It's been a while since the last report on my now not-so-new 20" telescope project so, here's a long overdue update. Make yourself comfortable--this is a long one.

Last fall I posted the progress I made to get the scope mounted along with my first raw images verifying that my cameras were well focused and that the telescope was at least roughly aligned. After that post, I still had to finish mounting my new EL flat panel so that I could calibrate some images. Little did I realize that the new flat panel would lead to cascading problems with both the telescope and my camera package that would bring the whole project to a screeching halt.

As most of you know, I've been operating a remote telescope in New Mexico for over three years and that experience has taught me a number of important lessons. First, Murphy's law is magnified by the distance between the telescope and the operator. If a scope works perfectly in your backyard, that's no guarantee that it will work as soon as it's a few states away. Things that you never thought of or ever experienced will fail at the worst possible times. The things that you guess might be the first to fail will run reliably for years while the most bullet-proof components will go tango-uniform only to work perfectly when you travel all the way out to middle of nowhere to fix the problem. Obviously, the problem is that when something fails at a remote site, the difficulty of getting it fixed is far beyond fixing anything in the backyard.

For those reasons, I consider deploying a new remote telescope to be the amateur imaging equivalent of launching Hubble. It better be well tested before it ships because the difficulty and expense of getting it fixed in the field can feel like launching a repair mission into space. And that issue gets magnified by a couple of orders of magnitude when you are shipping a big scope 8,000 miles to another hemisphere in another country. Even so, many folks totally underestimate the importance of carefully configuring and testing new equipment before it ships. I can't tell you how many folks have asked me why I didn't just have the new scope shipped directly from Planewave to the observatory in Chile! Hopefully this story will make it clear why it's a not a very good idea to just ship a pile of gear to a remote observatory with the idea that someone can put it together and that it will work.

So, let’s get back to those first calibration tests using my new 24” x 24” EL flat panel (shown in rev A). The moon was nearly full so I started by taking some narrowband images of the Bubble Nebula, which was in good place in the sky from my location in Oregon. I shot a series of 10-minute exposures through the Ha, O3, and S2 filters. Then I shot a series of darks, a sequence of bias data, and a series of flats taken with the flat panel positioned at the end of the scope—as it will be in the observatory. I shot all of the calibration data in the dark to avoid any problems with light leaks. When I calibrated the light data, I could immediately see that it wasn’t working correctly. The calibrated images all showed bright corners. I’ve been calibrating images for a long time but just to be sure, I triple-checked the calibration procedure to confirm that I was using the exact same method and settings that I always use with my C14 data.

When flat calibration doesn’t work, the very first thing to suspect is that unknown additive signals like stray light are screwing things up. But first, before I changed anything, I wanted to see if the new panel could somehow be the source of the problem so I took a sequence of sky flats at 90 degrees to the sun on a crystal-clear afternoon through the Ha filter. Same thing…the narrowband lights wouldn’t calibrate. The problem is pretty obvious in the raw calibrated frames but to make the problem VERY obvious here, I did a quick and dirty stack in the SRO palette which is shown in rev B. Obviously, the problem is extreme in the corners. Bright corners mean especially dark corners or an overly bright central region in the flat data. Rev C shows a comparison between a PW20 master Ha flat using the panel to a C14 master Ha flat using a flat panel. Interestingly, when I tried to calibrate LRGB data using the master sky flats, the results actually looked pretty close to right. However, trying to calibrate the same data with panel flats didn’t work at all. I tried a number of different configurations creating LRGB flats, moving the panel closer and further from the scope, using flat darks, and a number of other minor tweaks and nothing worked any better. Hmm…what’s going on here?

