Don Goldman, Ph.D.
August, 2009 - May 2011
TOPICS (click on link to answer your question)
What is a narrowband filter?
Why are they useful?
What is the most popular narrowband filter?
What about other popular narrowband filters?
What about H-beta or other narrowband filters?
What is Mapped Color?
What is the Peak Transmission of the Filter?
What is the Bandwidth, Width, FWHM of a Narrowband Filter?
What’s the Catch?
What about Astrodon Performance and Guarantee?
What about Astrodon Quality and Coating Durability?
How do I Select the Right FWHM for Me?
What About Using Narrowband Filters with Faster Optics?
What are the Negatives for Narrowband Imaging?
What about NII and H-a?
Can I mix narrowband filters of different FWHM?
OK, I Bought Your Narrowband Filters - Where's the Data?
Which End is UP?
Do I Need a Longer Exposure with 3 nm Filters?
When I began in astrophotography in 2002 many people were already imaging with H-alpha filters. I believe that my first H-alpha filter was a 9 nm Schuler filter. It was thicker than my LRGB filters and hence required manual refocusing. The same was true of Custom Scientfic H-alpha and LRGB filters that were sold by SBIG at the time, It was one issue that I resolved to fix when I came out with the first Astrodon filters which were all made to the same thickness and hence, did not require refocusing.
Early color narrowband images were made with H-alpha, OIII and H-beta, but since the H-alpha and H-beta contained the same information, they really could not be called tricolor. You could get the same result only using H-a and OIII by duplicating the H-a data, mapping the duplicated data to blue and decreasing its contibution relative to the red-mapped H-a. No need for H-beta. For astrophotography, H-beta provided redundant information, with a weaker signal and at a wavelength that has a lower quantum efficiency for red-sensitive CCDs, in addition to all that extra exposure time needed to collect H-beta in addition to H-alpha. These are the reasons why H-beta is not used in astrophotography today, even by those who initially used H-beta.
The first true tricolor (using 3 different elements) narrowband image in amateur astrophotography was made by Russell Croman on May 29 - June 8, 2003. It was an H-a/OIII/SII image of the Eagle Nebula, M16, done in mapped color using the Hubble color palette. It was similar to the famous "Pillars of Creation" from the HST. In a word, it was jaw-dropping, showing what amateurs could do with limited equipment from their backyards in city environments. It inspired me and many others to enter the world of narrowband imaging that uses the high contrast of these filters to bring out structure in a way that RGB filters simply cannot do. He used 3 nm filters, for reasons discussed below. Russ went on to produce some of the most impressive and beautiful narrowband images, even based upon today's high standards. He should be given credit, alongside others, for early development and popularization of amateur narrowband astrophotography.
So, with this preamble, let's delve into the technology, selection and use of narrowband filters for astrophotography.
A narrowband filter is designed for a specific astronomical emission line, such as hydrogen-alpha (H-a). It only passes a few wavelengths of light along with the emission line, rejecting all other light. As a result, the contrast improves dramatically, showing a great deal of structure. Think of it as a red laser compared to a red light bulb.
There are three major reasons. First, they bring out fine structure much better that RGB filters. Second, they allow you to image when the moon is up, thereby extending the time that you can use your equipment. Third, they allow you to image from light-polluted locations, where RGB imaging may be problematic.
You can see the enhancement of detail as the filter becomes spectrally narrower, from red filters to narrowband filters. You can even see more detail as the narrowband filter becomes narrower from 9 to 4 nm, not only in the bright object, but also in the background nebulosity.
H-a (hydrogen-alpha) at 656 nm (nanometers) is deep red in color and the most popular narrowband filter. Hydrogen is ubiquitous in the cosmos and is present in emission nebula (North American, Pelican), planetary nebula (Dumbbell, Ring), Wolf-Rayet objects (Crescent, Thor’s Helmet) and supernova remnants (Veil). Many imagers like to present just a black-and-white H-a image of an object. It is has a beauty all by itself, like an Ansel Adams photo. However, most imagers blend their H-a data into their red RGB data to enhance structural detail while maintaining a “natural” look. Therefore, the H-a filter should be your first narrowband addition to your LRGB filters. The basic imaging set of 5 filters becomes LRGBH-a.
