Mark Galer posted a second video (watch on top) explaining the A7II Phase Detection and other new Features (Thanks Mark!). And than there is that very cool video:
It shows how much(!) the sensor moves on the 5 axis stabilization. I am really surprised to see this doesn’t crop the image! That’s good news. And now the possible bad news: DSLRmagazine (translation here) tested the 5 axis stabilization on third party lenses. And well they got the “feeling” that on third party manual lenses where you have to manually set the focal length of the stabilization…well only the three axis stabilization works and not the full 5 axis! But keep that with a grain of salt because:
1) They are yet not 100% sure of that
2) Sony may fixes this in the final firmware version?
Can reviewers and testers please ask Sony about this matter? Thanks!
UDPATE: One more little good news: Imaging Resource compared the A7 vs A7II image quality at ISO 3200 and the A7II seems to have a slightly increased sharpness.
More A7II tests:
Spanish hands-on at Xatakafoto (translation here).
DC.view (google translation here) posted some ISO samples with A7II images on left and A7 images on right.
USA preorders:
Sony Alpha a7II (Body Only) for $1,698 at Amazon, BHphoto, Adorama, SonyStore.
Sony Alpha a7II with FE 28-70mm OSS Lens for $1,998 at Amazon, BHphoto, Adorama, SonyStore.
Sony 70-300mm f/4.5-5.6 G SSM II Lens for $1,148 at BHphoto.
EU preorders:
Sony Alpha a7II (Body Only) for 1.799 Euro at Sony Germany, Sony UK, Sony France, Sony Italy, Sony Spain, Sony Austria, Sony Switzerland, Sony Netherland, Sony Belgium, Sony Portugal, Sony Sweden, Sony Finland, Sony Norway.
Sony Alpha a7II with FE 28-70mm OSS Lens for 2.099 Euro at Amazon Germany and at Wex UK. In Finland at Verkkokauppa.
Asian/Australia preordersNote:
Sony A7II at Sony Store Australia (Click here) and Amazon Japan (Click here).
82 Responses
1st blood!
Well, some of Sony’s own lenses only get 3 axis stabilisation, so I would not be surprised that this is also the case for third party lenses as well.
yes, that’s what Sony said in their presentation.
This is just NOT true!
This happens only if you use A-mount lenses and Sony’s active (SLT) adapter and this is probably because if the sensor was moving this would affect the PDAF focusing as the later is not moving.
Aren’t the affected A-mount lenses Sony lenses then?
This is only if you use them with an active PDAF adapter as far as I know, probably since moving the sensor along 2 of these axis is getting it out of calibration with the PDAF sensor in the adapter that is fixed.
but the pdaf adapter has it’s own focussing, why should the sensor in the camera care about what the adapter is doing?
I didn’t understand the reasoning either.
Because with SLT/SLR cameras light reaches the sensor and PDAF sensor along two different routes and this has to be calibrated to very tight tolerances. So that the PDAF sensor knows exactly where the image sensor is to focus for it. Even when you have both firmly fixed in one body and it is easier to calibrate in SLRs they have been plagued with front/back focusing problems for ages.
Now take the PDAF sensor and put it in an adaptor (Sony active PDAF adapter) and now you have an even more difficult task, as the mount adds even more error.
Now start moving the image sensor along 5 axis while the PDAF sensor stays fixed and this probably ruins any calibration. Sony probably decided that 2 of these 5 axis cause most of the negative effect and switch them off when using the SLT adapter.
This is probably one of the best advantages of mirror-less cameras that focusing information comes off the same sensor that produces the image!
Which only prove what a fool you are.
“well they got the “feeling” that on third party manual lenses where
you have to manually set the focal length of the stabilization… only
the three axis stabilization works and not the full 5 axis”
That’s to be expected (not really “bad news” nor something you can hope will be fixed with firmware). The fact is you can’t stabilize left/right or up/down translational movements without knowing the subject distance (move your head from side to side and close objects will appear to move much more than distant ones). With native lenses this distance is provided by the autofocus system.