Next, I pulled the camera and pointed the scope at the flat panel so that I could look backwards through the system. I set up a 52 mm diameter ring in the same location as the sensor so that I could see what was happening at the very edge of the field (shown in rev D.) The images in rev E and F show what I could see when I looked in. Right away, I could see that the telescope baffles weren’t blocking direct light from the panel out past a radius of about 21 mm. Well…that’s not right! After some back and forth with the folks at Planewave, I learned that sure enough, the 20” was their first “big” CDK design and that they had only baffled it out to a 42 mm image circle. Oh, oh…that wasn’t my understanding of how the scope was designed when I ordered it! I got the impression that they had been meaning to fix it but had never gotten around to it. The good news is that they treated this as a customer service issue and quickly agreed to redesign the baffles and to supply a new set ASAP. So, PW gets high marks for product support and responsiveness. The photo in rev G shows the old and new primary mirror baffles. The new one is 3D printed and it includes well designed internal glare stops so that it is very black inside (as shown in rev H). I confirmed that the new baffles do indeed block light out to a 52 mm circle as shown in rev I. The one minor issue this created is that the original primary mirror cover wouldn’t fit over the new primary baffle so I had to carefully measure and cut a new hole in the mirror cover. I was pretty careful to measure multiple times but I was also really lucky that my hand cut hole came out with a near perfect fit on the first try (as shown in rev J.)

While I was waiting for the new baffles, I made a black foamboard baffle that mounted on the back of the secondary to block any external light from getting around the internal baffles and repeated my experiments with image calibration. Unfortunately, it didn’t help. I still couldn’t get any data to calibrate correctly with the flat panel. I never tried the one test that made a lot of sense, which was to put a 12.6” aperture over the front to see if the system would operate correctly at F/10.8. The one problem with that idea is that the obscuration ratio would have been about 70% with the larger central baffle but it still might have been a valuable thing to try.

I’ve never run into this situation and I was stumped. No matter what I tried, I couldn’t get calibration to work. So, I took a step back and reconsidered the advice that I often give others with this same problem: Fix the optics first before attacking other issues “downstream”. When I examined my light data, I could see right away that the amount of vignetting in the corners appeared to be pretty extreme.

So, let’s go back to how I laid out the whole camera/guider package in the first place. To make it easy, I simply took the design that worked so well on my C14 and figured out how to plug it into the back of the 20”. The diagram shown in rev K shows a meridional view of the light path and I knew that there would be some vignetting, which I figured would be easy to correct with flats. This layout worked perfectly on the C14, so what could possibly go wrong? And this is where I fell headlong into what should have been an obvious a trap.

What I hadn’t seriously considered was what was happening to the F6.8 ray bundles illuminating the corners of the sensor. My first diagram only looked at the rays in meridional plane and those rays only inscribe a 36 mm circle. The skew rays are the important ones! They are the rays than define the 52 mm image circle. At F/10.8 those skew rays are vignetted on the C14, but not by enough to matter. The problem becomes MUCH more significant when you add nearly two more F/stops to operate at F/6.8. Remember that image calibration divides one number by another and when the divisor is small, noise and with what might otherwise be minor offsets due to stray light and/or flat normalization, there may be large swings in the output. I have to admit that still I couldn’t say exactly what was causing the problem but it’s always good engineering practice to first fix the basic optical problems before going any further. Fixing the basics means minimizing mechanical vignetting, eliminating all potential sources of stray light, and carefully understanding the components, design, and layout to make sure that the system is working as it should be.

At that point, I did a full vignetting analysis of the design. My good friend Gaston Baudat at Innovation Insight kindly shared the dimensions of the XM guider so that I could do a careful analysis. Rev M shows the size of the openings compared with the un-vignetted beams at the ONAG input and output ports along with coverage over the internal dichroic beam-splitter. With the factory configuration, it looked bad and I initially thought that I’d have to abandon the whole idea of using ONAG on this system. I even designed an OAG solution and put together a parts list; but, after staring at the design for a while, it began to dawn on me that there might be a way to salvage ONAG on this system!

