DSLR battery discharge heat source

[Tommohawk]

Hi All.

I've been dabbling with DSLR imaging for a while and I notice that my winter images seem better than my recent May/June images.

This may simply be that in the wnter months we have darker skies and more time to image, but I wonder if camera heating is a factor. I've been looking at options to cool the camera, but does anyone know if battery discharge causes significant heat transfer to the sensor?

If so, would an external power supply give any benefit?

thanks

Tom

[Kropster]

It should give some benefit, but I think heat from the sensor and amplifiers is more of a problem.

[laser_jock99]

I always run off an external power supply- don't know if this keeps the camera any cooler or not? But it does stop me from running out of power mid session!

[Welrod50]

You will always generate internal heat with the battery, but depending on the internal structure of the camera itself this may not have a significant effect.  Some older Canons exhibit thermal noise from the battery, but I have not seen this myself with my 500D.  That said, I use an AC adapter anyway just so I don't need to fiddle with the rig once it's running by changing batteries.

You will find that the images are noisier in the summer months simply because the sensor itself warms up during exposures.  Warmer sensor = more noise.  With a DSLR, unless it has been specifically modified, there is no getting away from this.

[ronin]

I would say it would be that the sensor when in use sits say 15 degrees above ambient, so in summer the sensor is simply hotter and so generates more thermal noise within itself.

[sharkmelley]

Typically a Canon sensor will warm up by 12-15C when used continuously for 2 hours and this will approximately double the thermal noise seen in the image.

In Winter you are starting at a much lower temperature and this leads to a significant reduction in noise.  If your battery feels warm when you remove it then it could be contributing to sensor heat build up but I've never noticed a warm battery.

One other thing is to make sure the display on the the rear of the camera is disabled during imaging (hit the DISP button) because this is a significant source of heat - though it might vary from model to model.

Mark

[lensman57]

Hi All.

I've been dabbling with DSLR imaging for a while and I notice that my winter images seem better than my recent May/June images.

This may simply be that in the wnter months we have darker skies and more time to image, but I wonder if camera heating is a factor. I've been looking at options to cool the camera, but does anyone know if battery discharge causes significant heat transfer to the sensor?

If so, would an external power supply give any benefit?

thanks

Tom

The summer skies are noisy but the main  reason for DSLR noise is your sensor . I hardly do any DSLR imaging now but last winter the sensor on my modded 1100d was  11~13c while the outside temp was just above 0C, in the summer I would expect it to be about 30~40 C, for every 6 degrees increase in temp the noise nearly doubles, you can work the rest out for yourself.

A.G

[sharkmelley]

 for every 6 degrees increase in temp the noise nearly doubles, you can work the rest out for yourself.

It's the dark current that doubles for every 6 C temperature increase which in turn leads to the thermal noise doubling for every 12C temperature increase.

Mark

[Alien 13]

An alternative to an ext PSU is a battery grip, takes any heat away from the camera and makes changing batteries mid session easier.

Alan

[glowingturnip]

I have the same issue.  I've recently started using an external power source anyway, which has the benefit that I can run off more subs without running out of batteries, but it doesn't seem to make much difference to the amp glow.

My amp glow is in the top two corners of the frame, and also about 1/3rd along the top (so bottom of the camera ?)  The battery goes in at the side, so I don't think the battery is the main source of this glow anyway.

I make efforts to keep my target away from the top corners.  I'm also getting good at gradient removal post-processing, can usually get most of it apart from the very far corners that need to be cropped.

[Tommohawk]

Thanks for all the comments.

It should give some benefit, but I think heat from the sensor and amplifiers is more of a problem.

Yep, that seems to be the consensus.

I always run off an external power supply- don't know if this keeps the camera any cooler or not? But it does stop me from running out of power mid session!

And for that reason alone I think I will move over to external PSU. Look like I will be tied to laptop (with limited battery life) so mains supply will be required anyhow.

Typically a Canon sensor will warm up by 12-15C when used continuously for 2 hours and this will approximately double the thermal noise seen in the image.

In Winter you are starting at a much lower temperature and this leads to a significant reduction in noise.  If your battery feels warm when you remove it then it could be contributing to sensor heat build up but I've never noticed a warm battery.

One other thing is to make sure the display on the the rear of the camera is disabled during imaging (hit the DISP button) because this is a significant source of heat - though it might vary from model to model.

Mark

I wondered about display heat. Using a mobile phone as a comparison I notice when using GPS it gets mighty hot - but probably mostly due to the screen. My camera is a Nikon and the LCD isnt active once shooting starts so presumably no heat from this.

