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

Is Sony Really Alpha?

  • Please log in to reply
2186 replies to this topic

#76 alan.dang

alan.dang

    Viking 1

  • *****
  • Posts: 564
  • Joined: 15 Dec 2011

Posted 27 January 2015 - 12:28 AM

I don't like the Sony's because their RAW compression is lossy.  Lossy compression leads to posterization artifacts.

 

http://www.slrlounge...ull-frame-cmos/

 

I like my 60Da and given Roger Clark's data, I agree that the 6D and 7D Mark II are your best options for astrophotography.

 

I need to play with the NX1 some more before making a decision but so far I've been impressed.  I am able to pull out more data than I can with the 60Da for non-Ha sources.  I do need a bit more time with the sensor (and clearer/darker skies).



#77 mmalik

mmalik

    DSLR camera modifications

  • *****
  • Vendors
  • topic starter
  • Posts: 11,898
  • Joined: 13 Jan 2012
  • Loc: USA

Posted 27 January 2015 - 03:34 AM

Alan, what exactly is posterization? Are these color gradients? There is one line in your write-up, "Sony’s lossy compressed RAW format generates posterization artifacts that are not present on the Samsung’s format."

 

 

Plus you tested a7R; I think it is vastly different from a7/a7II/a7S:

 

 

a7s..................12MP

a7/a7II.............24MP

a7R..................36MP

 

 

Sony's that I am suggesting for astro mod are a7S and a7/a7II; could it be that there is selection bias in your experimentation? I.e., I was NOT referring to a7R (the one you tested...)

 

 

Here is DXOMark comparison of a7II/6D:

 

 

Overall Score......................90/82

Color Depth........................24.9/23.8 bits

Dynamic Range..................13.6/12.1 Evs

Low-Light ISO....................2449/2340 ISO

 

 

How is it possible that posterization, if true, would have been missed in DXOMark scores?

 

 

I have high hopes for NX1, but DXOMak data is not there yet. Regards


Edited by mmalik, 27 January 2015 - 03:49 AM.


#78 orlyandico

orlyandico

    Cosmos

  • *****
  • Posts: 9,862
  • Joined: 10 Aug 2009
  • Loc: Singapore

Posted 27 January 2015 - 04:16 AM

seems the 0.4e- read noise is a canard - because you get that at ISO 409600 as per sensorgen... at which the full well is a measly 55e-

 

who uses ISO 409600?

 

all sensorgen numbers: if we assume (measured) ISO 1600 equivalent here are some numbers:

 

Olympus E-PL5 (Sony IMX109) - ISO 3200, measured ISO 1668, read noise 2.1e-, full well 1463e-, DR 9.4, QE 60%

Sony A6000 - ISO 3200, measured ISO 2510, read noise 2.5e-, full well 1064e-, DR 8.7, QE 61%

Sony A7S - ISO 3200, measured ISO 2627, read noise 4.1e-, full well 5453, DR 10.4, QE 65%

Canon 6D - ISO 3200, measured ISO 2400, read noise 2.3e-, full well 2481,  DR 10.1, QE 47%

 

So the A7S is not "head and shoulders" above the 6D - it actually has almost double the read noise; but it also has double the full well.

 

It also looks to me that, if you stick with ISO 3200, the A6000 is actually a "better" deal than the A7S with its lower read noise. Of course the full well is only 20% so you'll get a lot of saturated stars.

 

It seems that at ISO below 12800, the 6D actually has lower read noise.  And I fail to see the benefit of ultra-high ISO because the full well is very shallow at those gain levels..


Edited by orlyandico, 27 January 2015 - 04:19 AM.


#79 mmalik

mmalik

    DSLR camera modifications

  • *****
  • Vendors
  • topic starter
  • Posts: 11,898
  • Joined: 13 Jan 2012
  • Loc: USA

Posted 27 January 2015 - 04:50 AM

Orly, some of what you are referring to about read noise, rightly so, plotted here...; there is more to the story in terms of QE... and other... parameters. True, ultra-high ISO is practically useless. Regards


Edited by mmalik, 27 January 2015 - 05:13 AM.


