Jump to content

  •  

CNers have asked about a donation box for Cloudy Nights over the years, so here you go. Donation is not required by any means, so please enjoy your stay.

Photo

Comparison Between DSLR and ASI1600mm-Cool Images

  • Please log in to reply
93 replies to this topic

#51 mmalik

mmalik

    DSLR camera modifications

  • *****
  • Vendors
  • Posts: 9535
  • Joined: 13 Jan 2012
  • Loc: USA

Posted 23 September 2016 - 11:53 PM

As for form factor, well, to each his own...

 

That each is a very large community & industry; that's the whole point. ZWO may be a competitor to CCD but it is not to DSLR. Just because it has CMOS chip, doesn't make it a contender; it is still in CCD domain for all practical purposes. It is the form factor at the end of the day that decides the domain, and/or its application. Regards


Edited by mmalik, 23 September 2016 - 11:58 PM.


#52 Jon Rista

Jon Rista

    ISS

  • *****
  • Posts: 23348
  • Joined: 10 Jan 2014
  • Loc: Colorado

Posted 23 September 2016 - 11:55 PM

 

As for form factor, well, to each his own...

 

That each is a very large community & industry; that's the whole point. ZWO may be a competitor to CCD but it is not to DSLR. Just because it has CMOS chip, doesn't make it a contender; it is still in CCD domain for all practical purposes. It is the form factor at the end of the day which decides the domain, and/or its application. Regards

 

I find that to be rather arbitrary. :p 



#53 whwang

whwang

    Mercury-Atlas

  • *****
  • Posts: 2776
  • Joined: 20 Mar 2013

Posted 24 September 2016 - 12:12 AM

 

I don't want to get into most of the debates.  Just let me point out one thing.  When we talk about Bayer sensors vs mono sensors, let's not forget that there is a processing technique called Bayer drizzle.  It allows you to obtain image resolution equivalent to what can be achieved with mono sensors of the same pixel size. This function has been offered by DSS since day one.  I see little reason to ignore it and then complain about the resolution from Bayer sensors.

Drizzling can help, however it is highly unlikely to overcome all the blurring caused by a low pass filter and demosaicing. I went from an average of 5-6" FWHMs with my 5D III, to 2-2.3" FWHMs with my ASI1600, on the same lens. I've drizzled a majority of my 5D III data, and I've come nowhere close to that kind of a resolution improvement. I do see improvements, particularly when using 2x scale with a drop shrink of 0.7 in PixInsight (which is quite superior to DSS's drizzling capabilities), however I usually still don't even get below 4". 

 

And the resolution improvement is only one of the benefits. The bigger benefit is really the full fill factor, higher sensitivity, and ultra low noise. For imagers stuck with light pollution, the efficiency improvements in gathering photons, and gathering as many photons as possible is the best way to overcome LP, are very worthwhile. Throw in the narrow band options...I see very little reason anyone who cannot visit a dark site on a regular basis to keep using a DSLR (or for that matter any OSC camera.)

 

Anyway. I'll leave it at that. I think the images will speak for themselves for the most part. ;) People can choose what's best for them. 

 

Hi Jon,

 

It looks like you are talking about is the normal "drizzle."  What I was talking about is "Bayer drizzle."  The normal drizzle is meant to deal with mono images.  Some people also use it for Bayer array images, but this isn't really what drizzle is for.

 

"Bayer drizzle" is a very different technique.  It still contains the word "drizzle" because it share some of the drizzle concept.  I like this name, because it shows that whoever came up with this idea (Dave Coffin, the author of dcraw, I believe) really understands what's behind it.  However, I agree that this name can make lot of confusion.  (Bayer drizzle is called "color in motion" in certain version of Nebuosity, and this name is not less confusing either.)

 

In Bayer drizzle, interpolation of the Bayer array is totally avoided.  There is no interpolation at all.  IF Bayer drizzle is implemented correctly, the resultant image should have no difference from an image taken with a mono sensor.