...... but last winter the sensor on my modded 1100d was  11~13c while the outside temp was just above 0C, in the summer I would expect it to be about 30~40 C, for every 6 degrees increase in temp the noise nearly doubles, you can work the rest out for yourself.

A.G

When you say modded 1100d I guess you mean filter modded, not cooled? Also, you say sensor temp was 11-13c - I'm wondering how you know that? Is there a temp readout - would be really useful to know.

An alternative to an ext PSU is a battery grip, takes any heat away from the camera and makes changing batteries mid session easier.

Alan

Yes, I thought about this - but then you add to the weight. .... but fewer trailing cables I suppose.

I have the same issue.  I've recently started using an external power source anyway, which has the benefit that I can run off more subs without running out of batteries, but it doesn't seem to make much difference to the amp glow.

My amp glow is in the top two corners of the frame, and also about 1/3rd along the top (so bottom of the camera ?)  The battery goes in at the side, so I don't think the battery is the main source of this glow anyway.

I make efforts to keep my target away from the top corners.  I'm also getting good at gradient removal post-processing, can usually get most of it apart from the very far corners that need to be cropped.

I dont notice amp glow at all and I also have Nikon camera - D3200. Actually its on loan from my son whilst I get my own sorted.  What length subs are you doing? Maybe this is because I'm only doing shortish subs - I've only just got guiding sorted, so have been limited to 90 seconds til now.

So:

1. Does anyone have a recommendation for external PSU please? Preferably one that will also give 12V 2Amp for the mount.

2. Also - has anyone tried any of the "cooling box" methods - DIY or otherwise? Do they work? Ive read a couple of threads on cold finger cooling but I think this is beyond my technical expertise.

TBH I'm still agonising over whether I should go cooled CCD. More expensive, but by the time you add on cooling system maybe not so different.

3. Does anyone know what percentage of noise from a modern Canon is fixed by doing darks? If darks fix the majority of noise then why worry? Ive just looked at my last set of darks from early June (pretty cool weather though) and the noise is identical on all - and it really doesnt look that bad. Theres no random noise that I can see.

4. Anyone know how DSS deals with noise - does it substitute interpolated pixels perhaps?

[Alien 13]

I dont know of a supply that will do the mount and camera so i built my own around 4 LM2596 dc-dc convertors this gives me 2x 7.3v for the camera and video light a 6V supply for my mount and 5V for charging my mobile (i use this with dslr controller app) the small box can be powered from a 12v battery or a plug in the wall 12v 6A adapter.

The result is no trailing cables (or a single 12v one if using the adapter.

Alan

[glowingturnip]

Thanks for all the comments.

I dont notice amp glow at all and I also have Nikon camera - D3200. Actually its on loan from my son whilst I get my own sorted.  What length subs are you doing? Maybe this is because I'm only doing shortish subs - I've only just got guiding sorted, so have been limited to 90 seconds til now.

So:

1. Does anyone have a recommendation for external PSU please? Preferably one that will also give 12V 2Amp for the mount.

2. Also - has anyone tried any of the "cooling box" methods - DIY or otherwise? Do they work? Ive read a couple of threads on cold finger cooling but I think this is beyond my technical expertise.

TBH I'm still agonising over whether I should go cooled CCD. More expensive, but by the time you add on cooling system maybe not so different.

3. Does anyone know what percentage of noise from a modern Canon is fixed by doing darks? If darks fix the majority of noise then why worry? Ive just looked at my last set of darks from early June (pretty cool weather though) and the noise is identical on all - and it really doesnt look that bad. Theres no random noise that I can see.

4. Anyone know how DSS deals with noise - does it substitute interpolated pixels perhaps?

I'm typically running 5 min guided subs, and the glow looks like below.  Fortunately it doesn't appear to be any more 'noisy' than the rest of the image, so is more a matter of a challenging machine-induced gradient to be got rid of.  Mine's an old D80, must upgrade at some point.

I use this external PSU - http://www.amazon.co.uk/gp/product/B005GQCR4U/ref=oh_aui_detailpage_o04_s00?ie=UTF8&psc=1 .  Don't get the 'Ex-Pro' one, I got that one first, and it failed on the 3rd try.

The key for noise is as many lights and darks as possible.  Noise is by definition random, whereas signal is always the same, so once you have many subs to work with, the random noise averages lower but the signal remains.  Noise decreases with the square root of total exposure time, though there's read noise from the DSLR every time you open the shutter too, so longer subs if you can, and plenty of them.  Darks also have noise, so do plenty of darks too, same exposure time, temperature and ISO as the lights.

