filter is not much different: it allows a selected electromagnetic spectrum to go through, and the rest is rejected. The major difference between a red filter and an H-alpha filter is the width of this spectrum, but the principle remains the same. The R filter has a very large spectrum (typically between 610nm and 700nm), and H-alpha filter has a very narrow one (typically between 653nm and 659nm). 

RGB filters

Have you ever wondered why some astrophotographers have different exposure times for Red, Green and Blue? There are usually two main reasons: Either they know about the theory, or... the last hours dedicated for the blue filter were cloudy :-)

Let's have a closer look to the spectrum of each filter:

Weird isn't it? It seems that the blue and red spectrum are larger that the green one. This would mean that for a "white" star, less green color than red will hit the sensor. But then, we should compensate it by having a longer exposure for Green, and less for Blue and Red?

Hmm... Maybe not. 

For the happy photons who will now pass through the filter, most of them will hit the sensor, and this is where quantum physics play its role. Here is an example of the quantum efficiency of the famous ZWO ASI 1600 for the visible and near infrared spectrum:

What now? It seems that each photodetector (or pixel if your prefer) is more sensitive to the green light (between 495nm and 570nm), than to red or blue. But then... would the spectrum width compensate this effect? 

Let's now compile the data in Excel, and multiply the Camera Quantum Efficiency by the Filter Transmittance: 

We can easily reproduce the same exercise with the Luminance filter, and we obtain the following Sensor transmission (Quantum efficiency x Luminance filter transmission):

The area below the Luminance transmission curve is now our reference. If we compare it with the area below the transmission curves of the Red, Green and Blue filters, we get following results:

• FILTER NORMALIZED TRANSMISSION
• Luminance: 1 (reference)
• Red: 0,34
• Green: 0,30
• Blue: 0,29

We can now calculate what would be the multiplicator of each filter exposure time in order to achieve the same luminous flux. By inverting the values above, we get the following results:

• FILTER 1/ NORMALIZED TRANSMISSION
• Luminance: 1 (reference)
• Red: 2,92
• Green: 3,33
• Blue: 3,47

If we consider that the 3 filters have exactly the same vignetting (which may or may not be true), we can now properly estimate the exposure time of each filter. In our example, 2,92h of exposure with the Red filter would correspond to 3,33h of Green, and 3,46h of Blue.

• OK, but what do you mean with "corresponds"?

This means that your Signal to Noise Ration (SNR) will be very similar with those exposure times. If one of the 3 filters is underexposed, we will need to mitigate the effect during the postprocessing or even worse, we'll add a background noise due to this underexposed color. 

Narrow-band filters

The usual combination of narrow-band filters is the magical SOH: Sodium II, Oxygen III and Hydrogen Alpha. If you practice astrophotography since more than a few month even with a digital camera, you surely have seen those magnificent pictures of the so-called Hubble pallet. 

This is not the subject of this article, but there are numerous other pallets, and plenty of different filters. However, we'll use the SOH combination as an example.

The estimation of the exposure time is the same as for the RGB. But unlike the RGB filters, narrow band filters have... a narrow bandwidth.

Let's have a look on the following diagram provided by BAADER:

What does this all mean?

• Bandwidth: 3,5 /4nm: the filters have different bandwidth to harmonize the exposure time from H-alpha with O-III / S-II. Same trick as for RGB, but for a much smaller bandwidth

• f/3: This type of filter is made for "fast" optics. For small F/D ratios, the "blue shift" needs to be compensated. Never heard of this blue shift? All technical details are very well explained in this article from Altair Astro:

https://www.altairastro.help/info-instructions/cmos/affects-of-telescope-focal-ratio-on-light-pollution-filter-performance/

• Ultra high speed filters: Can be seen as a marketing trick, but it indicates that the transmission ratio of each filter is high (in the example above: >90%)

Filter thickness

As mentioned at the very beginning, buying a consistent set of filters is a good decision. For the reasons described above, the exposure times are very likely to be harmonized even if they are not 100% identical.

But the bandwidth isn't everything. Ever looked at the filter thickness?

• Nope. But, does this have any influence?

 Yes. It slightly alters the backfocus, by adding approximately 1/3 of the filter thickness to the original value. Let's take an example: if the recommended backfocus of your favorite coma corrector is 55mm and the filter thickness is 3mm, you will need to consider 55+1mm of backfocus. And yes, 1mm can make a difference.

• So if I understand it well, different filter thickness results into a different backfocus. Does it modify the focus point?

The answer is yes. Now we start to understand why it is important to have a set of filters with identical filter thickness. Imagine that we start to image 3 frames of 5 minutes with OIII, and we continue with 3 frames of SII and 3 frames of H-alpha. With identical filters, you could theoretically perform only 1x focus for all filters, instead of one focus for each filter. Even when this procedure is automatized, it is time consuming and... each photon counts!

Narrow-band filter bandwidth

We are almost done. But one of the most controversial detail remains! What bandwidth shall we select? 2,5nm, 3nm, 4,5nm, 5nm, 6nm, 9nm or even 12nm?

OK let's relax. As we have seen before, a mix between 3,5nm and 4nm can help to harmonize the exposure times. Basically, the main categories are usually a multiple of 3 (3, 6, 9 and 12nm). Smaller bandwidth are usually more expensive, but are they so much better?

Here I can only speak about my own experience: With a Bortle scale between 5 and 6, our sky is extremely polluted by the city illuminations. Changing from 6nm to 3nm was significantly better in term of contrast, with much less background signal (and therefore, less background noise). 

Nevertheless, our telescope dealer didn't recommend to go lower. The redshift of far galaxies can be so high that the H-alpha is away from the filter "window". 

Otherwise, for any other object, this shouldn't be a problem if one important condition is respected: The filters spectrum must be perfectly "centered", which means that the emission ray it is supposed to isolate must be aligned with its maximum of transmission. 

In the case of H-alpha (656,3nm), the FWHM of a 3nm filter (see picture below) must be precisely set between 654,8nm and 657,8nm. A centering error of 1nm would have almost no influence on a 12nm filter, but it would ruin the effect of a 3nm filter.

Premium filter manufacturers deliver a measurement protocol with each filter. This is usually a guaranty of a serious manufacturing process.