 

I emphasized "IF" above, because I have some doubt on whether Bayer drizzle is correctly implemented in PI.  In most (if not all) of my tests with PI, its Bayer-drizzled image is not too much sharper than a normal VNG debayered image, and also a bit less sharp than DSS's Bayer-drizzled image.  Juan attributed this to the superior VNG implementation in PI, but I am not fully convinced.  (I believe PI's VNG is very good, but I don't believe it can be as good as a properly Bayer-drizzled image.)  He also failed to answer some of my concerns when he explained PI's Bayer drizzle implementation.  So I have some reservation toward PI's Bayer drizzle.  On the other hand, DSS's Bayer drizzle really works well in achieving high resolution.  You should give it a try.

 

The low-pass filter you mentioned is indeed a factor.  It can kill the resolution and it cannot be recovered in post-processing.  Fortunately, recent DSLRs tend to not have this filter.  Also, astro-modification often removes this filter at the same time.  So I do not consider it as a major threat any more.

 

Cheers,

Wei-Hao



#54 mmalik

mmalik

    DSLR camera modifications

  • *****
  • Vendors
  • Posts: 9535
  • Joined: 13 Jan 2012
  • Loc: USA

Posted 24 September 2016 - 12:22 AM

Wei, could you describe few basic steps how to actually do "Bayer drizzle" in PixInsight and at which point in the processing it needs to happen? Some basic conceptual and procedural details will be appreciated. Regards



#55 whwang

whwang

    Mercury-Atlas

  • *****
  • Posts: 2776
  • Joined: 20 Mar 2013

Posted 24 September 2016 - 12:26 AM

Wei-Hao, I have been using a Nikon for a while now, and still enjoying it, I am on the fence about getting one of the new cooled cameras, but I am wondering if I should just get my D5300 full spectrum modified.  Do you think it would be of much benefit?

Hi,

 

Is your D5300 modded at all?  If it had gone through an astro-mod, changing it to a full-spectrum mod will not help much (if any).  On the other hand, if it has never been modified, an astro-mod will significantly boost its capability on deep-sky images. On red objects, it should save you at least 2x of your exposure time.  Even on subjects not dominated by red nebulas, such as Milky Way and some large spiral galaxeis, the additional red on gas nebulas will give your images a very nice extra punch.  As long as your D5300 is primarily used for astrophoto, an astro-mod will be totally worthwhile.

 

That being said, there is still a good reason to consider a cooled camera: lower thermal noise. If your imaging environment is hot (>30 degC), then there is almost no doubt that you should go for cooling.  If it is just warm (>20 degC), then it is marginal.  If it is always below 15 degC (which is my case), then I do not believe cooling will help much.

 

My suggestion is to first modified your D5300, if it was not modified before.  It's the easiest/lowest-risk route to enhance your imaging capability.  And it is not that expansive, comparing to getting a new set of cooled mono camera plus filters.  If you find this still not enough after a few tries with the modified D5300, and if you find the major limit in your image is thermal noise, then you can consider a cooled camera more seriously.

 

Cheers,

Wei-Hao



#56 whwang

whwang

    Mercury-Atlas

  • *****
  • Posts: 2776
  • Joined: 20 Mar 2013

Posted 24 September 2016 - 12:35 AM

Wei, could you describe few basic steps how to actually do "Bayer drizzle" in PixInsight and at which point in the processing it needs to happen? Some basic conceptual and procedural details will be appreciated. Regards

 

Hi Mike,

 

I don't think I can make a clearer description than Juan's:

http://pixinsight.co...hp?topic=7184.0

I always read this when I forget how to do it.

 

Conceptually, in Bayer drizzle, one does not interpolate the Bayer array at all.  The consequence is that there will be empty holes in each of the RGB channels.  Using R as an example, in every four pixels, there will be only one pixel with real R data, and the other three pixels are empty.  This is of course not acceptable in any kinds of day-light photography.  However, in deep-sky photography, we take many exposures, and we often dither the exposures either intentionally (by using the dither function of the guiding and image acquisition softwares) or unintentionally (by having unavoidably tracking errors).  The dithered exposures can fill in the empty holes and eventually gives every pixel real R data without doing any interpolation.  This is the whole idea behind Bayer drizzle.