DSS has various routines for combining subs, most will statistically disregard outlying pixels from the stack and average the remainder - I like the 'adaptive sigma' routine, which does just that, similar to Pixinsight's Windsorised Sigma Clipping I think.

[Tommohawk]

Hi Stuart.

that glow does look a bit of a problem. I'll try my D3200 with 5 minute subs and see what happens - not sure when I'll get a chance to do this, but apparently some clear sky this evening. So long as I can watt for sundown - about 11.30pm!

Take your point about the number of subs - but have to disagree about random noise. I've looked closely and the noise on each of my last set of darks is identical.

That being so, makes you wonder just how much of a problem noise actually is? If DSS sees the same noise on each dark it should be a doddle to adjust the final image.

I'll try some 5 min subs and darks to check for amp glow and increased noise and post the results.

[lensman57]

I use APT for control and capture and it does report the sensor temp from the EXIF file data. Modded usually refers to filter mod, mine is a Baader replacement filter but with the AA filter removed. I requested this from Juan at Cheap astro photography and he did a great job of it for me. Dslr coolling boxex are effective but the best ones are with a cold finger attached to the back of the sensor as these are almost entirely DIY it requires major camera surgeory and steel nerves.

A.G

[glowingturnip]

Hi Stuart.

Take your point about the number of subs - but have to disagree about random noise. I've looked closely and the noise on each of my last set of darks is identical.

That being so, makes you wonder just how much of a problem noise actually is? If DSS sees the same noise on each dark it should be a doddle to adjust the final image.

ahh, that's different - the brightly coloured speckles you can see on a dark frame are a map of the bad pixels (hot or cold ones) that should be the same from frame to frame (at equal temperature and exposure time).  That's the point in doing darks, so you can subtract those from your lights and take away those systematic read errors.

However, lights and darks also have random noise, which differs from frame to frame, and needs to be averaged out.  You probably won't really notice the noise until you start stretching the histogram.  Looks like this:

ugh !  it's amazing I can get anything done really isn't it !

I guess that goes back to your original question - the hot pixels do depend on temperature and will be worse in the summer, and I suspect an external power supply won't particularly help in that respect

[Tommohawk]

Hi Stuart.

Admittedly I have only done 2 minute subs, so might expect more probs on longer subs, but if my darks all show the same light pixels, then I have duff/hot pixels, but no random noise at all???

Also, if the darks show no random noise, will the same be true for the lights? Obviously I cant visually check for noise on the lights.

I need to do some more tests - need a few more cloudy nights... (kidding)

[Tommohawk]

I use APT for control and capture and it does report the sensor temp from the EXIF file data. Modded usually refers to filter mod, mine is a Baader replacement filter but with the AA filter removed. I requested this from Juan at Cheap astro photography and he did a great job of it for me. Dslr coolling boxex are effective but the best ones are with a cold finger attached to the back of the sensor as these are almost entirely DIY it requires major camera surgeory and steel nerves.

A.G

Hey thats really useful info - I didnt realise EXIF data had sensor temp info. That will be invaluable if experimenting with cooling systems. Does anyone know how to see temp info though? I dont have APT - i have Nikon ViewNX2 but can only see basic data.

[IanL]

Regarding the amp glow problems some are having, just double check there is no light leaking in to the camera through the viewfinder.  A small amount of light bleeding around the edges of the sensor can look exactly the same as amp glow but is easy to cure - stick some black tape over the viewfinder or use the rubber plug that some cameras come with for the job.  Not saying it isn't amp glow as some DSLRs do suffer from it; on some Canon models software like APT or BYE use live view to control mirror lock-up and it can leave the amplifier turned on so you might need to test for that.

Noise: You can't subtract noise; by definition it is random.  If you take a light frame containing noise and subtract a dark frame also containing noise, you will add the noise from the dark.  It may seem counter-intuitive, but it is true!  That is why you make master calibration frames (bias, darks and flats).

As signal increases, noise increases as the square root of the signal, so stacking more frames improves the ratio of signal to noise (SNR).  Using a higher SNR light or calibration frame doesn't subtract the noise, it 'overwhelms' the increasing noise with signal that increases faster!

The following graph shows the relative improvement in SNR an extra sub each time:

Note that this is a relative measure, not an absolute one, so we're saying that a stack of 100 frames will have an SNR that is better than a single frame from the same source. We are not saying that a stack of 100 frames is always better than 1 frame, e.g. a single frame from a well cooled CCD might have less noise than a stack of 100 frames from a hot DSLR.

