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

Gain and Sub-exposure calculation spreadsheet for the ZWO ASI533

  • Please log in to reply
56 replies to this topic

#1 StevenBellavia

StevenBellavia

    Viking 1

  • *****
  • topic starter
  • Posts: 749
  • Joined: 11 Mar 2014
  • Loc: New York

Posted 04 December 2019 - 03:41 PM

Hi,

 

I made an attempt at a Gain and Sub-exposure calculation spreadsheet for the ZWO  ASI533.  It is on the shared Google drive with the others:

 

https://drive.google...U9CKj5GilnZW8VJ

 

I am missing data for the recommended offsets for the various gain settings.  For now I made it about the same as the 294. If someone could provide the pre-determined offsets from the ASCOM driver, or actual testing,  that would be great.

 

Otherwise I believe it to be complete and free of errors, but as always, please check it and report back to me. Once I get better offsets, I will make a revision, but I don't think there will be a significant difference in the results.

 

  • Use only the main (first) worksheet, and enter data in the blue boxes.
  • Like the other spreadsheets, it is just a guide to help choose the best gain and sub-exposure time, based on achieving a desired Swamp Factor, which most agree 10 is good for LUM and RGB.  A swamp of 10 may not be achievable for narrow-band at dark sites. Other posts have discussed this topic.
  • And, as all the others, the "Transmission Efficiency" is a loose parameter and depends on your optical system, the QE of the camera, etc. I have 0.7 as the default. Let me know what value works for you.
  • It can also be used as a crude Sky Darkness meter, if you use actual image data. If you have an SQM, you can use it to tweak the Transmission Efficiency.

 

I do not own this camera yet, but the specifications look very good, so I hope to purchase one in the future.

 

Steve


  • mtc, elmiko, RTJoe and 5 others like this

#2 MapleEve

MapleEve

    Vostok 1

  • -----
  • Posts: 180
  • Joined: 04 Nov 2019
  • Loc: Hangzhou

Posted 04 December 2019 - 07:24 PM

ASCOM driver for ASI533MC

 

DR: 0 gain 70 offset

UG: 100 gain 70 offset

LRN: 360 gain 70 offset

 

for the QE part, I suspect it's over 80%, and IR cut is a must, since my IC434 also has huge IR impact of the color balance and the star HFR. It's IR pass is much stronger than IMX294.


Edited by MapleEve, 04 December 2019 - 07:27 PM.

  • StevenBellavia likes this

#3 jdupton

jdupton

    Soyuz

  • *****
  • Posts: 3,743
  • Joined: 21 Nov 2010
  • Loc: Central Texas, USA

Posted 04 December 2019 - 10:23 PM

Steve,

 

   I would suggest a much lower effective QE as the default. In my opinion, a more realistic default might be around 0.279 which assumes Peak QE of 81% for the sensor and a light transmission of 90% for a telescope. Keep in mind that the Effective QE of a sensor is always much less than its Peak QE quoted by the manufacturers and vendors.

 

   See the example below for the IMX533 sensor.

 

ASI533MC_Effective_QE.png

IMX533 Color Sensor QE response curves (From ZWO)

 

   Note that the Peak QE of just over 80% peaks in the Green at about 525 nm. The Red and Blue responses are lower. The effective (Averaged) QE of a sensor is the area under the curves between 400 nm and 700 nm. The yellow dashed line on the chart shows the averaged effective response from 400 nm to 700 nm for one Red, two Green, and one Blue pixels. The area under that Yellow curve (divided by the wavelength range of interest) is the overall Average Effective QE for the sensor in the visible light range. That number is 31.03% in the case of this sensor.

 

   A default QE as used by your spreadsheet should assume 31% QE for the sensor multiplied by the transmission factor for the telescope used. For good refractors, that might be 90%+ while for something like the EdgeHD SCTs (with their correction lenses) used with the five element EdgeHD focal reducer can be as low as 52.7% transmission. Other telescopes' transmissions can be estimated by figuring 97.5% to 98.5% transmission at each air-glass interface and 92% to 94% at each mirror reflection in the optical system.

 

 

John


Edited by jdupton, 04 December 2019 - 10:26 PM.

