240 replies to this topic

[Shiraz]

Hi

The ASI1600 allows the user to choose almost any combination of settings -
question is, which is best.

 

To try to answer the basic question of how to choose a sub-length, an end-to-end astrograph model was used to generate tables of suggested sub-lengths for the ZWO1600 under a wide variety of conditions – broadband (luminance) and narrowband. These could be helpful in navigating between different combinations of camera gain, fNo and sky brightness.  Should be good enough to find a starting point if new to the camera, or to see what might happen if you try something different – attempted to keep assumptions reasonable, but your system optics/sky/filter etc may differ, so don't expect extreme accuracy.

 

To use the tables, first determine your sky brightness. If possible, measure with an SQM, but otherwise, find your colour zone from a sky brightness atlas – maybe try this as a starting point http://www.skyandtel...e-your-skyglow/

 

When you know your sky brightness/zone, go to the appropriate attached table and identify the column headed by your scope fNo. Then choose a gain from the first column – lower gain for dynamic range, higher gain for low noise short subs or narrowband. If you are new to the camera, suggest that you start out at gain 50 or 75 and that you use the original ZWO intermediate offset default of 21.  Now  look directly across the row for your chosen gain setting and find the sub-length that is also in the fNo column for your scope. If you cannot find the exact fNo for your scope, or want to use a gain value that is not listed, look at the surrounding data and guess – that will be good enough.

 

The value from the table is the shortest sub length that you can use (for that scope, sky and gain) and still be shot noise limited (5%RN criterion) - this is the most efficient way to image. The suggestion is definitely not a hard and fast rule though - you can use longer subs with no loss of signal to noise ratio, but more stars will be saturated. You can use shorter subs, but the total exposure time to get to a given image quality will be longer.

 

To refine the sub length, first take a recommended sub and use the cursor of your acquisition software to look at the ADU values in a background sky region of the scene (no stars and no hot pixels, preferably away from vignetting). Dither the cursor about slightly to estimate the average ADU background value in the clear region. From this background ADU value, subtract the bias value for your camera (again - use the cursor to estimate the average ADU signal in a bias frame taken at the same gain and offset as the light sub). Then compare your sub-minus-bias value with the “expected ADU value above bias” in the final column of the table. If your sub-minus-bias is lower than expected, increase the sub length to compensate – if you get a much higher value, consider reducing the sub length.

 

The Narrowband tables should be used in the same way as the Broadband ones, although there are more unknown variables to upset the accuracy – so use as a guide only. Because of the wide possible variability in parameters, I chose to generate tables only for the extremes of dark sky and full moonlit conditions - for other conditions, try to estimate from the extreme examples provided. As a general rule though, use subs as long as possible for narrowband.

 

You could use the same gain setting as for broadband, but you may get slightly better results using higher gain – 200 would be a good value to try at some stage. If you do change the gain, you will need new dark and bias calibration and could also set the offset higher (maybe use 50).

I assumed 7nm filters, so if you have narrower filters, you could use longer subs – (or vice versa). 
The estimates should be reasonably applicable to commonly used Ha, O3 and S2 filters, although the Ha and S2 will probably need a bit more exposure than O3.

 

 

Some comments:

Don't go by the visual appearance of a sub – with the 1600, they can look very thin, but when stacked, the final result will be fine – so ignore the fact that subs from the 1600 can look weak.

 

The modelling assumes an average optical efficiency. If you use a high quality refractor, you could possibly get away with shorter subs and if you use a highly obstructed scope, or one with standard Al mirrors, you will need longer subs.

 

the broadband calculations are luminance based. use longer subs for RGB filters and aim for the same expected ADU values.

 

The tables may possibly also be applicable to the QHY163. I don't have access to one, but found a couple of posts on other forums/threads suggesting that 139 gain on the 1600 corresponds to gain 12 on the 163. If so, and if the mapping is the same, maybe divide the values in the 1600 gain columns by 12 and go from there – eg, assume that gain 75 on the 1600 corresponds roughly to gain 6 on the 163. Anyone know for sure what the gain settings on the 163 indicate?