Note that number of frames is just one way of measuring the number of samples you have taken of your target.  So you could substitute 'Number of Frames' with 'Length of Exposure', so a 100 second exposure would have better SNR than a 1 second exposure (assuming you haven't saturated the sensor on a bright target). You can see that it is a game of diminishing returns, so you get a big improvement early on, but the more frames (or longer the exposure) the less improvement in SNR you get for each additional frame or minute.

This relationship works for light frames, but it also works for dark frames (and other calibration frames).  In the case of the dark frame, the signal is the fixed pattern of dark current that you want to subtract (which isn't random), and the noise is the dark current noise (which is random).  Of course in the case of a dark frame you need to match the amount of dark current in the master dark to the amount of dark current in the light; if you don't you will subtract too much or too little dark current and will either make the light frame worse, or not improve it as much as you might.

Here's a real example plotting the standard deviation of different sized stacks of bias frames. (Standard deviation is a rough measure of the amount of noise in an image though there are better ones).  You can see it is remarkably close to the theory above I hope!

I'm doing a whole load of research on DSLR darks because there is a lot of debate about whether they are useful at all, and if they are the best way to take and apply them.  It isn't ready to be published online yet (not least because I am having to learn PixInsight scripting so I can create a script to take all the statistics I need from my library of darks - I'm OK with JavaScript coding, but the documentation on the PI object model is sparse to say the least so a fair bit of guesswork and trial-and-error is needed).

That said here are a few things I can tell you:

1. You might as well forget about (Canon) EXIF temperatures for matching DSLR dark frames to light frames.  From what I have learned, the main purpose of the temperature sensor in Canon DSLRs is to monitor the processor chip used for liveview display.  It can get very hot with prolonged use and the sensor shuts the camera down to prevent damage.  Thus it is not particularly indicative of the actual sensor temperature.  The EXIF temp is also captured at the start of the exposure and doesn't tell you how hot the sensor got thereafter.  Here is a graph of my dark frame library:

It plots the mean brightness of each dark frame against the EXIF temperature.  Now if all was well with the world, the points would form a nice line sloping from bottom left to top right, but they don't!  (As the sensor temperature increases, the dark current should also increase and that would increase the mean brightness of the image). The thing to note from this graph is that the points form a 'wedge'; so you can see there are dark frames at 14C that have the same mean brightness as darks at 6C. So basically forget about using EXIF temperatures as some kind of precise measure to match darks together or match darks to lights.  Unless you are going to do more work than that, you might as well stack all your dark library regardless of temperature and hope for the best.  (That is not meant to be a joke or insult, I've done exactly that for the time being as it is the best I can manage until I have finished my tests).

2. Canon pre-processes the images on camera, and basically reduces the appearance of dark current in images (but not dark current noise).  This was discovered by Craig Stark, but I have done my own tests to confirm it (gnomes standing on the shoulders of giants and all that):

This is basically the same graph as above, but cleaned up a bit by taking the medians of the mean brightnesses (and switching the axes just to confuse you!)  The upshot is that at first, as the temperature of the sensor (roughly) increases, the mean brightness of the dark frames reduces, which is the opposite of what should happen.  Then at some point (16-18C) it stops happening and starts going in the right direction again.  Abandon hope all ye who pass this point!

Again this makes it very difficult to match dark frames and light frames.  Note that you could substitute 'Temperature' for 'Exposure Length' and get the same sort of curve (which is the test that Craig Stark did).  The camera isn't looking at the temperature or exposure length, it's looking at the 'optical black' which are a set of special pixels that have been masked off from light.  If they appear too bright, the camera subtracts a bit of brightness from the image (but after a point it stops doing so - kind of weird; nobody except Canon knows why and they aren't saying).

3. Instead you need to directly measure the amount of noise in your image somehow.  The dark current noise will increase as the square root of dark current (as we discussed above).  Also, remember you can't subtract noise from an image (not even those smart-alec Canon engineers can do it  ).  So by measuring the noise in a dark, we should be able to match dark frames together - same amount of noise = same amount of dark current (roughly).  I started out using standard deviations to measure the noise, but it is not a 'robust' estimator of noise (as I have discovered with the help of people smarter than me).  I plotted standard deviations against mean brightnesses:

I got what I expected, which is the same whacky curve as we have above - the standard deviations and means should form a nice bottom-left to top-right line, but they don't as we know Canon are suppressing the dark current up to a point, but it does prove that the the standard deviation is related to dark current noise as we'd expect.

