The Camera Store

[spacemn]

The curious thing about the comparison, is that the mirrorless cameras are all APS-C or MFT size sensors. I know there are more of those to test than mirrorless FF choices today, but that is a bit of an unfair comparison.

It will be more impressive when FF mirrorless comes this close on focus speed tests.

... lowlight capability (-4EV) of A7S...

I see several tests for the Canon 6D but none for the A7s. Can you post a link to this?

I haven't seen any test yet. It is what Sony advertise (seen here for example: http://www.bhphotovideo.com/c/product/1044728-REG/sony_ilce7s_b_alpha_a7s_mirrorless_digital.html)

The A7(R) AFs down to 0EV and the A7S all the way to -4EV, just like the GH4 (also advertised).

When that is said, I don't think that any mirrorless does have cross type PDAF AF. Not sure how essential it is in real life, for most scenarios the best mirrorless cameras does nail focus very consistently, unless you have something very low contrast, but then a DSLR would be in trouble as well.

 

[spacemn]

The curious thing about the comparison, is that the mirrorless cameras are all APS-C or MFT size sensors. I know there are more of those to test than mirrorless FF choices today, but that is a bit of an unfair comparison.

It will be more impressive when FF mirrorless comes this close on focus speed tests.

... lowlight capability (-4EV) of A7S...

I see several tests for the Canon 6D but none for the A7s. Can you post a link to this?

the 6D has been on the market for 1 year and a half, let's wait for the A7S to be released to see those tests.

Agree, we can't go with manufacturer's marketing. We need to wait for proof/tests.

Sure, but it would be pretty foolish of Sony if it didn't work. Panasonic advertised it and reports are that it AFs virtually in darkness. I don't think it would be a problem for Sony, now that Pana seem to have implemented it in the GH4.

 

[spacemn]

They forgot to mention video AF by the way. The A6000 would win by a margin. What I've seen the GH4 do during video is not very impressive in comparison.

 

[Bart Hickman]

One key area would be DOF. That is actually my only issue with otherwise an excellent comparison but I guess since Fuji does not have a faster telephoto zoom yet, DOF could not be properly equalized.

The other area will be file processing, where m43 are still at 16 MP, although in this particular test, only a6000 is higher than 16MP (all others are that, including D4s).

And even with single point AF, the circuitry for a sensor 4x as large (m43 vs FF) is still going to be more complex and longer.

I'm pretty sure they don't stream the entire image for doing AF--just a very small subset of data are downloaded. And the extra distances involved with a larger sensor (which would only be 2x as long on average) will have negligible impact on AF processing update rate.

Bart

 

[SHood]

The curious thing about the comparison, is that the mirrorless cameras are all APS-C or MFT size sensors. I know there are more of those to test than mirrorless FF choices today, but that is a bit of an unfair comparison.

It will be more impressive when FF mirrorless comes this close on focus speed tests.

... lowlight capability (-4EV) of A7S...

I see several tests for the Canon 6D but none for the A7s. Can you post a link to this?

Look at the specs.

And where do you find AF speed comparison at -4EV? Perhaps you can point us to this demonstration of very fast AF at -4EV in GH4 that you claimed above?

Here is a comparison of the GX7 which supports -4EV against the X-T1 and A6000. The GX7 is noticeably faster than the A6000.

Since Panasonic is using DFD for continuous focusing, they need to continue pushing the boundaries for CDAF. My concern with the on-sensor solution is that companies will stop developing CDAF and just focus on PDAF from the sensor. Low-light results will likely suffer as a result.

--
My Flickr Birds

 

[Lab D]

The curious thing about the comparison, is that the mirrorless cameras are all APS-C or MFT size sensors. I know there are more of those to test than mirrorless FF choices today, but that is a bit of an unfair comparison.

It will be more impressive when FF mirrorless comes this close on focus speed tests.

... lowlight capability (-4EV) of A7S...

