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

Sensor summary table

  • Please log in to reply
11 replies to this topic

#1 jimthompson

jimthompson

    Apollo

  • *****
  • topic starter
  • Posts: 1087
  • Joined: 06 Oct 2009
  • Loc: Ottawa, Canada

Posted 27 October 2016 - 04:35 PM

Hi All,

 

Throughout my journey through EAA I have been maintaining a spreadsheet that summarizes what data I have been able to find on the various sensors that are being used.  I recently updated it as part of another discussion on another group, and thought some might find it useful here.  Note that where there is a blank entry in my table I was not able to find any data.  If you are able to fill in a blank for me, please do!

 

Of special interest to me was the way in which sensitivity is quoted.  Sony is a little sneaky in this regard.  I have added a column labelled "Effective Pixel Average Sensitivity" that attempts to correct for all the differences in how sensitivity is quoted in the available materials (ie. thickness of IR cut filter, fratio of measurement lens, accounting for Bayer colour matrix).

 

cheers,

 

Jim T.

Attached Thumbnails

  • camera sensor summary_27Oct2016.png

  • Ptarmigan, nytecam, barbarosa and 9 others like this

#2 Astrojedi

Astrojedi

    Soyuz

  • *****
  • Posts: 3760
  • Joined: 27 May 2015
  • Loc: SoCal

Posted 27 October 2016 - 09:04 PM

Hi All,

 

Throughout my journey through EAA I have been maintaining a spreadsheet that summarizes what data I have been able to find on the various sensors that are being used.  I recently updated it as part of another discussion on another group, and thought some might find it useful here.  Note that where there is a blank entry in my table I was not able to find any data.  If you are able to fill in a blank for me, please do!

 

Of special interest to me was the way in which sensitivity is quoted.  Sony is a little sneaky in this regard.  I have added a column labelled "Effective Pixel Average Sensitivity" that attempts to correct for all the differences in how sensitivity is quoted in the available materials (ie. thickness of IR cut filter, fratio of measurement lens, accounting for Bayer colour matrix).

 

cheers,

 

Jim T.

Very useful. Thanks for posting. You should also normalize the sensitivity for pixel size or mv/micron as that is a more accurate measure of sensor sensitivity.

 

Also many of your sensitivity numbers look off. Can you explain your normalization methodology a bit more. Note that the sensitivity number for the 224 sensor is 2350mv (F5.6 Green signal 1/30s) and 825 is 2000mv (F5.6 Green signal 1/30s). The 285 is 1240mv again for the same measure.


Edited by Astrojedi, 27 October 2016 - 09:05 PM.

  • 42itous1 likes this

#3 Astrojedi

Astrojedi

    Soyuz

  • *****
  • Posts: 3760
  • Joined: 27 May 2015
  • Loc: SoCal

Posted 27 October 2016 - 09:26 PM

You should also add a column for dynamic range (fullwell / read noise). More useful than stating full well alone which is a little meaningless without taking into account read noise. Ultimately what you care about is dynamic range.

 

For CMOS sensors you can also list max frame rate at a stated resolution. Very useful and a unique advantage of the CMOS architecture (I.e. faster readout does not mean higher read noise like CCDs)


  • DSO_Viewer and 42itous1 like this

#4 jimthompson

jimthompson

    Apollo

  • *****
  • topic starter
  • Posts: 1087
  • Joined: 06 Oct 2009
  • Loc: Ottawa, Canada

Posted 28 October 2016 - 10:44 AM

Hi Hiten,

 

I believe the 825 and 285 sensitivities in my table match what you have said.  The 2350mV value quoted for the IMX224 is with the HCG (high conversion gain) mode on.  In normal mode the gain is half or 1175mV as shown in my table.  Do you happen to know at what gain setting this sensor switches into HCG mode?  Also, I understand that the well depth is affected by the HCG mode.  Do you have an idea by how much?  Anyway, I'll add a note about this in the table at the next revision.

 

The two main corrections to the sensitivity values are:

 

1.  Monochrome sensor sensitivity is measured using f/8 optics but the colour sensors are tested at f/5.6.  This is a marketing decision by Sony so that their colour sensors don't look so bad compared to the monochrome version.  To be able to fairly compare the sensitivity values you need to compensate for the different fratios, roughly a factor of 2.