  • DrGomer likes this

#4 DrGomer

DrGomer

    Ranger 4

  • *****
  • Posts: 374
  • Joined: 07 Aug 2019

Posted 05 December 2019 - 11:56 AM

John, is your using of average QE, binning everything together as if it is a mono sensor a good way to look at it?  If I do the same with mono plus narrow band....  a 5nm wide filter with 90% efficiency on an 80% efficient camera looking from 400-700 nm has a "QE" of 5/(700-500)*0.9*0.8= 1.8%  

 

-just learning here. 



#5 bortle2

bortle2

    Ranger 4

  • -----
  • Posts: 318
  • Joined: 18 Sep 2019

Posted 05 December 2019 - 01:09 PM

a 5nm wide filter with 90% efficiency on an 80% efficient camera looking from 400-700 nm has a "QE" of 5/(700-500)*0.9*0.8= 1.8%

Not if most of the signal is in those 5nm.



#6 DrGomer

DrGomer

    Ranger 4

  • *****
  • Posts: 374
  • Joined: 07 Aug 2019

Posted 05 December 2019 - 02:27 PM

Not if most of the signal is in those 5nm.

agreed but that doesn't change at all quantum efficiency. 



#7 jdupton

jdupton

    Soyuz

  • *****
  • Posts: 3,743
  • Joined: 21 Nov 2010
  • Loc: Central Texas, USA

Posted 05 December 2019 - 02:32 PM

DrGomer,

 

John, is your using of average QE, binning everything together as if it is a mono sensor a good way to look at it?  If I do the same with mono plus narrow band....  a 5nm wide filter with 90% efficiency on an 80% efficient camera looking from 400-700 nm has a "QE" of 5/(700-500)*0.9*0.8= 1.8%  

 

-just learning here. 

   I don't want to derail Steve's thread but will post this answer for you. If you want to delve into the methodology more, I suggest you post for more details in the thread where I outline this methodology.

 

   In answer to your questions:

  1. Yes, to determine the average QE of a color sensor, I use what I call a Quad of pixels, one red, two green, and one blue. I find the Response Value (QE) at the wavelength of interest of each of the filtered pixels to incoming photons and then add them up and divide by 4. This is equivalent to doing a DeBayer operation using the SuperPixel method.

    For a mono sensor (which doesn't exist yet for the IMX533), I would take the same four pixels (which have no Bayer filters) and do the same -- find the QE at the wavelength of interest and then average them (even though all will be same regardless). This is the same as resampling the image to same size as the OSC version.
     
  2. In regards to being a good way to look at it, it puts all sensors on equal footing when doing these sorts of calculations. I use the method to bring some rigor to the calculations of effective QE. Even if you don't DeBayer your images with the "SuperPixel" method, you get very similar results with VNG. For measuring the background sky level in one of my images, I found the difference between the color channels for each method to be 0.1 % for Red, 0.005% for Green, and 0.2% for Blue (SuperPixel to VNG). I think this Effective QE calculation method is more than close enough regardless of how you deBayer your color images.
     
  3. Regarding your example, you are close but not quite there.
    The difference is that your calculation must include the QE from the response curve at the wavelength in question rather than the peak QE of the sensor. I have never seen a sensor for which the response curve is flat -- ie has the same QE at all wavelengths. For example, the mono sensor calculation your did with a 5 nm narrow band filter would be accurate if and only if the QE curve showed 80% at the wavelength of your narrow band filter.

    Lets say that the QE curve for your mono sensor showed 80% at 500 nm and you were using an Oiii 5 nm filter. Then as you showed in your example, the mono sensor would have an effective QE of 5 / (700 - 400) * 80% * 90% = 1.2%. This translates to English as "with my telescope passing 90% of all light hitting the aperture using a 5 nm Oiii filter and my sensor with 80% QE at 500 nm, 1.2% of all photons entering the telescope will be captured by the sensor."

    If you do the same thing now with an Ha filter, you will get a different answer. The reason is that at 656 nm, the QE curve of your sensor might show a value of 64%. Now the calculation is 5 / (700 - 400) * 64% * 90% = 0.96%

   When using a spreadsheet like Steve's you need to account for the variability of the "QE" of the sensor with respect to wavelength. That way, you can enter a set of numbers that will give a pretty good estimate of the photon rate from your own sky on the night the image was taken.