I realized that by simply opening the holes in the input and output ports on just a single part in the XM ONAG unit, the beam splitter was just big enough to pass almost the entire beam with only a very small part of the light blocked at the top and bottom of the full ray bundle. The input port would require a square opening with rounded corners and the output port would need only a slightly larger circular opening to pass the rays over the full field with nearly zero vignetting—and all of that could be done with no change to the XM ONAG physical size or camera spacings! Since this change would eliminate the dovetail mounts on the ONAG unit, mounting the unit to the scope and mounting the FLI filter wheel to the output port required a new way to connect everything together. Making the system user friendly and avoiding interferences between all of the existing bolt holes required a LOT of care and some pretty serious attention to detail but I eventually got it all worked out with everything firmly bolting to the main ONAG body.

The next detail was how to get all this stuff made? Fortunately, Gaston was totally on board to help me out (Thanks again Gaston…you are the best!) He even offered to make the parts but I didn’t want to take advantage of his good will. Still, I accepted his offer to re-assemble and align the parts once I had them made. (I suggested that this modification should enable ONAG on any Planewave system as well as on other faster systems using big sensors for wide field work so hopefully it will make sense for IFI to offer something like this as a product at some point.) I only needed four new parts and I had a lot of untried options to get them made. Optec was swamped so that possibility was off the table. I finally settled on an online manufacturing service called Xometry (pronounced “Zometry”) at www.Xometry.com. They offer both CNC and 3D printing services and it looked like a good fit for what I needed. The one thing that I had to do before submitting my drawings was to get good at 3D design—but how hard could it be? Fortunately, as an adjunct professor, I had access to AutoCad and Inventor so I set about expanding my 2D CAD skills into 3D. (BTW: For any of you who are EAA members, an educational copy of Solidworks is available for free as a membership benefit!) After about a week I had the models ready and Rev N shows the 3D model for the whole assembly.

One thing that I especially like about Xometry is that you can submit a 3D model along with a 2D drawing to show tolerances. You can call out materials, finishes, anodizing, and pretty much anything else that you would specify with a local machinist. You can also have the choice to have the parts made in the US or in China at a lower price with slightly longer delivery time. Those of you who might think that adapters from Precise Parts are pricy are going to be completely stunned at the cost of custom CNC machined parts. Just one of my black anodized parts with a 63 micro-in finish from a US supplier was being quoted at nearly $1,000. In my experience as a professional engineer, that’s a really high price, but using Xometry’s auto-quoting tool, I found that I could get the same part made in China at just a bit over $400. For comparison, a similar part through Precise Parts might cost “only” $250. Based on my work experience, the one-off price from a local shop might have split the difference so while the one-off ~$400 price was a bit high; it wasn’t completely crazy either. So, the next time anyone feels compelled to complain about the high price of a custom part from Precise Parts, go get it quoted at Xometry…or even a local machine shop before ranting too loudly about it online.

The quality of the parts, including the quality of the black anodize coatings that I received from Xometry’s Chinese shops have been superb. Unfortunately, one of my parts arrived damaged and the customer support folks were instantly on top of it. They quickly ordered a replacement part and apologized profusely—and that’s the ultimate measure of a company. It adds yet another delay to my project but they did the right thing and they did it with a smile on their face. So far, my experience with Xometry has been excellent and I give them my highest recommendation. Revision N shows the assembled parts that I’ve had made. I’m still waiting on the replacement part for the one that got damaged so stay tuned on how well this whole thing works.

