TheSkySearchers

Hi AP'ers.

As most of you know, I am moving from a dslr to a dedicated astronomy camera (ZWO ASI 533) for imaging.

I am trying to understand exposure time to get things right or close without saturating stars that require lots of post processing to mitigate.

I am an Astrophotography Tool user and connecting a dedicated astronomy camera brings up a whole new set of tools. One tool that I was just reading about is the Pixel aid.

I know that some of you use APT and others use N.I.N.A, SGP and other imaging applications. They probably all have similar tools.

The APT Pixel Aid enables the user to examine an exposure (the whole image or a single pixel) to determine the image ADU. I assume that the purpose is to evaluate the image as a whole and to evaluate star exposure.

Looking at the FW graphics for the camera below, I see that Full Well (e-) should be about 16000 shooting at 100 gain.
Looking at the GAIN graphic at 100, I see that the there is 1 e- for each ADU.

So, does that mean that shooting at 100 gain, I should have a full well of 16000 and should expose for a time that measures the ADU of bright stars at no greater than 16000?

I think that something is flawed with my thinking, because I loaded an image from a few nights ago where I mistakenly had the gain at 300. I measured the whole image with the Pixel Aid and it showed that my maximum image (as a whole) is 65000.

If I chose to measure a single pixel, the value hits 65000 when I place the crosshairs on a bright star center.

Is there something wrong about what I think that this means, or did I grossly overexpose the stars in the image?
As you can see in the image above, the Histogram in the lower left side of the screen looks good (to me).

I appreciate any advice/comments! Thanks!

Here you go Jim,

An ASI533 Sub exposure calculator courtesy of Steve Bellavia over at CN Cheers,
JT

JayTee wrote: ↑Tue Sep 29, 2020 12:06 am Here you go Jim,

An ASI533 Sub exposure calculator courtesy of Steve Bellavia over at CN
ZWO ASI533-20200929T000349Z-001.zip

Cheers,
JT

Dang!!! That took an amazing amount of work!!!!

I will choke on that spreadsheet for awhile JT!!!. Awesome amount of information!

Thanks!

Here is another calculator from the Gibraltar Astronomical Society (GAS). Here is the link that allows you to download the spreadsheet.
http://www.gibastrosoc.org/sections/ast ... calculator

Cheers,
JT

EDIT: I just updated the link to be the correct link.

JayTee wrote: ↑Tue Sep 29, 2020 12:43 am Here is another calculator from the Gibraltar Astronomical Society (GAS)
https://docs.google.com/spreadsheets/d/ ... 2049877418

Cheers,
JT

Thanks JT!

When you get a chance, can you grant me access to the sheet? Thanks!

Juno16 wrote: ↑Mon Sep 28, 2020 11:12 pm Looking at the GAIN graphic at 100, I see that the there is 1 e- for each ADU.

So, does that mean that shooting at 100 gain, I should have a full well of 16000 and should expose for a time that measures the ADU of bright stars at no greater than 16000?

I think that something is flawed with my thinking, because I loaded an image from a few nights ago where I mistakenly had the gain at 300. I measured the whole image with the Pixel Aid and it showed that my maximum image (as a whole) is 65000.

If I chose to measure a single pixel, the value hits 65000 when I place the crosshairs on a bright star center.

Is there something wrong about what I think that this means, or did I grossly overexpose the stars in the image?

There are three sets of numbers involved in the imaging system. First is the well depth which is how many electrons each pixel can store - 16000 in this case.
Then there is the Analogue to Digital Converter (ADC) bit depth which for the ASI533 is 14 bits which therefore has a maximum value of 16383. The output of the ADC is the number of electrons in the pixel divided by the gain (in e-/ADU) plus the offset and strictly speaking this is the ADU value. However, the driver usually normalizes the ADU value to a 16 bit number by multiplying by 4 in the case of a 14 bit ADC. This gives rise to the 65000 value you are seeing as 16 bits has a maximum value of 65535. This is what the imaging software sees and is also called the ADU value.

In terms of how long you should expose for, that depends on what you are trying to achieve. But assuming you are going to be stacking then you want the subs to be shot noise dominant rather than read noise dominant for the stacking to be most effective at reducing noise.That's where the various spreadsheets come in. You want the sky background to be 3 to 10 times the read noise squared which, depending on the gain, could cause some stars to saturate. If that is not desirable the first option is to reduce the gain and recalculate the exposure time.

There are two places where saturation can occur. First is on the pixel tself when it captures more electrons than it can store. The other is in the ADC when the gain is above unity so that when the electrons on the pixel are multiplied by gain, they exceed the maximum value of the ADC. This limits how much of the well depth of the pixel is usable.