 

 

John


Edited by jdupton, 05 December 2019 - 02:38 PM.

  • Jon Rista, DrGomer and bortle2 like this

#8 StevenBellavia

StevenBellavia

    Viking 1

  • *****
  • topic starter
  • Posts: 749
  • Joined: 11 Mar 2014
  • Loc: New York

Posted 05 December 2019 - 03:56 PM

ASCOM driver for ASI533MC

 

DR: 0 gain 70 offset

UG: 100 gain 70 offset

LRN: 360 gain 70 offset

 

for the QE part, I suspect it's over 80%, and IR cut is a must, since my IC434 also has huge IR impact of the color balance and the star HFR. It's IR pass is much stronger than IMX294.

70 for all.

 

With a 50,000 elelctron full well and 14 bit ADC,  I know I am nit-picking, but I like to squeeze out every bit of dynamic range that I can.  Lower gains can have lower offsets.

 

Do you know if you can change those settings by checking "advanced" in the ASCOM driver, as is shown here for the 071?

Attached Thumbnails

  • ZWO_ASI_071_ASCOM_settings.png


#9 DrGomer

DrGomer

    Ranger 4

  • *****
  • Posts: 374
  • Joined: 07 Aug 2019

Posted 05 December 2019 - 04:27 PM

70 for all.

 

With a 50,000 elelctron full well and 14 bit ADC,  I know I am nit-picking, but I like to squeeze out every bit of dynamic range that I can.  Lower gains can have lower offsets.

 

Do you know if you can change those settings by checking "advanced" in the ASCOM driver, as is shown here for the 071?

Steven,  I just took receipt of my 533.  Under the advanced option, there is the ability to change the offset just like with your 071.


Since I'll have clouds for at least a few days, I've started taking darks.  Thought this histogram might be of interest to some (single capture)

Gain 100 (unity), 70 offset, -5C, 120 sec
 

Attached Thumbnails

  • Capture.JPG

Edited by DrGomer, 05 December 2019 - 05:07 PM.

  • StevenBellavia likes this

#10 StevenBellavia

StevenBellavia

    Viking 1

  • *****
  • topic starter
  • Posts: 749
  • Joined: 11 Mar 2014
  • Loc: New York

Posted 09 May 2020 - 07:22 PM

I too now own a ZWO ASI 533.

 

So far I only have used Gain 100, and I settled on an offset of 15 for that gain, which yields around 590 ADU's (in 16 bit), or 37 electrons.  I could have gone a little lower, but with so many electrons available, I didn't see the need to risk clipping any dark signal.

 

And with all due respect to J.D. Upton, who has done a fabulous job characterizing sensors and filters, I am finding that a Transmission Efficiency of 0.4 total (camera, + optics), is matching the sky background, verified with an SQM.  So I believe that to be correct, for now, but will look into it further. It is possible I have something awry in all my spreadsheets, regarding the sky background and transmission efficiency.  But other users tell me that the spreadsheet recommendations for gain and sub-expsoure match that of SharpCap Pro's Smart Histogram / Brain function.  So if this spreadsheet is anything close to the genius of Robin Glover... well why fix something that is not broken? But I'll take a look.

 

I also hope to do more bias signal studies at various gains, but it might not be any time soon. So far, only the Gain 100, Offset 15, at -5C is the only measured value there.  The others are estimated.

 

The latest spreadsheet, using actual values from some recent images, are in the same shared file, which I will repeat again here:

 

https://drive.google...U9CKj5GilnZW8VJ

 

Also note that I am only cooling to -5C, as it seems to struggle to go lower.  But this is still very low dark current, and in fact still lower than my ZWO ASI 183's at -15C.  And no amp glow!  Yay!

 

Steve

 

First and Second Light images from my new ZWO ASI 533MC Pro:

 

https://www.astrobin.com/nfp1kt/

 

https://www.astrobin.com/wxreri/


  • RTJoe and mAnKiNd like this

#11 Umasscrew39

Umasscrew39

    Apollo

  • *****
  • Posts: 1,454
  • Joined: 07 Dec 2016
  • Loc: Central Florida

Posted 10 May 2020 - 09:38 AM

I too now own a ZWO ASI 533.

