And what does Canon do?

[ULFULFSEN]

I want something better jrista!
Canon will not get my money for the Cameras they produce over the last couple of years.

70% of the time i shoot Portraits, the rest Landscapes.
I am happy with my old 5D MK2.
I looked long at the 5D III and it´s not worth it for me.

But IF Canon had produced a D800 equivalent i had bought it.
Do i need it... no.
Do members here need a 1 DX for a FLICKR account and cat or child photos?
I doubt it.

 

[ULFULFSEN]

I mean, the evidence is clear, see below...the first 2 are tiny crops of the bottom left and right corner...I mean who would want to buy that---it's so shadowy...I know...if only i had an exmor......don't even look at the 3rd, i mean, the 2 super crops should be enough to tell the story right?

I would not buy it because it´s overprocessed.

 

[scyrene]

Who's to say A7s level performance won't be the norm with the next round of APS-C or high res FF releases? I am still a firm believer that total sensor area and Q.E. are the primary factors that affect noise. In the case of the A7s...it's clear that they are doing some rather intense noise reduction in the BionzX chip as well (much like the DIGIC 6 does). All three of those factors, IMO, are more important than pixel size in the long run. So I see no reason you couldn't have that level of ISO performance in a sensor with smaller pixels.

For now, if you need cropping ability, then pixel size is probably a key factor. Doesn't mean that A7s level performance will stay only in sensors with bigger pixels forever. (Or even necessarily for all that long.)

Why all the optimism for other brands but not this one? I imagine all cameras will get more capable as time goes on. But it seems rather optimistic that e.g. Sony will produce the perfect device for your needs.

Astro is a very different beast. SNR, DR, and RN levels are all important. In astro, final signal strength or signal power is critical. For best results, you need to find the sweet spot that maximizes them all. I'd say before getting a different camera, start guiding your mount. More reliable tracking will probably do more for you than a better sensor, at least to start. I think that the A7s could certainly do better at high ISO than any current Canon sensor. It would certainly allow for the use of higher ISO settings...but I'm not sure that will really buy you much in the end. The key benefit of the increased DR of the A7s is simply a lower chance of clipping stars. In the long run, even with an A7s, guiding is going to allow you to expose for twice as long or longer than your current 2 minute exposures. (And, at 100mm, I would say even with a mount's periodic error, you should be able to expose for four minutes easily...unless the mount has some seriously wicked PE!)

In general ISO 3200, IMO, is rather high for astro unless your doing ultra wide field work (i.e. 16mm milky way imaging). For anything DSO related, you should be using the lowest ISO you can get away with and still have the minimal or nearly minimal read noise. That will maximize your DR and SNR, limiting the chance that stars will clip, maximizing your ability to increase the total signal power, and usually at no cost to your actual REAL exposure.

Thanks for the detailed reply.

I shot at the exposures that seemed right at the time. I have to admit, it's more of an intuitive thing, rather than working out the principles in advance. My expectations are still low, by your standards I guess, but I still find it fun

 

[jrista]

Who's to say A7s level performance won't be the norm with the next round of APS-C or high res FF releases? I am still a firm believer that total sensor area and Q.E. are the primary factors that affect noise. In the case of the A7s...it's clear that they are doing some rather intense noise reduction in the BionzX chip as well (much like the DIGIC 6 does). All three of those factors, IMO, are more important than pixel size in the long run. So I see no reason you couldn't have that level of ISO performance in a sensor with smaller pixels.

For now, if you need cropping ability, then pixel size is probably a key factor. Doesn't mean that A7s level performance will stay only in sensors with bigger pixels forever. (Or even necessarily for all that long.)

Why all the optimism for other brands but not this one? I imagine all cameras will get more capable as time goes on. But it seems rather optimistic that e.g. Sony will produce the perfect device for your needs.

That's a fair question, although I think I've answered it many times before. I've been watching...and waiting on...Canon for many years now. I didn't like their read noise in my first camera, the 450D. That was like 2008 timeframe...a LONG time ago now as far as DSLR cycles go.

I never liked the read noise of the 5D II, and the rumors that were coming out before the 5D III release indicated it would probably be 28mp with improved DR. So, I skipped the 5D II, and waited.

I got the 7D, then the 5D III...both STILL have Canon's signature nasty read noise.

I've been waiting for the 7D II for years...and the rumors now, although they could be dead wrong for sure...indicate that it is likely the 7D II won't even get a newly designed sensor.

