Let's drop the first bomb

[Erwann Loison]

How good is the actual DFD-based AF and how good can we expect it to be(come)?

 

[starfly]

How good is the actual DFD-based AF and how good can we expect it to be(come)?

I too am curious about this

 

[Adielle]

We can expect pretty much anything, and it's probably as good as it is.

 

[Erwann Loison]

We can expect pretty much anything, and it's probably as good as it is.

Fire burns and water wets. Anything else? Dfd is already in use in other panasonic cameras so so there must be opinions

 

[TN Args]

How good is the actual DFD-based AF and how good can we expect it to be(come)?

The key issue is the artificial intelligence (AI) aspect.

The announcement confirmed the rumours that Panasonic is developing next-gen DFD by using AI to recognize aspects of the scene and better inform where DFD 'leaps' to.

Some kind of "deep learning" AI is mooted.

Of course, that sounds all very well, but implementation could range from a mickey-mouse slight improvement on scene recognition already built into cameras, to something serious that really kills it and allows CDAF to match or exceed the best in OSPDAF.

 

[solarider]

How good is the actual DFD-based AF and how good can we expect it to be(come)?

Depends how well the engineers can tweak the DFD-AF algorithm. And if it's possible to tweak them well enough to get over the hill so to speak.

 

[Wellington100]

How good is the actual DFD-based AF and how good can we expect it to be(come)?

More to the point I wonder why Panny did not go the PD route when they had a clean slate sensor and camera. PD is basically bulletproof whereas DFD has never been very good.

 

[TN Args]

How good is the actual DFD-based AF and how good can we expect it to be(come)?

More to the point I wonder why Panny did not go the PD route when they had a clean slate sensor and camera. PD is basically bulletproof whereas DFD has never been very good.

PD reduces IQ.

 

[StevoPhilo]

DfD works really well in afing in low-light AF-S and it's really fast even in AF-C, but there are a few problems with it that don't really affect me too much as a photographer.

DfD wobbles in the AF system as it seems like it focus "breathes" a lot.

DfD isn't very good on the Video side in comparison to Canon's DPAF and Sony's AF.

DfD in AF-C can sometimes shift focus if there is a busy background and lose focus. (This may be addressed in the next firmware for the G9/Gh5 in October.

The bottom line is that it is one of the most responsive and fastest AF's I have used. It just lacks some consistency that you may find in like a Sony A9 or Nikon's 3d tracking.

 

[TN Args]

DfD works really well in afing in low-light AF-S and it's really fast even in AF-C, but there are a few problems with it that don't really affect me too much as a photographer.

DfD wobbles in the AF system as it seems like it focus "breathes" a lot.

DfD isn't very good on the Video side in comparison to Canon's DPAF and Sony's AF.

DfD in AF-C can sometimes shift focus if there is a busy background and lose focus. (This may be addressed in the next firmware for the G9/Gh5 in October.

The bottom line is that it is one of the most responsive and fastest AF's I have used. It just lacks some consistency that you may find in like a Sony A9 or Nikon's 3d tracking.

Sure, but this is about next-gen DFD. With AI.

 

[Kurgo]

How good is the actual DFD-based AF and how good can we expect it to be(come)?

More to the point I wonder why Panny did not go the PD route when they had a clean slate sensor and camera. PD is basically bulletproof whereas DFD has never been very good.

PD reduces IQ.

How though? Can't say I've seen terrible IQ out of Nikon, Sony and all other manufacturers that use PD or am I missing something?

 

[TN Args]

How good is the actual DFD-based AF and how good can we expect it to be(come)?

More to the point I wonder why Panny did not go the PD route when they had a clean slate sensor and camera. PD is basically bulletproof whereas DFD has never been very good.

PD reduces IQ.

How though? Can't say I've seen terrible IQ out of Nikon, Sony and all other manufacturers that use PD or am I missing something?

Depends how demanding you are.

OSPDAF 'robs' pixels from imaging duty, and they get interpolated (guessed). Also, their physical construction sometimes causes lines or stripes to occur on images.

Canon's DPAF version of OSPDAF seems to affect dynamic range by reducing the light-gathering proportion of the total sensor area. And they don't seem to be world-leading at AF either. >>gloves on LOL<<

 

[rashid7]

Panasonic will work on this -improving DFD performance- like their life depends on it

(because it does

 

[MayaTlab0]

How good is the actual DFD-based AF and how good can we expect it to be(come)?

