Review of the Sony A7S

[sharkmelley]

I recently bought a brand new Sony A7S full frame mirrorless camera with the intention of finally replacing my ageing but trustworthy modified Canon 350D. The A7S is the camera famed for seeing the Milky Way in liveview.  Put a 50mm F1.8 lens on it and through the electronic viewfinder in liveview you'll see hundreds of stars invisible to the naked eye. A 1 second exposure makes the spiral arms of galaxies being to appear.

There's a good review of it here: http://petapixel.com/2014/07/30/sony-a7s-astrophotography-review
And here it is imaging auroras in real time in Norway at 25 frames per second: http://www.cloudynights.com/topic/480866-fast-moving-real-time-auroras-in-norway/

These are certainly its party tricks - enabled by its ridiculously low read noise which drops below 1 electron (RMS) at high ISOs. Real time integration of images in software (Electronically Assisted Astronomy) makes this a great instrument for public outreach events.

But I'm jumping ahead.  Rather than repeating other reviews my intention is to present some (limited) personal experience and other technical information not easily found elsewhere.

In this post I'll present a quick summary.  More detailed information later. Apologies in advance for the technical nature of this but this is the stuff I wish I'd known in advance in order to make a more informed decision before buying the camera.

As an overview of the Sony A7S, it's an interchangeable lens, full frame camera with relatively large pixels - only 12.2 megapixels on a full frame - bucking the trend towards tens of megapixels.

The main advantages are these:

• Compared with DSLRs (especially full-frame DSLRs) the camera body is very compact and light in weight.
• In silent mode the shutter does not activate - meaning no shutter wear.
• At ISO 2000 and above, read noise is 1.3e and falling to well under 1e at very high ISOs.  Thierry Legault has done a very useful analysis of read noise here http://www.astrophoto.fr/sony_a7s_measures.html  I have also verified these figures with my own experiments.  It is believed that the sensor has 2 modes for analogue to digital conversion and it switches to the low noise mode at ISO 2000 and above - hence the step in the curve.
• Very low thermal noise means that for astrophotography in the UK (i.e. the complete absence of warm nights!) even on a slowish scope such as F6, skyglow will always dominate thermal noise (unless you are using narrowband filters).  This opens up the possibility of using the A7S on a whole range of scopes that previously were not that suitable for DLSR or mirrorless cameras.  My own experiments indicate a dark current of 0.15e/pixel/sec after 2 hours of solid imaging at an ambient temperature of 20C.  At around 8C this drops to 0.04e/pixel/sec.  So at 8C ambient temperature, even a long 20 minute exposure would have a thermal noise of 7e which is less than the read noise of a typical cooled astro-CCD
• With Wi-Fi switched off (airplane mode) and rear LCD deactivated (FINDER/MONITOR set to Viewfinder) battery life is 6 hours for an imaging session of continuous long exposures.  Low power consumption = low generated heat.
• Bias frames have a non zero-offset.  This means that proper dark frame calibration can be used - just like Canon but unlike Nikon DSLR cameras which clip dark frames making them unusable (unless you have installed the Nikon "hack").  Many people say that the noise levels are so low that dark frames are unnecessary.  But I would counter this by saying that good dark frame calibration allows even more faint data to be stretched out of the image.
• It is probably the most sensitive low light consumer camera in existence at the current time. But if you know of a better one, please tell me!

The main disadvantages are these:

