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

LENR or taking darks/temp matched dark library

  • Please log in to reply
35 replies to this topic

#26 B 26354

B 26354

    Vanguard

  • *****
  • Posts: 2,237
  • Joined: 05 Jan 2017
  • Loc: Southern California semi-desert (NELM mag 5.3)

Posted 21 September 2018 - 10:41 AM

Thanks, Alen! I figured as much... but I always like tossing in personal experiences and bits of info as food for thought for the readers out there who may not have previously encountered them. And I very much enjoy reading your posts and comments, for exactly those reasons.  waytogo.gif



#27 seigell

seigell

    Vostok 1

  • -----
  • Posts: 127
  • Joined: 10 Nov 2010
  • Loc: Clouds of Central Florida

Posted 21 September 2018 - 01:39 PM

I have also found that it has had zero impact on my productivity since my imaging is driven by when an object climbs above the trees and when it reaches the meridian, not by how many subs that I take.

Huh??  If you laid off the LENR, between those trees you could acquire 2x the number of Lights.  And if you continued Imaging on the other side of the Meridian Flip, another 2x.  Overall, you'd achieve 2x the SNR.



#28 Jon Rista

Jon Rista

    ISS

  • *****
  • Posts: 24,990
  • Joined: 10 Jan 2014
  • Loc: Colorado

Posted 21 September 2018 - 02:12 PM

Huh??  If you laid off the LENR, between those trees you could acquire 2x the number of Lights.  And if you continued Imaging on the other side of the Meridian Flip, another 2x.  Overall, you'd achieve 2x the SNR.

No, this is not necessarily true. It entirely depends on the magnitude of the FPN. Dark current is not just shot noise. It also leads to DFPN (dark fixed pattern noise) due to the non-uniform response of each pixel to dark current. There are also glows that can affect the sensor in various ways.

 

At higher temps, DFPN can become quite extreme. The "noise" introduced by DFPN, beyond just the larger scale (but often very difficult to remove) glow structures, there is also banding, hot and cold pixels, as well as smaller scale pixel-to-pixel variations. DFPN may not be totally consistent from frame to frame. Temperatures fluctuate, and an unregulated sensor can fluctuate quite a lot, which makes getting well-matched darks after the fact a nearly impossible task. Even if you manage to match the EXIF temps, that really doesn't tell you anything about the actual sensor temp, since EXIF temps are usually from temperature sensors removed from the image sensor itself...often far removed. EXIF temps in Canon cameras are from the off-sensor DSP units (DIGIC chips). EXIF temps in Sony cameras are from a sensor off near the battery compartment!

 

Subtracting a poorly matched dark master, made with frames acquired from again an unregulated sensor temp, are rarely ever going to match the temperature of the individual light frames, which themselves will vary in temps. As such, calibration will leave behind remnant FPN, and often quite a lot of it. If the scale and magnitude of the remnant FPN, not just the glows and banding, but right down to that pixel-to-pixel variation, is much larger than the dark current shot noise (a temporally random noise), then LENR could actually help improve your SNR, despite the fact that you get 1/2 the total exposure time.

 

Whether this is true for any given camera at any given location is not something I can tell you. There is no formula for determining it. The only way to know is to experiment and try. That would mean, yes, spending a night acquiring LENR data on an object or two, and another night acquiring non-LENR data on the same object or two. And comparing the results. Images from light polluted skies are likely to benefit less from LENR, while images from darker skies are more likely to benefit from LENR, especially on warm nights. On colder nights, especially near freezing during the winter, LENR would become largely useless, since at such temps dark current could be a tiny fraction of what it is at 20, 30, 40 degrees C. So LENR is not going to be worth it's exposure time cost all the time. It has a specific use case, on warmer nights, under darker skies.


  • B 26354 likes this

#29 seigell

seigell

    Vostok 1

  • -----
  • Posts: 127
  • Joined: 10 Nov 2010
  • Loc: Clouds of Central Florida

Posted 21 September 2018 - 02:16 PM

You cannot measure the Sensor Temperature with BYEOS.