 

If you change any of temperature, gain or offset, you will need new bias and dark calibration data.

 

your bias value should be at least a couple of hundred ADU - if you measure less, then increase the offset value.

 

If you decide to use very short subs, be aware of the processing load – it can take hours to process a few hundred 16mpix subs.

 

Thanks for looking, hope the above is useful - very grateful for any feedback. Regards Ray

Attached Thumbnails

• 
• 

Edited by Shiraz, 16 April 2017 - 04:15 AM.

[Shiraz]

narrowband

 

Attached Thumbnails

• 

[roofkid]

Hi Shiraz,

 

I have done my own calculations and can confirm that your numbers are in the right ballpark.

 

One thing of note for others: I believe this table is a great starting point but determining the optimal exposure is dependant on a few other things. Especially the filters you are using and your location. Just a few examples where it may be prudent to do your own additional research starting from the great table above. Personally I "only" have narrowband filters in the 7nm ballpark. The numbers above do not match my own calculations. I obviously need to expose shorter. Maybe these tables are calculated for 3nm or 5nm filters?

The broadband tables seem to be for L filters. However color filters also come into play at different exposure lengths. To complicate matters further I found out that I needed to expose my B channel about 50% longer (6m instead of 4m at gain 0) than my R and G channels what may have to do with the light pollution and my main location or the blue sensitivity of the ASI1600. That means that in the past I had underexposed my blue channel by just using the same value for RGB.

 

Hope that helps to put things into additional context and give you some thoughts on the next steps if you have worked with the tables above.

 

Great work

Sven

Edited by roofkid, 16 April 2017 - 05:41 AM.

[Shiraz]

thanks for the helpful feedback Sven. Have modified the post to make it clear that the broadband calculation is for luminance.

 

The narrowband calculations are for 7nm, but sky brightness varies so much over the range of conditions used for NB that I am not surprised you need different sub-lengths for your situation/equipment.

Edited by Shiraz, 16 April 2017 - 04:24 AM.

[sparkyht]

Regarding the narrowband table, I thought that exposures should be kept below 10 minutes because of the increasing amp glow? It's appearing to contradict a lot of other information I've read on CN, especially the sage advice from Jon Rista. I thought high gain/short exposures was the formula for narrowband? There have been others who've imaged with the ASI1600 like they would with a CCD, but I was led to believe that doing so was the "wrong" way to use the camera.

[Shiraz]

Regarding the narrowband table, I thought that exposures should be kept below 10 minutes because of the increasing amp glow? It's appearing to contradict a lot of other information I've read on CN, especially the sage advice from Jon Rista. I thought high gain/short exposures was the formula for narrowband? There have been others who've imaged with the ASI1600 like they would with a CCD, but I was led to believe that doing so was the "wrong" way to use the camera.

I don't see any inconsistency with Jon's advice. I understand that he images with a fast scope at high gain and from the tables, his system would only need short subs to be sky limited under almost all conditions (eg 1.4 minutes at gain 200 under moonlight with the chosen assumptions).

 

The tables also show the other side of the coin - what happens if you opt for a slow scope and low gain. You need longer subs and run into all of the problems of long subs, such as guiding difficulty, lost subs due to transient events, dark current issues etc. The text recommends higher gain for narrowband, as does Jon, but the tables also show what sub lengths you need if you ignore that advice and choose to go with low gain on a slow scope - I can't think of a reason why anyone would want to use that approach, but it is the owner's call and it will still work.

 

The tables were not intended as recommendations, they just tell you what sub length you need to be sky limited for any combination of sky, scope and gain that you choose. The bit that this adds to the discussion is how the scope and sky influence camera gain choices and sub lengths - you cannot make optimium gain and sub choices without considering sky and scope.

Edited by Shiraz, 16 April 2017 - 07:54 PM.

