FF Mirrorless and IBIS -- comparison testing?

[cbeasley]

I shoot both stills and video, almost entirely hand held. So IBIS is important for me. I have been using it for about 7 years as it has evolved on MFT cameras, specifically Olympus. There are a number of comparisons around for MFT comparing the different offerings (mostly Olympus vs Panasonic) and there is general agreement that Olympus has stayed at the front. Recently I have seen comparisons of Sony ff with MFT IBIS but with all the new FF offerings, and considering perhaps moving to ff mirrorless (with IBIS), I am wondering if anyone knows of reliable comparison testing of the these -- specifically Sony, Panasonic and Nikon -- as far as the effectiveness of their IBIS. IMHO, CIPA standards are suspect in translating to real-world performance, particularly in the case of video.

Appreciate any thoughts, comments and references.

[Manfred M]

I shoot virtually all my video using a standard video tripod that has a mid-spreader and a fluid head, so the argument between IBIS and in-lens stabilization (I shoot a Panasonic video camera) are irrelevant. As much as possible, I shoot my stills with an extremely sturdy tripod that I will use hang weights on the hook. I generally use a heavy-duty ball head (RRS BH-55) and recently I have started using a geared head in certain shooting situations. These methods beat IS every time.

I try to avoid hand-held video as much as I can but obviously with your underwater work, that is not possible. I have hand-held at times when I had to and in those situations I generally shot only wide angle to minimize the impact of camera movement.

CiC is primarily a site for still photographers, so if you are interested in video, may I suggest going to:

http://www.dvxuser.com/V6/

It is an excellent video resource and I spent a fair bit of time there when I was more into video work.

[cbeasley]

I shoot virtually all my video using a standard video tripod that has a mid-spreader and a fluid head, so the argument between IBIS and in-lens stabilization (I shoot a Panasonic video camera) are irrelevant. As much as possible, I shoot my stills with an extremely sturdy tripod that I will use hang weights on the hook. I generally use a heavy-duty ball head (RRS BH-55) and recently I have started using a geared head in certain shooting situations. These methods beat IS every time.

I try to avoid hand-held video as much as I can but obviously with your underwater work, that is not possible. I have hand-held at times when I had to and in those situations I generally shot only wide angle to minimize the impact of camera movement.

CiC is primarily a site for still photographers, so if you are interested in video, may I suggest going to:

http://www.dvxuser.com/V6/

It is an excellent video resource and I spent a fair bit of time there when I was more into video work.

Of course I agree a tripod is preferable and we use them sometimes underwater -- but scaled down a little . But this is not a video question and it is not about the lens vs in-body stabilization discussion -- I think that is pretty much settled. I was specifically asking about in-body stabilization comparisons for mirrorless ff cameras because the implementations are different and have different levels of efficacy. I am just curious if others have more information on this because I have not been able to find it yet. Thanks for the info and the link!

[lunaticitizen]

Jim Kasson's tests are reliable.

Sony A7II

Sony A7RIV & Sony A7RIII

Nikon Z 7

Fuji GFX100

I believe he doesn't own a Panasonic camera.

[Manfred M]

Of course I agree a tripod is preferable and we use them sometimes underwater -- but scaled down a little . But this is not a video question and it is not about the lens vs in-body stabilization discussion -- I think that is pretty much settled. I was specifically asking about in-body stabilization comparisons for mirrorless ff cameras because the implementations are different and have different levels of efficacy. I am just curious if others have more information on this because I have not been able to find it yet. Thanks for the info and the link!

Based on my experience shooting video and shooting stills, I'm not convinced that the IS works identically when used in still mode versus video mode. This is why I suggested checking with video shooters, rather than still shooters.

When I look at single video frame shot with IS turned on, it does not look nearly as sharp as a still image. If I consider the additional processing required to create a video at a given frame rate and the additional tracking that IS and more intensive duty cycle the camera has to deal with in video mode, I would not be at all surprised if this functionality ended up being less effective than when shooting stills. Look at the reaction time required in a 60 fps video tracking versus what is required in a single still frame. Also look at the power draw to keep a sensor moving in IBIS for minutes at a time, versus a relatively short time, even in high speed burst mode using a still camera...

[rpcrowe]

I don't shoot much video but, I did a quickie test with my A7iii and 70-200mm f/4 G lens. I chose this combination because i can shut off / turn on the stabilization of the lens easily by useA f the switch on the side of the lens.