Doing so with fully manual (no electronic communication between lens and camera) lenses is impossible, unless you use image analysis techniques to try and calculate the subject distance.
Wrong!
So when you move you head side to side distant objects move the same amount as close ones?
Yes. Because they don’t move; it’s your head that moves, duh.
Across the plane of your vision (and hence the sensor) they do!
Doesn’t matter, stabilization not as good as OMD.
Not at all – it works for all. The focal length is needed to correct for pitch and yaw (angular movement round the X&Y axis). For roll motion (angular movement round the Z axis), the focal length is immaterial.
As has been pointed out on other threads the term 5 axis is misleading. There are 3 axis. Along two of them (X & Y), the image is stabilised for angular (round the axis) and linear (along the axis) movement whilst, for the third Z axis it corrects only for roll motion.
So you might consider it “full” stabilisation on 2 axis, and partial stabilisation on the 3rd.
In practice, for most photography, pitch & yaw (rotation round the X & Y axis) are the most important followed by roll (rotation around the Z axis). Linear movements are only significant with close up photography.
Apologies, having read what you said it might be true for lateral movement. I’ll have to get out the pen & paper and work it out. However, I suspect for distant object lateral movement is not very significant…
Wrong, you listed only 3 axis.
The 5 are
X Shift – as you’ve mentioned
Y Shift – as you’ve mentioned
X Roll
Y Roll
Z Roll – as you’ve mentioned
There are only 3 axis in a 3D world. It’s simply wrong to talk of 5 axis. There’s just rotational and linear on each of three, and the so-called 5 axis stabilisation does not correct for linear movement on the Z.
Also, what you call “X roll” is normally termed “pitch” and “Y Roll” is yaw.
So I cover them all.
Your right, we should all refer to it as a system that does 3-axis stabilization of rotational movement and 2-axis stabilization of translational movement.
Or, we could not be pedantic and say 5-axis stabilization, since this is standard jargon in the camera world.
No you didn’t. You didn’t cover Y shift and X shift. I made a mistake saying that you’ve covered Y Shift and Z Shift, when in fact you covered Pitch and Yaw(thanks for being anally-retentive).
Sure, the correct way of calling it would be a 3 Axis stabilization with 5 degrees of freedom, but if we start going down this path, then there is no such thing as a translucent mirror, rather a semi-transparent mirror.
“Along two of them (X & Y), the image is stabilised for angular
(round the axis) and linear (along the axis) movement whilst, for the
third Z axis it corrects only for roll motion.”
Covers both rotation around and movement along the X&Y axis and rotation round the Z axis. That;’s 5 of the 6 possible degrees of freedom. As for “anally retentive”, I’d just call it an attempt to match your terminology with the usual one. But if you wish to deal in insults, then that’s up to you of course.
Their nomenclature is indeed slightly inaccurate, it’s not 5 axis stabilisation, it’s 5 degrees of freedom (out of a possible 6).
However “Degrees of Freedom” abbreviates to DoF, which would be very confusing to customers!
You can correct for pitch / roll and yaw fairly easily if you know the focal length.
You can’t perfectly correct for translation due to parallax effects (different parts of the scene appear to move different amounts depending on distance). You can however try and minimise the blur at a specific distance by moving the sensor to compensate. This is why distance information from the lens is required.
And you’re more than slightly wrong, but the amount of obfuscation been used confuses people into thinking you actually know what your talking about.
I spent 6 years researching camera movements and their impact on images, publishing several papers, it’s the one area relating to photography where I really do know what I’m talking about.
Here are some links if you’d like to read any of them:
http://scholar.google.co.uk/scholar?q=grum+bors&btnG=&hl=en&as_sdt=0%2C5
Perhaps you could explain where I’m “more than slightly wrong”?