 

Cheers,

Wei-Hao



#57 Jon Rista

Jon Rista

    ISS

  • *****
  • Posts: 23348
  • Joined: 10 Jan 2014
  • Loc: Colorado

Posted 24 September 2016 - 12:56 AM

I am aware of what Bayer Drizzle is. PI supports it as well. Interpolation is avoided in bayer drizzle because it is still drizzling. It's just drizzling in a color-aware manner. I have tried DSS's drizzle features...however they seem to be very unstable. I was never able to drizzle full 5D III frames no matter what I did. To actually get it to work well, I usually had to choose an ROI less than 50% of the field, which was certainly not ideal. DSS is also 32-bit, and doesn't seem to implement LBA, so it would never use over 2GB of memory either, which limited the amount of information it would work with anyway. I don't know if DSS really has a superior implementation of bayer drizzle or not...given how basic DSS seems to be in general, I'm a bit skeptical. (Also, IIRC, DSS development has been abandoned, and it's definitely got some bugs that haven't been fixed in the last few years...and if development has been abandoned, it won't be updated to add compatibility for new DSLRs in years to come.) 

 

I don't think it is really safe to say that you wouldn't have any difference with bayer drizzle vs. a mono sensor. The problem is the signal information from a bayer sensor is sparse. It does not have a truly full constituent of information across all the color channels. It IS possible to recover a lot of the missing information with drizzling, however, with bayer drizzle, you are effectively actually drizzling three images, not from data with a 100% fill factor...but from data with 50% and 25% fill factors.

 

You really need a LOT of subs to effectively recover enough information for bayer drizzle to work well because you have to fill in so much information, and with that many longer subs, information will still spread out more, same as with normal stacking of many, many subs. Resolution is favored by stacking carefully culled subs that only contain the best information/smallest stars...which is really easy to do when your mono LRGB subs can be 10-15s at a high gain setting. 

 

Thing is, I am still undersampled with the ASI1600, at a 1.3"/px image scale. I can drizzle even that data fairly effectively...and improve resolution even more. FWHMs around 1.8-1.9". So even with drizzling, that can be employed with both types of cameras.

 

The pixels on the ASI1600 are about 1/3 the area of a 6D or 5D III, 1/10th the area of an A7s,  a little over 1/2 the area of a D810's pixels. I've never seen a 50% FWHM improvement from any kind of drizzling... I do get some decent improvements...but I find it is a lot more useful for rounding up stars:

 

K0RMepn.gif

 

The low pass filter is really the primary issue. I always knew it was an issue, however I wasn't quite aware of how significant until I used the ASI1600. Even on a night of bad seeing, my FWHMs are still around 2.3" with the ASI....a night of average seeing and I would easily be 5-6" on the 5D III. In part I'm sure some of that blur is from seeing over longer subs, but with nights of excellent seeing my FWHMs never go much below 4"...about the best individual sub I might have seen was over 3.8". 

 

If you really want a resolution benefit with a DSLR, I think one of the current generation APS-C cameras with an astro mod, to remove that darn low pass filter, is probably the best way to do it. However, you lose the large FoV, and an APS-C is marginally larger than 4/3...a 2x1 mosaic with a 4/3 sensor will produce a larger FoV.

 

Anyway. This is just the beginning of the mono cmos astro cam age. It won't be long before APS-C sized sensors, possibly even Sony Exmor Rs, find their way into one of these new cameras. We may even see a full frame Exmor sensor in one of them someday in the not so distant future. ZWO has a distinct mission to keep their equipment in the "affordable" range, which I think they have done a great job of so far...so I consider it all to be quite game changing. Outside of the rare individual who has easy access to a good dark site, I don't know if I can continue to recommend DSLRs to very many people...not even beginners...given how effective NB imaging can be with these modern CMOS cameras with very, very short exposures. Short enough to be done unguided on beginner-grade equipment. That is totally a game changer for city-bound beginners...with just the camera and a single Ha filter screwed directly into the 1.25" filter holder that comes with the camera, a beginner could get started with high quality data even in a white zone with an AVX and a fast refractor, or even a small 6" newt. That is unheard of, and would actually be easier to do, and the data would be easier to process, than fighting all the LP with a DSLR. I think that is really exciting. 