KenGS wrote: ↑Tue Sep 29, 2020 3:15 am

There are three sets of numbers involved in the imaging system. First is the well depth which is how many electrons each pixel can store - 16000 in this case.
Then there is the Analogue to Digital Converter (ADC) bit depth which for the ASI533 is 14 bits which therefore has a maximum value of 16383. The output of the ADC is the number of electrons in the pixel divided by the gain (in e-/ADU) plus the offset and strictly speaking this is the ADU value. However, the driver usually normalizes the ADU value to a 16 bit number by multiplying by 4 in the case of a 14 bit ADC. This gives rise to the 65000 value you are seeing as 16 bits has a maximum value of 65535. This is what the imaging software sees and is also called the ADU value.

In terms of how long you should expose for, that depends on what you are trying to achieve. But assuming you are going to be stacking then you want the subs to be shot noise dominant rather than read noise dominant for the stacking to be most effective at reducing noise.That's where the various spreadsheets come in. You want the sky background to be 3 to 10 times the read noise squared which, depending on the gain, could cause some stars to saturate. If that is not desirable the first option is to reduce the gain and recalculate the exposure time.

There are two places where saturation can occur. First is on the pixel tself when it captures more electrons than it can store. The other is in the ADC when the gain is above unity so that when the electrons on the pixel are multiplied by gain, they exceed the maximum value of the ADC. This limits how much of the well depth of the pixel is usable.

Hi Ken and thanks for the explanation. You answered some questions that I had, but couldn’t word very well. One was normalizing the ADU by the driver going from 14 bits to 16 bits. Makes sense for sure and I had read somewhere that the conversion was going on, but I didn’t know where. Thats where the 65000 single pixel star ADU comes from. Thanks.

I played around with the spreadsheet that JT sent me and using the APT Pixel Aid, I entered values from the rocky session the other night. Using the single pixel mode to get a rough idea of the sky glow background from a light frame (6000 ADU), dark signal from a gain mismatched dark frame (2800 ADU), and bias signal from a dark flat frame (2800 ADU), with an exposure duration of 120 seconds. The spreadsheet calculates my sub exposure time at a gain of 100 to be 0.23 sec with an N of 1022. Clearly I am way off somewhere, but that is another issue.

Is the histogram useful in any way to get a rough idea of exposure time? I know that is very rough, but in the image in the above post, the histogram range is roughly 4000-65000 and the peaks are in the first 1/4 of the left side of the graph. To me, the range shows that the sky background to be around 4000 ADU and the saturated stars to be 65000 ADU.

I know that this is a super simplistic approach, but the spreadsheet has me a bit twisted.

Thanks again for the informative response!

Hi Jim
A slight correction to what I said. The full well of the ASI533 is 50,0000 (not 16,000) so each pixel can store up to 50,000 electrons. But at unity gain the usable well depth is still 16000 due to the ADC.
Your bias values seem odd at 2800 ADU as this would correspond to an offset of 700 whereas I understand the default offset to be 50 for this camera. That would give a bias of 200. That would partly explain why the spreadsheet gives the values you got. Also, 120s exposures in Bortle 7/8 skies would almost certainly be way too much. At unity gain with the ASI1600MM with RGB filters I'm using 30s expsoures in Bortle 5 sky. You will be using up valuable dynamic range which is critical in LP skies. FYI Dynamic range is the usable well depth divided by read noise and is pretty much optimal at unity gain on the 533. You can get a bit more at 0 gain but not much.
Assuming an offset of 50, the desired 16bit sky background you would want for 10 times RN squared would be about 300 at unity gain (the 300 being the 16 bit value). Read noise at unity is 1.5e- so [10 * (1.5)^2 + 50]*4 = 290

You can use the histogram BUT caution:
1) I understand the DSLR histogram has a log scale so the usual advice to set the peak at 1/4 or 1/3 does not apply on a linear scale
2) Read noise from astro cameras is often much less than a DSLR so you can push the peak further to the left

From what I calculated above, on a linear scale of 0-16000 the desired value for the peak of 290 would be 1/55 from the left !

To work out the exposure, I take a sample sub of, say 30s, and check the median value. The offset contributes a fixed amount of 200 to the desired value so only 90 is down to expsoure time. So if, say, the median was 240 that means the exposure is about 40/90 what you want so doubling the exposure to 60s would get you to about the right value. Its not necessary to be overly precise as folks use anything from 3x to 10x RN squared. But generally I aim to be a bit higher rather than below the desired value.
If instead the median was 500 then the desired expsoure time would be 90/210 x 30s or about 13s (or round up to 15s)

Note that for all these calculation, knowing the camera offset is important. Like I said, I think ZWO defaults to 50 and typically they set it to 7x the maximum read noise at any gain value. On the 533 the maximum read noise is a bit less than 4.0e- at zero gain which would give an offst of 30 or so by that calculation. So I would expect the 533 offset to be in that ball park and certainly not 700.