[Jon Rista]

I also don't see any discrepancies between Shiraz's tables and my own experiences. The only thing his tables don't show is gain settings 250 and 300, as 300 is the maximum useful gain on the ASI1600 (not sure exactly what it would be on the QHY, but since the ratio is ~11:1 ASI:QSI, I suspect a gain of 27-30 is probably the max useful on it.) 

 

The conversion gain ratio is .48e-/ADU @ Gain 200, and 0.15e-/ADU @ Gain 300. That's a pretty big difference. Read noise is about 1.3e- @ Gain 200, and 1.1e- @ Gain 300. Given his exposure recommendations in these tables for high f/ratio or 21.5mag/sq" skies (blue zone, BTW, quite dark!), I would say there is no reason not to use Gain 300 in these cases. Until you hit the read noise floor, so long as you aren't clipping (and this would be more true for high f-ratio than dark site, where image scale can spread a star out over many pixels and reduce star saturation rates), there is no reason not to use the highest gain you can to get the lowest read noise you can. An f/7 or higher f-ratio at a dark site would benefit quite a bit from Gains 200 though 300 I think, and your subs would still be short enough to allow high sub count stacking. 

 

At Gain 190 and up, though, your conversion ratio is ~0.5e-/ADU or lower, which means you'll be sampling the noise well, if not really well. That reduces quantization error to extremely small levels, effectively rendering the difference between a 16-bit and 12-bit ADC moot. So, even though you could, you wouldn't necessarily need, to stack lots of short subs at high gain. For those who have dark skies and want long exposures, with narrow band your all set. Gain 200 will give you 16 minute or longer exposure times, with well sampled noise, just like a CCD. However, there is one critical fact that really matters here: You COULD get a skyfog limited dark sky sub in 16 minutes per sub. With any CCD on the market save maybe the QSI6120, it's basically impossible to get skyfog limited subs anywhere, especially at a dark site. 

[Shiraz]

 

 

 Gain 200 will give you 16 minute or longer exposure times, with well sampled noise, just like a CCD. However, there is one critical fact that really matters here: You COULD get a skyfog limited dark sky sub in 16 minutes per sub. With any CCD on the market save maybe the QSI6120, it's basically impossible to get skyfog limited subs anywhere, especially at a dark site. 

to put this point in some perspective, with an f4 scope and under similarly dark sky, the model indicates that an 8300 would require 6 hour sub exposures to get to the same sky-limited state as the 1600 with 16 minute subs - sub length scales ~with the square of the read noise. Not wishing to knock the 8300, but rather emphasise just how revolutionary the new cameras really are for deep NB.

 

will redo the tables and incorporate gains 250 and 300.

Edited by Shiraz, 16 April 2017 - 10:52 PM.

[GIR]

to put this point in some perspective, with an f4 scope and under similarly dark sky, the model indicates that an 8300 would require 6 hour sub exposures to get to the same sky-limited state as the 1600 with 16 minute subs - sub length scales ~with the square of the read noise. Not wishing to knock the 8300, but rather emphasise just how revolutionary the new cameras really are for deep NB.

 

 

 

Was just wondering how this is possible if read noise is such a crucial factor....

 

http://www.astrobin....220132/?nc=user

 

Scope f/7, camera even noisier than 8300 (KAF 16803), subs 20 min, intergation time less than 6 h, narrow band (5nm).

[Jacques Kieffer]

very interesting topic

 

Jon, Shiraz,

 

do you have experience with asi1600 OSC ?

 

Are there similar settings available for the ASI1600cooled COLOR version ?

 

My SQM readings are between 19.60 and 20.00 and I'm a bit struggling with the settings for my ASi1600

I'm experiencing lots of amp glow and my darks seem not really to correct the problem.

My actual settings for deep sky, are unity gain, -20°C, and exp time of 300 sec, doing ca 50 subs, 50x darks, 50xbias.

 

I followed some other topics ie 'starting pointers for asi1600', but these are more for the monochrome versions.