Edited by Jon Rista, 24 September 2016 - 01:10 AM.

  • scott4comp likes this

#58 whwang

whwang

    Mercury-Atlas

  • *****
  • Posts: 2776
  • Joined: 20 Mar 2013

Posted 24 September 2016 - 01:19 AM

Hi Jon,

 

A lot of the problems you speculated about Bayer drizzle is really an implementation issue, not an issue of Bayer drizzle itself.  Not just Bayer drizzle, for any processing technique, poor implementation can destroy everything.  So I wouldn't use poor implementation to criticize the technique itself.

 

Is your 5D3 modified?  My 5D2 is modified and its low-pass filter is removed.  I can see dramatic improvement in resolution brought by Bayer drizzle:

http://www.asiaa.sin...c/fig20-2.2.tif

 

On the other hand, I have no doubt that going to small pixels can bring improvement.  I saw this improvement after I switch from 5D2 to D800.  I believe there is still room for improvement if I go from D800 to even smaller pixels, even though my pixel size on D800 is already 0.9"/pix.  Unfortunately, in most cases, substantially smaller pixels mean smaller sensors.  It is almost always a battle between resolution and FoV.  I know that you argued FoV can be recovered by mosaicking, if the sensor is substantially more sensitive.  Unfortunately, I am not a believer to that a mono sensor can be so much more sensitive, so sensitive that the disadvantage in smaller FoV can be compensated totally, unless we are talking about narrow-band imaging exclusively.  

 

I myself has a mono CCD.  So I am quite aware about the limit of mono sensors.  CMOS is definitely better than CCD, so the ASI camera is making an important step forward.  But the advancement is not enough yet, at least not enough for those who really need the FoV.  If any one can make a mono CMOS camera (better with a Sony CMOS, not a Canon one) with FF size and 30M+ pixel counts, and if the price is less than D810A, I will jump the fence too.  I don't even need cooling.

 

Cheers,

Wei-Hao


Edited by whwang, 24 September 2016 - 01:26 AM.


#59 bsavoie

bsavoie

    Viking 1

  • *****
  • Posts: 613
  • Joined: 30 Nov 2013
  • Loc: North Alabama 35747

Posted 24 September 2016 - 04:26 AM

thank you both for having different views. I am a beginner, so I am not sure exactly what is being said. I do find it exciting to read both points and try to find my own opinion. I think to fully understand this thread, it will take me another year or two.. Thanks fellas.

 

Bill



#60 vdb

vdb

    Apollo

  • -----
  • Posts: 1445
  • Joined: 08 Dec 2009

Posted 24 September 2016 - 07:07 AM

Hmm I disagree on the statement that even a mono cold finger dslr cannot compete ASI 1600, I have Herra's D600CAM and  it easily follows the ASI 1600 and has twice the FOV, even my D810A can easily follow the ASI 1600, I dropped canon several years ago after my Canon 6D, never looked back ...



#61 bobzeq25

bobzeq25

    Hubble

  • *****
  • Posts: 16382
  • Joined: 27 Oct 2014

Posted 24 September 2016 - 07:29 AM

Wei-Hao, I have been using a Nikon for a while now, and still enjoying it, I am on the fence about getting one of the new cooled cameras, but I am wondering if I should just get my D5300 full spectrum modified.  Do you think it would be of much benefit?

Consider getting it H alpha modded instead.  Full spectrum isn't really full spectrum, you need a UV-IR cut filter in the optical train to get good stars, with H alpha - it's in there.  Makes life a lot simpler.

 

As Jon Rista will be happy to tell you <grin> the benefits of cooled, and mono plus filters go far beyond Ha sensitivity.  But I still use my Ha modded D5500 as much as my Atik 694.  Just a lazy peon, I guess.  <smile>  Getting SGP fully up and running may change that.

 

Unsolicited advice.  Have fun with an Ha modded D5300 until the next big thing happens in CMOS, which will make the 1600 old news.  Tis the season, Orion, the Rosette, the Horsehead...


Edited by bobzeq25, 24 September 2016 - 07:32 AM.


