Line skipping

[John McCully]

I have developed an allergy. Well, it is not entirely new but recently my suffering level has increased. Old age I suppose. While not life-threatening it is somewhat annoying.

My allergy is toward marketing BS regarding gear, the latest manifestation being 'no line skipping'. The line skipping I'm referring to is not what is happening in a certain Latin American country right now where people, most everyone, is lining up to purchase food, and line skipping - jumping in front of someone else - is likely to be your very last move. No, the line skipping I'm referring to is what happens when the data flow off a many pixiled sensor can not be captured fast enough to keep up with the data flow and therefore, horror of horrors, lines of data are skipped which results in the twin evils of moire and aliasing.

But heavens to Betsey, guess what; along came Sony with their Bionz X chip and the problem is solved! This chip of Sony magic takes in all the data, absolutely no bleeding line skipping, and spits out 'a sharp image devoid of major aliasing and moiré whether recording internally at 1080p or externally in UHD 4K' which is then comfortably saved to the card, not a bead of sweat involved.

That quote from the B&H website (my favourites DVINFO sponsor).

And as one might expect this technological breakthrough is prominently displayed wherever gear-heads might visit, and the informed discerning Sony gear enthusiast must have a device with a Bionz X inside (and an XAVC variant, but that's another story). But interestingly I can not find a reference to line skipping on a Sony Website.

So my question, as the Wicked Witch screamed at the baker and his wife; 'Who Cares'!

You will note that B&H say 'devoid of 'major' aliasing and moiré' and not 'none'. That's telling perhaps.

Furthermore, despite serious searching on the Internet I can not find anything official from Sony, a white paper or equivalent, that speaks to this 'breakthrough'.

Is this all more marketing BS or should us ordinary folks insist on having a Bionz X chip inside?

Thanks for your thoughts.

John

[Phil Goetz]

Full sensor readout is one of the terms used. The other option is an optical low pass filter. Another option is giving the shooter a choice between having the filter on or off. Or offering cameras with or without a filter. For example, Sony A7 has a filter. Sony A7r does not... I don't think the A7s does either.

[Mark Rosenzweig]

Originally Posted by John McCully

I have developed an allergy. Well, it is not entirely new but recently my suffering level has increased. Old age I suppose. While not life-threatening it is somewhat annoying.

My allergy is toward marketing BS regarding gear, the latest manifestation being 'no line skipping'. The line skipping I'm referring to is not what is happening in a certain Latin American country right now where people, most everyone, is lining up to purchase food, and line skipping - jumping in front of someone else - is likely to be your very last move. No, the line skipping I'm referring to is what happens when the data flow off a many pixiled sensor can not be captured fast enough to keep up with the data flow and therefore, horror of horrors, lines of data are skipped which results in the twin evils of moire and aliasing.

But heavens to Betsey, guess what; along came Sony with their Bionz X chip and the problem is solved! This chip of Sony magic takes in all the data, absolutely no bleeding line skipping, and spits out 'a sharp image devoid of major aliasing and moiré whether recording internally at 1080p or externally in UHD 4K' which is then comfortably saved to the card, not a bead of sweat involved.

That quote from the B&H website (my favourites DVINFO sponsor).

And as one might expect this technological breakthrough is prominently displayed wherever gear-heads might visit, and the informed discerning Sony gear enthusiast must have a device with a Bionz X inside (and an XAVC variant, but that's another story). But interestingly I can not find a reference to line skipping on a Sony Website.

So my question, as the Wicked Witch screamed at the baker and his wife; 'Who Cares'!

You will note that B&H say 'devoid of 'major' aliasing and moiré' and not 'none'. That's telling perhaps.

Furthermore, despite serious searching on the Internet I can not find anything official from Sony, a white paper or equivalent, that speaks to this 'breakthrough'.

Is this all more marketing BS or should us ordinary folks insist on having a Bionz X chip inside?

Thanks for your thoughts.