I see several tests for the Canon 6D but none for the A7s. Can you post a link to this?

Look at the specs.

And where do you find AF speed comparison at -4EV? Perhaps you can point us to this demonstration of very fast AF at -4EV in GH4 that you claimed above?

I and others actually did tests with DSLRs and the GH4. I was shooting at ISO6400, F/2.8 and 1/8th shutter speed in near total darkness and the GH4 was still usually almost instantaneous. I don't have a 6D, but the GH4 bests all the DSLRs I have tried. Add this to the near D4s C-AF focusing and we can see the real potential of mirrorless...and PDAF is no longer needed.

If you get an A7S, please post your experiences so we can see if it lives up to the marketing. IIRC, It does not have PDAF pixels, so it should be interesting.

 

[peripheralfocus]

I think the AF-C performance of the latest crop of mirrorless cameras is very impressive and encouraging, and compared to DSLRs of similar price/segmentation, they are competing very effectively.

But here ...

Add this to the near D4s C-AF focusing

... you're getting a little bit over your skis based on one test where the results were presented in very rough descriptions with no example images. I'm not criticizing this test, because no single test can cover everything and this one is a useful contribution to our knowledge. But some things to keep in mind:

We don't know the settings they used for each camera.

They don't give us even a rough comparison figure -- i.e. how many sharp images per second for each camera. If the GH4 has a hit rate of 80% and the A6000 has a seemingly much worse hit rate of 50%, they are actually both producing essentially the exact same number of sharp images per second.

Without example images we can't see the magnifications they were working at (and thus whether depth of field was hiding small focus errors), nor can we see how they were judging sharpness (it matters because very few images are either perfectly focused or completely blurry -- the tester has to make a judgment).

They shot only one type of subject in one set of conditions. No team sports like soccer, which are much different because the existence of multiple players who can confuse the focus system is a huge issue. No indoor sports like basketball or hockey, which are typically shot in much lower light levels.

Results will change with different lenses and different photographers.

Again, these aren't criticisms of this test; it's just a caution about reaching large conclusions that one test cannot support.

 

[Gary H]

I liked the video. I shoot my daughter's soccer games for her team. I also video, but I use a 1DMKIII for stills and an A7 with external microphone and monitor for video. What I got from this is that today I could use a GH4 to replace both cameras. I won't do this because the A7 provides advantages that I value when shooting stills. Also, I would want a 400mm FE and the next generation FF prior to making such a transition. Now, I know that such an all in one workhorse will be available within the next couple of years. Photokina should give us FF autofocus pdaf with better coverage than the A7 and perhaps faster processing. Of course, fast FF autofocus is more difficult to achieve than APSc, or 4/3.

 

[SHood]

I think the AF-C performance of the latest crop of mirrorless cameras is very impressive and encouraging, and compared to DSLRs of similar price/segmentation, they are competing very effectively.

But here ...

Add this to the near D4s C-AF focusing

... you're getting a little bit over your skis based on one test where the results were presented in very rough descriptions with no example images. I'm not criticizing this test, because no single test can cover everything and this one is a useful contribution to our knowledge. But some things to keep in mind:

We don't know the settings they used for each camera.

They don't give us even a rough comparison figure -- i.e. how many sharp images per second for each camera. If the GH4 has a hit rate of 80% and the A6000 has a seemingly much worse hit rate of 50%, they are actually both producing essentially the exact same number of sharp images per second.

Without example images we can't see the magnifications they were working at (and thus whether depth of field was hiding small focus errors), nor can we see how they were judging sharpness (it matters because very few images are either perfectly focused or completely blurry -- the tester has to make a judgment).

They shot only one type of subject in one set of conditions. No team sports like soccer, which are much different because the existence of multiple players who can confuse the focus system is a huge issue. No indoor sports like basketball or hockey, which are typically shot in much lower light levels.

Results will change with different lenses and different photographers.