The next filter to complement your LRGBH-a set is an oxygen filter. OIII (“oh-three”) emits light near 500 nm and is a blue-green- or teal-colored filter. Many of my images of planetary nebula and supernova remnants are taken only with H-a and OIII filters. They show great structural detail, but have natural colors, looking like an RGB image.
If you want the look of Hubble Space Telescope (HST) images, such as the famous “Pillars of Creation” (the Eagle Nebula, Messier 16), then the next filter to consider adding to your collection after H-a and OIII is the SII (es-two) or sulfur filter. The SII filter, like H-a, is a deep red filter near 672 nm. H-a, OIII and SII filters provide that Hubble look for many emission nebula. Again, tri-color narrowband imaging with these three filters can be done with the moon up, so your equipment is not sitting dormant for several weeks.
H-beta is of interest for scientific research, but is not useful for narrowband imaging. It is a blue filter at 486 nm. It has the same information as the H-a filter, but its signal is one-quarter to one-third as intense. Thus, it does not really add new information and provides a weaker signal. Helium (He) and Nitrogen (NII – “en-two”) filters can show different structures in certain objects like planetary nebula, but are more restricted in their use. Astrodon does offer a NII narrowband filter for those wishing to explore these differences (see section below).
It is obvious when you combine your RGB data into a colored image how the colors are assigned. Narrowbands are different, since OIII is a mixture of green and blue, and H-a and SII are both red. So, we need to decide on what “color palette”, or color mapping protocol to use. The Hubble palette assigns (maps) red to SII, green to H-a and blue to OIII (in order of wavelength). This generally results in those familiar central blue backgrounds surrounded by golden yellow shock fronts in HST images. This image of the Pelican Nebula that I took is mapped this way. The golden color results from mixing green H-a and red SII that are enriched in the shock fronts. However there are many other ways to map colors. So, it is up to the imager to decide how best to portray his/her object. Regardless of what color palette is used, the goal is to use color to emphasize structure in an aesthetically pleasing manner. Our hobby is very technical, but the end result is art – technical art.
This peak transmission of the filter is a very important parameter. It represents how much light passes through the filter. This is expressed in percent, such as 95%T (T=transmittance or transmission). This means at that wavelength, the filter passes 95 out of every 100 photons to the detector. Commercial (non-astronomy) filters typically transmit only 50-60%. Astronomy is much more demanding, and nowadays, narrowband astronomy filters transmit in excess of 80%. The transmission of the filter improves the signal in the signal-to-noise (S/N) of your image. So, you want this value as high as possible. Astrodon 5 nm narrowband filters typically have 95%T, the highest value available, for this reason. Even our ultra-narrow 3 nm filters are guaranteed to be >90%T.
This is also a very important parameter that we use to describe narrowband filters. Let’s envision the transmission of a narrowband filter as a bell-shaped curve. Refer to the above chart. There are two wavelengths at 50% of the peak (maximum) transmittance; one at a higher and the other at a lower wavelength than the central or peak wavelength. The wavelength difference of these 50% points is referred to as the full-width-at-half-maximum transmittance (FWHM). It is also referred to as the bandwidth or width of the narrowband filter. When you hear about a 5 nm narrowband filter, the 5 refers to the bandwidth or FWHM in nm (nanometers). Narrowband filters are considered to have widths of 10 nm or less, with most ranging from 4.5 to 7 nm. Astrodon also offers an ultra-narrowband with a 3 nm width. Narrower filters enhance contrast by reducing the broadband light which decreases the noise (N) in S/N. The background signal decreases linearly as the bandwidth decreases. Narrower filters make it easier to image with the moon out (broadband light) and from light-polluted locations.
The goal is to maximize S/N. Therefore you want the highest peak transmission with the smallest bandwidth. However, it is more difficult to make narrower filters without losing peak transmission. If you buy a narrower filter with lower peak transmission, then S decreases as N decreases and you will likely not see any improvement in S/N. Astrodon has solved this problem and provides 5 and 3 nm narrowband filters guaranteed with peak transmissions > 90%T. In fact, they are typically closer to 95%T. Thus you can be assured of maximum S and lower N than other filters.