#80 mmalik

mmalik

    DSLR camera modifications

  • *****
  • Vendors
  • topic starter
  • Posts: 11,898
  • Joined: 13 Jan 2012
  • Loc: USA

Posted 27 January 2015 - 04:57 AM

a7/a7II seems more in the

 

So the A7S is not "head and shoulders" above the 6D...

 

 

a7/a7II seem more in the league of 6D in terms if pixel size; a comparison above...


Edited by mmalik, 27 January 2015 - 04:59 AM.


#81 orlyandico

orlyandico

    Cosmos

  • *****
  • Posts: 9,862
  • Joined: 10 Aug 2009
  • Loc: Singapore

Posted 27 January 2015 - 05:14 AM

I did some number crunching, and got these (all data from sensorgen).  The Q1 factor is basically just the product of pixel area (pixel pitch ^2), quantum efficiency, and the measured effective ISO when ISO is set to ISO 3200.  I chose ISO 3200 because the super-high ISO's are practically useless.

 

In other words, Q1 is directly analogous to "light gathering ability" (image scale taken out of the equation).  Obviously the large-sensor cameras come out on top, and the A7S comes out on top by a large-ish margin.

 

Here's the graph sorted by "light gathering ability" alone:

 

Cameras-by-Light-Gathering.png.jpg

 

Sony A7S is way on top here due to the fat pixels. It literally has more than double the light-gathering of the second-ranking Canon 6D.

 

I also calculate a Q2, which is Q1 divided by the read noise.  Q2 corresponds almost 1:1 with DR.

 

Cameras-by-Q.png.jpg

 

Here again the A7S comes out on top, albeit by about 40% over the 6D. Still quite significant.  The 0.6e- read noise of the A7S doesn't factor here because we're not using ridiculously high ISO's.

 

It would be nice if I could also calculate a Q2/dollar to get the best bang for the buck, but life's too short..


Edited by orlyandico, 27 January 2015 - 05:17 AM.


#82 orlyandico

orlyandico

    Cosmos

  • *****
  • Posts: 9,862
  • Joined: 10 Aug 2009
  • Loc: Singapore

Posted 27 January 2015 - 05:33 AM

I added the other A7 bodies to my list.   Seems that if we go by sensorgen info, at ISO 3200 setting, the 6D still comes out ahead of the A7 and A7R due to lower read noise and fatter pixels.

 

Cameras-by-Q2.png

 

Well this has been an enlightening exercise for myself (based on the thread I started on new DSLR's).  Seems that yes, a newer DSLR than my old Canon 40D would provide about factor-of-2 improvement in SNR (3dB) which means double the signal, or half the integration time.

 

And the full-frame ones would give about factor-of-4.


Edited by orlyandico, 27 January 2015 - 05:37 AM.


#83 mmalik

mmalik

    DSLR camera modifications

  • *****
  • Vendors
  • topic starter
  • Posts: 11,898
  • Joined: 13 Jan 2012
  • Loc: USA

Posted 27 January 2015 - 05:54 AM

"light gathering factor" Q1 = (pixel pitch ^2) * QE * effective_ISO

"quality factor" Q2 = Q1 / Read_Noise

 

Good analysis! You read my mind; I was about to ask you to include a7/a7II which you did. I think best bang for the buck at this point is a7, followed by 6D. a7S of course is too pricey for most. More on prices here.... You may want to crunch 7DII as well, regardless of somewhat equivalent price as 6D. Regards


Edited by mmalik, 27 January 2015 - 06:04 AM.


#84 orlyandico

orlyandico

    Cosmos

  • *****
  • Posts: 9,862
  • Joined: 10 Aug 2009
  • Loc: Singapore

Posted 27 January 2015 - 06:09 AM

That D5200 sure looks nice... because BackyardNik.

 

And I don't want to pay the big bucks for an FSQ85 or 106.

 

I added the 7D II and a few other cheap models.. if sensorgen data is to be believed, the 7D II is the "best" APS-C DSLR for AP.  However it's "Q2 per buck" is a whopping 1/2 that of the 6D (considering that, for me at least, they cost the same).