4.  The next step is where I am right now.  I measured the standard deviations of all my darks, and then tried to create a set that gives the best possible SNR (ratio of dark current to dark current noise).  My tests did not work out though - instead of being able to create better darks, they appear to be the same or even slightly worse by adding more frames to the master dark.

The problem is the non-robust nature of standard deviation, (i.e. small imperfections in the data throw it off badly).  So what I need to measure is the Median Absolute Deviation of the darks instead.  This is a robust statistic and again it correlates well to noise in images.  Problem is I can only obtain it by measuring one image at a time (not much fun when you have to measure hundreds of them), so I am writing a batch statistics script for PixInsight to do the job for me (can't find any other free package that has this measure and can read FITS images in a batch).

5.  Assuming I can make the best SNR dark using the MAD measure, I will then use PixInsight's dark scaling routine to match the darks to the light frames.  This already works very well for my 'everything and the kitchen sink' dark I mentioned above.  It doesn't need to know about temperature, exposure lengths or anything else really. Instead it just takes the master dark and tries different scaling factors against each light frame until it produces the least noisy calibrated light it can manage.  (Can and will back this up when I do my final write-up on the subject), but it means all you have to focus on is producing the highest possible SNR master dark and not worry about matching the dark and light frames.  I think this is the best any stock DSLR imager is going to be able to manage; to do any better you would need a set-point cooled camera.

6. A few people have suggested just using Bias frames instead of darks, but I have satisfied myself that this doesn't produce as good a result as a scaled dark frame.  Your mileage may vary with camera model and processing software I guess, but the resulting images were much noisier (both visually and using statistical measures).

7. A few people have suggested using hot/cold/warm/cool pixel correction tools (like PixInsight's cosmetic correction process).  This did work pretty well and produced images that were similar to my scaled 'kitchen sink' dark, and actually suppressed more hot pixels that the dark did.  The main problem with the PI tool is that you need a good master dark to feed in to the tool to map the hot and cold pixels anyway, so it's a bit of a Catch-22 situation.

[Tommohawk]

IanL - a whole lot of useful info there!

OK so I will treat EXIF data with a pinch of salt. Will be interested to read your final conclusions re darks. I ran some slightly longer subs last night - still a modest 3 minutes - and no glow to be seen. The 3 minute darks show rather more noise than previous shorter ones.

[sharkmelley]

Ian,

Very nice analysis - looking forward to your conclusion about whether or not Canon darks are useful.  My own view is that they are useful for calibrating the non-uniformity of dark current of pixels i.e. which pixels are running consistently "warmer" and which are running consistently "colder" in terms of dark current.

At risk of going further off topic you mention how Canons pre-process the raws, subtracting the "average" dark current to leave a raw dark frame biased at 1024 or thereabouts.  For completeness it is worth pointing out that Nikons also also subtract the average dark current but leave a raw dark frame biased at 0.  This has the effect of truncating the values of half the pixels to exactly zero in a raw frame.  As a result, the pixels running consistently "colder" will almost invariably be truncated to zero in every dark frame.  This makes it virtually impossible to create a Nikon master dark that does a good job of calibrating the light frames. 

Mark

[glowingturnip]

interesting write-up Ian.  Seeing as my DSLR is essentially rubbish, so it seems, I think most of my readings will be in the upper right part of your graphs !  Must get a new one, wonder if my wife will mind...

Agreed re light leaking through the viewfinder.  Had a strange gradient a while back which took me ages to track down to an unexpected gradient on the dark frames.  Think I must have shot them with the outside lights on and viewfinder unobscurred.  Had to throw those darks away.  I've got a view finder cover thingy now.

[IanL]

interesting write-up Ian.  Seeing as my DSLR is essentially rubbish, so it seems, I think most of my readings will be in the upper right part of your graphs !  Must get a new one, wonder if my wife will mind...

Agreed re light leaking through the viewfinder.  Had a strange gradient a while back which took me ages to track down to an unexpected gradient on the dark frames.  Think I must have shot them with the outside lights on and viewfinder unobscurred.  Had to throw those darks away.  I've got a view finder cover thingy now.

If anyone is doing their darks during the daytime it is also worth knowing that most plastic camera/lens caps will let through a lot of IR which will be detected (especially if you have astro-modified your DSLR).  You should either use a metal cap, or if you don't have one you can cover the plastic cap in kitchen foil.