John

The proof is in the resolution chart. Slashcam.de tests cameras and camcorders and provides 1080 resolution charts. I downloaded one from the Sony RX10 - which avoids line skipping - and the Sony RX1000 III - which does not. You can clearly see the difference in aliasing. Allergy cured?

[David Heath]

Originally Posted by John McCully

You will note that B&H say 'devoid of 'major' aliasing and moiré' and not 'none'. That's telling perhaps.

I wouldn't read too much into that quote. Even a top end HD camera with 3 1920x1080 chips will have *SOME* aliasing. It's inherent in a digital sampling system. The only way to avoid it is to heavily oversample and have a very aggressive low-pass filter, but that will bring other issues. (Be either unacceptably soft, and/or lead to too small photosites and need far more complex processing.)

The only real way to check what's going on is with proper tests, and for aliasing that really means zone plates.

I've got a big trip coming up and am looking to buy a bridge type camera for that, so the subject is relevant to me at the moment. I haven't been able to find any "proper" zone plates of the cameras I'm interested in, but best I've seen is Sony Cyber-shot DSC-RX10 Review: Digital Photography Review and putting the test rectangle on the resolution trumpets. The RX10 seems to do far better than most other cameras of the type in those tests, and on the basis of what I can see there I think I tend to think the Sony claims are justified.

Or at least mainly. It does seem to be giving full 1080 resolution from a 20Mp chip (which it wouldn't if it was line skipping). But there is aliasing above at very high frequencies. I suppose that's the real problem of trying to build a camera for both still and video use. If it was solely for video the aliasing could be greatly minimised with an OLPF - but that would make it a very soft stills camera! :-)

But it does seem as if Sony have got away from line-skipping, which is a big step forward.

[Mark Rosenzweig]

Originally Posted by David Heath

I wouldn't read too much into that quote. Even a top end HD camera with 3 1920x1080 chips will have *SOME* aliasing. It's inherent in a digital sampling system. The only way to avoid it is to heavily oversample and have a very aggressive low-pass filter, but that will bring other issues. (Be either unacceptably soft, and/or lead to too small photosites and need far more complex processing.)

The only real way to check what's going on is with proper tests, and for aliasing that really means zone plates.

I've got a big trip coming up and am looking to buy a bridge type camera for that, so the subject is relevant to me at the moment. I haven't been able to find any "proper" zone plates of the cameras I'm interested in, but best I've seen is Sony Cyber-shot DSC-RX10 Review: Digital Photography Review and putting the test rectangle on the resolution trumpets. The RX10 seems to do far better than most other cameras of the type in those tests, and on the basis of what I can see there I think I tend to think the Sony claims are justified.

Or at least mainly. It does seem to be giving full 1080 resolution from a 20Mp chip (which it wouldn't if it was line skipping). But there is aliasing above at very high frequencies. I suppose that's the real problem of trying to build a camera for both still and video use. If it was solely for video the aliasing could be greatly minimised with an OLPF - but that would make it a very soft stills camera! :-)

But it does seem as if Sony have got away from line-skipping, which is a big step forward.

You seem to have ignored my post above yours, which supplied resolution charts showing exactly what you are saying (maybe you cannot see them?). Is there some reason we should ignore those charts from slashcam.de? They clearly show the high aliasing artifacts of the line-skipping RX100 III and the lack of those from the RX10.

[David Heath]

Originally Posted by Mark Rosenzweig

You seem to have ignored my post above yours, which supplied resolution charts showing exactly what you are saying (maybe you cannot see them?).

Didn't ignore your post - started to reply before you posted, but it was a while before I finished. So didn't see your charts until after I'd posted.

Are you sure the RX100 result is the Mk3 - maybe not one of the earlier models? Because according to dpreview the Mk3 DOES have the Bionz X processor - it's the MkI and II that don't? They have the Bionz processor. (See Sony Cyber-shot DSC-RX100 III Review: Digital Photography Review - scroll down a bit for the comparison chart.)