 

Jacques

[Seanem44]

This needs to be stickied to this forum.  Great job!!!

Edited by Seanem44, 17 April 2017 - 08:06 AM.

[Jon Rista]

 

to put this point in some perspective, with an f4 scope and under similarly dark sky, the model indicates that an 8300 would require 6 hour sub exposures to get to the same sky-limited state as the 1600 with 16 minute subs - sub length scales ~with the square of the read noise. Not wishing to knock the 8300, but rather emphasise just how revolutionary the new cameras really are for deep NB.

 

 

 

Was just wondering how this is possible if read noise is such a crucial factor....

 

http://www.astrobin....220132/?nc=user

 

Scope f/7, camera even noisier than 8300 (KAF 16803), subs 20 min, intergation time less than 6 h, narrow band (5nm).

 

We were talking about skyfog limited subs. Getting skyfog limited subs is not absolutely necessary, it's just ideal. 

 

Also keep in mind, image scale is also a factor. An f/7 scope with 9 micron pixels will have a 1"/px image scale. An ASI1600 on the same scope would have a 0.4"/px image scale. That does matter.

 

On a given scope, the KAF-8300 would have about twice the pixel area, so that would reduce time by a factor of two. However, there is still going to be a large discrepancy in how long it takes to get a skyfog limited sub. I think most CCD users who follow an exposure rule use the 3*RN^2 rule, which says to expose your background sky median level to three times the read noise squared. With 9e- read noise, you would need a 243e- background sky signal to be skyfog limited. With only 1.15e- read noise (ASI1600 @ Gain 300), you would only need a 4e- background sky signal to be skyfog limited. Assuming a 0.01e-/s flux, that would amount to about 3h22m30s exposures with the KAF-8300 and only 6m40s exposures with the ASI1600, on the same scope. 

[hnau]

The only thing his tables don't show is gain settings 250 and 300, as 300 is the maximum useful gain on the ASI1600 (not sure exactly what it would be on the QHY, but since the ratio is ~11:1 ASI:QSI, I suspect a gain of 27-30 is probably the max useful on it.) 

QHY's test release of their SDK has multiplied the gain range for the QHY163 by 10 so that it looks like we will have a range of 0-580 eventually; settings between the 1600/163 series could be a lot more similar in the near future.

Edited by hnau, 17 April 2017 - 12:26 PM.

[Jon Rista]

 

The only thing his tables don't show is gain settings 250 and 300, as 300 is the maximum useful gain on the ASI1600 (not sure exactly what it would be on the QHY, but since the ratio is ~11:1 ASI:QSI, I suspect a gain of 27-30 is probably the max useful on it.) 

QHY's test release of their SDK has multiplied the gain range for the QHY163 by 10 so that it looks like we will have a range of 0-580 eventually; settings between the 1600/163 series could be a lot more similar in the near future.

 

Ah! That will be very handy! 

[GIR]

 

 

to put this point in some perspective, with an f4 scope and under similarly dark sky, the model indicates that an 8300 would require 6 hour sub exposures to get to the same sky-limited state as the 1600 with 16 minute subs - sub length scales ~with the square of the read noise. Not wishing to knock the 8300, but rather emphasise just how revolutionary the new cameras really are for deep NB.

 

 

 

Was just wondering how this is possible if read noise is such a crucial factor....

 

http://www.astrobin....220132/?nc=user

 

Scope f/7, camera even noisier than 8300 (KAF 16803), subs 20 min, intergation time less than 6 h, narrow band (5nm).

 

We were talking about skyfog limited subs. Getting skyfog limited subs is not absolutely necessary, it's just ideal. 

 

Also keep in mind, image scale is also a factor. An f/7 scope with 9 micron pixels will have a 1"/px image scale. An ASI1600 on the same scope would have a 0.4"/px image scale. That does matter.

