View Full Version here: : Commercial Dragonfly array

24-08-2017, 10:27 AM
I've been wondering when someone would commercialise the Dragonfly array design:
https://www.aliexpress.com/item/Astroshop-telescope-ARRAY-TA-80/32816662435.html?spm=2114.12010108. 1000013.5.7d847a7TmW4uS&traffic_analysisId=recommend_2088_3 _81019_new2&scm=1007.13339.81019.0&pvid=fbc6cc27-b80d-469d-8308-751e314bf36f&tpp=1


24-08-2017, 10:44 AM
Thanks for sharing Jason. It looks like a neat package. I wonder to what extent adjustment is possible (shims perhaps) to align all 4 telescopes?

On another level the cynic in me might say .... just like the days of Mono > Stereo > Quadrophonic > Surround. It's great and sells more speakers (read telescopes, although at some considerable performance benefit)

Full disclosure - I'm collecting the bits / working on one myself


24-08-2017, 11:38 AM
Hi JA,
I agree alignment would be an issue, as with guidescopes some means of XY movement is needed. I'm not sure if these array cradles have the capability. Better pictures of the cradle can be found here:
https://www.aliexpress.com/item/ASTROSHOP-telescope-ARRAY-TA-70/32823628159.html?spm=2114.10010108. 1000013.7.4b24ae2bKrbj2I&traffic_analysisId=recommend_2088_4 _82199_new&scm=1007.13339.82199.0&pvid=d5f8d202-8578-4b92-9370-056a7485a9ac&tpp=1

Beat Kohler at AOKSwiss developed an alignable cradle for making binoscopes from refractors, so it's doable.

24-08-2017, 11:49 AM
Thanks AA,

Yes the better images make it clear that each of the scopes has perhaps 2mm clearance allround in the mounting hole and the surrounding screws (hopefully nylon/nylon tipped) are probably used to clamp and adjust the alignment.

I would prefer (although more complicated/costly) something that grabbed the tube more rigidly and then could be adjusted by an external screwdrive arrangement in the horiz & vertical, which could then be locked off.


24-08-2017, 01:39 PM
Built in dithering?

24-08-2017, 02:54 PM
Looks interesting for sure.

However, when I think about it, instead of four tiny f/5.4 refractors with possibly flimsy focusers, I would rather have one solid 10" f/3 Newtonian.

24-08-2017, 03:11 PM
4 refractors (with CF tubes!! POOR choice!), non-motorised focusers...meaning you'd need 4 robotic focus arrangements (circa $500 each). Then a USB hub capable of synchronous data transfer from 4 CCD's (unless you staggered the timing for each download).

Software may be an issue too.

Ah stoof it, just buy 12" RC.

24-08-2017, 04:07 PM
Yeah, I think this Aliexpress array might be trying to cash in on the sex appeal of the Dragonfly without much thought for performance. Still it would be good to see a genuine commercial product one day...
BTW, can you believe U Toronto want to extend the Dragonfly from the current 48 to 480 lenses! :eyepop:

That's a huge amount of dew straps! ;)

25-08-2017, 08:58 AM
Lee Spitler at Macquarie U is testing performance of a lens array v scope (see pic).

25-08-2017, 09:34 AM
We had a discussion on this forum earlier about dragonfly concept (using Canon 400mm f/2.8 lenses).
I don't think there is any benefit (apart from saving imaging time perhaps) that would justify the cost of such instrument.

I fully agree with Slawomir's comment below :thumbsup:

26-08-2017, 09:59 AM

Suggest you read their webiste and some of their papers and see the unbelieveable scientific advances and savings they have made.
Its a complete paradigm shift in discovery.

The single major scientific benefit is that whilst almost all other Earth telescopes are limited to 28 magnitude imaging, the DragonFly arrays are limited to 32 magnitude imaging
The other benefit is that for less than $1M they can have the benefits of a much larger telescope costing between $10m and $100m, but given they go another 4 magnitudes deeper - they can do what the others simply cannot do !
How much is that worth in science ?
The world is spending Billions on new projects and these guys are doing more for a miniscule fraction of the cost - under $1M !!!