 

Cameras-by-Q2.png

 

A used D5200 still seems to be the best bang for buck for astro alone (keeping in mind I have a lot of Canon gear).


Edited by orlyandico, 27 January 2015 - 06:23 AM.


#85 mmalik

mmalik

    DSLR camera modifications

  • *****
  • Vendors
  • topic starter
  • Posts: 11,898
  • Joined: 13 Jan 2012
  • Loc: USA

Posted 27 January 2015 - 01:30 PM

Q1, Q2 comparison at ISO1600:

 

 

Q1, Q2 definitions...

 

Data source: sensorgen.info...

Attached Thumbnails

  • Q1Q2.JPG
  • Q1.JPG

Edited by mmalik, 28 January 2015 - 05:57 AM.


#86 mmalik

mmalik

    DSLR camera modifications

  • *****
  • Vendors
  • topic starter
  • Posts: 11,898
  • Joined: 13 Jan 2012
  • Loc: USA

Posted 28 January 2015 - 06:04 AM

Here is what things look like (the same) after Orly's... LOG2 conversion of Q1 at ISO 1600:


Q1=LOG(PixelSize^2*QE*MeasuredISO,2)...

Data source: sensorgen.info...

Attached Thumbnails

  • Q1_Log2.JPG

Edited by mmalik, 28 January 2015 - 06:09 AM.

  • Jon Rista likes this

#87 mmalik

mmalik

    DSLR camera modifications

  • *****
  • Vendors
  • topic starter
  • Posts: 11,898
  • Joined: 13 Jan 2012
  • Loc: USA

Posted 28 January 2015 - 06:20 AM

Q2 derived from Q1 LOG2 at ISO 1600; interesting...
 
 
Q2=(LOG(PixelSize^2*QE*MeasuredISO,2))/ReadNoise...

Data source: sensorgen.info...

Attached Thumbnails

  • Q2_Log2Derived.JPG

Edited by mmalik, 28 January 2015 - 04:07 PM.


#88 GJJim

GJJim

    Skylab

  • *****
  • Posts: 4,024
  • Joined: 09 Sep 2006
  • Loc: más y menos

Posted 28 January 2015 - 09:21 AM

Frank Lloyd Wright cautioned that no house should be built at the top of a hill. Instead, the house should be of the hill, both living in harmony.  :grin:



#89 mpgxsvcd

mpgxsvcd

    Vanguard

  • -----
  • Posts: 2,056
  • Joined: 21 Dec 2011
  • Loc: Raleigh, North Carolina

Posted 28 January 2015 - 11:17 AM

I added the other A7 bodies to my list.   Seems that if we go by sensorgen info, at ISO 3200 setting, the 6D still comes out ahead of the A7 and A7R due to lower read noise and fatter pixels.

 

Cameras-by-Q2.png

 

Well this has been an enlightening exercise for myself (based on the thread I started on new DSLR's).  Seems that yes, a newer DSLR than my old Canon 40D would provide about factor-of-2 improvement in SNR (3dB) which means double the signal, or half the integration time.

 

And the full-frame ones would give about factor-of-4.

Your chart really depends on the Actual ISO versus Real ISO value more than anything else. I can only assume that you pulled that value directly from DXO Mark since the numbers match its data exactly.

 

When I first bought the Olympus E-PM2 for Astro Photography I asked the exact same thing about it. If the DXO Mark value is so low for real ISO vs reported ISO then the camera would be terrible especially at longer shutter durations.

 

The dpreview staff completely debunked that stat that DXO Mark is reporting. However, I wanted to confirm that with real world testing. I tested several cameras(Panasonic GF1, GH2, GH3, GH4, Olympus E-PM2) on the exact same scope in a single night.

 

The Panasonic GH2 has the best actual ISO vs reported ISO of any of those cameras and it actually exceeds the value of the Nikon D3100 which was tops in your list.