• Full frame sensor means you need a large corrected image circle on your imaging equipment unless you are prepared to crop the image
• Large pixel size means that many folk will suggest it is suitable only for longer focal lengths.  Personally I see no reason not to use it with a wide angle lens.  After all, this is what DSLR and mirrorless cameras are good at!
• The narrow throat of the E-mount lens fitting means that very severe vignetting can occur at the corners of the sensor on a slow F-ratio scope.  This is not such a problem for fast scopes and not a problem for lenses.  However, on a C11 with F8 flattener the extreme corners were actually in complete shadow.
• Most software (DSS, IRIS, PixInsight) will only extract 12 bits from the raw files (Sony ARW files) instead of the usual Canon/Nikon 14 bits.  Actually 13 bits are encoded in the raw file for most modes (but bulb mode and silent mode are 12 bit) but DCRaw used by DSS, IRIS and PixInsight throws away the least significant bit.
• Amp glow becomes apparent in prolonged liveview mode at high ISOs but I haven't noticed it in long exposure astrophotography.
• Sony raw files are not really raw - an 11+7bit compression is used which can cause artifacts in small areas of very high contrast e.g. startrails.  See http://diglloyd.com/blog/2014/20140214_1-SonyA7-artifacts-star-trails.html for an example.  However in typical deep sky imaging this is unlikely to be a problem - I haven't seen an example in my own imaging.
• Sony raw files use a tone curve to compress 13 bits of data into 11 bits before the 11+7bit compression.  This leads to an effect known as "combing" in the histogram.  For low pixel values there are no gaps in the histogram of a raw file, the gaps then get successively larger (2, 4, 8, 16, 32) as the pixel values increase.  More info here: http://www.rawdigger.com/howtouse/sony-craw-arw2-posterization-detection   In my opinion, this quantisation error is unlikely to be a problem in practice.
• To use bulb mode, silent shutter mode must be switched off (I am at a loss to understand why)
• In Bulb mode, the Sony firmware automatically implements an algorithm to reduce hot pixel noise, which appears to be more or less identical to the original infamous Nikon "star eater" algorithm.  For any semi-scientific instrumentation e.g. measurements of variable star intensities, star magnitudes, star profiles and FWHM this totally rules out Bulb mode.  For producing pretty pictures of the universe (my own main interest) "star eater" does not seem to matter much - see my analysis here: http://www.dpreview.com/forums/post/55734616  As a workaround 30sec exposures can be used instead of Bulb mode.  Read noise at ISO 2000 and above is so low that 30sec exposures become quite practicable in most circumstances.
• The Bulb mode "star eater" algorithm means a different master dark must be created for Bulb mode imaging since the hot pixel suppression algorithm changes the characteristics of the dark and light frames..
• Computer software support for tethered control using a laptop  is woefully lacking, though it is rumoured that APT might support Sony in the near future.  Sony appear to be concentrating on "PlayMemories" camera apps and Wi-Fi control using smart phone apps instead of producing a decent Windows SDK (software development kit)

In summary, these advantages and disadvantages make the Sony A7S very much like Marmite (a UK joke - Marmite is something you either love or hate).  For many, one or more of the disadvantages will be a deal-breaker.  For the more adventurous, those prepared to work around the disadvantages, this is an incredibly performant camera.

As for me I'm determined to really put this camera through its paces.  I've just DIY modified mine for H-alpha sensitivity and I'm looking forward to repeating this M51 Whirlpool image but with the enhanced H-alpha sensitivity:  http://stargazerslounge.com/topic/242439-m51-whirlpool-first-light-with-sony-a7s/

I hesitate to call the A7S a game changing camera or a "disruptive technology" but I do think it will become the reference camera that others are measured against.

To be continued ...

Mark

[pete_l]

That's a great review Mark, thank you for sharing your well-considered views and the sort of astro related info that none of the mainstream reviewers would ever think of publishing.

I was amazed to read that this camera has a CMOS sensor. It would appear that we needn't worry about the CCD vs. CMOS "debate" any more.

[ronin]

Do DSLR's actually have a "bulb" mode these days?

That mode was when you pressed and held the release button down then the shutter remained open until you let go.

Meant that timing was/is questionable.

[Yoddha]

Great review Mark!