Here I am, around six years later...still waiting for Canon to do something about their read noise. I've been defending Canon's high ISO performance for years as well...however, it seems even that stronghold is being toppled by...again, Sony technology. Canon has long had excellent ergonomics, perfect button placement, they have one of the best AF system on the market (although that is another territory that is in a close race between them an Nikon & Sony). When you get right down to it, though...fundamentally...all the other features of a camera are tools that assist the photographer in getting light focused on the sensor so the SENSOR can create a photograph. High ISO, low ISO, doesn't matter where you live...other companies (particularly Sony) are getting better IQ out of their sensors...and the support technology that helps photographers get that light focused on their sensors is also improving (very quickly, thanks to shorter iterations than Canon uses).

After a point, you just give up waiting and hoping for a company to do something they have demonstrated, time and time and time again, that they simply have no interest in addressing. I don't know if it's simply that they don't have the ABILITY to do anything about their sensor technology (i.e. they would need to build a new billion-dollar fab, and simply cannot afford the investment?), or that they don't have the DESIRE to do anything about their sensor technology (too comfortable in their dominant position?) Either way...nothing's been done.

The 7D II may change that. If it hits the streets with a REAL stop or two of more DR...then I might have more hope in Canon again. Early rumors about pending Canon releases tend to be wrong, sometimes wildly (28mp high DR 5D III?), but the late rumors tend to be much closer to reality. We haven't had a CR3 yet, but I think the chances of the 7D II being radically different than the current rumored specs is lower than chances of the 7D II being pretty darn close to the current rumored specs.

So...I have no faith in Canon to do anything about their read noise. It's been many years and nothing has been done about it. Therefor, I have no reason to think that Canon will suddenly, overnight, become a leader in that area again.

Astro is a very different beast. SNR, DR, and RN levels are all important. In astro, final signal strength or signal power is critical. For best results, you need to find the sweet spot that maximizes them all. I'd say before getting a different camera, start guiding your mount. More reliable tracking will probably do more for you than a better sensor, at least to start. I think that the A7s could certainly do better at high ISO than any current Canon sensor. It would certainly allow for the use of higher ISO settings...but I'm not sure that will really buy you much in the end. The key benefit of the increased DR of the A7s is simply a lower chance of clipping stars. In the long run, even with an A7s, guiding is going to allow you to expose for twice as long or longer than your current 2 minute exposures. (And, at 100mm, I would say even with a mount's periodic error, you should be able to expose for four minutes easily...unless the mount has some seriously wicked PE!)

In general ISO 3200, IMO, is rather high for astro unless your doing ultra wide field work (i.e. 16mm milky way imaging). For anything DSO related, you should be using the lowest ISO you can get away with and still have the minimal or nearly minimal read noise. That will maximize your DR and SNR, limiting the chance that stars will clip, maximizing your ability to increase the total signal power, and usually at no cost to your actual REAL exposure.

Thanks for the detailed reply.

I shot at the exposures that seemed right at the time. I have to admit, it's more of an intuitive thing, rather than working out the principles in advance. My expectations are still low, by your standards I guess, but I still find it fun

I think you will find that shooting at ISO 400 and shooting at ISO 3200 does not really result in any major difference in what you've actually exposed. The key difference is whether the stars get clipped or not. You would need to stretch the image more manually...but ISO 3200 is really just doing most of that stretching for you (and, probably not doing it as well).

Astrophotography is a highly manual form of photography. You take dozens or even hundreds of light frames, along with dark, flat, and bias frames for calibration. You calibrate then integrate (stack) to reduce noise. At that point, you want as linear a signal as you can get, as a LOT of the processing to clean up the image, deconvolve, reduce stars, and denoise, should generally be done in linear space before you stretch (basically...you want to maximize your dynamic range as much as humanly possible before stretching). The image your working with at this point usually looks nearly pitch black (unless you have really deep exposures...which you can usually only get at over five minutes, or up to 20 minutes with narrow band filters).

After maximizing SNR and deconvolving, then you stretch, and the image, almost in all it's final beauty, is finally revealed.

Canon sensors have long had a fairly linear signal. Even Canon, like most manufacturers, seems to do a minimal amount of "signal cooking", so the signals are not completely linear like a CCD. Nikon cameras have long had a black point that simply clipped any negative signal below it, turning it all to black. With the black point hack, Nikon cameras behave much more like Canon cameras...they have more noise in the negative signal below the black point, and all the information there, after you calibrate and integrate, is recoverable. It seems that once you hack the black point, Nikon cameras with Exmor and Toshiba sensors are even more linear than Canon cameras...so all the edits you perform in linear space before stretching (stretching uses non-linear curves to lift shadows without affecting highlights...since the star centroids might be very close to white) are more effective.

