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

"rookie" question about gain

  • Please log in to reply
11 replies to this topic

#1 Stargazerman

Stargazerman

    Sputnik

  • -----
  • topic starter
  • Posts: 31
  • Joined: 17 Jul 2020
  • Loc: Mid Michigan

Posted 22 October 2021 - 04:13 PM

Hi all

 

list of my equipment, avx mount 6" newt scope an asi294mc pro camera baader markIII CC and a eaf auto focuser.

my question how to figure out what gain to use on my DSO targets. I have tried but not much luck with higher or lower gain ! By the way I am guiding too. To run my setup I'm using cpwi,sharpcap pro and phd2. Is the gain just trial and error or is there a formula. I'm a rookie at this ! I have had luck at 300 gain but my pics are poor and i'm not good at processing yet.

 

thanks in advance

Ken

 



#2 jgreif

jgreif

    Explorer 1

  • -----
  • Posts: 77
  • Joined: 09 Jun 2016

Posted 22 October 2021 - 04:40 PM

Gain vs Read Noise.jpg

my question how to figure out what gain to use on my DSO targets. I have tried but not much luck with higher or lower gain ! By the way I am guiding too. To run my setup I'm using cpwi,sharpcap pro and phd2. Is the gain just trial and error or is there a formula. I'm a rookie at this ! I have had luck at 300 gain but my pics are poor and i'm not good at processing yet.

Most camera manuals will have charts showing optimal gain (vs noise, for example).  Here's the chart for my camera, ZWO ASI533MC Pro.  Notice the incredible dip in noise at Gain 100.  That is the optimal gain, and what I use for imaging.  Hope that helps.


  • Sylon likes this

#3 JCDAstro

JCDAstro

    Vostok 1

  • -----
  • Posts: 106
  • Joined: 25 Mar 2021

Posted 22 October 2021 - 07:39 PM

I think it is 117-120 gain for the 294, that is where the dynamic range and the read noise have the best points in the curves.

Higher gains would only be useful for luck imaging on targets using short exposures. Lower gains would produce more well depth, but you pay a read noise penalty so you would need long exposures to "swamp the noise" if you search that term you will find several good posts on this topic here on CN.
  • jgreif and OldManSky like this

#4 sbharrat

sbharrat

    Apollo

  • -----
  • Posts: 1,203
  • Joined: 28 Nov 2020
  • Loc: NJ, USA

Posted 23 October 2021 - 08:18 AM

attachicon.gifGain vs Read Noise.jpg

Most camera manuals will have charts showing optimal gain (vs noise, for example).  Here's the chart for my camera, ZWO ASI533MC Pro.  Notice the incredible dip in noise at Gain 100.  That is the optimal gain, and what I use for imaging.  Hope that helps.

Curious. Anyone know *why* these charts look like this? What is the reason for the sharp transition? Is is something like the hardware itself supports from 100 gain on and below that it is just software multiply? What causes this step function?


  • jonnybravo0311 likes this

#5 jgreif

jgreif

    Explorer 1

  • -----
  • Posts: 77
  • Joined: 09 Jun 2016

Posted 23 October 2021 - 08:26 AM

Curious. Anyone know *why* these charts look like this? What is the reason for the sharp transition? Is is something like the hardware itself supports from 100 gain on and below that it is just software multiply? What causes this step function?

Truth is, I've never thought about "why?"  Some cameras I've owned aren't quite that sudden in drop, but most have a clear inflection beyond which additional gain doesn't provide much benefit.  Perhaps another CN'er knows.


Edited by jgreif, 23 October 2021 - 10:26 AM.


#6 unimatrix0

unimatrix0

    Apollo

  • -----
  • Posts: 1,109
  • Joined: 03 Jan 2021

Posted 23 October 2021 - 09:02 AM

Curious. Anyone know *why* these charts look like this? What is the reason for the sharp transition? Is is something like the hardware itself supports from 100 gain on and below that it is just software multiply? What causes this step function?

It's the chip characteristics, it is optimized to a certain level of gain by the manufacturer.  Not all sensor readouts like this, many has a more linear chart with no optimal gain settings.  It's really the same thing with DSLRs, they just don't go into details for the regular customers, because they aren't interested or care, but it's there. 
Think of it, like cars with RPM.  Certain cars will consume the least amount of gas and still produce plenty of power at certain RPM, they are optimized for - example -  2500RPM.  The transmission will either upshift or downshift to try to keep the engine running in that range. 
At least, that's how I look at it. 