#62 entilza

entilza

    Soyuz

  • *****
  • Posts: 3716
  • Joined: 06 Oct 2014
  • Loc: Oakville, ON, Canada

Posted 24 September 2016 - 09:47 AM

 

Wei-Hao, I have been using a Nikon for a while now, and still enjoying it, I am on the fence about getting one of the new cooled cameras, but I am wondering if I should just get my D5300 full spectrum modified.  Do you think it would be of much benefit?

Hi,

 

Is your D5300 modded at all?  If it had gone through an astro-mod, changing it to a full-spectrum mod will not help much (if any).  On the other hand, if it has never been modified, an astro-mod will significantly boost its capability on deep-sky images. On red objects, it should save you at least 2x of your exposure time.  Even on subjects not dominated by red nebulas, such as Milky Way and some large spiral galaxeis, the additional red on gas nebulas will give your images a very nice extra punch.  As long as your D5300 is primarily used for astrophoto, an astro-mod will be totally worthwhile.

 

That being said, there is still a good reason to consider a cooled camera: lower thermal noise. If your imaging environment is hot (>30 degC), then there is almost no doubt that you should go for cooling.  If it is just warm (>20 degC), then it is marginal.  If it is always below 15 degC (which is my case), then I do not believe cooling will help much.

 

My suggestion is to first modified your D5300, if it was not modified before.  It's the easiest/lowest-risk route to enhance your imaging capability.  And it is not that expansive, comparing to getting a new set of cooled mono camera plus filters.  If you find this still not enough after a few tries with the modified D5300, and if you find the major limit in your image is thermal noise, then you can consider a cooled camera more seriously.

 

Cheers,

Wei-Hao

 

Thanks Wei-Hao, my camera is not modified.  So you are suggesting an H-alpha modded camera I was always confused about this, I thought full spektrum was that.  Would this modification throw off the colours so reflection nebula would be different? Or does this modificatin only benefit?  The reason is, I often see extremely red-images from some modified cameras missing some of the blues or natural colours of the nebulas, where even the stars look red... I've processed only a few of others modified camera-data where it seemed to be more of a processing issue than due to the modified camera.  I just don't want to harm whatever I am doing now, if it is only a benefit and a  plus then I am not sure why I am waiting to get this done as it may be exactly what I am looking for.

 

What I've found is on the bright objects (even elephant's trunk) the stock camera is quite capable (I have not uploaded this image yet).  It was going after faint nebula such as the Cave nebula (SH2-155) where the stock camera really had difficulty.



#63 555aaa

555aaa

    Vendor (Xerxes Scientific)

  • *****
  • Vendors
  • Posts: 1355
  • Joined: 09 Aug 2016
  • Loc: Lynnwood, WA, USA

Posted 24 September 2016 - 10:23 AM

An unmodified DSLR has a far higher SAF ( spousal approval factor) than the ASI1600.
  • Pax2You likes this

#64 vdb

vdb

    Apollo

  • -----
  • Posts: 1445
  • Joined: 08 Dec 2009

Posted 24 September 2016 - 10:25 AM

Mono Cold finger D600CAM (Herra's construction)

8x5min 3 panel sub iso 800, would have needed 6 and probably more 8 with ASI 1600 would it have been better, maybe ...

 

29600380290_08ff96999f_b.jpgPropellor-Crescent by Yves, on Flickr



#65 whwang

whwang

    Mercury-Atlas

  • *****
  • Posts: 2776
  • Joined: 20 Mar 2013

Posted 24 September 2016 - 10:42 AM

I am not taking side here.  Pictures shown so far are all great, no matter taken by a DSLR or a cooled CMOS.  What we learn from them is that all these cameras are quite capable, which is good of course.  What we do not learn is which one is better and how much better.  To answer this, it really requires carefully arranged side-by-side on-sky tests or lab tests. 



#66 Ron359

Ron359

    Apollo

  • *****
  • Posts: 1145
  • Joined: 21 Apr 2008
  • Loc: -105 +39

Posted 24 September 2016 - 10:56 AM

I am not taking side here.  Pictures shown so far are all great, no matter taken by a DSLR or a cooled CMOS.  What we learn from them is that all these cameras are quite capable, which is good of course.  What we do not learn is which one is better and how much better.  To answer this, it really requires carefully arranged side-by-side on-sky tests or lab tests. 