Their huge improvement in observation means they are now seeing stuff that nobody has ever seen before and in doing so they they have upturned a huge number of conventional paradigms and continue to provide discoveries such as Galaxies that (on conventional theory) are comprised of 98% dark matter etc etc
Galaxies that shouldnt exist, new types of Galaxies - the list is endless

Literally hundreds of scientific papers have come from this small teams efforts and its opened the doors to entirely new science and new understanding of cosmology - and raising significant questions about most of the current theory of galactic evolution and formation.

Its worth the time to read.


26-08-2017, 01:38 PM
I still don't understand how this think work... and what is the difference between multiple astrographs used simultaneously on single object and single astrograph taking multiple subframes - except saving time, of course.
The explanation on the website is very unclear to me...

"Dragonfly is designed to reveal the faint structure by greatly reducing scattered light and internal reflections within its optics. It achieves this using commercially available Canon 400mm lenses with unprecedented nano-fabricated coatings with sub-wavelength structure on optical glasses.
Also, Dragonfly images a galaxy through multiple lenses simultaneously—akin to a dragonfly’s compound eye—enabling further removal of unwanted light. The result is an image in which extremely faint galaxy structure is visible.''

I have (only one) and I use Canon FD 400mm f/2.8 L for AP, optically it is the same as later EF model, and while it is exceptionally good, it is not THAT good...

26-08-2017, 04:43 PM

Then you need to read further and deeper - its very clearly explained.

I think your assumption that your lens is the same as their lenses may be incorrect.
Or maybe it isnt but you havent been aware of the technology in your lens - Im not sure exactly which lens design you have.
But is is also about the difference between the limiting magnitude resolving capacity of a refractor and reflector.

It has much to do with the almost complete loss of internal reflection due to the technology and process used to create the lens surface.

The concept of using multiple lenses is as you say, used to increase light and decrease integration times, its also about simultaneously capturing all bands (filters) concurrently.
So that is no different to normal - but the significant ability of their optics to see 4 orders of magnitude further than the previously accepted limit is what sets them apart.
Also by having multiple systems, they are in effect getting the benefit of all the tiny offsets in positions thereby reducing data collection noise in the one sitting as opposed to multiple sets of integration times.
Inbuilt static drizzling and dithering if you wish !

You might be interested to start reading some of their many papers (approaching 300) to see that their claims are all well documented, measured and proved.
Its all peer reviewed.

Once you have read a few - you will appreciate just what a paradigm shift their approach is and how it could conceivably revolutionise this entire field of astronomical imaging.
Their tiny little system can see what all the worlds biggest observatories cannot see for a mere fraction of the cost, compexity and logistics !
If they had enough funding and Im guessing that will come - they could duplicate their system 50 times over in the space of weeks and months rather than decades and billions$

In fact its the perfect example of a major scientific advance that well funded amateurs could jump on in the space of months and contribute to the science !
This is such a brilliant opportunity for amateur funded discovery.
They simply cannot look at the entire sky in the space of months or years - amateurs can beat then to it with "modest" 10 lens systems !



26-08-2017, 04:58 PM
Depending on where you read the site it either sounds like they're using the off the shelf lens' or ones that have been designed for this project specifically.

The usefulness of arrays largely depends on what your task is. The Dragonfly is low resolution faint material where you have a large sensor and reasonably short focal length. 48x400mm F/2.8 is far easier to make than a single 400mm F/0.404 lens.

Officina Stellare make 800mm F/3.8 astrographs. Building 100 of these, dual mounted on ASA DDM160s in a large shed may cost about $100 million Aussie. Would this be cheaper than a single 8m F/0.38? Maybe, maybe not. Designing such an instrument would be a nightmare beyond all comprehension.