 

Here is my qhy183C characteristics. Note at unity gain at 11, although I never use it, because the "bump" isn't really giving me any huge advantage in dynamic range increase or anything, it's just that gain settings, where it's 1e/ADU

and I never use that gain settings, because there is hardly any advantage using it, but more like oversaturating my stars too quickly. I use gain 10, if I'm taking like less than 1 minute exposures (sometimes it's like 30 second or 20 second exposures)  but with narrowband, my best settings were around gain 3 or 4 and 5 or gain 0, then I can I can take up to 240 second shots, depending on my sky quality, but I usually just go for 180 seconds still, because I'm still swamping the readout noise.  I just follow what sharpcap recommends, but I only deviate a bit from it, knowing that very small offset it recommends creates more horizontal banding. These stuff is something you can get to know from trial and error. 




 

Attached Thumbnails

  • qhy183C gain.JPG

Edited by unimatrix0, 23 October 2021 - 09:18 AM.

  • sbharrat likes this

#7 jonnybravo0311

jonnybravo0311

    Mercury-Atlas

  • *****
  • Posts: 2,941
  • Joined: 05 Nov 2020
  • Loc: NJ, US

Posted 23 October 2021 - 09:39 AM

Curious. Anyone know *why* these charts look like this? What is the reason for the sharp transition? Is is something like the hardware itself supports from 100 gain on and below that it is just software multiply? What causes this step function?

It's the point at which the sensor switches from low conversion gain to high conversion gain. In terrestrial photography, it's referred to as "dual ISO".


  • sbharrat likes this

#8 Stargazerman

Stargazerman

    Sputnik

  • -----
  • topic starter
  • Posts: 31
  • Joined: 17 Jul 2020
  • Loc: Mid Michigan

Posted 23 October 2021 - 09:45 AM

Thanks all for the information. Time for some research.



#9 sbharrat

sbharrat

    Apollo

  • -----
  • Posts: 1,203
  • Joined: 28 Nov 2020
  • Loc: NJ, USA

Posted 23 October 2021 - 09:45 AM

It's the point at which the sensor switches from low conversion gain to high conversion gain. In terrestrial photography, it's referred to as "dual ISO".

Well... I still didn't know what this meant. But it was a good starting point for Googling! Thanks!

 

> The High Conversion Gain (HCG) allows the acquisition of images with better quality under low light conditions. This is achieved by increasing the conversion gain of the pixels while keeping the noise contribution of the analog circuit constant.


  • jonnybravo0311 likes this

#10 JCDAstro

JCDAstro

    Vostok 1

  • -----
  • Posts: 106
  • Joined: 25 Mar 2021

Posted 23 October 2021 - 12:13 PM

Don't quote me, but this is how I understand it to work. There are several types of noise that contribute to a signal, but much of the noise floor is due to the read out circuit itself. That is, if one develops a pixel that has a higher signal level in it before the readout, then the resulting readout of the signal will have a better SNR. These modern CMOS sensors have extra mosfets or other amplification means directly inside a pixel that increase the conversion gain of a pixel at the cost of dynamic range. The loss of dynamic range is due to it being necessary to limit the full well capacity in order to not exceed the available voltage supplied to the sensor.

So no free lunch. And designers need to limit the power consumption of these devices to make them reasonable to cool. All these trades make the curves look the way they do.

#11 jdupton

jdupton

    Soyuz

  • *****
  • Posts: 3,817
  • Joined: 21 Nov 2010
  • Loc: Central Texas, USA

Posted 23 October 2021 - 01:29 PM

Shaun,
 

Curious. Anyone know *why* these charts look like this? What is the reason for the sharp transition? Is is something like the hardware itself supports from 100 gain on and below that it is just software multiply? What causes this step function?

   Below is an explanation from Jon Rista which covers HCG and LCG modes. The discontinuity is where the sensor switches from LCG to HCG.
 

This is "conversion" gain. There are different stages of readout in a CMOS sensor, and gain applied at more than one point in the process. 
 
The gain you usually adjust is the amplifier gain. Amplifiers (source followers in CMOS sensors, usually) have what is called input-referred noise. That means the noise of the amp itself is amplified, along with the input signal. Now, the amp noise is usually very very low, significantly lower than ADC noise, so the fact that this noise is amplified doesn't matter until you reach higher gains, where the noise curve flattens out. (Note that when the noise curve flattens out, that is the point at which input-referred amp noise has been amplified enough to swamp the ADC noise...meaning the amp noise is much larger than the ADC noise, in relative terms.) 
 