Completely agree.  What I find incredibly ironic in this whole discussion of 14 bit v. 12 bit, etc., is all the fine images shown in this thread are reduced both in resolution and bit depth to 8 bit jpegs viewed mostly on 8 bit screens.   


  • bobzeq25 likes this

#67 vdb

vdb

    Apollo

  • -----
  • Posts: 1445
  • Joined: 08 Dec 2009

Posted 24 September 2016 - 11:06 AM

While it is an Apples to oranges comparison, it still shows that DSLR's and especially Nikon have caught up with ccd/cmos dedicated camera's.

So yes to really see the difference in numbers, because I'm pretty sure onscreen you will see NO difference ... between ASI and Nikon,

which is not a real surprise because they use similar imaging sensor (yes I know not exactly the same ...)

 

Maybe I'll buy an ASI 1600 to do the test myself ...

 

/Yves



#68 bobzeq25

bobzeq25

    Hubble

  • *****
  • Posts: 16382
  • Joined: 27 Oct 2014

Posted 24 September 2016 - 11:30 AM

While it is an Apples to oranges comparison, it still shows that DSLR's and especially Nikon have caught up with ccd/cmos dedicated camera's.

So yes to really see the difference in numbers, because I'm pretty sure onscreen you will see NO difference ... between ASI and Nikon,

which is not a real surprise because they use similar imaging sensor (yes I know not exactly the same ...)

 

Maybe I'll buy an ASI 1600 to do the test myself ...

 

/Yves

One big deal remains.  With mono+filters, all the pixels are working all the time.  That gives an advantage in signal to noise ratio (equal imaging times) that's hard to beat.

 

I have and use, both an Atik 694 mono, and a Nikon D5500, Ha modded.



#69 Jon Rista

Jon Rista

    ISS

  • *****
  • Posts: 23348
  • Joined: 10 Jan 2014
  • Loc: Colorado

Posted 24 September 2016 - 11:53 AM

Mono Cold finger D600CAM (Herra's construction)

8x5min 3 panel sub iso 800, would have needed 6 and probably more 8 with ASI 1600 would it have been better, maybe ...

 

29600380290_08ff96999f_b.jpgPropellor-Crescent by Yves, on Flickr

The data here is very, very noisy at full size. A lot noisier than a single sub from the ASI1600. You could stack just 4 subs per panel from the ASI1600 and have better noise characteristics. Sure, you would definitely need more panels to cover the same region of sky...but on the IQ front, those panels could be 360 seconds rather than 2400 seconds long. You could then easily pack 12 panels in there, and have time to spare.

 

As Wei-Hao says, you really need to compare like images from the same optical system side-by-side to get a true idea of the differences. However, this is a single sub from the same general region of space, Cygnus, with the light spread out even more due to a longer FL, and even this single sub's noise characteristics appear superior to 8x300s subs from this mono-modded, cold-fingered D600:

 

68w2Uw5.jpg

 

As I said. A mono-mod is certainly going to get you closer in comparison...however, a mono mod rips off the microlenses as well as the color filters. So your quantum efficiency is going to take a hit. Between the Q.E. loss, and 3.8e- read noise vs. 1.3e- read noise (compare the squares though......14.44e- vs. 1.69e-!), and who knows what the dark current is of the cooled D600? 


Edited by Jon Rista, 24 September 2016 - 11:55 AM.


#70 whwang

whwang

    Mercury-Atlas

  • *****
  • Posts: 2776
  • Joined: 20 Mar 2013

Posted 24 September 2016 - 11:55 AM

 

One big deal remains.  With mono+filters, all the pixels are working all the time.  That gives an advantage in signal to noise ratio (equal imaging times) that's hard to beat.

 

Only true for narrow-band imaging.  For RGB imaging, all pixels on a Bayer array are indeed working, just on different colors.  On mono sensors, all pixels work on the same color at once, but you will need at least three exposures to get the color that a Bayer array can get in one exposure.  If all you need is a B/W image for a fixed wavelength, then a mono sensor is much more effective.  If what you need is an RGB color image, then the difference is much smaller, if both arrays are given the same amount of total exposure time.