 

So far I only have used Gain 100, and I settled on an offset of 15 for that gain, which yields around 590 ADU's (in 16 bit), or 37 electrons.  I could have gone a little lower, but with so many electrons available, I didn't see the need to risk clipping any dark signal.

 

And with all due respect to J.D. Upton, who has done a fabulous job characterizing sensors and filters, I am finding that a Transmission Efficiency of 0.4 total (camera, + optics), is matching the sky background, verified with an SQM.  So I believe that to be correct, for now, but will look into it further. It is possible I have something awry in all my spreadsheets, regarding the sky background and transmission efficiency.  But other users tell me that the spreadsheet recommendations for gain and sub-expsoure match that of SharpCap Pro's Smart Histogram / Brain function.  So if this spreadsheet is anything close to the genius of Robin Glover... well why fix something that is not broken? But I'll take a look.

 

I also hope to do more bias signal studies at various gains, but it might not be any time soon. So far, only the Gain 100, Offset 15, at -5C is the only measured value there.  The others are estimated.

 

The latest spreadsheet, using actual values from some recent images, are in the same shared file, which I will repeat again here:

 

https://drive.google...U9CKj5GilnZW8VJ

 

Also note that I am only cooling to -5C, as it seems to struggle to go lower.  But this is still very low dark current, and in fact still lower than my ZWO ASI 183's at -15C.  And no amp glow!  Yay!

 

Steve

 

First and Second Light images from my new ZWO ASI 533MC Pro:

 

https://www.astrobin.com/nfp1kt/

 

https://www.astrobin.com/wxreri/

Steve- your images look good and I think you are on the right track.  I have had the ASI533 since it first became available and after trying multiple settings from my home observatory (not great skies; Bortle 6 with many nights of average/below average seeing and transparency), I settled on a 200 gain; 10 offset; either 90 or 120s sub exposures. 

 

Bruce


  • StevenBellavia likes this

#12 RTJoe

RTJoe

    Lift Off

  • -----
  • Posts: 13
  • Joined: 24 Sep 2014
  • Loc: BW, Germany

Posted 11 May 2020 - 04:58 AM

Steven, thank you for your spreadsheet!

 

I try to understand some of the calculations – could you please explain why mapping from 16 bit ADU’s to 14 bit ADU’s is calculated through division by 16 in tab ASI533-specs-DR-SNR (Bias), but through division by 4 (2^14/2^16) in tab Gain-Sub-Exp (that is what I would have expected)?

 

Joachim



#13 StevenBellavia

StevenBellavia

    Viking 1

  • *****
  • topic starter
  • Posts: 749
  • Joined: 11 Mar 2014
  • Loc: New York

Posted 11 May 2020 - 07:38 PM

Steven, thank you for your spreadsheet!

 

I try to understand some of the calculations – could you please explain why mapping from 16 bit ADU’s to 14 bit ADU’s is calculated through division by 16 in tab ASI533-specs-DR-SNR (Bias), but through division by 4 (2^14/2^16) in tab Gain-Sub-Exp (that is what I would have expected)?

 

Joachim

Oh!  That's a mistake.  Not much effect, luckily..  It has been fixed and the new version is in the same shared drive.

Thanks for catching it!



#14 ippiu

ippiu

    Mariner 2

  • -----
  • Posts: 264
  • Joined: 06 Dec 2010
  • Loc: Italy

Posted 12 May 2020 - 06:11 AM

There is a little little error in the folder title, too... smile.gif

 

Immagine.jpg

 


  • StevenBellavia likes this

#15 jemarcks

jemarcks

    Sputnik

  • -----
  • Posts: 35
  • Joined: 18 Dec 2019

Posted 12 May 2020 - 12:29 PM

Steven,

Thanks for this spreadsheet!  Great work!

I bought a 533mc a few weeks ago and have been testing it out on clear nights.  I am pretty new to AP having completed my scope rig in janurary so i could start imaging with a dslr.  In late April, i upgraded to a 533mc pro.  I am still learning but my biggest issue is determining the best gain/offset/sub length for my target.

Can you give a overview on how best to use this spreadsheet to get the suggested gain and exposure length?