The rumors are right and I really work to add support for the alphas in APT. On this stage it is still on design level... Maybe more than two months ago I had a conversation with a nice guy in Sony discussing what is needed to unveil the full power of their cameras for the astro-photography. I have figured the needed features that has to be added to the WiFi API and also suggested to consider making "wire" SDK. These suggestions have been discussed in their IT department and for our sorrow are declined at this stage. The main reason is that Sony consider this market niche as very small to invest into the needed development

The biggest "show-stopper" in the moment is the lack of possibility to control bulb exposure in their WiFi API... The good news that there is still an option to use a serial cable like in the old Canons. Definitely it will be strange setup - WiFi control everything (draining power) and serial cable for shutter. But who knows - there is still hope that Sony will extend their API at later stage

[Mick J]

What an interesting thread, thanks Mark.

[StuartJPP]

Yep...cheers Mark I have seen this as a potential "game changer" but as usual there are things that could be better...such as a decent bulb mode.

Also glad to see you modified it...is there a guide to do it or did you just wing it? How difficult is the mod?

It was interesting watching a video recently where the MW was clearly visible...would make framing easy

[Coco]

This an a Takahashi Epsilon 180 ED would make a great combo, seeing stars and spiral arms is one thing seeing faint narrow band emissions another.

[sharkmelley]

Also glad to see you modified it...is there a guide to do it or did you just wing it? How difficult is the mod?

There's a series of photos on a Cloudy Nights thread but it doesn't show every single step: http://www.cloudynights.com/topic/489617-a-deeper-look-inside-sony-a7s…/

Some screws are not obvious to find.  I did take photos, so if there is interest I can easily put together a guide.  I found it slightly easier than the Canons (350D, 400D, 550D) I've modified.

Look at the amount of copper sheet used inside - I'm sure that helps to dissipate the heat:

This an a Takahashi Epsilon 180 ED would make a great combo, seeing stars and spiral arms is one thing seeing faint narrow band emissions another.

Indeed - I'm really looking forward to putting it on my Tak Epsilon.

Mark

[ollypenrice]

Great review and I always like your posts.

But, erm, well, why not CCD? 

Don't shout at me, I'm asking nicely!!!

lly

[michael8554]

Can't wait to see the new M51 image.....

Michael

[sharkmelley]

Great review and I always like your posts.

But, erm, well, why not CCD? 

Don't shout at me, I'm asking nicely!!!

lly

A very good question!  I admit that part of the reason is that I'm just stubborn and want to do things differently.  However, there is also some logic behind that madness.

I live somewhere reasonably dark - an average night has SQM readings of 20.7-21.0 This relatively low level of light pollution makes my main interest of "natural" colour RGB images quite viable.

I have to concede that, all other things being equal, LRGB imaging with a mono sensor is the best way to acquire full colour images. But all other things are never equal and I've been waiting quite some time for a mono CCD with at least an APS-C sized sensor (22x15mm).  A larger sensor will collect more photons and produce better quality images than a small sensor.  Currently, the best choice I could see was a CCD camera based on the full frame Kodak KAI-11002 e.g. an Atik 11000.  But these work out to be very expensive, especially once you add filter wheel and filters.  Also, their Quantum Efficiency is not high by modern standards, peaking at 50% and quite a lot lower at H-alpha.

Then along came the Sony A7S with a peak QE of 65%, if www.sensorgen.info is to be believed. I don't yet know that the H-alpha sensitivity is (when modified) but I'm guessing it's also going to be higher than the KAI-11002 - I'm trying to think of a way to measure it.  Of course H-alpha response is restricted to the red pixels on the sensor, which is a certainly a limitation, compared with mono. Also there is no possibility of recording luminance data. So whilst I'm not arguing that the Sony A7S will give the same image quality as the mono KAI-11002, I don't think it will be very far short and I do think the Sony A7S would be superior to the OSC CCD version, unless imaging in very warm conditions where the cooling becomes important.

Where the Sony A7S scores very highly is in its versatility. It's a consumer camera and doesn't need external power sources or computer connections. It's easy to fit a lens.  It's totally portable. The low read noise means that much shorter exposures are possible without loss of quality, once images are stacked.  So there is far less need for accurate guiding. It also takes very good quality video, though that's one aspect I haven't yet tried out.