Yes, it maybe limited to Mag 28, but it would hit it pretty damned quick.

26-08-2017, 05:35 PM

The world has lots of fast telescopes, but they still only see what we've been seeing for a while - but they still dont have the optical ability to detect anything better than Mag28.
Dragonfly detects down to Mag32 and Im pretty sure they hinted at better in the future in one of their papers.

So thats a 4000% increase in detection improvememnt - and not at some incredible cost of the technology but maybe as low as 1/1000th of the cost ! - with off the shelf components that any of us could buy and assemble if we had the money.

Its the reduced scattering and reduced reflections of what they refer to as "nano fabricated coating" - not sure how different this is to other lens coatings but it certainly seems to work for them.
As far as I know this is just an off-the-shelf Canon lens - but of their latest design.

Multiple lenses doesnt really count for anything more than more light gathering and some further noise reduction - so they can detect more in one session.

Im suprised they are getting the results with just a Kodak 8300 series sensor !
Its hardly a sensor known for its low noise or large well depth !!!

They are talking about using some small pixel cameras in order to get some finer resolution for more detail, but Im guessing they are coaxing a fair bit out of their drizzling and dithering anyway.

I guess when Canon offer them some 800mm lenses at a discounted price - (which Im sure they would for a 480 unit sale !!) - we will see some more interesting discoveries !!

Its pretty exciting stuff to see such a huge jump in sensing with an even greater order of reduction in savings and technology.

Ive been reading more of their papers and what they are finding is literally turning existing theories on their heads - that often can lead to new theories and understanding.

They have done with pennies what hasnt been done with $billions

26-08-2017, 06:55 PM
Heres one that gives some insight

27-08-2017, 10:06 AM
Interesting discussion.

I think that KAF 8300 is not as bad as we may think, and I believe in some cases it may surpass some more modern sensors, namely Sony's newer CCDs. I may be wrong, but on the same telescope KAF 8300 could give a higher SNR than ICX 814 , purely because of its pixels having surface area over twice as big as ICX 814 (29.2 micron squared vs 13.6 micron squared), thus in spite of having somehow lower QE and a higher RN, significantly larger pixels should make up for it on the same telescope, and there is no doubt that KAF 8300 gives a greater real estate, at a similar price tag.

28-08-2017, 06:17 AM
I am looking into paper...

The math mentioned in the paper is suggesting that for the extended objects F-ratio is more important than aperture (we all know that).
Also, according to their math (and our experience) the spatial resolution of the earth based telescope is limited by atmosphere turbulence, so 25cm aperture is the same as 2 metres (if there is no adaptive optics involved).

I am still trying to complete the whole thing in my mind, but my question still remains: what is the difference (apart from saving time) between stacking subframes taken by array of telescopes and stacking frames taken sequentially by a single telescope?

We all know that SNR is the key factor for detecting the low surface brightness objects... and because of LP and sky glow etc. SNR can be improved by two ways: imager dynamical resolution depth (number of bits per pixel) and stacking a as large number of subframes as it is possible (stacking is essentially averaging process). Then, aggressive stretching can bring out the extended objects which brightness is lower than sky glow (so their contribution is small). Currently, their achieved dynamical depth (detectability of extended objects) is 28 m/arc.sec.
Now, stacking of frames taken by other telescopes (from array) can average out differences between them (and between sensors as well), thus increasing the SNR (in the next couple of days I will try do do some math of my own to understand this in more detail).

Now about Canon lenses, 400mm f/2.8 in particular...
This is a very big lens, it is not cheap (2-5k$ second hand), very rare these days... and it is very heavy - my specimen (FD, from '80s) weights 5.5kg.
EF model is a bit lighter, ~3,5kg.
(some more details here (http://www.mir.com.my/rb/photography/companies/canon/fdresources/fdlenses/400mm.htm)). Marcus Keniath (http://www.4photos.de/camera-diy/Canon-NewFD_400_2.8-EF-Bajonett-Konvertierung.html) told me only ~2500 specimens were manufactured.
So, using this lens (or EF model) is somewhat strange (expensive!) choice for Dragonfly.. unless Canon decided to give them what was left from the sales of those models or what was returned for refurbishment...