2.  RGBG sensors have their sensitivity quoted as that of the green channel only.  To get an average sensitivity of the sensor you have to use the R:G and B:G sensitivity ratios to calc. the sensitivity of R and B pixels, and average that with the green pixels (average = (R + G + B + G)/4).  For the older sensors there is Sony data available for the R:G and B:G ratios, but nothing for the newer sensors esp. CMOS.  Where I was missing the R:G/B:G data I assumed typical values from other RGBG sensors, which works out to a correction of 0.75.

 

A number of years ago I looked at the spectral response plots for several sensors and did my own calculation of sensitivity from first principals.  One interesting outcome of this work was that in general RGBG one-shot colour sensors are on the order of 1/3 the sensitivity of the corresponding monochrome sensor.  Similarly CMYG sensors are roughly 1/2 the sensitivity of the corresponding monochrome sensor.  These rules-of-thumb are consistent with the Effective Pixel Average Sensitivity values in my table.

 

The sensitivity per unit pixel area is easy enough to do, I can add that to the table.  I can do the same for Max SNR as you suggest (well depth/read noise).  The max fps is also heavily camera dependant, plus there is only data for the CMOS sensors so I think I'll leave that one out.

 

cheers,

 

Jim T.


  • 42itous1 likes this

#5 Astrojedi

Astrojedi

    Soyuz

  • *****
  • Posts: 3760
  • Joined: 27 May 2015
  • Loc: SoCal

Posted 28 October 2016 - 12:15 PM

Jim,

 

You should not discount the HCG mode. That is a function of the sensor architecture which enables higher sensitivity. Nothing 'fake' about it. It allows the realization of the full intrinsic sensitivity of the low noise sensor.

 

Note that the frame of reference you are using is very CCD centric which is why you are calling the non-HCG mode 'normal'. In reality there is nothing 'abnormal' about HCG mode in CMOS. CCDs just don't have this capability.

 

Also where did you see that 2350mv sensitivity is in HCG mode? Is it in the Sony materials?

 

I think ZWO's website has some analysis on well depth. Also note as I mentioned earlier well depth alone is not meaningful. You need to see it in context of the read noise to understand dynamic range produced.

 

Ultimately what matters is dynamic range. CCDs require much larger full wells to produce the same dynamic due to the much higher read noise.

 

Hiten


Edited by Astrojedi, 28 October 2016 - 12:57 PM.


#6 mega256

mega256

    Surveyor 1

  • *****
  • Posts: 1540
  • Joined: 10 May 2007
  • Loc: N of Tampa

Posted 28 October 2016 - 12:25 PM

Thanks Jim....very nice to have this chart...



Bob

#7 Relativist

Relativist

    Cosmos

  • *****
  • Posts: 8151
  • Joined: 11 Oct 2003
  • Loc: OC, CA, USA

Posted 28 October 2016 - 01:19 PM

I agree with Hiten. For example with the A7s I always use an ISO higher than 2000 which puts the camera in the 'HCG' mode. In fact if the chart is for cameras to use as part of an electrotelescope then those values like sensitivity while in 'HCG' mode are preferable over settings an APer might use.



#8 jimthompson

jimthompson

    Apollo

  • *****
  • topic starter
  • Posts: 1087
  • Joined: 06 Oct 2009
  • Loc: Ottawa, Canada

Posted 28 October 2016 - 02:02 PM

Jim,

 

You should not discount the HCG mode. That is a function of the sensor architecture which enables higher sensitivity. Nothing 'fake' about it. It allows the realization of the full intrinsic sensitivity of the low noise sensor.

 

Note that the frame of reference you are using is very CCD centric which is why you are calling the non-HCG mode 'normal'. In reality there is nothing 'abnormal' about HCG mode in CMOS. CCDs just don't have this capability.

 

Also where did you see that 2350mv sensitivity is in HCG mode? Is it in the Sony materials?

 

I think ZWO's website has some analysis on well depth. Also note as I mentioned earlier well depth alone is not meaningful. You need to see it in context of the read noise to understand dynamic range produced.