More to the point I wonder why Panny did not go the PD route when they had a clean slate sensor and camera. PD is basically bulletproof whereas DFD has never been very good.

PD reduces IQ.

How though? Can't say I've seen terrible IQ out of Nikon, Sony and all other manufacturers that use PD or am I missing something?

Various problems, often related to striping in areas affected by flare, or to striping in underexposed areas.

An example of the sort of thread you often see about the impact of masked PDAF pixels on IQ :

Loading…

www.dpreview.com

DFD and DPAF should be immune to these issues.

I don't think that anyone will defend the idea that masked pixels are the future (at least as long as these issues can't be effectively solved). But right now in 2018 it seems to provide the best balance in terms of AF performance in various scenarios.

I guess that no one can know for sure how Panasonic's S1 and S1R version of DFD will effectively focus in various scenarios as it seems to heavily rely on sensor readout speed and even more so on processing power and algorithms.

One question that I'm wondering about though is : Panasonic only supports DFD with Panasonic lenses in the m43 format. Will it still be the case with the L mount, or will DFD be also supported with Leica and Sigma lenses ?

 

[Nomoreheroes]

How good is the actual DFD-based AF and how good can we expect it to be(come)?

We will se what new algoritms for DFD does in AF performance for the GH5 in October.

New FW coming

Loading…

www.panasonic.com

 

[Erwann Loison]

From your answers so far it doesn't look too bad. I'm mostly concerned if it can be snappy and track ok (or even well). I don't really care at all about smooth video focusing at the moment. I'm not asking for Nikon-like 3D tracking, although I would want the focus acquisition to be fast. Tracking correctly is a plus but I could live with it if it is just good (not the best but definitely usable).

It's interesting to see Panasonic go the deep learning route, because the possibilities are really worth doing it (if I was a billionaire I would love to start a computational photography firm actually, I definitely see that as the short/mid term future of imaging devices to go beyond some physical limits of the current technology), but my experience with deep learning stuff is that it entirely rely on the data used for the training. It can do wonders or fail miserably in situations you'd judge pretty darn identical with your brain.

If I were Fuji I'd think of opening a website where user can contribute data so the system improve with real field cases rather than lab tests and what they could manage to pull off by themselves.

We'll see when the cameras are released, my #1 interest for this camera is to not have to choose between old and new mounts from the other companies and their respective lenses lineup. I would feel terrible when I finally reach the point of having enough saving to buy the gear of a lifetime (sometime in 2019) that my super expensive lens that I had to buy in EF, F or A mount doesn't have a new body to attach to in 5 years, forcing me ot use adapters. Then finally, there is a "new mount" version but my old mount version is now worth nothing. Interesting times ahead but also probably disgusting times for some.

 

[starfly]

From your answers so far it doesn't look too bad. I'm mostly concerned if it can be snappy and track ok (or even well). I don't really care at all about smooth video focusing at the moment. I'm not asking for Nikon-like 3D tracking, although I would want the focus acquisition to be fast. Tracking correctly is a plus but I could live with it if it is just good (not the best but definitely usable).

It's interesting to see Panasonic go the deep learning route, because the possibilities are really worth doing it (if I was a billionaire I would love to start a computational photography firm actually, I definitely see that as the short/mid term future of imaging devices to go beyond some physical limits of the current technology), but my experience with deep learning stuff is that it entirely rely on the data used for the training. It can do wonders or fail miserably in situations you'd judge pretty darn identical with your brain.

If I were Fuji I'd think of opening a website where user can contribute data so the system improve with real field cases rather than lab tests and what they could manage to pull off by themselves.

We'll see when the cameras are released, my #1 interest for this camera is to not have to choose between old and new mounts from the other companies and their respective lenses lineup. I would feel terrible when I finally reach the point of having enough saving to buy the gear of a lifetime (sometime in 2019) that my super expensive lens that I had to buy in EF, F or A mount doesn't have a new body to attach to in 5 years, forcing me ot use adapters. Then finally, there is a "new mount" version but my old mount version is now worth nothing. Interesting times ahead but also probably disgusting times for some.

Yeah, I'm in a similar situation where I have Oly m4/3 now and want to switch to FF. Currently Nikon Z is my top contender. Ruled out Canon because of no IBIS, and I think I'll prefer the larger body of the Nikon to the Sony (but haven't had a chance to hold one yet, maybe sometime next week in a store somewhere as the Nikon Z7 will start shipping this week).