The results were:

1. best was with lens OSS and camera IBIS on
2. second was lens OSS on and camera IBIS off
3. third (but, still decent) was lens OSS off and camera IBIS on

Obviously, Manfred is correct in that using a sturdy tripod and or, a bit less efficiently, using a monopod correctly is the absolutely best way to get a steady video image. However, this will also result in rather static video. Lately, it seems, the choice of videographers (for better or worse) is a more fluid video presentation with a lot of camera movement. I am not a fan of a lot of hand-held camera work...

There is another way in which a hand-held camera can be effective and that is in judicious choice of shots. When you are shooting a static image hand-held, every camera bump and jiggle is apparent to the viewer. However when you are shooting a scene with lots of movement (as an extreme example: some close-ups of a Mardi Gras Parade) the camera bumps and joggles are not noticed to any great degree.

Another factor is the lens focal length. When I was shooting 16mm motion pictures, the normal angle lens was considered 25mm. My hand-held camera (a Bell and Howell Model 70 Filmo) had 10mm. 16mm and 25mm lenses on the turret. I never hand-held a lens longer that 25mm. The shorter focal lengths were more forgiving regarding camera bumps and shakes...

Proper use of a multi-axis gimbal another consideration. I cannot find the YouTube video, but I watched a video (when the Sony A6600 was introduced) comparing hand held video using an IBIS equipped A6600 against a non-IBIS A6400, mounted on a gimbal. The premise of this video was that the non-IBIS camera with gimbal and the IBIS camera were at about the same price point.

The gimbal equipped A6400 beat the pants off the IBIS equipped A6600 in the line of steady camera presentation. Camera movement was buttery smooth with the gimbal but, just barely acceptable with IBIS...

Using a gimbal that simply depends on camera balance is a cheaper (but, not quite as effective) way to get smooth hand held moving camera shots.

[DanK]

I assume that is supposed to be 7 R III. Interesting. So at least with that particular combination, lens-based stabilization is superior to IBIS. I wonder if that is generally true. If so, it let's some of the air out of the IBIS balloon.

[rpcrowe]

I assume that is supposed to be 7 R III. Interesting. So at least with that particular combination, lens-based stabilization is superior to IBIS. I wonder if that is generally true. If so, it let's some of the air out of the IBIS balloon.

I meant to type A7iii and have changed this on my post.

I cannot confirm or deny this but, I have read that other brands of camera IBIS might be superior to the IBIS in the Sony cameras.

However, the test that I did was a one time quickie and I would not rely on this to purchase or not purchase any specific camera. If I wanted to do a lot of hand-held mobile camera work, I would purchase a good gimbal and learn how to use it. I would also tend to use a camera/lens that can be accommodated by the gimbal.

There are IMO, many other factors besides IBIS that should be taken in consideration when choosing a camera for video work. Among these are but, not limited to, focus capability and the ability to capture video that can be edited well in what ever post-processing program you choose for editing...

[cbeasley]

I don't shoot much video but, I did a quickie test with my A7iii and 70-200mm f/4 G lens. I chose this combination because i can shut off / turn on the stabilization of the lens easily by useA f the switch on the side of the lens.

The results were:

1. best was with lens OSS and camera IBIS on
2. second was lens OSS on and camera IBIS off
3. third (but, still decent) was lens OSS off and camera IBIS on

Obviously, Manfred is correct in that using a sturdy tripod and or, a bit less efficiently, using a monopod correctly is the absolutely best way to get a steady video image. However, this will also result in rather static video. Lately, it seems, the choice of videographers (for better or worse) is a more fluid video presentation with a lot of camera movement. I am not a fan of a lot of hand-held camera work...

There is another way in which a hand-held camera can be effective and that is in judicious choice of shots. When you are shooting a static image hand-held, every camera bump and jiggle is apparent to the viewer. However when you are shooting a scene with lots of movement (as an extreme example: some close-ups of a Mardi Gras Parade) the camera bumps and joggles are not noticed to any great degree.

Another factor is the lens focal length. When I was shooting 16mm motion pictures, the normal angle lens was considered 25mm. My hand-held camera (a Bell and Howell Model 70 Filmo) had 10mm. 16mm and 25mm lenses on the turret. I never hand-held a lens longer that 25mm. The shorter focal lengths were more forgiving regarding camera bumps and shakes...

Proper use of a multi-axis gimbal another consideration. I cannot find the YouTube video, but I watched a video (when the Sony A6600 was introduced) comparing hand held video using an IBIS equipped A6600 against a non-IBIS A6400, mounted on a gimbal. The premise of this video was that the non-IBIS camera with gimbal and the IBIS camera were at about the same price point.