From the outset, Astrodon made a decision to provide the highest performing narrowband filters made to professional research standards for the imaging community. The above discussion explained why providing the highest peak transmission with the smallest FWHM is paramount in this regard, but we went one step further. We guarantee >90%T peak transmittance at the emission line as measured in an f/9 system and we put this guarantee on every box. To our knowledge, we are the only filter provider to guarantee this performance. This is expensive to do, but it guarantees to you the performance that you expect from Astrodon without the need to place scans of the filter in each box. Here is an example of why less expensive filters cannot make this guarantee.
This is a close-up spectral comparison of our Astrodon 5 nm H-a filter and a competitor's 7 nm H-a filter taken on a high-resolution spectrophotometer. We tested 3 different 7nm H-a filters. The Astrodon filter reaches a peak transmittance of 96% at H-a. The competitor's filters reach or exceed 90%T but not at the H-a emission line. At the H-a emission line, the competitor's wider filter reaches 86%, 74% and 71%, or 9% to 25% less than the Astrodon. The greater, but more costly manufacturing control assures the Astrodon filters will perform as advertised. Without such guarantee from Astrodon, you simply will not know if your lower cost H-a filter has low efficiency and will not know why it takes so long to achieve good signal-to-noise.
The above discussion only addressed the signal (S) in signal-to-noise (S/N). The FWHM of the filter affects the noise, N. The wide filter, as shown in the sequence of the Crescent nebula above, loses contrast compared to our narrower filters. Let's examine how much.
These data were taken with a 12.5" RCOS RC telescope, Astrodon 3, 5 and 6 nm H-a filters, and Apogee U16M CCD camera. Three 10 minute exposures were taken and calibrated. The same region in each frame that lacked emission nebulosity and stars were analzyed and the average ADU signal was measured. The trend is shown above and is linear.
The competitor's 7 nm H-a filter would then have 24% greater background signal than our Astrodon 5 nm H-a filter based upon the above analysis, while having 9% less signal. Our Astrodon 3 nm H-a filter has nearly 50% less background signal. In summary, this translates into better contrast for finer detail, clarity in fainter extensions of emission regions and distinction of background nebulosity.
In addition to providing superior optical performance for highest contrast, Astrodon also provides the highest quality substrates and coatings. Our substrates are straie-free, polished to 1/4-wave propagated wavefront and both sides are parallel to better than 30 arcseconds. You may hear others use terms like "planeoptically polished", but it means the same thing. Our substrates are single pieces of highest quality fused silica or other clear optical glasses and are not laminates made by gluing multiple, thinner pieces of glass together. We totally rely on the coatings to provide our bandpass and out-of-band blocking performance, and do not need absorbing colored glass substrates.
There is a misconception that filters cut from larger coated plates are somehow inferior to individually coated pieces both in optical quality and sealing of the coatings at the cut edges. This is simply incorrect and may suggest a problem with adhering their coatings to their substrates if their coatings are cut. Cutting filters from larger plates does not degrade 1/4 wave or 30 arsecond parallelism specifications. This is why all major coating companies have fully equipped glass fabrication facilities on-site. They all have the need to cut filters from larger coated plates or special shapes for their various clients.
Let's put some numbers to this. Every large plate prior to coating is precisely measured to ensure that our research-grade specifications are achieved and exceeded. Among six different lots, the thicknesses were measured to be 3.010 +/-0.001 mm. Our specification is 3.000+/-0.025. The centers of these large plates were measured to be 0.045+/-0.006 per inch propagated wavefront and the edges were 0.044 +/-0.006, whereas our specification is 0.25/inch (1/4 wave). We cannot find thickness specifications of the 7nm H-a or wider 8.5 nm OIII and SII filters. This means that you can rely on Astrodon quality in every single filter.