If you select the page I linked to for the RX10, it gives you the option to compare side by side a portion of the test chart from differing cameras. I'm just not sure I'm seeing the vast differences between the EX10 and RX100 (all Mk's) that your screen shots suggest?

[John McCully]

Originally Posted by Mark Rosenzweig

The proof is in the resolution chart. Slashcam.de tests cameras and camcorders and provides 1080 resolution charts. I downloaded one from the Sony RX10 - which avoids line skipping - and the Sony RX1000 III - which does not. You can clearly see the difference in aliasing. Allergy cured?

No, if anything you have made it worse :-). According to both B&H and DPREVIEW the Sony RX100(0) III has a Bionz X as has the Sony RX10 and if the Bionz X processor eliminates line skipping as many (other than Sony) claim then your post has truly deepened the muck and mystery! What you are saying (I think) is that based on the Slashcam.de charts the RX100 III exhibits aliasing while the RX10 chart does not therefore the RX10 avoids line skipping.

I must be missing something here.

[Mark Rosenzweig]

The point is that the RX10 does not do line skipping and the RX100 III does. It is true that the BIONZ X processor evidently has the processing power to avoid line skipping, but whether that is implemented is another issue. The processor is necessary but not sufficient.

Maybe there are heat issues - and the tiny RX100 III (I am sure the chart I obtained is the RCX100 III) would overheat if the processor were run at full power.

Remember that this processor is used in the AX100 4K camcorder as well. So, it can support 4K video. The RX10 does not shoot 4K video. Again, the processor is necessary, but its existence is not sufficient for any particular video goody.

So, we can say that the BIONZ X processor really does allow good things for video - avoiding line skipping for 1080 video (and that really matters) and allowing 4K video. But its mere existence in a camera does not guarantee either will be implemented.

[John McCully]

What evidence, what science is behind your claim that 'The point is that the RX10 does not do line skipping and the RX100 III does'?

It does seem very odd to me that the manufacturer does not make such a claim. I agree that 'we can say that the BIONZ X processor really does allow good things for video' but that's no big deal as so do other manufacturer's processing engines in their cameras.

Your claim without serious supporting evidence is interesting but not particularly compelling particularly in view of the fact that Sony make no such claim, that I can find.

Just to be clear I'm not saying that the Bionz X processor does not enable the avoidance of line skipping but in the absence of sound science supporting the claim I remain highly sceptical.

[Mark Rosenzweig]

Originally Posted by John McCully

What evidence, what science is behind your claim that 'The point is that the RX10 does not do line skipping and the RX100 III does'?

It does seem very odd to me that the manufacturer does not make such a claim. I agree that 'we can say that the BIONZ X processor really does allow good things for video' but that's no big deal as so do other manufacturer's processing engines in their cameras.

Your claim without serious supporting evidence is interesting but not particularly compelling particularly in view of the fact that Sony make no such claim, that I can find.

Just to be clear I'm not saying that the Bionz X processor does not enable the avoidance of line skipping but in the absence of sound science supporting the claim I remain highly sceptical.

You are correct I do not have the proprietary schematics to show you what Sony cameras are doing. I do have a brain and some knowledge.

Here is how we make inferences:

It is known that sampling from the entire sensor requires a high-power processor. Sony now has that processor power (those specs you can look up).

It is known that line skipping does produce moire.

When we both see no moire *and* very high resolution (see the chart, and also stills from the camera)) and bragging about a powerful processor, we are inferring that line skipping is not going on since anti-aliasing filters (another means of reducing aliasing) would degrade resolution. And the resolution is great.

No moire+high resolution+powerful processor->no line skipping. David is drawing the same conclusion.

Could be something else no one knows about. If you have some other hypothesis or information for how the absence of aliasing is going on tell us.

Camera companies do not typically talk about how they achieve results. For example, Panasonic obviously and evidently takes a central crop from a large sensor to get 4K video rather than sampling, while Sony does not. I have not seen either manufacturer discuss this.