 

We can extend this discussion to LRGB imaging, which is 2x to 3x more efficient than RGB imaging on continuum sources.  This gives a huge edge to mono sensors.  On the other hand, on emission line objects, L do not often give you much more signal than R or G or B (after you subtract the sky background, which itself is a source of noise).  The advantage of LRGB on emission line objects is smaller than what many people have believed, probably well within 2x.  Some people have realized this for a long while (by lot of practicing of CCD LRGB imaging), but most people don't.

 

In short, mono sensors do have a big advantage.  However, this is not because its array is fully populated as oppose to Bayer arrays (except for narrow-band imaging).  This is because of the possibility of adopting the LRGB technique.

 

Cheers,

Wei-Hao



#71 vdb

vdb

    Apollo

  • -----
  • Posts: 1445
  • Joined: 08 Dec 2009

Posted 24 September 2016 - 12:04 PM

Well the loss of micro lensing on paper seems a big deal in reality it is not, these cmos sensors are pretty efficient ...

The only thing modded mono camera's without an UV/IR cut filter have is that they are so sensitive they pick up any IR emitting from the electronics ...



#72 vdb

vdb

    Apollo

  • -----
  • Posts: 1445
  • Joined: 08 Dec 2009

Posted 24 September 2016 - 12:13 PM

For me an ASI with FF would be the real deal ... I like big wide mosaics and the 1600 just doesn't cut it ...

Would I prefer the ASI 1600 over any APS DSLR, of course! 

But I stand by my claim, a processed image will be hardly different when you take a mono DSLR cooled or ASI.

 

Noise evaluation, it's Herra's;

http://www.cloudynig...eristics/page-3



#73 Jon Rista

Jon Rista

    ISS

  • *****
  • Posts: 23348
  • Joined: 10 Jan 2014
  • Loc: Colorado

Posted 24 September 2016 - 12:13 PM

Hi Jon,
 
A lot of the problems you speculated about Bayer drizzle is really an implementation issue, not an issue of Bayer drizzle itself.  Not just Bayer drizzle, for any processing technique, poor implementation can destroy everything.  So I wouldn't use poor implementation to criticize the technique itself.
 
Is your 5D3 modified?  My 5D2 is modified and its low-pass filter is removed.  I can see dramatic improvement in resolution brought by Bayer drizzle:
http://www.asiaa.sin...c/fig20-2.2.tif


I'm not sure it's fair to say the PI implementation is poor. Different than DSS? Sure. The real question would be, is the DSS implementation sharing any luminance information across color channels. If it is, then it would still effectively be interpolatng, but at the same time, the sharing of luminance information would result in better results. The PI implementation, IIRC, effectively generates three separate drizzles, and does not cross-share any luminance information across channels. That means you have to fight a 75% fill factor deficiency in the red and blue, 50% deficiency in green, in order to properly fill in the higher resolution image with bayer drizzle.

On the other hand, I have no doubt that going to small pixels can bring improvement.  I saw this improvement after I switch from 5D2 to D800.  I believe there is still room for improvement if I go from D800 to even smaller pixels, even though my pixel size on D800 is already 0.9"/pix.  Unfortunately, in most cases, substantially smaller pixels mean smaller sensors.  It is almost always a battle between resolution and FoV.  I know that you argued FoV can be recovered by mosaicking, if the sensor is substantially more sensitive.  Unfortunately, I am not a believer to that a mono sensor can be so much more sensitive, so sensitive that the disadvantage in smaller FoV can be compensated totally, unless we are talking about narrow-band imaging exclusively.  
 
I myself has a mono CCD.  So I am quite aware about the limit of mono sensors.  CMOS is definitely better than CCD, so the ASI camera is making an important step forward.  But the advancement is not enough yet, at least not enough for those who really need the FoV.  If any one can make a mono CMOS camera (better with a Sony CMOS, not a Canon one) with FF size and 30M+ pixel counts, and if the price is less than D810A, I will jump the fence too.  I don't even need cooling.