It's certainly not without its disadvantages, as I already pointed out in the original post.  It's also not a cheap camera by DSLR standards but it is far cheaper than a cooled CCD with the same size sensor.  All in all, it seemed worth a punt!  I'm very interested to discover just what can be achieved with it.

Mark

[ollypenrice]

A very good question!  I admit that part of the reason is that I'm just stubborn and want to do things differently.  However, there is also some logic behind that madness.

I live somewhere reasonably dark - an average night has SQM readings of 20.7-21.0 This relatively low level of light pollution makes my main interest of "natural" colour RGB images quite viable.

I have to concede that, all other things being equal, LRGB imaging with a mono sensor is the best way to acquire full colour images. But all other things are never equal and I've been waiting quite some time for a mono CCD with at least an APS-C sized sensor (22x15mm).  A larger sensor will collect more photons and produce better quality images than a small sensor.  Currently, the best choice I could see was a CCD camera based on the full frame Kodak KAI-11002 e.g. an Atik 11000.  But these work out to be very expensive, especially once you add filter wheel and filters.  Also, their Quantum Efficiency is not high by modern standards, peaking at 50% and quite a lot lower at H-alpha.

Then along came the Sony A7S with a peak QE of 65%, if www.sensorgen.info is to be believed. I don't yet know that the H-alpha sensitivity is (when modified) but I'm guessing it's also going to be higher than the KAI-11002 - I'm trying to think of a way to measure it.  Of course H-alpha response is restricted to the red pixels on the sensor, which is a certainly a limitation, compared with mono. Also there is no possibility of recording luminance data. So whilst I'm not arguing that the Sony A7S will give the same image quality as the mono KAI-11002, I don't think it will be very far short and I do think the Sony A7S would be superior to the OSC CCD version, unless imaging in very warm conditions where the cooling becomes important.

Where the Sony A7S scores very highly is in its versatility. It's a consumer camera and doesn't need external power sources or computer connections. It's easy to fit a lens.  It's totally portable. The low read noise means that much shorter exposures are possible without loss of quality, once images are stacked.  So there is far less need for accurate guiding. It also takes very good quality video, though that's one aspect I haven't yet tried out.

It's certainly not without its disadvantages, as I already pointed out in the original post.  It's also not a cheap camera by DSLR standards but it is far cheaper than a cooled CCD with the same size sensor.  All in all, it seemed worth a punt!  I'm very interested to discover just what can be achieved with it.

Mark

Gotcha. However, I use three of the big Kodak chips here and, despite the numbers, they seem pretty good to me. The thing about the Ha is that, as you say, the response per pixel may not be great but you get it on all pixels which matters both for resolution and sensitivity. In nebular imaging I really couldn't live without Ha. When you compare a red layer with an Ha layer all the interest and structure is in the Ha.

The other thing is that, with the big chip, I can image a target in the TEC140 which might need the shorter FL of the Taks to frame on a smaller chip. The TEC images are way better in terms of resolution.

Olly

Edit, I always trot this one out to 'sell' Ha;

[StuartJPP]

Olly, probably a stupid question (its late) but why haven't you captured any (or very little) Ha in the RGB image above?

Don't R filters cover Ha wavelengths?

Assume it is due to exposure length of the Ha subs that you added?

There is an image on Scott Rosen's site of Simeis 147 taken with a modified DSLR and clip in Ha and OIII filters. Look at the excessive exposure lengths though they should be reduced with newer cameras/faster optics.

http://www.astronomersdoitinthedark.com/index.php?c=147&p=526

[ramdom]

Yoddha, so if I understand you right, the Sony A7s API is posted somewhere and anyone can write a program that directly accesses the API to control the camera through WiFi?

--Ram

[Fieel]

Never thought that the Sony A7S was this great for astrophotography. What's that "Electronically Assisted Astronomy" you're talking about? The camera it's so astrophorography oriented that there is an astronomy software integrated in the firmware? If so, my mind is blown.

[sharkmelley]

What's that "Electronically Assisted Astronomy" you're talking about?