[John McCully]

I am correct in questioning your claim which you now admit is just an inference and you do not have supporting data. As it happens I also have a brain and some knowledge so that proves nothing much. I don't see that you and David are drawing the same conclusions. Re-read his post.

Whatever, you are certainly not the only one brazenly converting an inference into a blatant statement of fact. David certainly did not do that. My scepticism is well placed methinks. And yes, it could be something you don't know about. Having said that let me add that you might be correct - I simply don't know.

Suffice to conclude that at this point in time in the absence of hard facts all this talk about line skipping is conjecture and perhaps another example of marketing BS, perhaps not. The jury is still out.

[Mark Rosenzweig]

Originally Posted by John McCully

I am correct in questioning your claim which you now admit is just an inference and you do not have supporting data. As it happens I also have a brain and some knowledge so that proves nothing much. I don't see that you and David are drawing the same conclusions. Re-read his post.

Whatever, you are certainly not the only one brazenly converting an inference into a blatant statement of fact. David certainly did not do that. My scepticism is well placed methinks. And yes, it could be something you don't know about. Having said that let me add that you might be correct - I simply don't know.

Suffice to conclude that at this point in time in the absence of hard facts all this talk about line skipping is conjecture and perhaps another example of marketing BS, perhaps not. The jury is still out.

This is what David said: "But it does seem as if Sony have got away from line-skipping, which is a big step forward."

Same as I said - you cannot get the RX10 results and line skip (I could have added unless there is some mysterious technology no one has ever heard of).

Your skepticism is not well-placed; it is not based on anything. There is good reason to believe there is no line skipping by the RX10; there is no good reason to doubt it. But you are free to do so. You can also doubt the sun will rise tomorrow, and post a thread expressing your skepticism about that and demand absolute proof the sun will rise.

I enjoyed your first post in this thread, and I am certainly sympathetic with your dislike of marketing hyperbole. But, again, the RX10 really delivers a good 1080 video result, whch is not marketing BS, and that is ultimately what matters.

[David Heath]

Originally Posted by John McCully

Whatever, you are certainly not the only one brazenly converting an inference into a blatant statement of fact. David certainly did not do that.

Suffice to conclude that at this point in time in the absence of hard facts all this talk about line skipping is conjecture and perhaps another example of marketing BS, perhaps not. The jury is still out.

Well..... what I would say is that I'd like to see more definitive charts, and know exactly the conditions they were shot under. For example, I believe the GH4 can work in different modes, one using the whole sensor (or a 16:9 section with the full width) or a 3840x2160 central crop. They will give different results, and different angles of view, and by knowing the basic sensor dimensions and with calibrated zone plate results it's possible to get a pretty good idea of what the camera is fundamentally doing.

As far as this case goes, then I'm minded to agree that the RX10 is indeed probably *not* line skipping, but I'd want to see more definitive tests.

And really such tests can far more easily show up definitively that line-skipping IS what a given camera is doing than prove the negative. And in such a latter case, the charts are likely to show a good result, so then it becomes more a case of "so what"? If it is line skipping, so what if it produces a good result?

The trouble is that "line skipping" has become to be seen as an undesirable end in itself, to be denied whatever the technical truth is - something that leads some marketing people to talk the BS that you so deplore, John! :-) Whereas in practice it's far more complex and there's line skipping and line skipping.....

Basically, if we're talking about a Bayer type sensor, what any camera can do really boils down to a relatively small no of basic choices. Ideally for 1080 video you want a sensor of 2x the pixel count of the output in each axis - so 3840x2160 - to give high quality output with simple processing. Unfortunately, that count is not really considered high enough now for decent stills work, so in the sort of camera we're talking about we have to start with a much higher count - so what to do?

And there are really only 3 basic decent choices nowadays:

1/ Take a 3840x2160 crop in the centre. High quality with simple processing and likely minimal aliasing - but at the expense of a crop factor - narrowing the angle of view.