Anyway, this is generally what my images look like when I start (in a program called PixInsight):

This is the linear mode of a stack of 116 light frames calibrated with 30 flat frames and 200 bias frames. This is a "screen stretch" that shows the image signal buried deep within the black shadows:

LOT of information buried WAY deep in those shadows. At this point, with 116 frames stacked, my random noise has been lowered quite a bit, however the averaging actually tends to enhance Canon's banding, so you will probably notice a bit of that in the subsequent closeups here.

This is a 100% crop of Elephant Trunk, before any NR has been applied:

This is the same crop, after TGVDenoise has been applied:

TGV stands for Total Generalized Variation, a highly effective means of normalizing differences between adjacent samples of a signal. It effectively wipes out, completely, high frequency noise (as you can see). This leaves behind a bit of blotchiness, however it's easier to clean that up in non-linear, post-stretched space. I usually do quite a bit more work in linear before stretching, but just to show what a couple applications of noise reduction algorithms can do, here is the Elephant Trunk after stretching and an application of ACDNR:

ACDNR is applied with a mask to protect the higher SNR areas, primarily relegating the NR to lower SNR areas. ACDNR stands for Adaptive Contrast-Driven Noise reduction. It's great for reducing mid-frequency noise that often looks like "blotchiness", and is quite effective at reducing red-dominated color noise. This is the mask I used with it:

Finally, here is the image after a little bit more stretching and curves:

Don't be afraid of shooting at a lower ISO. Once you calibrate (which uses darks, biases, and flats to subtract out as much read noise as possible, and flatten the field to remove vignetting and dust spots) and stack (which averages together a bunch of frames to reduce random noise...the reduction is basically the SQRT(frameCountStacked)...so, stack 25 frames, you reduce noise by a factor of 5x, stack 100 frames, you reduce noise by a factor of 10x), you have a LOT of usable information buried in what will initially appear to be total, pitch black pixels. They aren't black, though...the SNR of a properly calibrated stack of images is far higher than any single frame that you may be used to using for regular terrestrial photography.

You might notice the larger stars in the image above look a bit bloated and very white. I shot that image at ISO 800 (elephant trunk is a very dim DSO)...so my stars actually clipped. I was having trouble getting 600s exposures without having problems with my stars getting egg-shaped (tracking considerations...discussion for another day), so I didn't really have any option but to use ISO 800 and use 300s exposures. But...it goes to show you how valuable dynamic range is when it comes to astrophotography...it is super easy to clip stars. I probably could have gotten away with ISO 400 and 360-420 second exposures...but it would have taken a little more experimentation to make sure the red channel exposed above the RN floor...and I had a very short window of opportunity between storms here in Colorado to get the number of sub frames I needed. Those are the kinds of considerations you should make when your in a pinch...otherwise...I pretty much use ISO 400 for the majority of my astro work.

 

[scyrene]

That's a fair question, although I think I've answered it many times before. I've been watching...and waiting on...Canon for many years now. I didn't like their read noise in my first camera, the 450D. That was like 2008 timeframe...a LONG time ago now as far as DSLR cycles go.

I never liked the read noise of the 5D II, and the rumors that were coming out before the 5D III release indicated it would probably be 28mp with improved DR. So, I skipped the 5D II, and waited.

I got the 7D, then the 5D III...both STILL have Canon's signature nasty read noise.

I suppose I'm just a lot more laid back about it all. I started with a 300D. Upgrading to a 50D was amazing. Going from that to a 5DIII was even more amazing. I don't feel like I'm limited by the camera body yet. If and when I upgrade, it may or may not be to a Canon body (although as with many others, it's easier to stick with them because of the lenses I have). But I'm sure my next camera will be even better, although I suspect the difference won't be as big. But I don't feel any anger or frustration with any given company. I look at th products on offer and choose what fits my needs and budget.

As for read noise, I guess I'm a lot less discerning - your work must be a lot more technical. There are limitations in every area, but that's life. I work with what I've got.

Astrophotography is a highly manual form of photography. You take dozens or even hundreds of light frames, along with dark, flat, and bias frames for calibration. You calibrate then integrate (stack) to reduce noise. At that point, you want as linear a signal as you can get, as a LOT of the processing to clean up the image, deconvolve, reduce stars, and denoise, should generally be done in linear space before you stretch (basically...you want to maximize your dynamic range as much as humanly possible before stretching). The image your working with at this point usually looks nearly pitch black (unless you have really deep exposures...which you can usually only get at over five minutes, or up to 20 minutes with narrow band filters).