 

thanks!

jeff


  • StevenBellavia likes this

#16 StevenBellavia

StevenBellavia

    Viking 1

  • *****
  • topic starter
  • Posts: 749
  • Joined: 11 Mar 2014
  • Loc: New York

Posted 13 May 2020 - 08:37 AM

There is a little little error in the folder title, too... smile.gif

 

attachicon.gifImmagine.jpg

I fixed it.

And it is funny, because ZWO themselves had it wrong too, the same way on their web page.

You would think after I told them about it, I would have been more careful too?

Thanks!


Edited by StevenBellavia, 13 May 2020 - 08:37 AM.


#17 StevenBellavia

StevenBellavia

    Viking 1

  • *****
  • topic starter
  • Posts: 749
  • Joined: 11 Mar 2014
  • Loc: New York

Posted 13 May 2020 - 09:12 AM

Steven,

Thanks for this spreadsheet!  Great work!

I bought a 533mc a few weeks ago and have been testing it out on clear nights.  I am pretty new to AP having completed my scope rig in janurary so i could start imaging with a dslr.  In late April, i upgraded to a 533mc pro.  I am still learning but my biggest issue is determining the best gain/offset/sub length for my target.

Can you give a overview on how best to use this spreadsheet to get the suggested gain and exposure length?

 

thanks!

jeff

I am afraid to give an overview, as it might completely derail the thread, which is really just to point to the location of the spreadsheet.

 

But I will anyway.

 

The original goal of the spreadsheet was twofold:

1. To determine a gain and subexposure that results in swamping the read noise by 10X

2. To be able to use actual image data to determine your sky background

 

Part 2 is much more difficult, as it requires actual sky survey data, in various wavelengths, and then many assumptions regarding how much of the sky photons makes it into a digital signal.  So quantum efficiency, optical transmission efficiency, etc, all have an effect.  But I did the best I could (so far).

 

Part 1 is much easier, if you already know your sky background. (via an SQM, or using dark sky maps, etc).

Why 10X?  That is just a good rule of thumb, and there are multiple CloudyNights and other threads on this.  Put simply, 10X means you are at 95% of the best Signal-to-Noise (SNR) you can get for that particular sky.

So once you have agreed to the 10X rule of thumb, it's easy:

Sky Background = 10 x (Read Noise)^2

So if you pick a region in your image of blank sky, the ADU values (in whatever system you are in, typically 16 bit), should be 10X greater than the square of the read noise in that same system.

 

Note that SharpCap Pro does this too, using actual image data while in the filed.  I think the Pro version is about $10 a year.  So if you don't like my spreadsheet, which is perfectly fine, there are alternatives.

 

Also note, the 10X rule might not be practicable for narrowband and/or true dark sky sites.

 

Regarding the spreadsheet:

Only enter values in the blue boxes (or bluish-green boxes, if you know those values)

If One Shot Color (OSC), or RGB filters, enter 100nm for filter bandwidth (Cell D15) wavelength, and for filter wavelength (Cell D14) enter 550 for OSC, or 650, 550 and 450 for Red, Green and Blue, respectively.

For straight luminance / monochrome, enter 300 nm for filter bandwidth (or whatever the bandwidth is for you luminance filter).  This has the effect of reducing the subexposure by 3X (and hence why people like to shoot in luminance, another long topic on CN...)

 

Transmission efficiency is still uncertain, but I am finding 0.3 to 0.4 is now working for me (I have data from a Bortle 3-4 area, from last night, I want to try).

 

If you know your sky background and dark current you don't need to use the middle portion of the spreadsheet.

 

The right portion is the recommendations to achieve Swamp 10X for that scope and camera, in that sky.

 

I hope that helps a little.

 

Steve


Edited by StevenBellavia, 13 May 2020 - 09:16 AM.

  • jemarcks likes this

#18 StevenBellavia

StevenBellavia

    Viking 1

  • *****
  • topic starter
  • Posts: 749
  • Joined: 11 Mar 2014
  • Loc: New York

Posted 13 May 2020 - 09:39 AM

I should also mention:  I AM NOT AN EXPERT.