IIRC, you have the KAF-16803? I don't think you can really use that as a gauge of the performance of the ASI1600. I've worked a number of large data sets from both KAI-11002 and KAF-16803 cameras. At a pixel level, both are still pretty noisy. Particularly the KAI-11002...the KAF-16803 is better, but every one of those cameras seems to have a lot of defects, lots of hot pixels, and are just noisy in the fainter details. Very noisy. You really need to work some ASI1600 data to understand how clean it is. I'd be happy to share some, if you are interested.

Regarding FoV. I do fully understand if someone is uninterested in mosaicing to get a larger FoV. Mosaicing does involve more processing effort, no question there. For some people, that just isn't what they want, and I definitely understand that.

It is possible to use a shorter scope with a smaller sensor to achieve the same FoV, though. And when you have smaller pixels on the smaller sensor, you are not necessarily losing much on the image scale front. So it is still possible to get a wide FoV with a small sensor, and it may just require adding a reducer. That said, on the flip side, I also understand that a larger sensor used on a larger scope with a larger aperture can acquire more signal in a given amount of time for the same FoV. Assuming that signal is not counteracted by high noise, then you could end up with a higher SNR as well. There are certainly benefits to big sensors, and I'll be first in line when a mono Sony FF Exmor is put into one of these cmos astro cameras. However, in the mean time, as someone who likes big fields myself, I have little doubt that I can create images of much higher SNR, and the same FoV, by mosaicing 2x2 with the Asi1600...and, I have little doubt I could create those mosaics in less time than it would take me to create just one reasonable panel with the 5D III, astro modded or otherwise, from my red zone back yard. The narrow band option is really the big one here. I would never mono-mod any of my DSLRs...I use them for other things, and mono-modding isn't an option. Not to mention I'm not really willing to take the risk of damaging the sensor in the process.

#74 Jon Rista

Jon Rista

    ISS

  • *****
  • Posts: 23348
  • Joined: 10 Jan 2014
  • Loc: Colorado

Posted 24 September 2016 - 12:16 PM

Well the loss of micro lensing on paper seems a big deal in reality it is not, these cmos sensors are pretty efficient ...
The only thing modded mono camera's without an UV/IR cut filter have is that they are so sensitive they pick up any IR emitting from the electronics ...


Without the microlenses, they are actually not that efficient. Without microlenses, Q.E. can drop to 30% or less. It was the addition of the microlenses in the first place that improve their Q.E. about a decade ago...and the move to gapless microlenses improved it more...and now in recent years, the move to offset microlenses at the periphery of large sensors that has reduced vignetting caused by the angle of incidence of light, producing more normalized Q.E. across the frame. Removal of the microlenses absolutely has an impact. I'll chop up some 100% crops for comparison...there is still no contest here. ;)

 

EDIT:

 

Comparison of 100% crops...2400s 8x300s stacked from the D600 cold finger mono mod:

 

SCdnkh0.jpg

 

Vs. a single 90s sub from the ASI1600, stretched to a similar intensity:

 

wjmt4k8.jpg

 

A difference factor of 26:1 here, just going by exposure time. Even if  you figure there may be a stop worth of f-ratio difference here, this is still over a factor of 10:1. Just for reference, here is a 2250s integration of the same crop from the ASI1600:

 

dGiRb9L.jpg

 

I do indeed believe the loss in Q.E. matters. The ASI1600 is mono and still has it's microlenses...I believe that is one of the advantages of mono cmos astro cams over mono-modded DSLRs. 


Edited by Jon Rista, 24 September 2016 - 12:33 PM.


#75 JukkaP

JukkaP

    Viking 1

  • -----
  • Posts: 559
  • Joined: 14 Aug 2016

Posted 24 September 2016 - 12:27 PM

The convercation is not informatic enymore. Please, stop. You are overtalking each other and confusing everybody.

There could be some conclusion to achive in here if people would think this trough.
  • Ron359 likes this


CNers have asked about a donation box for Cloudy Nights over the years, so here you go. Donation is not required by any means, so please enjoy your stay.


Recent Topics






Cloudy Nights LLC
Cloudy Nights Sponsor: Astronomics