After maximizing SNR and deconvolving, then you stretch, and the image, almost in all it's final beauty, is finally revealed.

The mount I use is a budget model, designed for cameras. It's not great, but it still improved what I could achieve massively. Maybe 4 mins is possible, although there are so many aircraft and satellites that I have to discard more frames the longer they are. I've done the whole flats, lights, darks, bias, etc. I actually do it entirely manually, stacking and aligning by hand, because the stacking software (as I think I've mentioned) tends to be PCs-only. It's quite therapeutic actually

But tbh it doesn't bother me that other manufacturers are coming up with better solutions for this. Astro work will always be a minor pursuit for me. The number of days when it's not cloudy, or too windy, or too humid, and I'm in a location with low enough light pollution to do useful work, are very few indeed. So it's not worth me spending tonnes of money on it. If I did, I'd probably go down the route of a dedicated astro camera, either a modified DSLR or a CCD, but that's way too expensive for something I do every few months. I'll never be able to compete with you astro enthusiasts, especially mountain dwellers

As for blowing stars... I don't see the problem. I think it looks okay. Maybe I'm missing something (unless the subject is the stars themselves, like in globular clusters, then I guess you expose for them).

 

[jrista]

That's a fair question, although I think I've answered it many times before. I've been watching...and waiting on...Canon for many years now. I didn't like their read noise in my first camera, the 450D. That was like 2008 timeframe...a LONG time ago now as far as DSLR cycles go.

I never liked the read noise of the 5D II, and the rumors that were coming out before the 5D III release indicated it would probably be 28mp with improved DR. So, I skipped the 5D II, and waited.

I got the 7D, then the 5D III...both STILL have Canon's signature nasty read noise.

I suppose I'm just a lot more laid back about it all. I started with a 300D. Upgrading to a 50D was amazing. Going from that to a 5DIII was even more amazing. I don't feel like I'm limited by the camera body yet. If and when I upgrade, it may or may not be to a Canon body (although as with many others, it's easier to stick with them because of the lenses I have). But I'm sure my next camera will be even better, although I suspect the difference won't be as big. But I don't feel any anger or frustration with any given company. I look at th products on offer and choose what fits my needs and budget.

As for read noise, I guess I'm a lot less discerning - your work must be a lot more technical. There are limitations in every area, but that's life. I work with what I've got.

Astrophotography is a highly manual form of photography. You take dozens or even hundreds of light frames, along with dark, flat, and bias frames for calibration. You calibrate then integrate (stack) to reduce noise. At that point, you want as linear a signal as you can get, as a LOT of the processing to clean up the image, deconvolve, reduce stars, and denoise, should generally be done in linear space before you stretch (basically...you want to maximize your dynamic range as much as humanly possible before stretching). The image your working with at this point usually looks nearly pitch black (unless you have really deep exposures...which you can usually only get at over five minutes, or up to 20 minutes with narrow band filters).

After maximizing SNR and deconvolving, then you stretch, and the image, almost in all it's final beauty, is finally revealed.

The mount I use is a budget model, designed for cameras. It's not great, but it still improved what I could achieve massively. Maybe 4 mins is possible, although there are so many aircraft and satellites that I have to discard more frames the longer they are. I've done the whole flats, lights, darks, bias, etc. I actually do it entirely manually, stacking and aligning by hand, because the stacking software (as I think I've mentioned) tends to be PCs-only. It's quite therapeutic actually

But tbh it doesn't bother me that other manufacturers are coming up with better solutions for this. Astro work will always be a minor pursuit for me. The number of days when it's not cloudy, or too windy, or too humid, and I'm in a location with low enough light pollution to do useful work, are very few indeed. So it's not worth me spending tonnes of money on it. If I did, I'd probably go down the route of a dedicated astro camera, either a modified DSLR or a CCD, but that's way too expensive for something I do every few months. I'll never be able to compete with you astro enthusiasts, especially mountain dwellers

As for blowing stars... I don't see the problem. I think it looks okay. Maybe I'm missing something (unless the subject is the stars themselves, like in globular clusters, then I guess you expose for them).

Blown stars can completely dominate an image. There are star reduction routines, but once stars start to clip (which is pretty much guaranteed at ISO 3200 unless your on the A7s), then they "bloat". Stars should generally only be about 3 pixels in size except for the largest, and the largest should look more like gaussian spots rather than solid white discs.