 

On a given scope, the KAF-8300 would have about twice the pixel area, so that would reduce time by a factor of two. However, there is still going to be a large discrepancy in how long it takes to get a skyfog limited sub. I think most CCD users who follow an exposure rule use the 3*RN^2 rule, which says to expose your background sky median level to three times the read noise squared. With 9e- read noise, you would need a 243e- background sky signal to be skyfog limited. With only 1.15e- read noise (ASI1600 @ Gain 300), you would only need a 4e- background sky signal to be skyfog limited. Assuming a 0.01e-/s flux, that would amount to about 3h22m30s exposures with the KAF-8300 and only 6m40s exposures with the ASI1600, on the same scope. 

 

Thanks for the explanation Jon although the comparison to the Gain 300 and  dark skies might  be a bit extreme,

...and the total integration time needed would still be the same  with 1600(163) and 8300 ?

[Jon Rista]

 

 

 

to put this point in some perspective, with an f4 scope and under similarly dark sky, the model indicates that an 8300 would require 6 hour sub exposures to get to the same sky-limited state as the 1600 with 16 minute subs - sub length scales ~with the square of the read noise. Not wishing to knock the 8300, but rather emphasise just how revolutionary the new cameras really are for deep NB.

 

 

 

Was just wondering how this is possible if read noise is such a crucial factor....

 

http://www.astrobin....220132/?nc=user

 

Scope f/7, camera even noisier than 8300 (KAF 16803), subs 20 min, intergation time less than 6 h, narrow band (5nm).

 

We were talking about skyfog limited subs. Getting skyfog limited subs is not absolutely necessary, it's just ideal. 

 

Also keep in mind, image scale is also a factor. An f/7 scope with 9 micron pixels will have a 1"/px image scale. An ASI1600 on the same scope would have a 0.4"/px image scale. That does matter.

 

On a given scope, the KAF-8300 would have about twice the pixel area, so that would reduce time by a factor of two. However, there is still going to be a large discrepancy in how long it takes to get a skyfog limited sub. I think most CCD users who follow an exposure rule use the 3*RN^2 rule, which says to expose your background sky median level to three times the read noise squared. With 9e- read noise, you would need a 243e- background sky signal to be skyfog limited. With only 1.15e- read noise (ASI1600 @ Gain 300), you would only need a 4e- background sky signal to be skyfog limited. Assuming a 0.01e-/s flux, that would amount to about 3h22m30s exposures with the KAF-8300 and only 6m40s exposures with the ASI1600, on the same scope. 

 

Thanks for the explanation Jon although the comparison to the Gain 300 and  dark skies might  be a bit extreme,

...and the total integration time needed would still be the same  with 1600(163) and 8300 ?

 

Yes, you would still need the integration time. 

 

I'm not sure why Gain 300 is extreme, though. I guess I might not have been clear, but we aren't talking about LRGB here, we are talking about narrow band. With narrow band, especially at a higher f-ratio, ABSOLUTELY use Gain 300. You want the lowest read noise you can get. For LRGB, you wouldn't need Gain 300, but LRGB isn't really the issue when it comes to getting skyfog limited subs. ;P

 

And here is where the ASI1600's advantage comes in. You CAN expose for longer than minimally necessary. The moment you do that, then it would take you less integration time to get the same SNR than a camera with higher read noise. If you used 10 minute subs rather than 6, or 15 minute subs rather than 10, you would be able to get much better SNR on faint details, even at a higher gain setting (any higher gain setting, actually, as the maximum read noise on the ASI1600 is 3.5e-, much lower than the KAF-8300). 

[fnord123]

 I'm not sure why Gain 300 is extreme, though. I guess I might not have been clear, but we aren't talking about LRGB here, we are talking about narrow band. With narrow band, especially at a higher f-ratio, ABSOLUTELY use Gain 300. You want the lowest read noise you can get. For LRGB, you wouldn't need Gain 300, but LRGB isn't really the issue when it comes to getting skyfog limited subs. ;P

 

And here is where the ASI1600's advantage comes in. You CAN expose for longer than minimally necessary. The moment you do that, then it would take you less integration time to get the same SNR than a camera with higher read noise. If you used 10 minute subs rather than 6, or 15 minute subs rather than 10, you would be able to get much better SNR on faint details, even at a higher gain setting (any higher gain setting, actually, as the maximum read noise on the ASI1600 is 3.5e-, much lower than the KAF-8300). 