I still have to paint the interior surfaces of the new parts and I’ve never been very impressed with most black paint. A good spray paint is Krylon Caouflage Ultra-Black (https://www.amazon.com/gp/product/B00176TH8C/ref=ppx_yo_dt_b_asin_title_o05_s00?ie=UTF8&psc=1). I’ve also looked at Stuart Semple’s Black 2.0 (which in my opinion isn’t very black) and Black 3.0 acrylic paint. The ultimate black coatings come in the form of engineered carbon nanotube materials. Vanta Black™ is one such material that is virtually unobtainable and Singularity Black™ made by Nano Labs is another. Both of these materials specify reflectivity values under 1%. Singularity Black is available, but it is a bit tricky to apply, it requires heat curing, it’s fragile, and it’s quite expensive. I’ve recently discovered Musou Black from a company in Japan that claims a similar reflectivity (under 1%) using a water-based pigment-based paint. It was developed for optical applications, it is much easier to apply and it is much less expensive than Singularity, though it’s only available from a single US distributor. The three biggest concerns are that it forms a very fragile coating that cannot be handled, there doesn’t seem to be any data on durability under temperature extremes, and it is not supposed to perform very well in the NIR. I have some on order (along with an air brush) so I’ll try to file a report on how it works once I have a chance to try it. With a super black coating, you cannot see the form of the coated structure—even under bright light and that’s the ultimate visual indicator of the blackness of the coating.

Getting back to the scope: When I took my first raw images, something else jumped out in the raw data. If you look at rev ‘O’, you’ll see that the off-axis star images show the primary diffraction pattern due to the secondary spider along with another set of diffraction spikes at 45 degrees. A quick look at the scope revealed why. I have a Planewave supplied spandex cover over the OTA truss structure. This kind of cover is valuable to control dust accumulation when a scope sits parked—essentially most of the time outside for years at a remote location. I found that about 8-10 inches behind the front support ring, the fabric cover gets stretched to form a straight line between the widely spaced truss supports. This forms a square aperture at 45 degrees to the spider supports and the cover is intercepting the off-axis ray bundles. That’s what causes those the extra field dependent diffraction patterns.

To address this problem, I made a 3D model of the OTA truss and designed a 3D printed support to hold the cover out of the light path. Four of these are needed around the truss and Rev ‘P’ shows how it’s wire-tied to the structure. BTW, my newly acquired 3D skills paid off here. Frankly, I can’t even imagine designing this kind of part in 2D! The stray diffraction patterns should be completely eliminated with this simple solution. The guys at PW liked this idea so much that they tell me that it might appear in their catalog one of these days.

The last thing that I’ll share here is a photo of the final version of my control box. I’ve added a system to mount filters to the glass door on the front. The image in Rev ‘Q’ shows the neutral density filter in place to limit light from the internal indicator panels. I also have an opaque black-out filter for use in the observatory. The folks at DSW go nuts over any stray light—no matter how minor and I suspect that that’s the case at any hosting facility.

Stay tuned for the next installment when I try the new configuration to see how it all works. All of this stuff it taking forever and I feel like I’m swimming through molasses to get this scope running, but I’m enjoying the journey. I just hope that I live long enough to get it working well enough to ship off to Chile ASAP! It might be slow going during Q1 since I’ll be mostly in Arizona, but stay tuned…

John

Addendum: Black is Black (with apologies to Los Bravos)

As I've said, a key goal of this part of the project is to reduce vignetting and eliminate (or at least greatly reduce) stray light. In general, the topic of reducing and measuring stray light is a big one so I simply want to add some specific results and a few more comments about reducing low angle of incidence strays. This isn’t simply an academic exercise. Stray light can have implications beyond basic image calibration. My priorities with my C14 didn’t include very careful stray light management and as a result there are certain interesting objects near bright stars that I simply cannot image with that system. There is simply too much stray light that makes its way to the sensor and cannot be removed from the data. In addition, I’ve seen a lot of discussions about this topic among ATMs and imagers so it may be valuable to leave a record of what I’ve done so that others can reference it.

My biggest concern with the redesigned ONAG unit and my adapters are the relatively large, flat surfaces surrounding the input port. In principle, those surfaces mostly lie within the baffle shadows but a specular return from any low angle stray could be significant due to the size of the surfaces. The best way to handle strays on a round surface is with a fine, sharp threads but that’s not possible on these flat surfaces. The second-best thing is to bead blast the surface, incorporate glare stops, and paint it with a good black coating. In my design, I have two low-angle glare stops but due to space constraints, they are pretty shallow—only 0.050”. I didn’t have the surfaces bead blasted so I resorted to bonding and trimming 220 grit sand paper to the critical surfaces.