EAA, at least my understanding of it, is to use a camera (typically a video camera such as the Mallincam) instead of an eyepiece to give real time or Near Real Time views through the telescope out to a screen.

I'd like to do 10sec subs, for instance, and incrementally stack them in real time as they download to the laptop.  A very high quality image will then be built up in real time in front of the "audience".  This will introduce folk to the principles of imaging and how stacking multiple sub exposures improves an image.  I believe DSS can already do something along these lines.  I'm thinking of writing a PixInsight script do do this.  I already have a script that auto-loads and displays a user chosen crop at any magnification of images as they arrive at the computer - I use this for focusing with Bahtinov Grabber because the Sony is currently unsupported by BackyardEOS, APT, Nebulosity etc.  Just a small change would allow these to be integrated on-the-fly.  Note that the Sony RCC application will control the USB-tethered camera and download them.  The PixInsight script then automatically picks them up.

Of course any DSLR could be used in a similar way but the low read noise of the A7S makes it especially attractive.

Mark

[Yoddha]

Yoddha, so if I understand you right, the Sony A7s API is posted somewhere and anyone can write a program that directly accesses the API to control the camera through WiFi?

--Ram

Hi Ram,

Yes it is available from the Sony's developer site. In fact it is just one PDF file describing the JSON commands sent trough WiFi...

[gorann]

Thanks for the great reviiew and interesting comments!

 I am rather new to astrophotography but I am already thinking about the next step from my Canon EOS 60Da, I wonder if I should wait a year or two  for CMOS to finally take over from CCD. I get the impression that the CCD era might soon be over (Kodak has apparently given up) and CCD developent has stopped. The problem at the moment seems to be that no dedicated astrophotography company has got hold of the latest large low-noice CMOS chips to adapt them to astrophotography and sell them in monochrome versions (e.g to allow high resolution narrow band imaging). The CMOS low noice should put less demand on cooling and I assume that prices would be quite a bit lower than for comparable large CCD AP cameras. As a novice I of course wonder if I am all wrong in my analysis of the situation?

[Yoddha]

Kodak is out of the CCD game, but Sony are making very interesting sensors so the race between CMOS and CCD in AP will continue at least for several years

[pixueto]

Based on that analysis, wouldn't a canon 6D be superior to the sony A7S for astro work? The 6D has a very low noise and canon prouces real raw files. Also, its price is much lower.

[sharkmelley]

Unfortunately, I don't have a Canon 6D handy to make direct comparisons.

I'm sure the Canon 6D is very competent and it is definitely much better supported by astro-software than the Sony.  Producing proper raw files is certainly an advantage with the Canon but the workaround with the Sony (to prevent "star eater") is to switch off bulb mode and use 30sec (or shorter exposures) - though this is only practical on a fast scope.  For deep sky imaging, dithering mitigates the star eater issue though stars are certainly dimmer.

Other things to consider are:

1) QE - Sensorgen suggests the QE of the A7S is higher than the 6D (65% vs 47%)

2) Sony A7S read noise is a lot lower than the 6D - which is great for short exposures and video.

3) Background patterning & banding. I don't know if the 6D suffers from background patterning and banding but it is completely absent in the A7S at ISO 2000 and above (though present at slower ISOs).  Processing an image from the A7S is pure joy compared with my old Canons 350D and 550D.

4) Thermal noise is incredibly low on the A7S.  I don't know about the 6D - I'd love to see comparable figures i.e. dark current at ambient 20C after 2 hours of solid 5 minute exposures.

Liveview on the A7S is astonishing, especially when modded. I'm imaging NGC 7000 tonight on my Tak Epsilon and I was able to see the red H-alpha nebulosity in liveview.

Mark

[StuartJPP]

Hi Mark...I have got a 6D that I can run a test on...Just give me instructions on what you want and I can do it for you.