Again, these aren't criticisms of this test; it's just a caution about reaching large conclusions that one test cannot support.

There are also differences in lenses that impact performance, size and cost. The results we see in this test doesn't necessarily translate to other lenses. And while the A6000 is by far the cheapest and smallest body, the body/lens combination tested was comparable in cost and much larger in size to the other mirrorless combinations tested.

We need to remember this was a body/lens test not just a comparison of camera bodies.

 

[stevo23]

Why? Focus speed should be independent of frame size I would think. And what this tells us is that they're packing more and more processing and control speed into smaller and less expensive packages.

Focus speed depends on CPU processing speed. A FF shot contains a lot more information and takes a lot more CPU power to process than an APS-C shotor an m4/3 shot . So focus speed is very much dependent on frame size, which is why Nikon's mirrorless was able to achieve very quick focus quite some time ago.

I could be wrong, but I think focusing speed is dependent on the number of points being used and my impression is that the focusing systems are not very aware of the size of the format. If you're using a single point right in the center, what does it matter the sensor is the tiny 4/3 or full frame?

One key area would be DOF. That is actually my only issue with otherwise an excellent comparison but I guess since Fuji does not have a faster telephoto zoom yet, DOF could not be properly equalized.

How would DOF influence focusing speed?

The other area will be file processing, where m43 are still at 16 MP, although in this particular test, only a6000 is higher than 16MP (all others are that, including D4s).

And even with single point AF, the circuitry for a sensor 4x as large (m43 vs FF) is still going to be more complex and longer.

Keep in mind, I'm only talking about autofocus speed which seemed to be the primary point of discussion in the Camera World article. I don't think he touched very much on buffer and sustained cycle rate - all of which are influenced heavily by processing power.

But just processing a focus point or two doesn't seem to have anything to do with format size when comparing the kind of focusing systems he was using in the article. And I wonder if the Nikon has even less to worry about since the TTL focusing is independent of the sensor itself.

 

[Gary H]

Regarding the right lens/body combination and the GH4...

I believe that the Panasonic 35-100mm lens provides a real speed advantage compared to other less expensive long zoom lenses, or non Panasonic zooms on the GH4. So, you need to shell out some real money on the lens prior to going out and shooting a sports event with the GH4. I think the same can be said of the A6000 and the FE 70-200mm lens. Just prior to the posting of this video, I had recommended these two combinations to a friend. I'm pretty sure he wasn't looking to spend this kind of money to provide him with the ability to video and shoot stills of his daughter's soccer games. That said, I've talked to other parents that shoot with low end DSLRs and have lots of out of focus shots. Both the GH4 and the A6000 when combined with their respective lenses represent good value when you consider that previously we were looking at purchasing a 1DMKIII and Canon glass. I'm not saying that the A6000 is a better camera, but the combination is certainly smaller and much less expensive.

 

[quezra]

Why? Focus speed should be independent of frame size I would think. And what this tells us is that they're packing more and more processing and control speed into smaller and less expensive packages.

Focus speed depends on CPU processing speed. A FF shot contains a lot more information and takes a lot more CPU power to process than an APS-C shotor an m4/3 shot . So focus speed is very much dependent on frame size, which is why Nikon's mirrorless was able to achieve very quick focus quite some time ago.

I could be wrong, but I think focusing speed is dependent on the number of points being used and my impression is that the focusing systems are not very aware of the size of the format. If you're using a single point right in the center, what does it matter the sensor is the tiny 4/3 or full frame?

One key area would be DOF. That is actually my only issue with otherwise an excellent comparison but I guess since Fuji does not have a faster telephoto zoom yet, DOF could not be properly equalized.

How would DOF influence focusing speed?