There is another reason for this misconception about using larger plates. They use colored glass substrates to block light outside the bandpass either toward higher or lower wavelengths, thereby reducing the need for more coatings. This lowers their production costs. However, this also lowers their peak transmission and also creates more problems in making narrower, higher contrast filters, such as those provided by Astrodon. Furthermore, it may be that larger plates of colored glasses contain straie, forcing the need to individually select small substrates that are straie-free that can be individually coated. Astrodon relies solely on coatings to provide the highest contrast in terms of peak transmission and out-of-band blocking. Our out-of-band blocking is 1-3 orders of magnitude better in our 5 nm H-a filter at wavelengths longer than the bandpass than their 7 nm H-a.
Individually coated pieces are often held in a coater with a metal container having a "lip" that covers the top edge of the glass. This often leads to a thin region at the edge of the filter that is minimally coated or uncoated because the "lip" prevents the coating from being deposited underneath. Sometimes, this region is ground down, ensuring that the coating goes completely to the edge. This is expensive. If an uncoated rim exists, then that region needs to be blocked by the filter wheel or other means to prevent light leakage. The image below is of the same low-priced 7 nm H-a 50.8 mm diameter, 3 mm thick filter showing this rim both in reflected light and in transmission in front of a 70W incandescent light bulb. Please note that the edge of this 7nm has been blackened, which does not mitigate light leakage through the rim.
By coring out round or cutting square filters from larger plates, coatings are ensured to go completely to the edge, as shown in the Astrodon filters below..
Furthermore, Astrodon uses the hardest oxide coatings on both sides of the substrate from the latest coating technologies, including ion beam and magnetron sputtering. The edges can be cut and do not have to be sealed against moisture attack. We have a 5-year limited warranty against coating delamination.
Our 5 nm narrowband filters will be appropriate for most people. Reflected moonlight peaks near the OIII wavelength. OIII will pick up more moonlight and show more gradients than H-a or SII filters. Since we tend to use narrowband filters when the moon is up, a 3 nm OII filter may be a better choice to reduce the effects of moonlight, especially if there is a blanket of reflective snow on the ground. It is OK to mix narrowband filters of different bandwidths. If you live in a light polluted location, all 3 nm filters may be a better choice. I personally prefer all 3 nm filters for highest contrast and detail.
You may have heard that a narrowband filter shifts its bandwidth to shorter wavelengths with faster optics. As this shift occurs the transmission at the emission wavelength may decrease, making the filter less efficient. This is not significant with our 5 nm narrowband filters for most scopes slower than about f/3. The 3 nm filters may lose perhaps 15%T at f/3, but is still useable.
Please note that Astrodon guarantees >90%T at the emission line for slow (e.g. f/9) systems for BOTH its 5 and 3 nm narrowband filters. 3 nm filters from one other vendor are specified at >70%. Thus, a loss of 15%T from a starting point of 93%T is still much better than a loss of 15% from a 75%T filter.
To demonstrate, below is a comparison of 10 min exposures of our 5 and 3 nm H-a filters taken on a Takahashi f/3 Epsilon-180 telescope with an SBIG STL11000 camera from suburban Sacramento, CA. Good signal is still achieved with the narrow 3 nm H-a filter at f/3 even with a 15% loss of signal due to the spectral shift because the lower background signal helps to compensate this.
We can take this one step further. Gary Gonella provided a single, uncalibrated 20-minute exposure from a QSI583 CCD camera on a C14 telescope outfitted with a Starizona hyperstar, making the overall system f/2. We simply processed this exposure in Photoshop and present it below, showing that even at f/2, there is still reasonable signal for a 3 nm H-a filter. Note that the vignetting in the corners has not been corrected with flat frames yet.
There are several reasons why the 3 nm filters still work well at f/3. First, the decrease in %T at the emission line as a function of incident angle is actually less than predicted by the simple equation found in most texts because the bandpass also changes shape at greater incident angles. You can see that in an angular study of our 3 nm OIII filter shown below.
The %T at the 500.7 nm emission line decreases from 93% to about 70% at 8 degree off-axis, corresponding to an f/3.5 system. This trend can be seen in the graph below. This shows why 3 nm narrowband filters can be used on an f/5 refractor without loss, which starts to occur faster than f/4.5.