BTW, to reduce internal reflections, I had to remove the two front elements from my specimen - they were just flats (coated of course) and used for mechanical protection of the inner ED element.

28-08-2017, 07:53 AM
They specifically chose the EF 400mm f/2.8L IS II USM for its optical quality - not just any old 400m f/2.8 lens. The noise they are trying to minimise is optical noise from scattering and internal reflections and apparently this lens excels in this regard. BTW it weighs "only" 3850g.
By combining in a large array they create an effective f-ratio much lower than a single lens but without introducing the optical problems of a single large lens of the same f-ratio: total aperture divided by focal length. I didn't see it said explicitly but assumed that to obtain the effective f-ratio the cameras must all be synchronised so that stacking of the simultaneous images would work to reduce the optical noise.

28-08-2017, 08:11 AM
FD and EF 400mm F/2.8 are optically very similar (same optical formula.. EF model is newer (however still obsolete) so possibly better...)

I still think (until I am proven wrong... and so far nowhere in their math I was able to see the evidence of it) that the "only" benefit of Dragonfly concept is the reduction in time used for observation (and averaging across the number of optics-camera systems).

BTW.. lower F/number reduces the exposure time for extended objects... but once the noise floor is hit (and that is skyglow and/or LP) you don't have anywhere to go from there - the only way further higher dynamic resolution and noise averaging (both camera noise and optical noise).

28-08-2017, 08:14 AM
If stray light is something so concerning why dont they have baffled extention tubes, like the Hubble approach (and mine) to reduce stray light reaching the lens.
I still have no comparrison shots, with and without, because I am busy still setting up.
But in my case at least the unit must help stop stray light from the street light nearby.
If Hubble uses this approach is that not a decent pointer to there being something to the idea.

28-08-2017, 08:15 AM
Hi Bojan,

What they claim is that the almost total lack of reflections and scattering that you would otherwise get on every other optical surface of a reflector or ordinary refractor telescope in existence gives rise to a theoretical limit on the resolving capability of 28 magnitudes.

That is the SNR gain they are taking advantage of.
Their optics dont have this problem so they can go down to 32 magnitudes.

That is to say that the optical noise produced by the minute amount of reflection and scattering at the wavelength level of imperfection on ordinary glass surfaces is enough to limit their ability at 28 magnitudes
Theirs doesnt have this thanks to Canons process.

Thats their SNR magic bullet !

The result of Canon's "nano fabrication coating" is that the surface of the optical elements is to all intents and purposes perfectly "smooth" because it is well below the length/height of any of the wavelengths of light they are interested in and thus in effect invisible to the optical system.
As a result of this they have reduced the inherent optical noise so that they can resolve down to 32 magnitudes.

All the rest of the stacking just helps with ordinary sensor and data collection noise reduction and improved integration time - more light for in same time. Its nothing special at all - but just as necessary as normal to imporve SNR, remove sensor and processing noise etc etc

As soon as others are capable of coating their surfaces in the same manner, I guess their technical advantage disappears and Im sure that might happen, but there may well be some manufcaturing and technical limitations (and patents) that make this difficult or impossible for larger elements such as big refractors and mirrors.

In answer to your question "what is the difference (apart from saving time) between stacking subframes taken by array of telescopes and stacking frames taken sequentially by a single telescope? "
There is no difference whatsoever ! - you seem to miss the point - the gain they claim in their papers (and their published results appear to support it) is their ability to avoid the increased noise of surface imperfections that exist on other optics.