There is no temperature probe attached to Image Sensor.

The temperature probe is inside the Digic Processor - the "CPU" Chip.

Whenever you use the Digic Processor, it gets hotter and hotter.

The temperature of the Image Sensor, is not the same temperature as the CPU chip.

 

You are matching your Dark Library to the temperature of the Canon Digic Processor Chip, not the Image Sensor.

 

There is ONLY one source of Internal Temperature Measurement available for Images taken via a Canon DSLR (or NIKON), and it is a value provided via the Camera Maker's SDK and not written by the DSLR into the EXIF data.

So:

1) One must use BYE/BYN or APT or SGP (or one of the other PC Apps that tether the DSLR via the SDK), in order to have ANY reference value regarding Camera Temperature.

2) Both the Image Sensor and the DIGIC Chip heat-up when they are used to generate an Image.  Actually, both generate most of their heat at the End of an Exposure, when the Image is being Read Off of the Chip and Digitized and then the RAW Processing and File Storage tasks occur on the DIGIC Chip.  As these DSLR Sensors are CMOS, there is a Maintenance Current utilized throughout the duration of the Exposure (this heats the Image Sensor - and is a component source of the FPN).

3) The Image Sensor and the ADC and the DIGIC Chip are packaged tightly together, so most any heat generated on the DIGIC or ADC is going to partially transfer to the Image Sensor.  In fact, DSLR industrial designs rely upon that, as only the Image Sensor is exposed to External Cooling.

4) It doesn't matter that much where the Temperature Sensor is Located, as long as it is the same Reference used for both the Light and the Dark. (And Canon hasn't confirmed the location of the Temp Sensor.)

5) Since the DSLR has little idea whether the User is generating Lights or Darks or Bias or Flats, the Temperature Value is equally valid in any Image Type.

 

You are matching your Dark Library to the temperature of the Canon DIGIC Processor Chip, not the Image Sensor.  Yes.  But matching Temperatures is the only Critical Element.



#30 seigell

seigell

    Vostok 1

  • -----
  • Posts: 127
  • Joined: 10 Nov 2010
  • Loc: Clouds of Central Florida

Posted 21 September 2018 - 02:32 PM

You are matching your Dark Library to the temperature of the Canon Digic Processor Chip, not the Image Sensor.

MVAS:  Your profile doesn't have a Signature Line to list what Cameras you use.  But odds are that it is either a DSLR, or a CMOS Astro Camera sourced from a DSLR Sensor Maker.  In both cases, the "Sensor Temp" that you access is most likely exactly the same "from the Processor Chip" as you decry above.  Most AP Camera Designs are simply re-purposed DSLR or Machine Vision Sensors from Sony or Pentax or Panasonic.  The only difference from the original DSLR is that a few kilobytes of fresh code was written to the Processor so as to enable USB output instead of SD Card and to enable the TEC Control.



#31 seigell

seigell

    Vostok 1

  • -----
  • Posts: 127
  • Joined: 10 Nov 2010
  • Loc: Clouds of Central Florida

Posted 21 September 2018 - 04:19 PM

Even if you manage to match the EXIF temps, that really doesn't tell you anything about the actual sensor temp, since EXIF temps are usually from temperature sensors removed from the image sensor itself...often far removed. EXIF temps in Canon cameras are from the off-sensor DSP units (DIGIC chips). EXIF temps in Sony cameras are from a sensor off near the battery compartment!

As the Interior Components of a DSLR (or CMOS AP Imaging Camera) are essentially a Closed System, there is very little "randomization" of Temperatures.  That is not to say that the Reported EXIF Temp is going to be EXACTLY the Image Sensor Temp.  But rather that the Heating / Cooling / Heat Conduction is going to be Very Repeatable within that Closed System.  Given the Metal plus Polycarbonate Camera Body Frame and the Poly and Rubber Cover all acting more as Insulation than as Radiator, the only Variable is the Exposure to Ambient Air when the Shutter(s) are Open.  And, given that is either a closed body of air in the case of any AR Window or Filter or Corrective Optics or Camera Lens found nearby in the Optics Train, or a nearly closed large mass of Ambient Temp Air in the case of any other Optics configuration, the Cooling Effect on the Image Sensor is going to be Present but Very Repeatable.