KenGS wrote: ↑Tue Sep 29, 2020 12:29 pm Hi Jim
A slight correction to what I said. The full well of the ASI533 is 50,0000 (not 16,000) so each pixel can store up to 50,000 electrons. But at unity gain the usable well depth is still 16000 due to the ADC.
Your bias values seem odd at 2800 ADU as this would correspond to an offset of 700 whereas I understand the default offset to be 50 for this camera. That would give a bias of 200. That would partly explain why the spreadsheet gives the values you got. Also, 120s exposures in Bortle 7/8 skies would almost certainly be way too much. At unity gain with the ASI1600MM with RGB filters I'm using 30s expsoures in Bortle 5 sky. You will be using up valuable dynamic range which is critical in LP skies. FYI Dynamic range is the usable well depth divided by read noise and is pretty much optimal at unity gain on the 533. You can get a bit more at 0 gain but not much.
Assuming an offset of 50, the desired 16bit sky background you would want for 10 times RN squared would be about 300 at unity gain (the 300 being the 16 bit value). Read noise at unity is 1.5e- so [10 * (1.5)^2 + 50]*4 = 290

Hi Ken,

Yes, the offset is 70 in the ascom driver. Very close to your estimation. I do plan to use unity gain and leave the offset at 70.

KenGS wrote: ↑Tue Sep 29, 2020 12:29 pm
You can use the histogram BUT caution:
1) I understand the DSLR histogram has a log scale so the usual advice to set the peak at 1/4 or 1/3 does not apply on a linear scale
2) Read noise from astro cameras is often much less than a DSLR so you can push the peak further to the left

From what I calculated above, on a linear scale of 0-16000 the desired value for the peak of 290 would be 1/55 from the left !

Yes, I see what you mean. The histogram does not work as it does for the dslr. Definately not 1/4 to 1/3 from the left!

KenGS wrote: ↑Tue Sep 29, 2020 12:29 pm
To work out the exposure, I take a sample sub of, say 30s, and check the median value. The offset contributes a fixed amount of 200 to the desired value so only 90 is down to expsoure time. So if, say, the median was 240 that means the exposure is about 40/90 what you want so doubling the exposure to 60s would get you to about the right value. Its not necessary to be overly precise as folks use anything from 3x to 10x RN squared. But generally I aim to be a bit higher rather than below the desired value.
If instead the median was 500 then the desired expsoure time would be 90/210 x 30s or about 13s (or round up to 15s)

Note that for all these calculation, knowing the camera offset is important. Like I said, I think ZWO defaults to 50 and typically they set it to 7x the maximum read noise at any gain value. On the 533 the maximum read noise is a bit less than 4.0e- at zero gain which would give an offst of 30 or so by that calculation. So I would expect the 533 offset to be in that ball park and certainly not 700.

Excellent! Great recommendations that give me something to start with. I really appreciate you taking the time here to help out. Thank you Ken!

[mention]Juno16[/mention] and [mention]Graeme1858[/mention]
Sorry guys. I have corrected my earlier post so the correct link is displayed. No permission needed.

JT

JayTee wrote: ↑Tue Sep 29, 2020 8:32 pm @Juno16 and @Graeme1858
Sorry guys. I have corrected my earlier post so the correct link is displayed. No permission needed.

JT

Thanks JT.

Unfortunately, I am a bit snowed in by this and I will just play around tonight. The near full moon doesn't help, but I have the L Enhance installed and hopefully, that will reduce the moon glow.

I am just going to shoot for much lower exposure and see how things work out. Absolutely not a good night to image with an OSC, but I don't get many, so might as well play around some.

Hopefully, I will get up to speed soon and be able to put those spreadsheets to use.

Thanks again!

Remember, as KenGS said it really depends on the target you're imaging. Take the Horsehead, for example, it's a dark nebula sitting right next to a really bright star. You don't have much choice there but to allow Alnitak to get "blown out" if you even hope to capture the Horsehead.

And don't forget that [mention]STEVE333[/mention] showed us a method to reduce star size in PI. So, theoretically, you could solve most if not all your blown-out stars in post-processing so you don't need to intentionally deny yourself a greater signal depth on your object of interest just to avoid achieving full well depth from the brighter stars.