 

Ultimately what matters is dynamic range. CCDs require much larger full wells to produce the same dynamic due to the much higher read noise.

 

Hiten

Hi Hiten,

 

From what I have been able to find, only the IMX224 sensor is confirmed to have HCG mode out of all the Sony sensors in my list.  That to me means that HCG mode is not "normal", it is a special feature of that sensor.  I have found some discussions online about the IMX290 also having HCG mode but have not been able to confirm it.  I think it is presumptuous to assume that everyone will use this feature all the time.  For solar system imaging for example I would prefer the lower noise and larger SNR of non-HCG mode (note that read noise is not the only source of noise in the image).  Do I understand correctly that in HCG mode each pixel is generating twice as many e- for every photon, thus filling up the charge well in half the time?  I know read noise is slightly lower in HCG mode, but is not the end result still a reduction in SNR?  Agreed, stacking can help to overcome this but in turn you are now committing to using stacking when HCG mode is on.

 

Cheers,

 

Jim T.


Edited by jimthompson, 28 October 2016 - 02:04 PM.


#9 Relativist

Relativist

    Cosmos

  • *****
  • Posts: 8151
  • Joined: 11 Oct 2003
  • Loc: OC, CA, USA

Posted 28 October 2016 - 03:01 PM

Jim, take a look at the linked graph here for the 224 and here for the 290. The DR and read noise graph has a jump where the HCG comes into effect. The other similar graphs for other cameras will tell you where that point is. This is also one of the fundamental advantages to going to these CMOS sensors for EAA. While the DR is imacted with higher gain that has little to do with the actual use.

 

We have a good start here, to all be on the same page reading from here for a bit might put us on the same page. You'll note that they quickly come to the conclusion that the A7s, because of larger pixel size, Higher QE and HCG mode, can have equivalent read noise to the compared cameras that don't have those features at much shorter exposures. In terms of DR, were doing EAA not AP, so giving up things like DR and full resolution for exposure speed is a common compromise, so we often bin and push up gain whereas the AP won't.



#10 Astrojedi

Astrojedi

    Soyuz

  • *****
  • Posts: 3760
  • Joined: 27 May 2015
  • Loc: SoCal

Posted 28 October 2016 - 03:54 PM

 

Jim,

 

You should not discount the HCG mode. That is a function of the sensor architecture which enables higher sensitivity. Nothing 'fake' about it. It allows the realization of the full intrinsic sensitivity of the low noise sensor.

 

Note that the frame of reference you are using is very CCD centric which is why you are calling the non-HCG mode 'normal'. In reality there is nothing 'abnormal' about HCG mode in CMOS. CCDs just don't have this capability.

 

Also where did you see that 2350mv sensitivity is in HCG mode? Is it in the Sony materials?

 

I think ZWO's website has some analysis on well depth. Also note as I mentioned earlier well depth alone is not meaningful. You need to see it in context of the read noise to understand dynamic range produced.

 

Ultimately what matters is dynamic range. CCDs require much larger full wells to produce the same dynamic due to the much higher read noise.

 

Hiten

Hi Hiten,

 

From what I have been able to find, only the IMX224 sensor is confirmed to have HCG mode out of all the Sony sensors in my list.  That to me means that HCG mode is not "normal", it is a special feature of that sensor.  I have found some discussions online about the IMX290 also having HCG mode but have not been able to confirm it.  I think it is presumptuous to assume that everyone will use this feature all the time.  For solar system imaging for example I would prefer the lower noise and larger SNR of non-HCG mode (note that read noise is not the only source of noise in the image).  Do I understand correctly that in HCG mode each pixel is generating twice as many e- for every photon, thus filling up the charge well in half the time?  I know read noise is slightly lower in HCG mode, but is not the end result still a reduction in SNR?  Agreed, stacking can help to overcome this but in turn you are now committing to using stacking when HCG mode is on.

 

Cheers,

 

Jim T.

 

Jim,

 

Quite a few CMOS sensors have HCG mode or something similar, not all are used in Astro applications. Irrespective, HCG mode is an intrinsic feature of the sensor architecture and you cannot exclude it just because other sensors don't have it. Every camera/sensor applies gain... you should think of HCG mode as gain on steroids only possible because of the low read noise. If you tired anything similar with CCDs you would just be amplifying noise.