 

I can definitively say that all of the cameras I tested produced exactly the same exposure with the same settings on the exact same scope on the same night. However, the GH2 was the only one of them that produced a noticeably horrible image. It had excessive noise and serious banding issues at all ISO values. I sold it the next day because it really was not a good stills camera especially for AP.

 

The GH3 and Olympus E-PM2 were rock solid in every aspect. Both are exceptional cameras as the QE value(62%) in your chart indicates. Both cameras were rumored to share the same or a similar sensor. The Olympus one was confirmed to be the excellent Sony sensor while Panasonic would never admit what they used. However, both cameras perform nearly identically in every situation so I have to assume the sensors are comparable.

http://www.dxomark.c..._664_677#tabs-2

 

Personally I think your Q1 and Q2 numbers are bogus. I don’t believe the DXO Actual vs. Reported ISO values at all and the testing I have done leads me to believe that. In addition the expert staff at dpreview has confirmed the same thing. The other numbers look right though.

 

If we look at everything but your Q1 and Q2 numbers then the only real difference is that the Olympus E-PM2 rockets up the chart from nearly last to at least top 5 if not 2nd depending on what stat you think is the most important.

If I was you I wouldn’t go around touting your Q1 and Q2 numbers as the ultimate statistic especially since all other testing contradicts those conclusions. You are much better off evenly weighting each camera based on all of the other columns in your chart. From my experience that will give you a much better idea of real world performance.


Edited by mpgxsvcd, 28 January 2015 - 11:21 AM.


#90 Jon Rista

Jon Rista

    ISS

  • *****
  • Posts: 24,854
  • Joined: 10 Jan 2014
  • Loc: Colorado

Posted 28 January 2015 - 03:26 PM

seems the 0.4e- read noise is a canard - because you get that at ISO 409600 as per sensorgen... at which the full well is a measly 55e-

 

who uses ISO 409600?

 

all sensorgen numbers: if we assume (measured) ISO 1600 equivalent here are some numbers:

 

Olympus E-PL5 (Sony IMX109) - ISO 3200, measured ISO 1668, read noise 2.1e-, full well 1463e-, DR 9.4, QE 60%

Sony A6000 - ISO 3200, measured ISO 2510, read noise 2.5e-, full well 1064e-, DR 8.7, QE 61%

Sony A7S - ISO 3200, measured ISO 2627, read noise 4.1e-, full well 5453, DR 10.4, QE 65%

Canon 6D - ISO 3200, measured ISO 2400, read noise 2.3e-, full well 2481,  DR 10.1, QE 47%

 

So the A7S is not "head and shoulders" above the 6D - it actually has almost double the read noise; but it also has double the full well.

 

It also looks to me that, if you stick with ISO 3200, the A6000 is actually a "better" deal than the A7S with its lower read noise. Of course the full well is only 20% so you'll get a lot of saturated stars.

 

It seems that at ISO below 12800, the 6D actually has lower read noise.  And I fail to see the benefit of ultra-high ISO because the full well is very shallow at those gain levels..

 

Comparing everything at ISO 3200, and indeed, the A7s has higher read noise. Because the A7s has twice the FWC, you could always bump it up to ISO 6400, get lower read noise, still have a higher FWC, and have the higher Q.E. to boot.

 

There is something else about Q.E....it affects your imaging rate. The 65% Q.E. of the A7s means you can reduce your exposure times by almost 40% at any given exposure setting relative to say the 6D. The 7D II is a great AP camera for the same reason, it's 59% Q.E. That lets you expose for less time at the same exposure settings. Or, conversely, expose at the same time and get a higher SNR as a result. 

 

Throw in the huge pixels of the A7s on top of it's high Q.E., and your per-pixel SNR just gets that much better. I think that's why it tops the charts in both your Q1 and Q2 factor tests. 



#91 mmalik

mmalik

    DSLR camera modifications

  • *****
  • Vendors
  • topic starter
  • Posts: 11,898
  • Joined: 13 Jan 2012
  • Loc: USA

Posted 28 January 2015 - 03:55 PM

A look at Orly's... Q1 LOG2 at ISO 6400:


Q1=LOG(PixelSize^2*QE*MeasuredISO,2)...