Cheers,
JT

JayTee wrote: ↑Tue Sep 29, 2020 9:31 pm Remember, as KenGS said it really depends on the target you're imaging. Take the Horsehead, for example, it's a dark nebula sitting right next to a really bright star. You don't have much choice there but to allow Alnitak to get "blown out" if you even hope to capture the Horsehead.

Cheers,
JT

Thanks JT.

Yeah, and it doesn't help with the moon in the way!

If I understand Ken correctly, I should set the exposure time so that my light frame median ADU is much, much lower than my last light frames at 120 seconds that had a median ADU of 14k. I will just mess around and see what goes!

Thanks!

Juno16 wrote: ↑Tue Sep 29, 2020 9:34 pm I will just mess around and see what goes!

That my modus operandi!

Juno16 wrote: ↑Tue Sep 29, 2020 9:34 pm

JayTee wrote: ↑Tue Sep 29, 2020 9:31 pm Remember, as KenGS said it really depends on the target you're imaging. Take the Horsehead, for example, it's a dark nebula sitting right next to a really bright star. You don't have much choice there but to allow Alnitak to get "blown out" if you even hope to capture the Horsehead.

Cheers,
JT

Thanks JT.

Yeah, and it doesn't help with the moon in the way!

If I understand Ken correctly, I should set the exposure time so that my light frame median ADU is much, much lower than my last light frames at 120 seconds that had a median ADU of 14k. I will just mess around and see what goes!

Thanks!

Yes it will be much lower than 14k!
Remember that the sky background will change significantly from full moon to no moon.
As another member of "just mess around and see" camp I take a preview sub at the start of each session to check the ADU level and adjust exposure if necessary.

KenGS wrote: ↑Tue Sep 29, 2020 11:01 pm

Yes it will be much lower than 14k!
Remember that the sky background will change significantly from full moon to no moon.
As another member of "just mess around and see" camp I take a preview sub at the start of each session to check the ADU level and adjust exposure if necessary.

Thanks again for all of your help Ken. You definitely helped me understand, but I have some way to go.

If anyone is interested, I will post some image links with data tomorrow or when I can. Just in case you or someone else might want to comment. Of course, this will only be an exposure test as the near full moon will prevent any slightly serious attempt. That is okay though. I need to "play" around with things.

I do have a L Enhance filter installed and not sure if it will help, but definitely shouldn't hurt. The filter isolates H-alpha, H-beta, and Oxygen iii Nebula Emission Lines
The bandwidths are 24 nm for OIII and H-beta, and 10 nm for Ha.

Thanks!

Hey Jim,

I notice you have SharpCap. Have you ever tried using the SmartHistogram (Brain) function? I've found it helps me get a ballpark idea of my exposure length on any given night. I'm not sure if the 533 is included in the default data set yet so you may need to run the sensor analysis.

--Steve

starfield wrote: ↑Tue Sep 29, 2020 11:46 pm Hey Jim,

I notice you have SharpCap. Have you ever tried using the SmartHistogram (Brain) function? I've found it helps me get a ballpark idea of my exposure length on any given night. I'm not sure if the 533 is included in the default data set yet so you may need to run the sensor analysis.

--Steve

I do have Sharpcap and I looked at the SmartHistogram in the Help section today. I have only used Sharpcap for PA and really like it. I plan to give it a try tonight just see what develops.

I will look into the sensor analysis. I appreciate the tip!

Not sure what to expect with the moonlight and the L enhance, but I have the clear night, so might as well try things.

Thanks Steve!

Juno16 wrote: ↑Wed Sep 30, 2020 12:20 am
Not sure what to expect with the moonlight and the L enhance, but I have the clear night, so might as well try things.

Thanks Steve!

If there is some HA, guarantee you will see it with the L-enhance filter regardless of the Moon being out.

Point it at the Veil Nebula!

Mac wrote: ↑Wed Sep 30, 2020 12:22 am

Juno16 wrote: ↑Wed Sep 30, 2020 12:20 am
Not sure what to expect with the moonlight and the L enhance, but I have the clear night, so might as well try things.

Thanks Steve!

If there is some HA, guarantee you will see it with the L-enhance filter regardless of the Moon being out.

Point it at the Veil Nebula!

Like this?
This is a 60 second exposure at a gain of 104 and offset of 70. Median ADU is 544.

A 30 second image at the same settings has a median ADU of 470.

A 120 second image with the same settings has a median ADU of 696.

Not sure what all of this means, but with the near full moon, it is amazing that the ADU is so much lower tonight than a few nights ago when I used a gain of 300 and the IDAS LPS D-1 filter and a median ADU of 14000!

I am going to shoot some light frames at 60 seconds and stack them up tomorrow (only because I have the darks).

Very interesting! I will post back tomorrow.

Thanks all!