Apologies for resurrecting an old thread, but I just came across it as a result of reading this thread.

 

Can someone elaborate on what the part about exposing longer?  I always had the understanding that if one assumes the same total integration time, the only benefit of using any gain lower than 300 with an ASI1600MMC was having less frames to process,  with the drawback being bit depth loss if I go under ~32 frames, plus greater dependency on having a good mount.    Is that not the accepted wisdom now?

Edited by fnord123, 08 June 2017 - 03:42 PM.

[Bart Declercq]

 

 I'm not sure why Gain 300 is extreme, though. I guess I might not have been clear, but we aren't talking about LRGB here, we are talking about narrow band. With narrow band, especially at a higher f-ratio, ABSOLUTELY use Gain 300. You want the lowest read noise you can get. For LRGB, you wouldn't need Gain 300, but LRGB isn't really the issue when it comes to getting skyfog limited subs. ;P

 

And here is where the ASI1600's advantage comes in. You CAN expose for longer than minimally necessary. The moment you do that, then it would take you less integration time to get the same SNR than a camera with higher read noise. If you used 10 minute subs rather than 6, or 15 minute subs rather than 10, you would be able to get much better SNR on faint details, even at a higher gain setting (any higher gain setting, actually, as the maximum read noise on the ASI1600 is 3.5e-, much lower than the KAF-8300). 

Apologies for resurrecting an old thread, but I just came across it as a result of reading this thread.

 

Can someone elaborate on what the part about exposing longer?  I always had the understanding that if one assumes the same total integration time, the only benefit of using any gain lower than 300 with an ASI1600MMC was having less frames to process,  with the drawback being bit depth loss if I go under ~32 frames, plus greater dependency on having a good mount.    Is that not the accepted wisdom now?

 

Using lower gain gives you a higher dynamic range in the individual frames, which helps avoid blowing out the brighter stars in the image - even at very short exposures, some stars in the FOV will be overexposed at gain 300 which may not be desirable, for example when imaging open clusters or nebulae which contain brighter stars.

Edited by Bart Declercq, 09 June 2017 - 04:41 AM.

[fnord123]

 

Can someone elaborate on what the part about exposing longer?  I always had the understanding that if one assumes the same total integration time, the only benefit of using any gain lower than 300 with an ASI1600MMC was having less frames to process,  with the drawback being bit depth loss if I go under ~32 frames, plus greater dependency on having a good mount.    Is that not the accepted wisdom now?

Using lower gain gives you a higher dynamic range in the individual frames, which helps avoid blowing out the brighter stars in the image - even at very short exposures, some stars in the FOV will be overexposed at gain 300 which may not be desirable, for example when imaging open clusters or nebulae which contain brighter stars.

Thanks, that makes sense.  Testing my understanding, are any of the following statements incorrect assuming I am imaging at F/6 in a red zone?

 

• At gain 139, I should expect < 10s LRGB (minimum) exposures and aim for ADU 450.  Since I use SGP, I should aim for a median pixel value of (16 * 450) 7200.
• If I want to maximize dynamic range, I should use a gain of 50, 75, or 100, since the target ADU is 340.
• Using 0 gain allows me to use 30s minimum exposures, which means less processing files, but causes the target ADU to go back up to 420, thus losing dynamic range. The reason for this is that the read noise is higher at this gain.

Thanks for all the explanation - I have a better grasp of how best to use this camera now

[Stelios]

 

 

Can someone elaborate on what the part about exposing longer?  I always had the understanding that if one assumes the same total integration time, the only benefit of using any gain lower than 300 with an ASI1600MMC was having less frames to process,  with the drawback being bit depth loss if I go under ~32 frames, plus greater dependency on having a good mount.    Is that not the accepted wisdom now?