All of this tends to make the EXIF Temp, regardless of the Physical Location, a "Rather Repeatable" and therefore "Usable" Reference Data.

 

In fact, in some ways, the Exposure and Shutter Sequence employed in a LENR Image Session has more "Variables".

1) At the end of a Normal Exposure, there is a 5-15 second "pause" at the end of every Exposure Sequence as the Camera pushes the Image Data across the USB to the PC.  Sure there is some Controller activity associated with this "data push", but usually it is Interrupt/DMA driven and for the bulk of the time the Processor is idling.  During all of this time, the Processor is radiating some of it's Thermal "Waste" into the nearby Image Sensor, and the latter is radiating into the Mirror Box.  But as this Time and Effort is Consistent regardless of the Type of Image or the preceding Exposure Time, the Thermal Impulse is essentially the same every Exposure.

2) There is an even longer "pause" if Dithering / Settling is used between Exposures.

3) At the end of a LENR Exposure, the Image is read off the Sensor and into Memory.  Then, while the Shutter remains closed BUT without any of the other End-of-Exposure Processing or USB Offloading, the Dark Exposure is performed.  While the Dark is "Exposing", the Thermal Impulse from the Processor is radiating into the Image Sensor.  This, already is a different Thermal Event profile than had occurred during the paired Light Exposure immediately before.  At the end of the LENR Dark Exposure, the Processor has a different / greater processing load as it applies the Dark Data and then performs the RAW Processing and the Data Offload.

 

Does the LENR approach produce better quality than a Diligently Created and Diligently Applied Dark Library??  I don't know.  And I don't know if it is consistent across DSLR Brands or even Models.

 

I'm more willing to attribute the relative success of LENR Images to "secret sauce" - the DSLR Engineers have massively more access to the gritty details of every aspect of Optical and Thermal Performance of their Chips as well as direct and expert access to the Code being executed (and have every bit or more access to the theoretical - most of the best Optical and Photographic Professors and Engineers work for Big Camera).  They have direct access to the Optically Masked Pixels outside the FOV, to feedback parameters in the Sensor and the ADC and the Processor - all of which can only possibly be inferred through the Processed Pixel Data contained in the RAW Image File.

Balancing that, the External Stacking and Calibrating Software have only the advantage of Double the Exposure Time and larger Processing and Storage Resources (and the Cleanest Image Data that the Manufacturer's Engineers could produce).

 

Toss-up??



#32 Jon Rista

Jon Rista

    ISS

  • *****
  • Posts: 24,990
  • Joined: 10 Jan 2014
  • Loc: Colorado

Posted 21 September 2018 - 05:02 PM

As the Interior Components of a DSLR (or CMOS AP Imaging Camera) are essentially a Closed System, there is very little "randomization" of Temperatures.  That is not to say that the Reported EXIF Temp is going to be EXACTLY the Image Sensor Temp.  But rather that the Heating / Cooling / Heat Conduction is going to be Very Repeatable within that Closed System.  Given the Metal plus Polycarbonate Camera Body Frame and the Poly and Rubber Cover all acting more as Insulation than as Radiator, the only Variable is the Exposure to Ambient Air when the Shutter(s) are Open.  And, given that is either a closed body of air in the case of any AR Window or Filter or Corrective Optics or Camera Lens found nearby in the Optics Train, or a nearly closed large mass of Ambient Temp Air in the case of any other Optics configuration, the Cooling Effect on the Image Sensor is going to be Present but Very Repeatable.

All of this tends to make the EXIF Temp, regardless of the Physical Location, a "Rather Repeatable" and therefore "Usable" Reference Data.