This is something I'm not very clear about either. In the thread in the Open Discussion, one guy claims the test is how many "DoFs" the object passes through in a given series of shots. And given the shots were all shot at F4, apparently that gives m4/3 a bigger advantage (and the test wasn't particularly tough on any camera). On the other hand, the D4S would have been focusing at f2.8 since DSLRs AF wide open when not in live view mode, which means its AF is even faster. But then if having a thin DoF stretches C-AF capabilities (since that means the target crosses "more" DoFs in a series of shots), then what's the balance of trade-offs actually like?

 

[SHood]

Why? Focus speed should be independent of frame size I would think. And what this tells us is that they're packing more and more processing and control speed into smaller and less expensive packages.

Focus speed depends on CPU processing speed. A FF shot contains a lot more information and takes a lot more CPU power to process than an APS-C shotor an m4/3 shot . So focus speed is very much dependent on frame size, which is why Nikon's mirrorless was able to achieve very quick focus quite some time ago.

I could be wrong, but I think focusing speed is dependent on the number of points being used and my impression is that the focusing systems are not very aware of the size of the format. If you're using a single point right in the center, what does it matter the sensor is the tiny 4/3 or full frame?

One key area would be DOF. That is actually my only issue with otherwise an excellent comparison but I guess since Fuji does not have a faster telephoto zoom yet, DOF could not be properly equalized.

How would DOF influence focusing speed?

This is something I'm not very clear about either. In the thread in the Open Discussion, one guy claims the test is how many "DoFs" the object passes through in a given series of shots. And given the shots were all shot at F4, apparently that gives m4/3 a bigger advantage (and the test wasn't particularly tough on any camera). On the other hand, the D4S would have been focusing at f2.8 since DSLRs AF wide open when not in live view mode, which means its AF is even faster. But then if having a thin DoF stretches C-AF capabilities (since that means the target crosses "more" DoFs in a series of shots), then what's the balance of trade-offs actually like?

Lighting conditions also play a big part in the final results. This test was done in ideal conditions. As light falls the GH4 will widen the gap over the other mirrorless solutions.

--
My Flickr Birds

 

[peripheralfocus]

How would DOF influence focusing speed?

It doesn't affect the speed, per se, but it affects the precision that the system must achieve. If you have a lot of depth of field, the AF system doesn't have to be as precise because its mistakes will be covered up -- or more likely partially covered up -- by the DOF. Another implication of this is that, when pictures have a large DOF, it's very difficult for the evaluator to determine exactly where the camera was focused.

So for camera tests, this is important because if you say the Flexomatic 1000 gave me 10 out of 10 sharp pictures of a running horse, and the Luxosnapper 5000 gave me only 7 out of 10 sharp pictures of the running horse, it means nothing if you shot the Flexomatic at f/11 and the Luxosnapper at f/2.8. The Luxosnapper's job was much, much harder. It's not a fair test.

I test a lot of cameras, and I run into this all the time because typically I get sent only the kit lens. Usually, it opens no wider than f/5.6 at its telephoto end (and the telephoto end of kit lenses is usually not all that long). When I shoot a normal action subject like a cyclist or runner at f/5.6, especially with a small-sensor camera, I can have a hard time seeing where the camera was really focused. There's just too much DOF and I can't reach any real conclusions.

 

[stevo23]

How would DOF influence focusing speed?

It doesn't affect the speed, per se, but it affects the precision that the system must achieve. If you have a lot of depth of field, the AF system doesn't have to be as precise because its mistakes will be covered up -- or more likely partially covered up -- by the DOF.

Well, I guess in practical application, that could come out as faster AF lock. However, the D4s still beats them all and it's the narrowest DOF of all. So, if you propose to "normalize" the results, then the M43 has to have a substantial correction applied to it's results.

Another implication of this is that, when pictures have a large DOF, it's very difficult for the evaluator to determine exactly where the camera was focused.

I'm not sure of that. It's got a focus point to work with.

So for camera tests, this is important because if you say the Flexomatic 1000 gave me 10 out of 10 sharp pictures of a running horse, and the Luxosnapper 5000 gave me only 7 out of 10 sharp pictures of the running horse, it means nothing if you shot the Flexomatic at f/11 and the Luxosnapper at f/2.8. The Luxosnapper's job was much, much harder. It's not a fair test.

