Strange Things Said XXI

"A customer is more agile than a manufacturer."

While I'm paraphrasing something I read, this is indeed true, and it's the core of many of our complaints about cameras. In recent years, camera companies have become less agile due to a confluence of factors both under and not under their control. Even the smartphone makers have become somewhat less agile in the COVID era, though I suspect some of that is that they’ve run out of the easy ideas and gains.

Thus, complaining about what companies haven’t done yet will be a feature of the Internet basically forever.

“For [phase detect autofocus] it is necessary to embed a sensor for distance measurement in the sensor, and since color information and luminance information cannot be obtained from the embedded location, the image processing engine must compensate for it… Conventional image processing has not been able to guarantee image quality that satisfied us.”

Also “…using [PDAF], pixel defects will inevitably occur.” — Panasonic manager interviews in Japanese press

Google translation didn’t do these statements any justice, but even the original Japanese didn’t explain things clearly. Moreover, there’s a lot of “avoid embarrassment” in the entire explanation. Nikon has been doing since 2011 what Panasonic said they couldn’t do, and both Nikon and Sony currently use the same approach that Panasonic is now adopting, and have for many years. I’d also suggest that an out of focus photo has way more “pixel defects” than an in-focus one captured with a phase-detect-on-sensor camera ;~). 

To clarify, most phase-detect-on-sensor is done via putting some number of secondary lenses in the microlens layer just above the image sensor. Nikon, Sony, and I believe Panasonic, do this on every 12th row of blue/green photosites. The tricky part of this approach is that some pixels on the image sensor have two potential outputs: focus data and exposure data. Because the microlenses above the PDAF photosites are different (to create the phase information), the exposure data these can attain has to be adjusted for the impacted cells to “match” the others. That was one of the things that caused banding in deep shadows on early Nikon Z6 and Z7 cameras: the adjustment was ever so slightly wrong on rows with high contrast in them.  

In short, Panasonic is trying to paper over the fact that other companies moved as much as a decade ahead of them. I and others have been saying for some time that Panasonic had their head in the sand vis-a-vis phase detect autofocus. They can try to come with any excuse that they’d like, but unfortunately, they’re trying to weasel away from the reality that they are just last to market with phase-detect-on-sensor. 

The Fujifilm X-H2S is as good as the Sony A1 in autofocus tracking. —general conclusion by many after viewing a test done by a German Photographer based upon his “bird test track.

That video and resulting discussion eventually produced two articles and a long discussion on FujiRumors. 

The problem here is trying to summarize from limited data, and then attempting to put a spin on the data, including the headline “How 61% AF Accuracy…are Better than 85% Accuracy…” Basically, the argument is that 61% at 40fps is as good as 90% at 30fps. And that ignores the fact that there’s still a 12.5% improvement in number of sharp images from the Sony A1. I guess 12.5% less is somehow better in some alternative universe ;~). 

There’s a lot to unpack here, including the original “bird test track,” which is relatively slow and linear compared to what happens in real life, and doesn’t really take into account varying flight patterns, backgrounds, or foreground/background exposure. 

The A1, meanwhile, is a 50mp camera and the X-H2S is a 26mp camera. So we’re also not accounting for any differences in “focus” we might decide upon if we downsized the Sony images to the Fujifilm size. All of the evaluation of “in focus” in these tests is subjective (what the tester sees), not objective (measured in some repeatable way). Setting up a faux measurable test doesn’t really change things.

At least dpreview has backed away from relying on their bike test to talk about autofocus accuracy, but I’ve seen plenty of other Web sites try to provide absolute numbers for focus. For instance, there’s one that’s been stating that the Nikon Z9 has a 96% hit rate. Uh, okay. I can’t replicate that exact number, so I’d say that the testing and measurement that produced that number is not sufficiently defined to hold up to any real scrutiny. 

And that’s the problem: the camera companies don’t even begin to disclose how they test and approve their autofocus systems, and to what standards they’re using to generate that. In most machine language (AI) techniques that are being used, you establish a goal level of performance and iterate until you’re above that, then see if you can go still further without creating new common situations that fail. It very well may be that in the test pattern each camera company has established, that they’re all hitting 93%+ using their standards in their labs. 

In the field, not so much. 