I do not believe they are using 1980's technology, my understanding is that its latest technology, some published articles list the actual model number of the lens, current retail pricing is about $10kUS and that corresponds with their costed model.

Hope that makes sense.

For an amateur maybe $2k or $10k for a lens is a lot but for a research project that is finding and discovering what they are uncovering - its not really a big investment, just a matter of them getting appropriate funding to go further and based on their findings - I would be surprised if they didnt get funding for their 480 lens system.

They did all this on a baby Paramount with photographic lenses not a Magellan or a Hubble !

I just started at their website and a few press releases and then started reading the links that progressively went further into the actual research papers. Its all discussed openly and details presented.

28-08-2017, 08:34 AM
Here is service manual for EF lens (https://www.google.com.au/url?sa=t&rct=j&q=&esrc=s&source=web&cd=1&cad=rja&uact=8&ved=0ahUKEwjUloSYx_jVAhXCJpQKHSGZDQ cQFggpMAA&url=http%3A%2F%2Ff20c.com%2Fstuff%2 Fcanon%2Fpartslist%2FEF%2520400%252 02.8%2520L%2520USM.PDF&usg=AFQjCNFvtIeoh61PaZ0I_4wSxotR033 oJA) , optical formula (number of elements, their shape etc.) is the same as (my) '80-es model. Of course there are improvements, most likely in details, like coatings. And focus motors etc.
I haven't tried anything with this (newer) model, but if it so good as they claim, it would have been known already in our community... and that is not the case (OK - the price may be the factor here... but because some of us are more than ready to open their valets for high quality, the whole story somehow doesn't fit).

Again, I appreciate the saving in time needed for single observation.
Everything else (to me) is .. well.. .a bit overblown.

BTW, the magnitude limit is 28 magnitudes per square arc sec. And they are talking about 32... this is 4 magnitudes more, 0r ~32x lower intensity difference. This means dynamical resolution must be 5-bit higher....

28-08-2017, 10:22 AM
Hi Bojan,

Its not the same technology today as it was nearly 40 years ago !
I dont need to read a service manual to know this.

They would have better glass, better rpoduction techniques, better quality control, better grinding equipment and better testing equipment and certainly better coating technologies - since that is the specific area that they attribute their claims.
Canon has many patents on this - Google them to read about it.
But one striking difference between this nano coating is that its a single coating - not a series of multi layer coatings like those used in normal optics. One coatiing does it all.

Simplified - Canons Nano Coating technology produces a crystalline surface coating that has a variable comparative density. The base is a high density, the surface is lighter.
Its done with holes that are approximately 1/4 the wavelength of the light of interest.
The coating starts out life as a fluid spin coating process of 5-10 nanometer sized Al2O3 which is dried and then is reaquafied and dried out again, during which stage it recrystallises and produces the variable density coating with the right sized and ratio of hole sizes.
The fluid and process they use is their black art !

Their Patents were only lodged in the last few years and this technology did not exist in 1980.

Old film lenses are well known and understood for their lack of antireflective coatings - its generally why a cheap kit lens of today performs better than most of the best glass from the film era (with some rare exceptions)

Film didnt reflect light to the same extent that CCD sensors do and hence why the old lenses didnt perform very well with CCDs
Standard film grain structure was also a limit they built their lenses around whereas CCDs can resolve down to smaller pixels/grains needing better resolving lenses and Im pretty sure that needs to be across a wider spectral bandwidth than film was sensitive to.

As to why one person didnt discover what another person discovered - I cannot tell you - the world is full of discovery - someone has to be first !

Its a question of recognising the merit in what you have discovered and then after all the unexpected discoveries they realise just what the ramifications are of their initial discovery.
They certainly didnt expect it and so I doubt anyone else could have.

You can't do that if you are stuck in the mud of old fashioned paradigms . . . and that is what they did - looked outside the box by going back to basics and using a ready made solution.