This is, I’m afraid, a little wrong. Or, rather, your understanding of how shake affects different distances is correct, but your conclusion is wrong.
The traditional X/Y anti-shake system doesn’t have to compensate for movement of objects on the image plane. It has to compensate for movement OF the actual image plane. The only thing that’s needed for that is focal length, as magnification defines how large the movements should be for a given amount of shake; focus has no part in that*), because if the system can accurately compensate for a certain amount of shake at a certain focal length, the sensor will be still relative to the scene and you don’t have to compensate for any “perspective shake” or whatever you want to call it.
When shake actually occurs, meaning the system has been driven outside its limits, then you have a point — it matters. But you can’t compensate for it, the system is already outside its limits.
(* Actually focus has (theoretically) a little part in that, because magnification changes with many lenses when focusing, but it’s of little overall importance as Minolta’s/Sony’s FL reporting from zoom lenses is crude anyway and that is a much bigger margin of error.
Magnus I’m afraid you are misunderstanding what I wrote, and how Sony (and Olympus’) “Five Axis” stabilisation works.
To [approximately] correct for pitch, roll and yaw you can simply move/rotate the whole sensor. Since rotations about the camera centre are distance independent no 3-D knowledge is required to calculate the correcting motion.
This is the 3-axis stabilisation that is enabled with 3rd party lenses.
The full 5-axis system corrects two further movements, translation (not rotation) in the vertical / horizontal axis. When you for example move the camera sideways objects in the scene move across the image plane (and would therefor become blurred). However now the amount objects move depends on distance. Imagine looking out of a train window, the foreground moves past very fast, the distance much slower.
You can attempt to correct this type of movement by moving the sensor, but you can only correct for one distance as each distance requires a different amount of movement.
In order to apply the correction for sideways movement at the plane of focus you need to know how far away that is, that is why this correction is only available if the lens communicates this information back to the camera.
Ah, I admit I missed the little word “translational” *oops*. Sorry for that. We are in complete agreement. 🙂
(OT: if that’s your site I’m loving the pics, but you should really have a landscape section too if the quality of the background pic is representative)
Cheers, yeah that’s my site. There is no landscape section as it’s hard to get paid to shoot landscapes!
any ideas what’s the life of the moving parts/system of the stabilization?
pretty boring life probably if you only shoot your cat. But off and on it might get to see a half naked woman or so, expect the moving parts to get a little bit more stiff then…
that whatever dreams 🙂 you are obviously dreaming for cats and girls 🙂
stop dreaming… go for a girl 🙂
d@mn. i want this camera! just to play with it.
That’s a substantial object which has to be moved around, and it has to be done rather fast. The actuators must be rather powerful. Comparing it with the known size of an FF sensor it looks like there’s about +/- 2.5mm or so of movement possible, or roughly +/- 40 minutes of arc (pitch or yaw) with a 200mm lens mounted, or +/- 20 minutes of arc with a 400mm lens.
As a guide, the Moon is roughly 30 minutes of arc across when viewed from the Earth.
“Imaging Resource compared the A7 vs A7II image quality at ISO 3200 and the A7II seems to have a slightly increased sharpness.”
Really? Sharpness favours one or the other, depending on where you look. That says there is a slight difference in focusing. Either way, the differences are miniscule and will have zero impact in the real world.
Typical hyperbole of recent posts on this site, properly associated with senseless fanaticism.
OSS > IBIS
Especially with all those primes. Oh wait, most FF primes aren’t (OS) stabilized and likely never will be.
So how do you correct for roll around the lens axis using OSS?
Too bad the stabilization is not great for video and Sony handicapped the video output.
See EOS-HD for his review.
The EOS-HD review is fanboy BS, and you are a troll. It sucks that you’re stuck with a crappy sensor and no IBIS, but such is Canon.
HAHA WOW. Read the review, in fact read all of Andrew’s stuff. He isn’t a Canon fanboy, in fact he OWNS an a7s and loves it. It’s just a damn review.