I would think shooting at a brighter aperture gives a lot of advantage, so I would have guessed you have it reversed. But then this also makes one start to think about the fact that a DSLR is generally opened at max aperture when using TTL focusing whereas the mirrorless is using microlenses on it's sensor and in the case of my A7, is stopped down. So Here's a correction to apply in the other direction for mirrorless.

I test a lot of cameras, and I run into this all the time because typically I get sent only the kit lens. Usually, it opens no wider than f/5.6 at its telephoto end (and the telephoto end of kit lenses is usually not all that long). When I shoot a normal action subject like a cyclist or runner at f/5.6, especially with a small-sensor camera, I can have a hard time seeing where the camera was really focused. There's just too much DOF and I can't reach any real conclusions.

Yes, this is true. It also comes into play when fine tuning autofocus.

 

[EinsteinsGhost]

The curious thing about the comparison, is that the mirrorless cameras are all APS-C or MFT size sensors. I know there are more of those to test than mirrorless FF choices today, but that is a bit of an unfair comparison.

It will be more impressive when FF mirrorless comes this close on focus speed tests.

... lowlight capability (-4EV) of A7S...

I see several tests for the Canon 6D but none for the A7s. Can you post a link to this?

Look at the specs.

And where do you find AF speed comparison at -4EV? Perhaps you can point us to this demonstration of very fast AF at -4EV in GH4 that you claimed above?

Here is a comparison of the GX7 which supports -4EV against the X-T1 and A6000. The GX7 is noticeably faster than the A6000.

It should be with deepest DOF of the three. But, you didn't provide answer to my question: a demonstration of very fast AF at -4EV (the test, based on a6000 settings is at +1EV for that camera).

Since Panasonic is using DFD for continuous focusing, they need to continue pushing the boundaries for CDAF. My concern with the on-sensor solution is that companies will stop developing CDAF and just focus on PDAF from the sensor. Low-light results will likely suffer as a result.

--
My Flickr Birds

 

[EinsteinsGhost]

The curious thing about the comparison, is that the mirrorless cameras are all APS-C or MFT size sensors. I know there are more of those to test than mirrorless FF choices today, but that is a bit of an unfair comparison.

It will be more impressive when FF mirrorless comes this close on focus speed tests.

... lowlight capability (-4EV) of A7S...

I see several tests for the Canon 6D but none for the A7s. Can you post a link to this?

Look at the specs.

And where do you find AF speed comparison at -4EV? Perhaps you can point us to this demonstration of very fast AF at -4EV in GH4 that you claimed above?

I and others actually did tests with DSLRs and the GH4. I was shooting at ISO6400, F/2.8 and 1/8th shutter speed in near total darkness and the GH4 was still usually almost instantaneous.

In other words, you don't have a video demonstrating this very fast AF at -4EV.

I don't have a 6D, but the GH4 bests all the DSLRs I have tried. Add this to the near D4s C-AF focusing and we can see the real potential of mirrorless...and PDAF is no longer needed.

If you get an A7S, please post your experiences so we can see if it lives up to the marketing. IIRC, It does not have PDAF pixels, so it should be interesting.

 

[EinsteinsGhost]

One key area would be DOF. That is actually my only issue with otherwise an excellent comparison but I guess since Fuji does not have a faster telephoto zoom yet, DOF could not be properly equalized.

The other area will be file processing, where m43 are still at 16 MP, although in this particular test, only a6000 is higher than 16MP (all others are that, including D4s).

And even with single point AF, the circuitry for a sensor 4x as large (m43 vs FF) is still going to be more complex and longer.

I'm pretty sure they don't stream the entire image for doing AF--just a very small subset of data are downloaded.

Only if the camera uses a dedicated processor only for AF. The DOF issue is still not addressed.