The gimbal equipped A6400 beat the pants off the IBIS equipped A6600 in the line of steady camera presentation. Camera movement was buttery smooth with the gimbal but, just barely acceptable with IBIS...

Using a gimbal that simply depends on camera balance is a cheaper (but, not quite as effective) way to get smooth hand held moving camera shots.

Thanks for the thoughts and agree. But some of my video is underwater. I have a tripod for that, but it is limited in utility. Subjects don't usually just pose for you. IBIS helps a lot in this situation as well as the ones you mention. Wide angle also works very well, normally. Another technique to disguise wobbles is to provide smooth camera movement such as slow swimming in (or out) on a subject. Wish I could get a gimble that I could take underwater! Hey, maybe that's a business opportunity.

[cbeasley]

I assume that is supposed to be 7 R III. Interesting. So at least with that particular combination, lens-based stabilization is superior to IBIS. I wonder if that is generally true. If so, it let's some of the air out of the IBIS balloon.

Yes, that is what I have read -- newer cameras offer the option to use both if available (or not). I think it is settling out that you really only need lens + in body for long focal lengths. Olympus is now making stabilized lenses for this reason. I don't have any stabilized lenses but I have hand held with nice results with a 600mm, shutter at 1/200 using IBIS.

[cbeasley]

Based on my experience shooting video and shooting stills, I'm not convinced that the IS works identically when used in still mode versus video mode. This is why I suggested checking with video shooters, rather than still shooters.

When I look at single video frame shot with IS turned on, it does not look nearly as sharp as a still image. If I consider the additional processing required to create a video at a given frame rate and the additional tracking that IS and more intensive duty cycle the camera has to deal with in video mode, I would not be at all surprised if this functionality ended up being less effective than when shooting stills. Look at the reaction time required in a 60 fps video tracking versus what is required in a single still frame. Also look at the power draw to keep a sensor moving in IBIS for minutes at a time, versus a relatively short time, even in high speed burst mode using a still camera...

Yes, pretty sure there are issues with video and IS that don't manifest with stills. For example, how well does it handle panning? Of course, the correct answer is turn it off in the direction you are going to pan, but in real shooting situations, this is time consuming and for quickly developing shots, not practical.

The other video/IS issues you mention are great questions -- with I could drag some information out of the manufacturer, but no luck on that. The frame quality is interesting -- pf course, there are the issues of exactly how is the sensor sampled which I suspect is a large part of it. I don't know about it being less effective than with stills - good question. Also, compression for video varies widely and I am certainly no expert. I think the power issue you mention is likely responsible for the long lead time to put IBIS into FF architectures. My MFT system definitely eats the battery much fast than a system without IBIS. I have not quantified how much it takes, but I expect it is substantial.

[rpcrowe]

Would something like these help???

https://www.ebay.com/sch/i.html?_fro...+camera+gimbal

[cbeasley]

Would something like these help???

https://www.ebay.com/sch/i.html?_fro...+camera+gimbal

At best, I think these things are "splash proof". Also, you need something really hefty to handle a MFT or FF underwater rig with housing, lights, etc. Maybe it will happen one day. Would be great, except for the weight, cost, hassle, etc. to get it underwater. But I signed up for this. Meanwhile, I'm working on other approaches to help the situation. I don't see why you couldn't take regular gimbal technology and essentially put a waterproof case around it. Just takes $$$ and the market is probably not that big.

Thanks for the link.

[cbeasley]

Would something like these help???

https://www.ebay.com/sch/i.html?_fro...+camera+gimbal

Actually, I have a few more words on this. If you think about ir, IBIS is similar to a gimbal, but applied only to the sensor. It is sort of an "inverse gimbal" if you will allow me invent a term. I say it is an inverse gimbal because it does not try to hold the entire camera still, as a gimbal does, rather it tries to cancel the body movement effect on the image by moving the sensor opposite to the camera body movement. Thinking about it in this way, you realize it is a much lower-power solution to the problem than a gimbal because the sensor has a much smaller mass than the camera body. But, of course there are limitations, some of which have been discussed.

So, like a gimbal, IBIS is not perfect and has limitations, but it can be a useful tool.

[Manfred M]

Yes, pretty sure there are issues with video and IS that don't manifest with stills. For example, how well does it handle panning? Of course, the correct answer is turn it off in the direction you are going to pan, but in real shooting situations, this is time consuming and for quickly developing shots, not practical.