Data source: sensorgen.info...

Attached Thumbnails

  • Q1_Log2-ISO6400.JPG

Edited by mmalik, 28 January 2015 - 03:55 PM.


#92 mmalik

mmalik

    DSLR camera modifications

  • *****
  • Vendors
  • topic starter
  • Posts: 11,898
  • Joined: 13 Jan 2012
  • Loc: USA

Posted 28 January 2015 - 03:56 PM

Q2 derived from Q1 LOG2 at ISO 6400; interesting...


Q2=(LOG(PixelSize^2*QE*MeasuredISO,2))/ReadNoise...

Data source: sensorgen.info...

Attached Thumbnails

  • Q2_Log2Derived-ISO6400.JPG

Edited by mmalik, 28 January 2015 - 04:07 PM.


#93 Jon Rista

Jon Rista

    ISS

  • *****
  • Posts: 24,854
  • Joined: 10 Jan 2014
  • Loc: Colorado

Posted 28 January 2015 - 04:02 PM

What I would be curious in seeing is a plot of cameras at the optimal ISO. At what ISO do you get an optimal balance of ISO, read noise, and well depth? We keep comparing cameras at the same settings...but I am not sure that is entirely valid. Clearly, the A7s can be used at a higher ISO setting, and you get a similar FWC, similar RN, to the 6D at a stop lower ISO. Is it possible to "normalize" different cameras to an "optimal" state, and compare them that way? Maybe we just start with a single factor...at what ISO does each camera end up with the same general FWC, say ~2500e-? 

 

Just taking three of the better AP cameras:

 

A7s: ISO 6400/2.9e- RN/2707e- FWC

7DII: ISO 1600/2.8e- RN/2509e- FWC

6D: ISO 3200/2.3e- RN/2481e- FWC

 

Are these cameras roughly equivalent at these settings? The theory of equivalence (not on a sensor-area basis, but on an absolute-area basis...pick any given area of sensor, say 5x5mm, and compare...these three camera would produce equivalent normalized results at the above settings) would generally say yes...but I'm curious what everyone thinks the case may be for astrophotography...

 

Technically speaking, with the above settings for these three cameras, the A7s would then be able to be exposed 1/4 as long as the 7D II, and 1/2 as long as the 6D, to get roughly identical results. If we factor in Q.E., the A7s could probably be exposed even less time than that, and get similar results:

 

A7s: ISO 6400/2.9e- RN/2707e- FWC/65%
7DII: ISO 1600/2.8e- RN/2509e- FWC/59%
6D: ISO 3200/2.3e- RN/2481e- FWC/47%


Edited by Jon Rista, 28 January 2015 - 04:08 PM.


#94 mmalik

mmalik

    DSLR camera modifications

  • *****
  • Vendors
  • topic starter
  • Posts: 11,898
  • Joined: 13 Jan 2012
  • Loc: USA

Posted 28 January 2015 - 04:05 PM

...much better off evenly weighting each camera based on all of the other columns in your chart. From my experience that will give you a much better idea of real world performance.

 

I tend to agree with you and I that's what I have mostly done in the past...; I see no harm giving different ways of number crunching a try. Open to seeing your number crunching methods. Regards


Edited by mmalik, 28 January 2015 - 04:11 PM.


#95 boardriderz

boardriderz

    Explorer 1

  • -----
  • Posts: 81
  • Joined: 04 Aug 2010
  • Loc: Brisbane, QLD, Australia

Posted 28 January 2015 - 04:07 PM

Hi guys, first post a a few years, but I have been ghosting for a while,

 

If you are really looking at bang for your buck, shouldn't you add the Nikon D3300 / D3200 and the Canon 100D to the list as these are the latest versions of those sensors in the cheapest package? Under $350 for a new D3200 online and just over $400 for the D3300?



#96 mpgxsvcd

mpgxsvcd

    Vanguard

  • -----
  • Posts: 2,056
  • Joined: 21 Dec 2011
  • Loc: Raleigh, North Carolina

Posted 28 January 2015 - 04:12 PM


Comparing everything at ISO 3200, and indeed, the A7s has higher read noise. Because the A7s has twice the FWC, you could always bump it up to ISO 6400, get lower read noise, still have a higher FWC, and have the higher Q.E. to boot.