The amp gain is a second gain. There is actually a gain that occurs before that...that is the conversion gain. Conversion is the process of converting a collection of electron charge into a voltage. That conversion can be done such that each electron represents a small amount of voltage...or, a large amount of voltage, or maybe you could have a range of possible discrete conversion gains that you could select from, converting each electron into some discrete amount of voltage.
 
In any case, a low conversion gain means each electron represents a smaller amount of voltage, whereas a high conversion gain means each electron represents a larger amount of voltage. Most sensors just have a single conversion gain, and it would be tuned to provide the best possible dynamic range AND SNR at lower gains (which is usually most important for digital cameras intended to be used in situations with plenty of light and the need for a lot of dynamic range). Remember that SNR in a single frame is better when you have more total signal. For digital cameras, this means lower gains, which with the kind of light we usually have in most terrestrial photography situations, are capable of handling significantly larger signals than we usually deal with in astrophotography. 
 
Cameras with two conversion gains, LCG and HCG, usually have an amplifier gain setting at which the conversion gain changes. A higher conversion gain means that fewer electrons can represent a larger signal (in terms of voltage), which means that the input signal to the amplifier starts out larger, meaning that you need less of that kind of gain, which also amplifies the input referred noise. That in turn means you have less read noise, for any given amount of electron charge. That, of course, means for any given amount of charge (real signal), you get more out of the electron capacity you have, which means you get more dynamic range. 
 
Something else that is important to note here, though, is that dynamic range is not SNR. For the best single-frame SNR, you still would want to use a lower gain. Even if the dynamic range is the same at say minimum gain with LCG as it is at the HCG gain, the fact that you could have a lot more signal (charge) at minimum gain means the frame will have a higher SNR. Now, more capacity for more charge means you will need longer exposures as well. For daytime photographers, this might mean a fraction of a second longer. but for astrophotographers it could mean minutes, even tens of minutes longer.
 
Now, if you have the same amount of dynamic range, then in the end you can achieve the same SNR with either minimum gain, or the HCG gain. Each frame could contain the same range of signal values. You would just have to stack fewer low gain frames, or more high gain frames. But in the end, you should be able to achieve the same result at either gain setting. This is a key difference with astrophotography, vs. your average brightly lit terrestrial photography. With astrophotography we stack, which is a great normalizer and can neutralize a lot of the differences between gains. There is usually some kind of tradeoff...one of them being sub length vs. sub count, and there are pros and cons to both of those.

 

 

John


Edited by jdupton, 23 October 2021 - 01:32 PM.

  • sbharrat likes this

#12 Jared

Jared

    Fly Me to the Moon

  • *****
  • Moderators
  • Posts: 7,007
  • Joined: 11 Oct 2005
  • Loc: Piedmont, California, U.S.

Posted 23 October 2021 - 01:53 PM

One important point, even for cameras that have dual conversion gain...

 

The crossover point is not necessarily the "optimal" point for shooting your exposures. It is often a good compromise between dynamic range and read noise, but that's it. For example, with my camera I get the largest dynamic range with a gain of zero. The read noise is around 3.5e- and the full well capacity is around 50,000e-. That works really well for me with broadband targets as, even under dark skies, it's not hard to swamp 3.5e- of read noise with relatively short three minute exposures. The story is a bit different for narrow band, though. By raising the gain to 56 on my camera (it will vary from one camera to the next) the read noise drops to about 1.6e-. With narrowband, getting the read noise low is much more important than the overall dynamic range since its unlikely one will be taking sub exposures long enough for read noise to become immaterial. So even though there is a slight hit in dynamic range due to the lower full well capacity, I generally choose this value for narrowband.

 

There is no single "optimal" value. It will depend on your skies, how long your subs are, what object you are shooting, whether you care about the total number of subs (processing power), etc.. For most people most of the time you either want the largest dynamic range possible (so you won't clip too many stars) or the lowest read noise that still has reasonable full well capacity (for narrowband). If you happen to have a dual gain chip, the point at which the read noise seems to suddenly drop in the manufacturer's graphs isn't a bad place to start, but to say it's the "optimal" value is misleading. It's rarely where you get the most dynamic range. 

 

Remember that the quantum efficiency of the chip--the odds that a given photon will be registered--does not vary with gain. All you are doing with gain is trading dynamic range for read noise.


  • sbharrat 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