The other video/IS issues you mention are great questions -- with I could drag some information out of the manufacturer, but no luck on that. The frame quality is interesting -- pf course, there are the issues of exactly how is the sensor sampled which I suspect is a large part of it. I don't know about it being less effective than with stills - good question. Also, compression for video varies widely and I am certainly no expert. I think the power issue you mention is likely responsible for the long lead time to put IBIS into FF architectures. My MFT system definitely eats the battery much fast than a system without IBIS. I have not quantified how much it takes, but I expect it is substantial.

When it comes to IS, the important factor to me is that it is available and can be used. Whether it is in-body (i.e. sensor movement) or in lens (i.e. optical element that deflects) is irrelevant in the larger scheme of things. In theory, IBIS has two significant advantages:

1. It should lens independent, i.e. any lens one fits should benefit; and

2. It should reduce the overall cost of system ownership and it will be less expensive to design and manufacture lenses without the stabilization. The cynical side of me suggests that the manufacturer's profit margins could be where this is all going. A DSLR is estimated to have about 30% more parts than a mirrorless camera; which reduces both the parts costs and the labour to assemble a camera. Are mirrorless cameras selling for about 30% less than an equivalent DSLR; no, in fact they are often selling at a premium...

The main downside of IBIS is that shifting the sensor package will also impact focus sharpness if the shift is made in a direction that is normal to the sensor plane. Movement of the sensor on the other two axes will not impact focus.

While the camera and lens manufacturers have not shared the specifics of how their implementation of IS works, the technology is fairly easy to understand. The cameras and / or lenses are equipped with solid-state accelerometers (sometimes incorrectly referred to as "electronic gyroscopes"). These devices measure acceleration (i.e. changes in velocity). With this data it is possible to calculate the probable direction and speed of camera movement and to shift the applicable anti-shake element to counter that movement by moving in the opposite direction and the correct angle and velocity. This needs to be done while the camera's shutter is open, so in most instances a fairly short time and distance. The mechanism needs to return to a neutral position for the next exposure once the image has been captured. If the motion is too great, i.e. outside the range of motion, the image will still be blurred but possibly less than if the technology is not used. When filming video this corrective motion needs to occur at the camera frame rate.

The discussion discounts the effect of inertia resulting from the shifting of the physical IS elements in the camera or lens. This is going to be less of an issue during a single shot, but could definitely be an issue in video where the frame rate will require appropriate damping to prevent overshoot or undershoot.

None of this is an issue when shooting with a tripod and the IS functionality has been turned off. I hadn't noticed any significant impact on battery life when shooting video, but then the video camera has a fairly massive battery versus what we find in still cameras, so I probably just did not notice.

Video compression, especially in newer HD cameras is quite efficient. I haven't looked at them in details over the past few years, but the last time I did, a very common technique was to record a full frame and then only record the differences between that master frame and the new frame for the next 14 frames. Space / processing savings are also implemented by reduced chroma subsampling (i.e. using a 4:2:2 or 4:2:0 pull-down).

[cbeasley]

When it comes to IS, the important factor to me is that it is available and can be used. Whether it is in-body (i.e. sensor movement) or in lens (i.e. optical element that deflects) is irrelevant in the larger scheme of things. In theory, IBIS has two significant advantages:

1. It should lens independent, i.e. any lens one fits should benefit; and

2. It should reduce the overall cost of system ownership and it will be less expensive to design and manufacture lenses without the stabilization. The cynical side of me suggests that the manufacturer's profit margins could be where this is all going. A DSLR is estimated to have about 30% more parts than a mirrorless camera; which reduces both the parts costs and the labour to assemble a camera. Are mirrorless cameras selling for about 30% less than an equivalent DSLR; no, in fact they are often selling at a premium...

The main downside of IBIS is that shifting the sensor package will also impact focus sharpness if the shift is made in a direction that is normal to the sensor plane. Movement of the sensor on the other two axes will not impact focus.

While the camera and lens manufacturers have not shared the specifics of how their implementation of IS works, the technology is fairly easy to understand. The cameras and / or lenses are equipped with solid-state accelerometers (sometimes incorrectly referred to as "electronic gyroscopes"). These devices measure acceleration (i.e. changes in velocity). With this data it is possible to calculate the probable direction and speed of camera movement and to shift the applicable anti-shake element to counter that movement by moving in the opposite direction and the correct angle and velocity. This needs to be done while the camera's shutter is open, so in most instances a fairly short time and distance. The mechanism needs to return to a neutral position for the next exposure once the image has been captured. If the motion is too great, i.e. outside the range of motion, the image will still be blurred but possibly less than if the technology is not used. When filming video this corrective motion needs to occur at the camera frame rate.