 

What people prefer (stacking vs. Single long exposure) is a seperate discussion and should not be mixed with sensor specs. I return to this point below.

 

Also I think you are missing the point of lower read noise (and any sensor) architectures. The full well size is irrelevant. What matters is dynamic range of the sensor.. Seems counter intuitive especially if you are coming from the CCD world because we have been trained to believe larger the full well the better. The nuance is that this is true only if read noise is the same. The dynamic range of the 829 sensor is exactly the same as the 224 despite the fact that the 829 FW is nearly 50k.

 

There is no reduction in SNR with shorter exposures using a very low read noise camera. SNR is a ratio. Total exposure time will also be lower but that is a more complex relationship. 

 

Remember if your read noise is <1e you can almost count individual photons but in a camera with 7e read noise you need 7 times the photons for the same SNR (I am dramatically simplyfiing the relationship here but it is directionally correct).

 

Wrt the preference  of stacking vs. single exposure that is a debate as old as stacking : )

 

You can argue it both ways based on personal preference. My view is that stacking is overall a positive development. Using low read noise cameras and stacking now you can generate much better captures and SNR with simple Alt Az mounts than you could with well polar aligned EQ mounts and long exposures using high read noise CCD cameras a few years ago.

 

Just look at my 290 thread. Many of those ARPs would have required longer sub exposures at longer focal lengths using traditional CCDs to produce the same detail. I know because I have imaged many of them over the years. I could not be more happier with the technological progress in this field (actually I want a lot more innovation but given the size of this market and the minuscule R&D dollars being invested the progress is quite good).

 

Hiten


Edited by Astrojedi, 28 October 2016 - 04:04 PM.


#11 jimthompson

jimthompson

    Apollo

  • *****
  • topic starter
  • Posts: 1087
  • Joined: 06 Oct 2009
  • Loc: Ottawa, Canada

Posted 28 October 2016 - 04:41 PM

So we are on the same page (been doing some Googling this afternoon...)

 

- Read Noise:  A measure of the noise attributed to small variations in the manufacture of each pixel and its associated analog-digital converter (ADC).  Also includes random variations in signal resulting from the physics behind how the sensor and other electronics is working.  The lower the read noise, the smaller the contrast difference that can be resolved by the sensor.

- Shot Noise:  A measure of the noise attributed to the discrete/probabilistic nature of photons and the electrons they generate within the sensor.  This is inherent to the problem of counting photons.  Impact is larger when exposure times are low.

- Dark Current:  A measure of the noise attributed to heat in the sensor.  The warmer the sensor, the more randomly generated electrons will be picked up by the sensor.

- Signal-to-Noise Ratio (SNR):  A measure of how small a signal is detectable relative to noise.  Often defined as:

 

                     SNR = 20 log (sqrt(FWC)), where FWC = Full Well Capacity

 

- Dynamic Range (DR):  A measure of how many discrete levels can be detected within the signal.  Is affected by both FWC and read noise...smaller the read noise the smaller the step change in contrast that is detectable.  Often defined as:

 

                     DR = 20 log (FWC/read noise)

 

cheers,

 

Jim T.



#12 Astrojedi

Astrojedi

    Soyuz

  • *****
  • Posts: 3760
  • Joined: 27 May 2015
  • Loc: SoCal

Posted 28 October 2016 - 05:17 PM

Jim,

 

These are very simplified but more or less accurate definitions.

 

That SNR formula is a little too simplified as it does not take into account read noise. The Full well capacity needs to be normalized for sensor noise before you apply that formula to determine SNR.

 

It also completely ignores stacking or Skyglow considerations. Skyglow builds up a lot faster than the signal you care about in extended exposures (I.e. Your sensor well could be full of photons/electrons you don't care about including dark current and skyglow). Further when you read out the well in high read noise sensors/cameras some of of the information is unfortunately lost.

 

Full well capacity is not the appropriate way to estimate SNR or dynamic range. The sensor well is just a way to collect electrons generated by the detector irrespective of what those electrons represent.

 

Hiten


Edited by Astrojedi, 28 October 2016 - 07:05 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