 

In fact, in some ways, the Exposure and Shutter Sequence employed in a LENR Image Session has more "Variables".

1) At the end of a Normal Exposure, there is a 5-15 second "pause" at the end of every Exposure Sequence as the Camera pushes the Image Data across the USB to the PC.  Sure there is some Controller activity associated with this "data push", but usually it is Interrupt/DMA driven and for the bulk of the time the Processor is idling.  During all of this time, the Processor is radiating some of it's Thermal "Waste" into the nearby Image Sensor, and the latter is radiating into the Mirror Box.  But as this Time and Effort is Consistent regardless of the Type of Image or the preceding Exposure Time, the Thermal Impulse is essentially the same every Exposure.

2) There is an even longer "pause" if Dithering / Settling is used between Exposures.

3) At the end of a LENR Exposure, the Image is read off the Sensor and into Memory.  Then, while the Shutter remains closed BUT without any of the other End-of-Exposure Processing or USB Offloading, the Dark Exposure is performed.  While the Dark is "Exposing", the Thermal Impulse from the Processor is radiating into the Image Sensor.  This, already is a different Thermal Event profile than had occurred during the paired Light Exposure immediately before.  At the end of the LENR Dark Exposure, the Processor has a different / greater processing load as it applies the Dark Data and then performs the RAW Processing and the Data Offload.

 

Does the LENR approach produce better quality than a Diligently Created and Diligently Applied Dark Library??  I don't know.  And I don't know if it is consistent across DSLR Brands or even Models.

 

I'm more willing to attribute the relative success of LENR Images to "secret sauce" - the DSLR Engineers have massively more access to the gritty details of every aspect of Optical and Thermal Performance of their Chips as well as direct and expert access to the Code being executed (and have every bit or more access to the theoretical - most of the best Optical and Photographic Professors and Engineers work for Big Camera).  They have direct access to the Optically Masked Pixels outside the FOV, to feedback parameters in the Sensor and the ADC and the Processor - all of which can only possibly be inferred through the Processed Pixel Data contained in the RAW Image File.

Balancing that, the External Stacking and Calibrating Software have only the advantage of Double the Exposure Time and larger Processing and Storage Resources (and the Cleanest Image Data that the Manufacturer's Engineers could produce).

 

Toss-up??

Trust me. I did DSLR imaging for years, under both dark skies and light polluted skies. Dark current and DFPN are one of the main reasons I switched to mono cooled cameras with regulated cooling.

 

Sensor temp can and will vary, and over the length of a night can vary a lot. Closed system or not, the entire camera will cool over time, and the sensor itself, being an electronic circuit of very high density and operating at very high frequency, will heat up and cool down during operation. A readout can increase temperatures a lot, particularly in the DSP chips or ADC units. That heat is released into this "closed system". With longer exposures, during which time these circuits are idling, that heat will dissipate, until the next readout. The moment you use live view for focusing, the amount of heat can be significantly more, rising temperatures even higher, but again, during a longer exposure that heat will dissipate. There is internal thermal management, and some heat is shunted to specific parts of camera bodies so it can be released to the air. Not a particularly efficient system in any camera I've used, but it does work to remove heat, so the system will definitely not remain at a constant temperature.

 

Remember, it only takes 4-5C for dark current to DOUBLE. It doesn't take doubling of dark current to produce a meaningful change in DFPN. DFPN is also the sole reason we calibrate at all. Hot and cold pixels. That is the most well known aspect of DFPN, but far from the only. Without DFPN, there is no reason to calibrate. If you cannot well-match the DFPN in your lights, correction with a master dark is what is a tossup. It may work, it may not, it may even make things worse!