 

I particularly like Post #77 on this thread, from my mentor and guru, Jon Rista, replying to another guru, freestar8n

 

https://www.cloudyni...-returns/page-4

 

If you don't feel like going there, here is the part I like:

 

"I would say that there IS indeed an optimal exposure for most cameras, but it is not just optimal from a noise standpoint. It is optimal from a usage of dynamic range/well capacity standpoint.

While noise is important, noise is not the only detrimental thing to our data, and IMO we need to consider the entirety of the signal...not just the darkest parts. Optimal is about everything, not just noise".

 

These guys know what they are talking about.  I am just a number cruncher.  I like decisions based on numbers.



#19 Astrojedi

Astrojedi

    Fly Me to the Moon

  • *****
  • Posts: 5,743
  • Joined: 27 May 2015
  • Loc: SoCal

Posted 13 May 2020 - 01:18 PM

Steven,

 

Excellent effort. In practice though I always use 100 gain and 120s or 300s exposures for broadband (depending on scope) and 300 gain and 30s or 60s exposures for astrometry. For narrowband is use 300 gain at 300s.

 

The reason is it is easier to maintain and use a dark library. As you can see that even with this limited selection I have 6+ sets of darks to deal with. While the SNR may not be 100% optimized it makes little difference as long as I don't lose any electrons to quantization error or clipping or read noise. The levels I can deal with in post process.

 

AJ


Edited by Astrojedi, 13 May 2020 - 01:19 PM.

  • StevenBellavia likes this

#20 StevenBellavia

StevenBellavia

    Viking 1

  • *****
  • topic starter
  • Posts: 749
  • Joined: 11 Mar 2014
  • Loc: New York

Posted 14 May 2020 - 10:14 AM

Steven,

 

Excellent effort. In practice though I always use 100 gain and 120s or 300s exposures for broadband (depending on scope) and 300 gain and 30s or 60s exposures for astrometry. For narrowband is use 300 gain at 300s.

 

The reason is it is easier to maintain and use a dark library. As you can see that even with this limited selection I have 6+ sets of darks to deal with. While the SNR may not be 100% optimized it makes little difference as long as I don't lose any electrons to quantization error or clipping or read noise. The levels I can deal with in post process.

 

AJ

Hi,

 

I ran a quick simulation, using as much real data as I have so far, and the remainder from the SONY IMX533 specs.

 

I see no benefit to choosing Gain 200 or Gain 300 over Gain 100. 

For Gain 200, it is almost a half a stop decrease in Dynamic Range, with only a 1% increase in integrated SNR.

For Gain 300, it is over 3 stops decrease in Dynamic Range, with only a 4.7% increase in integrated SNR.

 

Normally, Gain is increased to lower read noise, which in theory should increase SNR (at the expense of DR).  But when you are starting with only 1.5e- read noise, dropping to 1.3e- or 1.2e- has very little effect on SNR.

 

This is a summary of my simulation:

 

ASI 533 Gain 0, 100 and 200

Scope: f/6.9, 150mm aperture
Ha, 6nm
Sky = 20.0 (at 656 nm, 6nm bandwidth)
Subexposure = 300 sec
Total Exposure = 240 minutes

 

Gain     Swamp     DR       SNR
   0        0.71        16.0      2.20
100       4.77        14.9      3.00
200       6.35        14.5      3.03

300       7.40        11.7      3.14


Edited by StevenBellavia, 14 May 2020 - 10:15 AM.

  • jdupton, ippiu, acommonsoul and 1 other like this

#21 Astrojedi

Astrojedi

    Fly Me to the Moon

  • *****
  • Posts: 5,743
  • Joined: 27 May 2015
  • Loc: SoCal

Posted 14 May 2020 - 02:10 PM

Hi,

 

I ran a quick simulation, using as much real data as I have so far, and the remainder from the SONY IMX533 specs.

 

I see no benefit to choosing Gain 200 or Gain 300 over Gain 100. 

For Gain 200, it is almost a half a stop decrease in Dynamic Range, with only a 1% increase in integrated SNR.

For Gain 300, it is over 3 stops decrease in Dynamic Range, with only a 4.7% increase in integrated SNR.

 

Normally, Gain is increased to lower read noise, which in theory should increase SNR (at the expense of DR).  But when you are starting with only 1.5e- read noise, dropping to 1.3e- or 1.2e- has very little effect on SNR.