And the extra distances involved with a larger sensor (which would only be 2x as long on average) will have negligible impact on AF processing update rate.

Can we define what qualifies as negligible? One of the key developments with integrated electronics is shortest possible path. Making it longer is opposite of that, even you may quantify it as negligible.

Bart

--
http://bhimages.zenfolio.com

 

[EinsteinsGhost]

Why? Focus speed should be independent of frame size I would think. And what this tells us is that they're packing more and more processing and control speed into smaller and less expensive packages.

Focus speed depends on CPU processing speed. A FF shot contains a lot more information and takes a lot more CPU power to process than an APS-C shotor an m4/3 shot . So focus speed is very much dependent on frame size, which is why Nikon's mirrorless was able to achieve very quick focus quite some time ago.

I could be wrong, but I think focusing speed is dependent on the number of points being used and my impression is that the focusing systems are not very aware of the size of the format. If you're using a single point right in the center, what does it matter the sensor is the tiny 4/3 or full frame?

One key area would be DOF. That is actually my only issue with otherwise an excellent comparison but I guess since Fuji does not have a faster telephoto zoom yet, DOF could not be properly equalized.

How would DOF influence focusing speed?

Precision and contrast requirements. Start with complete blur, and you have a bit more to go than if you start with subject that is within DOF (but not full focus). Would you not expect any difference?

But, why not make DOF equal?

The other area will be file processing, where m43 are still at 16 MP, although in this particular test, only a6000 is higher than 16MP (all others are that, including D4s).

And even with single point AF, the circuitry for a sensor 4x as large (m43 vs FF) is still going to be more complex and longer.

Keep in mind, I'm only talking about autofocus speed which seemed to be the primary point of discussion in the Camera World article. I don't think he touched very much on buffer and sustained cycle rate - all of which are influenced heavily by processing power.

There is only one processor in these cameras (not counting D4s), which has to do every bit of processing.

But just processing a focus point or two doesn't seem to have anything to do with format size when comparing the kind of focusing systems he was using in the article. And I wonder if the Nikon has even less to worry about since the TTL focusing is independent of the sensor itself.

You still need processor to process AF information and in fact the best solution might be with Sony A-mount (where TTL focusing is independent of mirror-movement as well).

 

[stevo23]

Why? Focus speed should be independent of frame size I would think. And what this tells us is that they're packing more and more processing and control speed into smaller and less expensive packages.

Focus speed depends on CPU processing speed. A FF shot contains a lot more information and takes a lot more CPU power to process than an APS-C shotor an m4/3 shot . So focus speed is very much dependent on frame size, which is why Nikon's mirrorless was able to achieve very quick focus quite some time ago.

I could be wrong, but I think focusing speed is dependent on the number of points being used and my impression is that the focusing systems are not very aware of the size of the format. If you're using a single point right in the center, what does it matter the sensor is the tiny 4/3 or full frame?

One key area would be DOF. That is actually my only issue with otherwise an excellent comparison but I guess since Fuji does not have a faster telephoto zoom yet, DOF could not be properly equalized.

How would DOF influence focusing speed?

This is something I'm not very clear about either. In the thread in the Open Discussion, one guy claims the test is how many "DoFs" the object passes through in a given series of shots. And given the shots were all shot at F4, apparently that gives m4/3 a bigger advantage (and the test wasn't particularly tough on any camera). On the other hand, the D4S would have been focusing at f2.8 since DSLRs AF wide open when not in live view mode, which means its AF is even faster. But then if having a thin DoF stretches C-AF capabilities (since that means the target crosses "more" DoFs in a series of shots), then what's the balance of trade-offs actually like?

Lighting conditions also play a big part in the final results. This test was done in ideal conditions. As light falls the GH4 will widen the gap over the other mirrorless solutions.

I think scene brightness has a big result on the accuracy of the results, but why do you think the GH4 has an advantage over others?