 

LENR itself is not perfect, and as you can see from my examples above, even LENR can imperfectly match the prior light. At extreme temps, this is more of an issue than at high but not quite so extreme temps. Up to 30-35C, I find that LENR is basically perfect. Not once in the years I did DSLR imaging was I ever once able to create a master dark, no matter how masterfully I worked at it, no matter that I spent DAYS (literally, camera running, non stop, in the freezer, fridge, me manually changing both of their set point temps even, in the ambient air, etc. to acquire thousands of dark frames at a wide range of temps), and even did this repeatedly on multiple occasions throughout each year, that was actually able to properly calibrate even one light. There was always remnant DFPN. Every time...even on nights of cold temps, the 5D III I used most of the time had an amp glow, and I was never able to fully and properly match that glow. And...that glow shows up in a lot of my images. It's ugly, nasty, I hated it every time, but I could never do anything about it. 

 

So to assume that you even CAN create a well matched master dark is an assumption I would question. With low dark current, the chances that you might increase. With DSLRs or mirrorless cameras that are truly known to have very low dark current (far fewer do than you might think), might be easier to get well matched darks in very cold weather. But warmer nights? The chances are very slim. LENR, since the dark is taken immediately after the light, will always be better matched. Perfectly matched? Maybe not, but far closer than you are likely to get even with concerted effort at creating a good master dark library. The simple problem here is...we just plain and simply don't know the temp of our sensors! Never have, quite probably never will, because no one actually seems to measure the temperature of the sensor itself in DSLR or mirrorless cameras. Half of them (even more than half) don't even measure at all. It is all 100% assumption based...you assume that the temp was stable, you assume that the temp was some particular temperature, you assume that the EXIF temp is right, you assume that the EXIF temp actually even means anything... You cannot match a master to a light based on assumption...

 

Again. LENR is not a panacea. It has its time and place, and not every time and place is its.  When you image on a warmer night, particularly with Canon DSLRs but also many Nikon DSLRs and Sony mirrorless, especially if you use live view for...anything...LENR could produce better images, despite the loss of integration, because it calibrates so much better. It can usually remove the vast majority of the DFPN, and produce a totally flat dark signal (as you can see in my 40C example above). The remaining shot noise, whatever it is, even though it will increase by 40% with the LENR dark subtraction, if often (and maybe even usually) much smaller in scale than the DFPN itself, and often smaller than any remnant DFPN you would get with a poorly matched master dark. Further, even if it is not smaller than remnant DFPN, the fact that the remaining shot noise is gaussian and evenly distributed throughout the field means it is far more aesthetically pleasing, so even if your SNR is not the absolute max...your IMAGE may still look better...quite possibly a lot better. 


  • sharkmelley and t-ara-fan like this

#33 sharkmelley

sharkmelley

    Aurora

  • *****
  • Posts: 4,952
  • Joined: 19 Feb 2013
  • Loc: UK

Posted 21 September 2018 - 06:48 PM

 

As the Interior Components of a DSLR (or CMOS AP Imaging Camera) are essentially a Closed System, there is very little "randomization" of Temperatures.  That is not to say that the Reported EXIF Temp is going to be EXACTLY the Image Sensor Temp.  But rather that the Heating / Cooling / Heat Conduction is going to be Very Repeatable within that Closed System.  Given the Metal plus Polycarbonate Camera Body Frame and the Poly and Rubber Cover all acting more as Insulation than as Radiator, the only Variable is the Exposure to Ambient Air when the Shutter(s) are Open.  And, given that is either a closed body of air in the case of any AR Window or Filter or Corrective Optics or Camera Lens found nearby in the Optics Train, or a nearly closed large mass of Ambient Temp Air in the case of any other Optics configuration, the Cooling Effect on the Image Sensor is going to be Present but Very Repeatable.

All of this tends to make the EXIF Temp, regardless of the Physical Location, a "Rather Repeatable" and therefore "Usable" Reference Data.

 

 

Really? 

 

Here's some evidence that doesn't support your claim.  Look at Table 3 in Roger Clark's review of the Canon 7DII:

http://www.clarkvisi...ion-canon-7dii/

 

For an EXIF temperature of 23C there are 2 values of dark current, with a factor of almost 2x difference between them.  That's why I would never rely on Canon's EXIF temperature as a predictor of thermal fixed pattern noise.  It's why I regard building a comprehensive dark library at different temperatures as a fruitless exercise.