Again, it may just be the technical side of things...although I've been learning the in's and out's of astro processing from guys who are as technical as I am, and they care about minutia that I can't even recognize yet (some of thee really good ones can tell which color channel is bloating, even...some NB imagers will often tell you when they see your OIII channel bloating, which is creating a halo around your otherwise perfect Ha stars. )

Anyway, maybe it is just the difference in expectations...I don't like to fight against my hardware. I'll be dropping $4000+ on a CCD camera so I can stop fighting with the noise in my DSLRs. I do have quite a bit of lens lockin when it comes to regular photography...which is the large part of why I want Canon to improve their sensors. Now not just at low ISO...but at high ISO as well. I want that ~9 stops of DR at ISO 51200 (that is just frickin phenomenal)...it all just means I have more room within which to work, and I wouldn't have to worry as much about technical minutia like...am I going to clip my highlights if I ETTR, am I going to have noisy shadows, etc.

 

[ULFULFSEN]

From a consumers point of view it´s complete bollocks to defend technology that is worse than competing technology.

Even when you don´t need the improvements.
I don´t need 100 crosstype AF points and 10 FPS.
But i would never say it´s useless or create workarounds to show you can shot great images with worse technology too.



You don´t buy a Hamburger that tastes worse than another, right?
You don´t tell other Hamburger fans he is as good as the other because both offer the same carbs, vitamins, proteins etc.
You buy the Burger that tastes delicious.

So people should stop making excuses for Canon.
Canon still makes good cameras, but the sensors are not up to the competition in some cases. If you are affected by these cases doesn´t matter.

 

[privatebydesign]

From a consumers point of view it´s complete bollocks to defend technology that is worse than competing technology.

You don´t buy a Hamburger that tastes worse than another, right?
You don´t tell other Hamburger fans he is as good as the other because both offer the same carbs, vitamins, proteins etc.
You buy the Burger that tastes delicious.

So stop making excuses for Canon.
Canon still makes good cameras, but the sensors are not up to the competition.

That is a faulty analogy, you don't buy a hamburger if you want chicken; if your primary need is sensor output get one of the Exmor cameras, if you want Canon lenses, and they have many unique and class leading lenses, flashes (and nobody else makes a factory radio flash system) etc etc then a Canon IS the better buy.

For the billionth time, Canon does not hold the lead in sensor output especially at low iso, but it does have many other system advantages over competitors systems, and for many keen amateurs, semi pros, and pros those system advantages are bigger factors than the sensor differences. The low end is dominated by price and market share and everybody agrees Canon has been leader there for many years.

 

[ULFULFSEN]

That is a faulty analogy, you don't buy a hamburger if you want chicken; if your primary need is sensor output get one of the Exmor cameras, if you want Canon lenses, and they have many unique and class leading lenses, flashes (and nobody else makes a factory radio flash system) etc etc then a Canon IS the better buy.

What you are talking about is a menu.
I am talking about a single Hamburger (only the sensors)

For the billionth time, Canon does not hold the lead in sensor output especially at low iso, but it does have many other system advantages over competitors systems....

I did say that didn´t i?

Canon still makes good cameras, but the sensors are not up to the competition

 

[privatebydesign]

That is a faulty analogy, you don't buy a hamburger if you want chicken; if your primary need is sensor output get one of the Exmor cameras, if you want Canon lenses, and they have many unique and class leading lenses, flashes (and nobody else makes a factory radio flash system) etc etc then a Canon IS the better buy.

What you are talking about is a menu.
I am talking about a single Hamburger (only the sensors)

For the billionth time, Canon does not hold the lead in sensor output especially at low iso, but it does have many other system advantages over competitors systems....

I did say that didn´t i?

Canon still makes good cameras, but the sensors are not up to the competition

But if everything else on the menu is more appealing than a hamburger, it doesn't matter how "good" it is.

Look, people banging on about how bad/far behind Canon sensors are, and they are, seem to miss the point that for the vast amount of the time they are still more than good enough to actually achieve what you need.

Even if the burger is the best burger in the world if only need a single one, why would you order a triple burger?

Other factors are far more important to most people most of the time than more DR and more MP, and that is a fact supported by sales data.

 

[Loswr]

Look, people banging on about how bad/far behind Canon sensors are, and they are, seem to miss the point that for the vast amount of the time they are still more than good enough to actually achieve what you need.

^^ THIS ^^