The Panasonic offering seems interesting, though the body looks absolutely huge, and so do the lenses. Not sure if I'll wait another 6 months before the Panasonic becomes available, but at the same time I'm in no rush as I slowly try to sell my m4/3 gear first.

Anyway, on the topic of native vs adapted lenses, I too will prefer to only use native mount lenses for any system I end up buying in to. One thing I do like about the Nikon FF mirrorless offering is that if I want to buy a cheap old used lens for say the F mount, I can, and it should work flawlessly as the adapter is designed by Nikon and will be fully compatible. For example, I've never owned primes before, but I want to try using a 50mm f/1.8 prime, to see if I'd like it. Instead of buying the new native mount 50mm Z-mount lens at $600, I can buy a used 50mm f/1.8 F-mount lens for $150, and it should work without any issue using the Nikon adapter on a Nikon body. And if I don't really like primes after all, I can just sell it again, for probably the same price or slightly less, so very little financial risk there. A native lens I would have to buy new, so resale value will be a lot less of course.

So going Panasonic means that there will only be native lenses available, which will not be affordable per se. And yeah there might be 3rd party adapters released, but those never work perfectly. Overall not a bad thing per se, but a bonus you get with the Canon and Nikon eco-system, and not with a completely new eco-system.

 

[TN Args]

Anyway, on the topic of native vs adapted lenses, I too will prefer to only use native mount lenses for any system I end up buying in to. One thing I do like about the Nikon FF mirrorless offering is that if I want to buy a cheap old used lens for say the F mount, I can, and it should work flawlessly as the adapter is designed by Nikon and will be fully compatible. For example, I've never owned primes before, but I want to try using a 50mm f/1.8 prime, to see if I'd like it. Instead of buying the new native mount 50mm Z-mount lens at $600, I can buy a used 50mm f/1.8 F-mount lens for $150, and it should work without any issue using the Nikon adapter on a Nikon body. And if I don't really like primes after all, I can just sell it again, for probably the same price or slightly less, so very little financial risk there. A native lens I would have to buy new, so resale value will be a lot less of course.

So going Panasonic means that there will only be native lenses available, which will not be affordable per se. And yeah there might be 3rd party adapters released, but those never work perfectly. Overall not a bad thing per se, but a bonus you get with the Canon and Nikon eco-system, and not with a completely new eco-system.

It's better than that. Sigma are going to make an EF-L adapter. So put a Sigma EF mount lens on the adapter and it should work perfectly, since Sigma know exactly how they made their own lenses in EF mount.

 

[starfly]

Anyway, on the topic of native vs adapted lenses, I too will prefer to only use native mount lenses for any system I end up buying in to. One thing I do like about the Nikon FF mirrorless offering is that if I want to buy a cheap old used lens for say the F mount, I can, and it should work flawlessly as the adapter is designed by Nikon and will be fully compatible. For example, I've never owned primes before, but I want to try using a 50mm f/1.8 prime, to see if I'd like it. Instead of buying the new native mount 50mm Z-mount lens at $600, I can buy a used 50mm f/1.8 F-mount lens for $150, and it should work without any issue using the Nikon adapter on a Nikon body. And if I don't really like primes after all, I can just sell it again, for probably the same price or slightly less, so very little financial risk there. A native lens I would have to buy new, so resale value will be a lot less of course.

So going Panasonic means that there will only be native lenses available, which will not be affordable per se. And yeah there might be 3rd party adapters released, but those never work perfectly. Overall not a bad thing per se, but a bonus you get with the Canon and Nikon eco-system, and not with a completely new eco-system.

It's better than that. Sigma are going to make an EF-L adapter. So put a Sigma EF mount lens on the adapter and it should work perfectly, since Sigma know exactly how they made their own lenses in EF mount.

Ok, well that's a good start at least. Will be interesting to see how the non Panasonic lenses will perform on the Panasonic body though. With m4/3 non-Panasonic lenses don't perform as well on Pana bodies as Panasonic lenses.

 

[rashid7]

"It's better than that. Sigma are going to make an EF-L adapter. So put a Sigma EF mount lens on the adapter and it should work perfectly, since Sigma know exactly how they made their own lenses in EF mount."

-sounds EXTREMELY optimistic to me