However, this is not the entire story. The optics of the telescope must also be considered. Not all light rays are coming into the detector within the camera at 8 degrees off-axis at f/3.5. We have modelled an f/3.5 system with a 20% obstruction and generated a histogram of all rays of light hitting a 50 x 50 mm square detector, as shown below.
The chief ray does come in at 8 degrees but there are many other angles, also. One must convolve (multiply and sum) this curve with the %T vs angle curve above to determine the overall efficiency loss. When that is done, we find that the effective efficiency only decreases at the OIII emission wavelength to 80%, and not 70% as predicted just from the angular measurements of the filter alone. This is still very high performance for such a narrow filter.
This is why, even at f/2.92, the signal remains strong for the 3 nm filter.
This result is important, in that many imagers wish to use the same narrowband filters on more than one optical system. This great usability of 3 nm filters to f/3 makes this possible. This provides imagers a choice of a 3 nm filter for highly light polluted locations even for fast f/3.5 systems.
There are a few ultra-fast optical system that operate at f/2. In this extreme case, our 5 nm filters is probably a better choice.
Narrowband filters are more expensive than LRGB filters. Stars are much dimmer, making guiding behind narrowband filters more difficult. As a result, you may need to consider off-axis guiding, such as with an Astrodon MonsterMOAG, or a separate guide scope, where in both cases, the guide camera sees unfiltered light. Also, narrowband signals are inherently weak, requiring long exposures of 20 – 40 minutes. This requires a mount that is stable over these times with acceptable periodic error.
This is a bit complicated. It is not well known that most H-a filters pass both H-a and NII. H-a emits at 656.3 nm and NII emits most strongly at 658.4 nm (and weakly at 653.8 nm). These are very close together spectrally. Thus, most H-a filters are wide enough (e.g. 4.5 nm bandwidth and wider) to pass both emission lines as shown for the older Astrodon 6 nm filter above. Our 3 nm H-a begins to separate both emission lines and reduces the NII contribution significantly, also shown above (blue curve). In this example the 3 nm filter only transmits 15% at the NII 658.4 nm wavelength, whereas the H-a remains unchanged. As mentioned earlier, some objects are enriched in NII, such as planetary nebula and Wolf-Rayet bubbles. The Dumbbell Nebula, M27, is a good example, as shown below, taken with 3 nm narrowband filters. The wispy clouds in the core of M27 are dominantly NII. A tricolor narrowband image is also shown below, mapping OIII to blue, H-a to green and NII to red to produce a beautiful color image. This information provides you with a choice based upon your light pollution, desire for more detail, or simply wanting all the photons you can get out of your H-a filter.
Yes. Stars will be smaller with narrower filters, but the effects can be dealt with in processing. Processing will be required even if your bandwidths are the same. This arises because OIII and SII are much weaker signals than H-a. OIII and SII data must therefore be stretched during processing to match the intensity of the strong H-a. Stretching makes those stars larger, often even larger than the H-a stars. You may have seen magenta halos around bright stars in tricolor narrowband images. This is how they arise, since we often map OIII to blue and SII to red. Blue + Red = Magenta. Software methods such as deconvolution prior to color combining can offset some of these effects. Selective color correction that desaturate magenta is another means to reduce the problem.
If you are used to the strong signals that you get from your LRGB filters, you may be disappointed or even concerned with the relatively weak narrowband signals, especially from OIII and SII. They may only be 500 - 1000 ADU (counts - analog to digital units) as displayed on your computer screen. But, in most cases, these signals are there. Your first look at your object is from your data acquisition program just after a frame is downloaded. This is a raw frame, that is, it has not been calibrated with darks, biases or flat frames, and has hot pixels. The image is scaled by these hot pixels pushing the weak narrowband data into the background. So, it looks like there is almost nothing present. One trick that I use in MaximDL is to perform a Kernel/Median filter on the displayed image. The image below shows a 30 min. raw frame from an Astrodon 3nm H-a exposure of IC1795 (left) and after the application of the filter in Maxim (right). Notice how the H-a signal, especially in the fainter background regions, pops out after the filter is applied. This filter only operates on the displayed data. The file is not changed unless you ciick Save, so do not click on Save.