 

This is a summary of my simulation:

 

ASI 533 Gain 0, 100 and 200

Scope: f/6.9, 150mm aperture
Ha, 6nm
Sky = 20.0 (at 656 nm, 6nm bandwidth)
Subexposure = 300 sec
Total Exposure = 240 minutes

 

Gain     Swamp     DR       SNR
   0        0.71        16.0      2.20
100       4.77        14.9      3.00
200       6.35        14.5      3.03

300       7.40        11.7      3.14

Note that you need to account for quantization error when signal is low (short exposures e.g. astrometry or very light limited exposures e.g. narrowband). In those situations you need to scale the small portion of the full well used to use the full ADC bit range. In these situations reduction of overall dynamic range is not an issue - the bigger issue is loss of dynamic range to the ADC steps.

 

I see a significant difference when doing astrometry (which sometimes requires very short exposures to capture fast moving very faint targets) - at least half a magnitude improvement at minimum in the stack.


Edited by Astrojedi, 14 May 2020 - 02:13 PM.

  • StevenBellavia likes this

#22 StevenBellavia

StevenBellavia

    Viking 1

  • *****
  • topic starter
  • Posts: 749
  • Joined: 11 Mar 2014
  • Loc: New York

Posted 17 May 2020 - 12:47 PM

Note that you need to account for quantization error when signal is low (short exposures e.g. astrometry or very light limited exposures e.g. narrowband). In those situations you need to scale the small portion of the full well used to use the full ADC bit range. In these situations reduction of overall dynamic range is not an issue - the bigger issue is loss of dynamic range to the ADC steps.

 

I see a significant difference when doing astrometry (which sometimes requires very short exposures to capture fast moving very faint targets) - at least half a magnitude improvement at minimum in the stack.

I could be wrong, but with a 14-bit ADC, which can handle 16,384 electrons at 1:1, that the quantization error goes to zero when the full well is at or below 16,384, which looks to me occurs around unity gain (?)

 

Can you go into more detail about your half magnitude improvement?



#23 Astrojedi

Astrojedi

    Fly Me to the Moon

  • *****
  • Posts: 5,743
  • Joined: 27 May 2015
  • Loc: SoCal

Posted 17 May 2020 - 02:40 PM

Not quite. Remember (analog) gain multiples the number of electrons which is why read noise declines relative to signal. My goal in astrometry is to try and detect faint NEOs with sufficient SNR in short exposures for confirmation. 
 

Let’s say the faint source produces 1e of signal. At unity gain it will be swamped by the read noise. At a gain of 300 that signal will produce ~10e but the read noise will be close to 1e producing an SNR of almost 10.

 

(replace e above with ADUs which is technically correct)


Edited by Astrojedi, 17 May 2020 - 03:00 PM.


#24 ks__observer

ks__observer

    Surveyor 1

  • *****
  • Posts: 1,728
  • Joined: 28 Sep 2016
  • Loc: Long Island, New York

Posted 17 May 2020 - 03:02 PM

Not quite. Remember (analog) gain multiples the number of electrons which is why read noise declines relative to signal. My goal in astrometry is to try and detect faint NEOs with sufficient SNR in short exposures for confirmation. 
 

Let’s say the faint source produces 1e of signal. At unity gain it will be swamped by the read noise. At a gain of 300 that signal will produce ~10e but the read noise will be close to 1e producing an SNR of almost 10.

 

(replace e above with ADUs which is technically correct)

It will not produce 10 electrons.

It will produce 1 electron that looks 10e large.

The SNR is based on the total number of electrons counted -- not how you later magnify the size of each electron.


  • kingjamez likes this

#25 Astrojedi

Astrojedi

    Fly Me to the Moon

  • *****
  • Posts: 5,743
  • Joined: 27 May 2015
  • Loc: SoCal

Posted 17 May 2020 - 04:40 PM

Do you have another way of measuring SNR besides ADUs?

 

Also note it is not multiplying the single electron by 10x... rather it is producing more electrons from the signal. There is a subtle but important difference as this happens in the analog domain. And yes it does increase SNR as it brings more signal above the noise threshold of the camera.


Edited by Astrojedi, 17 May 2020 - 04:46 PM.



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