Black paint is the last line of defense. The big question is what is the best black paint that can be obtained within a reasonable timeframe and at a “reasonable” cost that would work for coating metal parts with an acceptable level of durability? Obviously flat black paint from the hardware store is quick, easy, cheap, and durable; but how black is it? In the ideal world, coating a part with 100% absorbing paint would render it completely void of all apparent depth or structure. It would appear as simply a black hole outlined against the background. Even the best flat black spray paint never looks like that and the easiest way to compare “blackness” is to put different paint samples side by side.

Artists have been in search of the ideal black paint but most aren’t all that black. I found that in addition to the paints that I listed in the text above, Acktar offers a number of different high-performance light absorbing coatings that can be applied in different ways. Edmund Optics is the distributor and you can find Acktar in their online catalog. They are a bit less expensive than Singularity Black but that doesn’t mean they are cheap. An 8" x 8" self-adhesive blackened sheet that claims to absorb over 99% of incident light costs a little more than $100. Acktar provides a lot of data on their coatings, which are aimed primarily at professional applications.

I mentioned Musou Black paint above and I was able to mail-order a 100 ml bottle online for $35 plus shipping. 100 ml is supposed to cover up to one square meter (which based on my experience is a bit optimistic.) Before receiving the Musou Black, I had spray painted a strip of 220 sandpaper with Ace Hardware Ultra Flat Black spray paint and I simply brushed a little Musou Black on the end for comparison. As you can see in Rev R, the Musou Black is significantly more light absorbing than the flat black spray paint. So, I primed the bare metal interior surfaces on my parts with flat black spray paint and lightly brushed Musuo over the pre-primed surfaces on 2 of the 3 parts that needed to be coated. Visually, it is very black, but it's not a perfect "visual black hole" since you can still see some of the underlying part structure. For the last part, I used an airbrush to gently apply multiple thin layers and that took the "blackness" to the next level. The manufacturer claims that sprayed on, Musou will achieve a 99.4% absorption level. Even in moderately bright light, features on the part almost completely disappear into a visual black hole. Visually Musou Black indeed appears to be extremely black so the 99.4% number is believable; but not confirmed. Measuring that level of absorption to an accuracy of 0.1% is quite difficult but it would be valuable to get some high quality BRDF data from a good lab. There is a statement on a website indicating that Musou might not be very absorbing in the NIR, but that one random statement doesn't include any data to back it up. In my case, I'm transmitting NIR light to the guide camera so it won't get to the imaging camera and it won't matter either way.

Unfortunately, the Musou coating is also fairly fragile. Touching the painted surface will "gloss" the finish and rubbing the surface will knock small flecks of paint off. I haven't done very much testing but it seems that the coating adheres reasonably well and will stay put in places where it can't be disturbed. Obviously changes in temperature and humidity are a concern but we'll see how it works. Years ago, I used candle black to eliminate strays from stainless steel spatial filters with 10 micron pinholes. Since even a tiny bit of stray light reflected from the pinhole retro-reflected back into the laser would cause it to go unstable, the coating had to be VERY black. The candle black trick worked like a charm and in spite of the extremely fragile nature of the soot coating, many of those blackened pinholes are still working and in use thirty years later! So, I'm hoping that Musou Black paint will hold up as well...but time will tell.

I've added a few pictures here to show what the coated surfaces look like. The surfaces are so black that it's a bit difficult to get good photos. I tried to position the parts so that light was being specularly reflected from an overhead light source so these photos show the worst case. Without a specular reflection, the part topography is essentially invisible! I don’t know if Musou Black is the ideal black paint for optical applications but my initial impression is that it is VERY promising. It is certainly VERY black, it’s easy to apply, it didn’t cost a fortune and it was easy to get. The big open question is durability. I’ve committed to using it so we’ll see how it goes.