Lastly, please realize that the OIII and SII data can be relatively weak and that you will have to "stretch" the data during processing to achieve a better match with the typically strong H-a data. These processing methods include, digital development (DDP) in programs such as MaximDL and CCDStack, and curves, shadows/highlights and others in Adobe Photoshop.
This is one of the most frequently asked questions for unmounted filters. Which way should the filters be oriented in the filter wheel? Which side should face the telescope? In short, the answer is that it does not matter.
There are two issues that need to be considered separately:
Halos are typically 100 pixels in diameter or less and are centered or nearly centered around bright stars. They arise from a design problem in the filter - optical "crosstalk" between the surfaces that are 2-3 mm apart, i.e. the thickness of the filter. As, such filter orientation will not have an effect on halos.
Larger scale reflections produce larger artifacts, such as exit pupil ghosts that may involve the filters. The distance between the filter and CCD window or sensor, or a rear optical element (field corrector) in the telescope are often several cm, so the size of the artifact is much larger, perhaps 500 pixels in diameter or larger. This is a system property within which the filter can contribute. They can occur between the highly reflective CCD sensor and the chamber window of the camera, for example. However, if there is a problem, it will occur regardless of filter orientation.
Please realize that 1.25" mounted filters for the smaller CCD sensors are mounted in screw-in holders by the filter supplier. No orientation of the actual filter is specified. These mounted filters screw into filter wheels made by different camera manufacturers differently. Some face the telescope and some face the CCD sensor. If there were major problems with filter orientation, you would have read about it on the internet. The general absence of such discussions suggest that filter orientation is not a prevalent problem.
A common myth is that only one side of modern astronomy filters has antireflective (A/R) coating. This is simply not true. The shinier coating having the bandpass has A/R built in. If not, we would not achieve 99% transmission within the bandpass. The same is true for the other side, which may be a high-quality, broadband A/R coating, or a combination of blocking layers (that prevent light from transmitting outside the bandpass) and A/R. In the latter case, the coatings on the two surfaces cannot be considered separately.
Some filters have an arrow on the side pointing toward the shinier side and a generic note suggesting to point the arrow to the light source. Filters made for microscopy, fluorescence work or other applications involving lasers place the shinier side toward the light source so that any heating will not be within the body of the filter if the laser enters from the other side. This is not the case for astronomy. Other astronomy filter suppliers simply suggest placing the shinier side toward the telescope without providing any rationale.
So, if you are still concerned about filter orientation, how can you tell which is the shinier side of the filter?
Astronomy filters are coated on both sides of the substrate. One side is commonly shinier than the other due to the overall design. You can tell which is the shinier side by looking at the filter at a glancing angle. You will likely not be able to see into the body of the filter when the shiny side is up. Conversely, you should be able to see into the body of the filter on the opposite side at a glancing angle. The picture below of an Astrodon 50 mm square 5 nm H-a fllter demonstrates this effect where you can only see the internal back and side walls on the side opposite to the shiny side.
The next image of these same H-a filters is very instructive. It show identical reflections of the camera from both sides, and that is the point. If there is an optical problem in the
telescope system, the side of the filter facing the telescope will not make any difference. You will have the same problem either way, such as a reflection between a rear optical element in a refractor and the filter, or between the filter and the detector in the CCD camera. It is akin to asking which is worse - a reflection from a first- or second-surface mirror. There may be slight differences in reflectivity, but you will still get a strong reflecton from both types of mirrors. This is demonstrated in the reflections above.
Therefore, again, it does not matter which side of the filter faces the telescope.
No, because the peak transmission is still above 90% for both 3 and 5 nm filters. If fact, it will take you longer to reach the sky noise limit with the narrower filter due to the lower background. I take 30 minute exposures with both filters.
Happy Narrowband Imaging!
Don Goldman, Ph.D.
Copyright 2010 Astrodon. No portion of this document may be reproduced without written permission from Astrodon Imaging