The discussion discounts the effect of inertia resulting from the shifting of the physical IS elements in the camera or lens. This is going to be less of an issue during a single shot, but could definitely be an issue in video where the frame rate will require appropriate damping to prevent overshoot or undershoot.

None of this is an issue when shooting with a tripod and the IS functionality has been turned off. I hadn't noticed any significant impact on battery life when shooting video, but then the video camera has a fairly massive battery versus what we find in still cameras, so I probably just did not notice.

Video compression, especially in newer HD cameras is quite efficient. I haven't looked at them in details over the past few years, but the last time I did, a very common technique was to record a full frame and then only record the differences between that master frame and the new frame for the next 14 frames. Space / processing savings are also implemented by reduced chroma subsampling (i.e. using a 4:2:2 or 4:2:0 pull-down).

That's a nice discussion on IBIS and how it works works. Another issue, similar to the overshoot issues you mention, is the tacit assumption that all movement of the camera body should be counteracted by the stabilization system. This is not always the case. Panning to follow a moving subject should not be counteracted by stabilization but this is difficult for the accelerometers to predict. It is only after the fact that this could be recognized. An idealized constant velocity pan movement would represent a spike in acceleration followed by no acceleration (assuming a constant vector velocity of the camera in panning) and then a negative spike when panning stops. There are noticeable differences in how systems handle this situation. I guess they put in some sort of damping, but this is tricky because any damping also affects accuracy for motion you do want to counteract. Of course the best thing is to turn stabilization off if you plan this type of shooting, and do it with a tripod if at all possible. In reality, if you are using stabilization, this will hit you more than you would like so you would like a system that handles it as well as possible, if you are going to use it.

Besides IPB compression (which you describe) there is also All-I (all inter) which does not compress along the time axis. only within individual frames. This gives the equivalent of a jpeg at each frame, but I have always been disappointed with the results, probably because on my camera, it was only available with 1080p, so the frames are low resolution to begin with and there are other factors like focus and shutter speed. I view trying to use a still out of video as sort of an act of desperation -- you just didn't get a still and you need it. My comment about video compression related more to varying quality of codecs -- but as you indicate, probably not so much of a concern nowadays on good cameras. Having said that, interpolation can be a very difficult problem when data are aliased. It all depends on what prior information you have.

[rpcrowe]

I don't think this is true but, shouldn't a crop sensor camera (given the same brand and quality level) exhibit a better IBIS capability because the sensor is smaller and should not need to move as much as the larger full frame sensor?

[cbeasley]

I don't think this is true but, shouldn't a crop sensor camera (given the same brand and quality level) exhibit a better IBIS capability because the sensor is smaller and should not need to move as much as the larger full frame sensor?

I'd say that it just requires engineering to make it happen on larger sensors. And more power. I don't think guys like Nikon, Sony, etc. would be putting it in their new flagship cameras if they couldn't do it right. It is especially significant that Nikon has launched a new line of pro cameras and lenses with this technology. Another way to say it is that it should be easier with smaller sensors.

[DanK]

Panning to follow a moving subject should not be counteracted by stabilization but this is difficult for the accelerometers to predict. It is only after the fact that this could be recognized.

Assuming the panning is on the horizontal plane, some Canon lenses have had an option to turn off compensation for panning for quite some time. I have no idea whether IBIS lenses have this. Since I don't do video and generally don't do stills that entail panning, I haven't checked.

[cbeasley]

Assuming the panning is on the horizontal plane, some Canon lenses have had an option to turn off compensation for panning for quite some time. I have no idea whether IBIS lenses have this. Since I don't do video and generally don't do stills that entail panning, I haven't checked.

I don't know either about all -- there are a number of implementations out there now -- but Olympus has the option to turn off horizontal or vertical so that you can pan either way (actually, the vertical so you can pan horizontally in portrait orientation) . Or you can totally disable it. Of course it is mostly for video and action/sports applications. BTW, IBIS is "In Body Image Stabilization" and is not in the lens. To make it more complex, you can also use it in conjunction with stabilized lenses. I don't have any stabilized lenses so I don't have any idea what happens in this case except that the settings allow you to use both or turn them off -- separately, I think.