2/ Line skip. The best way of doing this is to read the lines in pairs (read two, skip two) then ignore half the photosites in each line, again two at a time. Which means you get one usable 2x2 Bayer block in every 4x4 block of photosites, and values for R,G,B can be directly read from the corresponding colour photosites. It follows that the non-aliased resolution MUST be exactly a quarter of the sensor dimensions, the characteristics are likely to be symmetrical in H&V, it will have full resolution for each colour, and also that any aliasing is likely to be monochrome.

So start with a 16Mp sensor and the horizontal photosite count will be about 4600 photosites. With line skipping as above it predicts horiz res of 1150, which equates to just under 650 lpph for a 16:9 frame. When zone plates show exactly that figure and the other criteria above are met, you can be virtually certain it's line skipping!

3/ Read all the photosites each frame, do a full deBayer at frame rate, then downconvert to the output resolution. Theoretically the best way to do it - but it requires far, far more computing power than the above, and that has tended to mean more power and more heat. As said before, I'd like to see zone plates, but on the basis of the dpreview chart I tend to think that is what the RX10 is doing. It is giving good resolution - better than may be expected if it was line skipping - but if there is a "but" it may be that the downconversion stage leaves a lot to be desired.

Ideally, the downconversion stage would include a digital low pass filter to limit the resolution BEFORE doing the actual downconversion, and from the (scant) evidence from the dpreview chart it doesn't seem to be doing that very well. So my thought (with scant evidence) is that with this processor, it really is quite a quantum leap forward. Maybe not perfect, but you are really pixel peeping then, and a lot better than what's been the norm for DSLR type video off high count sensors for the last couple of years.

(One good thing about the line-skipping approach is that it tends to produce an intermediate output of LOWER res than the desired output - so you derive the latter via UPSCALING, not DOWNSCALING. And upscaling is far easier to do well. So the 650lpph intermediate is easy to upscale to both 720 and 1080.)

[John McCully]

Glad you enjoyed my first post. Did you see the movie 'Into the Woods'? I haven't yet but the various clips and trailers are great. Incidentally seems the colourists went big on the sulfurous greens which while not my cup of tea were well done, in my humble opinion.

And truth be told I have enjoyed many of your posts here and there. In particular I have appreciated the sharpness of your thinking hence my surprise at the dogmatic stand you have taken here, unlike David who moderated his comments with 'it does seem' which is a country mile away from your 'fact'.

To equate the sun rising with my questioning the line skipping is a tad far fetched, don't you think. Hardly relevant in my opinion. Whatever let's be done with this back and forth as this will get us nowhere fast. You are indeed free to conclude that the Bionz X processor eliminates line skipping. Me, I'm keeping an open (sceptical) mind, for now.

I have the RX10 and yes, the video is excellent, no doubt about that. And I can buy into the notion that the Bionz X processor is playing a part here. But I won't claim 'no line skipping', for now. I should also point out that just a moment ago I took the RX10 and the NEX 5n with the somewhat upsharp Sony E 18-200mm F3.5-6.3 OSS attached out and shot some tiled roofs and a corrugated iron building. No moire and aliasing with both cameras. In fact I have not infrequently cut footage from both the above together in Vegas Pro 12 and with a bit of fiddling here and there both in the cameras prior to shooting and in Vegas Pro 12 I would defy anyone to tell the difference when viewed on my Dell 24 inch monitor.

I agree, results are the bottom line and not the words of promise.

Cheers.

[David Heath]

Originally Posted by John McCully

What evidence, what science is behind your claim that 'The point is that the RX10 does not do line skipping and the RX100 III does'?

The trouble is that it's the old "prove you don't beat your wife" argument.

Bruises tend to be proof of violence, but the absence of them could just mean it happened long enough ago for them to have healed.

With line skipping, charts can almost definitively show when it IS happening, but when it's not they can never definitively prove it's not.