Up till now, Im guessing most amateurs werent specifically interested in this field of low surface brightness objects.
The fact that Dragonfly has now discovered so much that is new and a lot of what they have found simply defies the current theories and explanations of galaxy formation etc will no doubt stimulate some interest in the amateur fraternity.

Pretty sure 4 magnitudes is a factor of 39.8 ?
I rounded it up to 40x earlier

Im not sure they are necessarily claiming they have 5 bits of higher dynamic range - they are just able to see the extra 'bits' below 28mag that everyone else has lost, buried in the noise of their optics !!
They can get their dynamic range expansion same as we do by stacking different length subexposures.

Anyway - there is lots to read about what they claim and over 200 approaching 300 papers out there that document their discoveries arising from their observations in such a short space of time from grants only provided in 2013, 2014 and 2016 - so its an amazing amount of discovery in such a short time.
No matter how they have achieved it.

28-08-2017, 11:00 AM
First there was McGregor v Mayweather, now there is Bojan v Rally!
I'm finding the latter fight very informative and more interesting to follow...
Next Round! :rofl:

28-08-2017, 01:06 PM
No, it's not Bojan vs Rally..
It is just clearing the understanding of basics, as I want to know what is this about this dragon fly and reasons for going for it.
Personally, I am not someone who would take other people's things for granted, that is why I am posting the questions, after all this is what makes difference between science and "science".. Science correct itself through checking and re-checking, the other is not.

As to optical design differences between FD and EF 400mm f/2.8L lens, I only stated the optical system is the same - details like coatings etc may be different of course. Is this is THE detail that makes all the difference, that remains to be seen. Alex's baffling comes to mind..
And yes, Rally is right, 2.512^4 = 39.8, my calculation was very coarse.

I am reading further.. but it will take some time for the next round :)

28-08-2017, 01:26 PM
Goodness - people reading the posts because they think its a fight !!

Bojan and I have communicated on and off forum many times about lots of things - he wants the info to back it up and I'm happy to provide what I can and why - afterall Im the one making the claims !! - hardly adversarial.
I am sure he is enjoying the process as much as me, forces us both to question our beliefs.

But yes, I do agree its a really interesting development that seems to have floated under the radar for a while and the reason why I keep adding to it.
I reckon we should understand more about it and how they did it - because I reckon there will be some amateurs who will potentially be interested to go down the same route albeit more cheaply just so they can go 40x deeper than before !
A 3 lens system isnt necessarily out of the question and Im sure there will be other nano coated lenses coming out soon - or maybe Takahashi will borrow some technology from Canon as I believe they already do now ?

Most of everything I have referred to can be found on their website, but you need to follow to a few external links (and chains) and read a bit to get the full story.

I think the Dragonfly site just assumes everyone doesnt need to have all the relevant details and that if its in a paper or 200 then its all there.
Canons patents are about as descriptive as a Kodak CCD spec sheet really ! - Im sure its done that way deliberately - but one of the big-name photographic journalists got a factory tour and an interview at the Canon lens factory where they make these and he got a good scoop on it - so that was more interesting and enlightening to read !

If I can refind that link after so long I'll post it up too.

28-08-2017, 02:10 PM
They claim they are seeing intensity 4 mag below some previously established threshold.. and that threshold (probably camera noise, skyglow at dark site plus scatter in optics etc) can be reached very quickly with f/2.8 lens (couple of minutes exposure at ISO3200 will do.).
I made a quick calculation (adding magnitudes, 0 and 4... ) and mag difference (contribution) between old threshold (28m /arcsec sq) and new one is -0.027 mag, or ~2.45% brighter in linear terms.

If I am not mistaken, 'normal' camera with 14-bit resolution should be able to see 0.012% intensity difference. This is valid in the linear range of the sensor, and no camera noise accounted for.
So. yes... no need for 5 bits more, but averaging (stacking) is definitely a key here.

28-08-2017, 07:11 PM

Cant find the patent !
But here's the journalists Canon interview