Using lower gain gives you a higher dynamic range in the individual frames, which helps avoid blowing out the brighter stars in the image - even at very short exposures, some stars in the FOV will be overexposed at gain 300 which may not be desirable, for example when imaging open clusters or nebulae which contain brighter stars.

Thanks, that makes sense.  Testing my understanding, are any of the following statements incorrect assuming I am imaging at F/6 in a red zone?

 

• At gain 139, I should expect < 10s LRGB (minimum) exposures and aim for ADU 450.  Since I use SGP, I should aim for a median pixel value of (16 * 450) 7200.
• If I want to maximize dynamic range, I should use a gain of 50, 75, or 100, since the target ADU is 340.
• Using 0 gain allows me to use 30s minimum exposures, which means less processing files, but causes the target ADU to go back up to 420, thus losing dynamic range. The reason for this is that the read noise is higher at this gain.

Thanks for all the explanation - I have a better grasp of how best to use this camera now

 

Unless I've totally missed the boat, the answer is *no* to your first one. The 16 bit multiplication is already built in, and the SGP value has bias included. I believe what you're aiming at in SGP is around 912 16-bit ADU. You can go higher of course, clipping some stars in the process, but you will be very over-exposed at 7200 ADU.

 

My calculation is based on target 16-bit ADU = 16* (20*[ReadNoise]/[Gain] + [Offset]) so for Unity (139) gain we have 16*(20*1.8/1 )+ 21) = 16*(57)= 912 ADU.  

[fnord123]

 

Thanks, that makes sense.  Testing my understanding, are any of the following statements incorrect assuming I am imaging at F/6 in a red zone?

• At gain 139, I should expect < 10s LRGB (minimum) exposures and aim for ADU 450.  Since I use SGP, I should aim for a median pixel value of (16 * 450) 7200.
• If I want to maximize dynamic range, I should use a gain of 50, 75, or 100, since the target ADU is 340.
• Using 0 gain allows me to use 30s minimum exposures, which means less processing files, but causes the target ADU to go back up to 420, thus losing dynamic range. The reason for this is that the read noise is higher at this gain.

Thanks for all the explanation - I have a better grasp of how best to use this camera now

Unless I've totally missed the boat, the answer is *no* to your first one. The 16 bit multiplication is already built in, and the SGP value has bias included. I believe what you're aiming at in SGP is around 912 16-bit ADU. You can go higher of course, clipping some stars in the process, but you will be very over-exposed at 7200 ADU.

 

My calculation is based on target 16-bit ADU = 16* (20*[ReadNoise]/[Gain] + [Offset]) so for Unity (139) gain we have 16*(20*1.8/1 )+ 21) = 16*(57)= 912 ADU.  

 

I thought SGP showed stuff in 16 bit, while the ASI1600mm camera ADUs are in 12 bit, so SGP values are left shifted 4 bits (x16) vs. the camera?

 

Just to make sure we are talking about the same thing - in the attached image, is the 1824 median value the one you are saying should be (for unity gain) 912 ADU (assuming no 16bit/12bit compensation needed) or 14,592 (if one was to multiply by 16, which I understand you don't recommend).

 

 

Also, the table for a red zone at the top of this thread is recommending 450ADU as a target for LRGB at gain 139 as opposed to 912ADU. I couldn't find the formula they used for it, so am unsure of the reason for the difference from your recommendation?

Edited by fnord123, 11 June 2017 - 01:48 PM.

[Stelios]

Yes. It's that median value I'm talking about. I don't know where he gets the bias information to calculate the 450 ADU, I suspect "offset" is bias offset, although then it should be less (21*16=336).  In other words, the reason his is less is because he subtracts the bias offset before arriving at a recommended value. 

 

There was a huge thread (search for "Am I exposed right") about this subject, and related sub-threads. The formula I'm using came from there. Jon Rista I think had the first version.

Edited by Stelios, 11 June 2017 - 04:07 PM.