 

Also, for the avoidance of doubt, I don't agree with Clarkvision's claim that the dark current of a 7DII is 10x smaller than previous Canons.  I think he got  lucky with his sensor.  I tested a 7DII and found it was pretty much in line with most other Canons.   Unfortunately, for one reason or another there appears to be huge differences between individual Canon cameras of the same model.  For instance a Canon 200D/SL2 that I'm testing is showing much higher read noise levels than Bill Claff shows on this PhotonsToPhotos site.  Neither Bill or myself can explain the difference. It's delaying my detailed comparison of two entry level cameras - the Canon 200D vs Nikon D5300.

 

Mark


Edited by sharkmelley, 21 September 2018 - 07:03 PM.

  • Jon Rista likes this

#34 Petros

Petros

    Explorer 1

  • -----
  • Posts: 67
  • Joined: 04 Nov 2006
  • Loc: Germany / Greece

Posted 21 September 2018 - 11:54 PM

I currently use a (quite noisy) canon 100D for DSLR imaging and I used LENR for the first time last week on a warm night (camera temp around 30C). Well... my results are better. I had never ever used LENR in the past 12 years.

But I did 2 changes to my setup at once: stopped using my IDAS LPS filter and tried LENR, so I am not 100% sure which attributed more to the better results, the shorter exposures (because of not using the LPS filter) or the LENR.

 

I use both a dedicated CCD camera with regulated cooling (old SBIG ST-2K) and my various Canon cameras over the years. I clearly see the difference in noise when using my DSLRs, noise is just not removed properly no matter what I do (dithering, master darks, etc) but my feeling with LENR was that this time I could stretch the image a bit more.

 

I tend to think that the noisier the camera is, the better results LENR has, but I will try to do some dedicated tests next time I use my DSLR to confirm this.

 

Clear skies,

Petros



#35 mvas

mvas

    Apollo

  • -----
  • Posts: 1,337
  • Joined: 27 May 2015
  • Loc: Eastern Ohio

Posted 22 September 2018 - 12:11 AM

My 6D has a mode where it will take 4 lights, then one dark, and apply that dark to the previous 4 lights.  So that is only a 20% loss in light gather time, better than the 50% that some cameras offer.   I believe some 5D models, and the 7DMark II also do this.    The lower end crop sensors do not offer this feature.

 

From what I know, you have to shoot the sequence of 4 shots back to back, with perhaps a 1 second delay in between them.  The 4:1 lights:darks does not work if you trigger the shot by USB. You have to use an intervalometer.

 

I tried it just shooting 5min subs with the lens cap on, it definitely works. Will try in the field when I get a chance. 

I thought I read, a long time ago, that there were several modes when LENR setting = ON ...

 

Take 1 long exposure and then WAIT ... the LENR Dark will occur automatically

Take 2 long exposures and then WAIT ... the LENR dark will occur automatically

 

Save "RAW" only mode then the LENR Dark will be forced after the 3rd image, even if your intervalometer tries to trigger another image

Save "RAW + JPEG" mode then the LENR Dark will be forced after the 4th image, even if your intervalometer tries to triggers another image

 

So, it appears you can trigger a LENR after 1, 2, 3, or 4 exposures.



#36 Petros

Petros

    Explorer 1

  • -----
  • Posts: 67
  • Joined: 04 Nov 2006
  • Loc: Germany / Greece

Posted 30 September 2018 - 02:59 AM

Following up my previous post,

I did another test and I cannot confirm that LENR had a significant advantage.

In my case it seems that the light pollution filter was killing my DSLR setup.

 

I inspected some frames with and without LENR up close and without doing any analysis, I could not see any major positive advantage on the frames that had LENR enabled. There might be a difference on warm summer nights but on my last test I could not identify any major advantages and concluded that at least on my setup, the major advantage was removing the light pollution filter.

 

Clear skies,

Petros




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