 

There is something else about Q.E....it affects your imaging rate. The 65% Q.E. of the A7s means you can reduce your exposure times by almost 40% at any given exposure setting relative to say the 6D. The 7D II is a great AP camera for the same reason, it's 59% Q.E. That lets you expose for less time at the same exposure settings. Or, conversely, expose at the same time and get a higher SNR as a result. 

 

Throw in the huge pixels of the A7s on top of it's high Q.E., and your per-pixel SNR just gets that much better. I think that's why it tops the charts in both your Q1 and Q2 factor tests. 

 

 

Using that same logic the Olympus E-PM2 looks like an absolute beast for A.P. since its Q.E. is 62% and its pixel size is very close to the 7D MKII at 3.7 vs. 4.2. Remember the Olympus E-PM2 is about 1/8 the cost of the 7D MKII.



#97 Jon Rista

Jon Rista

    ISS

  • *****
  • Posts: 24,854
  • Joined: 10 Jan 2014
  • Loc: Colorado

Posted 28 January 2015 - 04:15 PM

I would think the smaller pixels of the Olympus would work against it, countering a little bit of it's Q.E. advantage. Larger pixels, higher Q.E. support a higher ISO. Imaging at a higher ISO means imaging FASTER. Imaging faster means getting more signal in any given total amount of time. More signal, better images. Now, that totally ignores cost... :p At 1/8th the cost of a 7D II, sure, there is definitely some bang for the buck there.



#98 mpgxsvcd

mpgxsvcd

    Vanguard

  • -----
  • Posts: 2,056
  • Joined: 21 Dec 2011
  • Loc: Raleigh, North Carolina

Posted 28 January 2015 - 08:16 PM

 

...much better off evenly weighting each camera based on all of the other columns in your chart. From my experience that will give you a much better idea of real world performance.

 

I tend to agree with you and I that's what I have mostly done in the past...; I see no harm giving different ways of number crunching a try. Open to seeing your number crunching methods. Regards

 

 

I did a few calculations by subtracting out the Equivalent ISO column. I also added in the Amazon price for each camera.

The first diagram below is considering all parameters except for Equivalent ISO and including price. This is the one that I think is the most realistic for the actual performance per dollar of each camera. However, fixed pattern noise and other noise pattern noise are not reflected in any of these tests so that could produce undesirable real world results.

 

I believe the takeaway from these charts is that the Canon 550D is never a good choice if you are only considering these parameters even if cost is a factor. With this data it never finishes ahead of the Olympus E-PM2. However, I completely understand that other parameters that are not in this chart like BYEOS could sway someone to use the 550D. However, it clearly is not a good choice for image quality even when price is considered as well.

The A7S is clearly the leader on paper. It just simply isn't even close when you only look at these parameters. However, there are reports of undesirable noise patterns and other anomalies that could make it not the best choice.

Attached Thumbnails

  • Camera Comparison Full with Price.jpg

Edited by mpgxsvcd, 29 January 2015 - 11:38 AM.


#99 mpgxsvcd

mpgxsvcd

    Vanguard

  • -----
  • Posts: 2,056
  • Joined: 21 Dec 2011
  • Loc: Raleigh, North Carolina

Posted 28 January 2015 - 08:19 PM

Here are a few more calculations I ran by subtracting off different criteria. I don't believe these are as realistic but they are still interesting.

Attached Thumbnails

  • Camera Comparison Full.jpg


#100 mpgxsvcd

mpgxsvcd

    Vanguard

  • -----
  • Posts: 2,056
  • Joined: 21 Dec 2011
  • Loc: Raleigh, North Carolina

Posted 28 January 2015 - 08:20 PM

Another one.

Attached Thumbnails

  • Camera Comparison without Read Noise and Sensor um.jpg

Edited by mpgxsvcd, 28 January 2015 - 08:21 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