[fnord123]

I guess that is a reference to this thread?

 

Excerpting some text of Jon's from there: "With the ASI1600, a 5000 DN background level would be about 313 in 12-bit ADU. Considering at unity that your FWC is only 4095 ADU, you would still be throwing away a ton of dynamic range with a 313 ADU background. To fully swamp the read noise at unity, you don't need more than a 35 ADU background sky level...which is 560 DN in 16-bit."  FWIW this is one of the things that made me assume the x16 multiplier to go from ADU to SGP numbers.

 

Edit: Reading your formula, I think I understand now - your formula has already multiplied by 16, converting the native 12 bit ADU to a 16 bit value sufficient for SGP.  Is that a correct interpretation?

Edited by fnord123, 11 June 2017 - 09:42 PM.

[Jon Rista]

He was using my formula, and yes, mine definitely includes bit depth conversion. Here is my original formula:
 
MinDN16 = (((ReadNoise * 20) / Gain) + BiasOffset) * 2^16/2^Bits
 
The relevant terms are bolded. Note that Bits is one of the relevant terms...that is your camera's ADC bit depth. In the case of the ASI1600 and QHY163, that bit depth is 12. For some other CMOS cameras, it can be 12 or 14, and with some that support hardware binning, it may even be 10. For most DSLRs it is 14. For most CCDs it is 16 (so the scale factor is 1). The scale factors between bit depths are:
 
 8->16 = 256x
10->16 = 64x
12->16 = 16x
14->16 = 4x
 
I broke my formula down once before, in another forum, but here it is again, for those who want to understand what it's doing:
 

The MinDN16 formula is mine. When you are converting a signal from electrons to 16-bit DN, you have to add the bias offset to the signal after it has been converted to ADU, since the offset is in ADU.

  

So, to break it down:

 

MinSignalADU = ((ReadNoise * 20) / Gain)

 

This takes the read noise of the camera at the specified gain setting, and multiplies it by 20. We multiply by 20 to determine a sufficient minimal signal level. This is the 20*RN rule. There is another rule...3*RN^2, which we could use instead:

 

MinSignalADU = ((ReadNoise^2 * 3) / Gain)

 

Now that you have converted the multiple of read noise to ADU via the gain setting, you still have to account for the offset that the ADC adds to the data. That is where the + BiasOffset comes from: 

 

MinSignalAndOffset = MinSignalADU + BiasOffset 

 

Once you have the signal and the bias offset in ADU, you can then convert that from the camera bit depth to 16-bit: 

 

MinDN16 = (MinSignalADU + BiasOffset) * 2^16/2^Bits

 

Edited by Jon Rista, 12 June 2017 - 12:31 AM.

[Shiraz]

The original tables were not computed using any of the shortcut formulas - they were based on results from a quite detailed end-to-end astrograph model and they show the effects of optics and sky brightness, as well as the camera

However, the "expected ADU above bias" results in the last columns are similar to those from the well established John Smith/Stan Moore formula (~10*RN*RN/gain)*16 and Jon Rista's approximation (20*RN/gain)*16. The exception seems to be that (as shown in the attached table) Jon's method tends to underestimate at low gain and overestimate at high gain, compared to the other methods.

My camera does not produce bias levels that can be reliably predicted from offset*16, so the expected ADU values given in the tables do not include the bias offset - recommend that the actual bias ADU be measured for the camera (at chosen gain, offset, temp) and added to the expected ADU value.

fnord123, don't sweat on getting it exactly right - the ADU values (16bit) in the tables are only based on the the ZWO indicative data - plus, there is some discussion over what the appropriate read noise effect should be. The worst that can happen is that, if you use subs that are too short, you will take longer overall to get to any given picture quality or, if you use excessively long subs, you will lose some dynamic range in your final stack.

regards Ray

table shows approx expected ADU(16bit) values from the three prediction methods - need to add bias ADU to these values.

Edited by Shiraz, 12 June 2017 - 05:43 PM.