Actually I hate canon. I like Sony, the comments he made were actually quite relevant. Sony just seems to push out stuff every year with only marginal improvements. Why did they not put the D750 sensor with better video processing? This camera could have been awesome.!!
for 5-axis stab lens has to report the distance on which it is focused. it might be not fixed at all.
Wrong, read the discussion below. Only focal length is necessary.
sigh.
Do you guys know that Manfrotto sells tiny steady tripods for chips? I used to be into these things too, then I improved and could not care less.
if the original a7 and a7r were announced in oct 2013 and the a7s in aprl 2014, anyone have an educated guess as to when we might see updated versions of a7r/a7s?
We probably won’t. It seems more likely the specialist high res and low light cameras will be in the A9 range, commanding a higher price point. We already know the high res A9(R) is coming, and I would expect an A9(S) to follow later.
well..still leaves me with no idea on how to move fwd. have been heavily considering picking up an a7s..but with the surprise updated a7II, i figured some refreshes are just around the corner in the release cycle..trying to avoid picking up a new camera and then feeling let down because i didnt wait a few months..
I would say it depends on if you need a new camera now or not. I don’t think an a7S replacement is on the cards for a while… probably won’t arrive until the summer. That’s my guess anyway.
so buy one now and then buy another one then is what you’re saying? Winky 🙂
I don’t care how much the sensor moves. I don’t care if it can dance the Macarena and bark like a dog. The only thing that matters is what it can do for me in practical terms.
I have successfully hand-held a 400mm lens at 1/10 second using the A7M2
As usual, these demo videos of mechanical limits mean absolutely nothing. Shake is pretty high-frequency and is in most cases in the sub-millimeter range. Minolta also had a great tech demo where they moved around a sensor for distances approaching half an image height, but that means nothing. It just means it’s a tech demo. A sensor vibrating in the 0.1mm range would be a really boring tech demo.
i’m gonna buy this camera to shoot my cat, cuz my iPhone 6 can’t zoom..
Maybe your cat can zoom closer to you.
good idea! i’ll ask her. thank you 🙂
or just feed your cat more so it grows big enough to fill the frame
thanks, but i don’t like fat cats or whatever.
I love my cat and I am not going to offend her with these toys 🙂
😀
given the freedom of movement – does the image get cropped? I noticed on Mark Galer’s full sized same with the FE16-35 rolling with IBIS turned on; the image size is less than 6000×4000 – more like somewhere around 20Mp (5562 x 3510).
No the image does not get cropped – I keystoned the image in Lightroom Lens Corrections which accounts for the slightly smaller size.
While all these Sony Artisans are testing this new camera poor Brian Smith is reduced to peddling camera bags on dpreview.
does anyone think it’s possible for the A7S mark 2 to have an SDI connection in the extra space under the new location of the connectors on the A7 mark2? SDI is a much better connection than HDMI (especially micro HDMI) and it should be able to run 4k out from it as well.
No.
But it would be pretty awesome if it did. (Take notes Sony =D)
It is pretty clear from the video that Sony only does shifts and rotation, not pitch and yaw. It is like 3-axis in E-M10, not 5-axis in E-M5, E-P5 and E-M1.
Are you saying the information on the Sony website is a lie?
I am saying I am seeing with my own eyes that the mechanism is incapable to tilt the sensor.
That would be a yes then.
It does not move in VIDEO MODE. Stabilization is like old A99, electronic and with heavy rolling shutter, especially with 1/50 shutter speed. Unusuable!!!
Not sure where this rumour started but the sensor movies in stills and movie mode.
No, it does not!
Care to cite your source?
Watch the movie on YouTube: http://youtu.be/0t0d8W7qX80
Oly stabilization in video mode also is electronic, sensor does not move!!!
I am currently holding the camera and my sensor is moving in Movie mode.
No, it does not!!!
http://youtu.be/0t0d8W7qX80