Gary Honis has done some review work on the 6D...perhaps he has already covered this...

http://dslrmodifications.com/6DReview/6DReview.html

[sharkmelley]

I'm currently working on a PixInsight script into which calibration frames (bias, flats, darks) can be loaded and it will do all the necessary calculations for working out gain, read noise and dark current.  This should make it a lot easier to compare various cameras.

Mark

[sharkmelley]

Hi Mark...I have got a 6D that I can run a test on...Just give me instructions on what you want and I can do it for you.

Gary Honis has done some review work on the 6D...perhaps he has already covered this...

http://dslrmodifications.com/6DReview/6DReview.html

PixInsight script can be found here:

http://www.cloudynights.com/topic/504543-pixinsight-script-for-calculating-dslr-sensor-characteristics/?p=6662387

At a given ISO, take 2 bias frames in quick succession, 2 flat frames in quick succession and take any number of long exposure dark frames in sequence.  Run the script on them and it will produce output like that below.  Be sure to make a note of the ambient air temperature because the dark current results will change significantly depending on outside temperature.  I'm hoping to shortly kick off a project that will enable the worldwide astro-community to collate and compare data on their DSLR cameras. 

The data below was acquired at a constant room temperature of 20C:

=============================

DSLR Sensor Parameters v0.0.2

=============================

Camera: Sony ILCE-7S

ISO speed: 2500

Image Size: 4256 x 2848

Bias frames exposure time: 1/8000 sec

Flat frames exposure time: 1/1000 sec

Dark frames exposure time: 302 sec

Light frames exposure time: 302 sec

Channel order:             [RGGB]

Gain per channel(e/ADU):   [ 1.651, 1.664, 1.669, 1.651 ]

ISO for unit gain:         [ 4128, 4161, 4171, 4129 ]

Read Noise(e):             [ 1.34, 1.34, 1.35, 1.33 ]

Dark Current (e/pixel/sec) [ 0.0919, 0.092, 0.0912, 0.092 ]

Dark Current (e/pixel/sec) [ 0.1219, 0.1202, 0.1207, 0.1246 ]

Dark Current (e/pixel/sec) [ 0.1349, 0.1328, 0.1332, 0.1353 ]

Noise estimates for 302sec exposure

Read Noise(e):             [ 1.34, 1.34, 1.35, 1.33 ]

Thermal Noise(e):          [ 6.38, 6.33, 6.34, 6.39 ]

Skyfog Noise(e):           [ 42.44, 44.19, 44.42, 30.19 ]

Mean values in ADU

Bias Frame 1: [127.9, 127.9, 127.9, 127.9 ]

Bias Frame 2: [127.9, 127.9, 127.9, 127.9 ]

Flat Frame 1: [748.6, 1148.4, 1147.5, 1053.3 ]

Flat Frame 2: [748, 1147.3, 1146.6, 1052.4 ]

Dark Frame 1: [ 126.4, 126.4, 126.5, 126.4 ]

Dark Frame 4: [ 126.7, 126.7, 126.7, 126.4 ]

Light Frame 1: [1347.2, 1631, 1632.1, 902.3 ]

Light Frame 2: [1464, 1779.1, 1780.1, 957.6 ]

DSLR Sensor Parameters v0.0.2 Completed.

[Howie_Oz]

Pretty cool stuff sharkmelley!

See how it goes and I am sure many others would be interested too Including me as I am interested in the A7S for exactly the same reasons as folk on this posting! Still hunting for info and hence just googled and happened upon this very post!

But FYI, just last week in my hunting for info on how to do this I just happened to go to NSN website when this guy was broadcasting and there it was ..... A7S stacking and rotating and being colorised in software called astrotoaster.

I thought it was pretty neat! Astrotoaster works similarly to how I think you are proposing your PixInSIght processing. IE camera saves via its normal Canon or Nikon or whatever software into a specified folder. Astrotoaster is set to automatically read any new shots going into that folder and stack as many as you request. It uses the DSS engine to stack and rotate and then allows colour tweaking and stuff on the result.

BTW I have no affiliation with the dudes who write Astrotoaster nor DDS.