However, the key point I’d like to make is completely missing in any of the discussions that are going around the Web right now: you have no control over what was missed. Yes, the X-H2S may get 24 sharp images per second, but that also means it’s getting 16 unsharp ones and you won’t know which ones those they are until you look at them later. It’s very possible that those 16 unsharp ones are the ones you really wanted.

I’d argue that I’d rather have 15 sharp images every second at 15 fps than what the X-H2S is doing. Yes, fewer images per second, but an assurance that all the images I obtained are usable. Indeed, that’s where Patrick’s (FujiRumors site owner) argument just falls on its face: The Sony A1 is producing 3 out of focus images a second in the stated test, while the Fujifilm X-H2S is producing 16. The likelihood that “peak moment” of what was collected is correct on the Fujifilm is only two-thirds as likely. I don’t like "two-thirds as likely” as a statistic ;~). I’d want to avoid that, as it means there’s far too much a chance that the image I want won’t be in focus. 

I’ll point out that I practice what I preach here. I set my Z6 II at 5.5 fps, not 14 fps, and for a reason: the way the viewfinder works and intrudes on autofocus performance on that camera. At 5.5 fps I can follow action and keep the autofocus system doing what I want it to, while at 14 fps I’ll lose framing and focus system positioning due to the slideshow viewfinder lag, and thus start getting out of focus images. Using 5.5 fps also forces me to not do long continuous bursts (though I can), but to be more selective about when I press the shutter release. As you probably have noticed, I get the image with my Z6 II. Pretty much always. Not 61% of the time, but somewhere >90%, just like I do on my Sony A1 and Nikon Z9. 

Everyone wants to declare winners and losers. Everyone wants their chosen brand to be the winner. In the end, what determines the winner is your understanding of a focus system and your consistency in driving (controlling) it to its best possible results. The “numbers game” that is being discussed on the Internet to declare winners and losers is not just misleading, but often wrong.

As if to further prove my point, today as I edit this article, Fujifilm has announced a firmware update to the X-H2S to improve subject detection and focus tracking. So it’s not as if they haven’t seen the same thing I did in my review sample.

“RAW means raw data” New York Times explanation

I don’t know why we’re still dealing with this problem, and I, too, sometimes fall prey to using the wrong capitalization due to varying uses by various companies. I think it was about 2006 when I started hearing complaints about referring to the capture of original data as RAW files. All caps is used to refer to acronyms, and there is no acronym involved here.

At some point not long after, I started trying to consistently refer to raw files as, well, raw files. My current standard in my books is to use “raw” to refer to such files generally, NEF to refer to a Nikon raw file specifically (and no, it’s not NEF raw, as NEF is always raw and doesn’t need the add-on). 

The problem is that camera and smartphone companies continue to misuse the language. For instance, the NYT didn’t catch that Apple actually refers to their iPhone raw file format as ProRAW. Why RAW is capitalized in that name, I don’t know, particularly because Apple’s video formats are referred to by Apple as ProRes. Sony, meanwhile, uses the term RAW generically for their ARW files, apparently to match the capitalization of JPEG and TIFF (both of which are acronyms, by the way). As does Fujifilm. The only company that might be justified in capitalizing their use of RAW is Panasonic, because that’s their actually file type (extension), and extension names are usually capitalized. 

Canon actually goes mansplaining on you, much like the NYT did: “It takes the name “RAW” because the data is raw.”  Yeah, really useful explanation there. It would be like me saying my name is THOM because my name is Thom. I call this concept negative data. There’s less information presented than needs to be presented ;~).

Which brings me, and likely the New York Times—though again, they missed what Apple calls their format—to have to come up with a standard by which to document things. Basically, we follow “style sheets” to try to come up with a consistent usage for our readers. 

Merriam-Webster is a common dictionary we writers and publications use, and they define “raw” as (2c) "not being in polished, finished, or processed form", which is exactly what a raw data file is. M-W’s definition is not capitalized. Adobe even gets this right: ACR is Adobe Camera Raw, not Adobe Camera RAW because you don’t put an acronym within an acronym (oh lord, why do I think some science fiction writer is now going to say to themselves, why didn’t I think of that, and thus come up with names that are names within names?*). Wikipedia gets this mostly right, though for some reason they call it “Raw” rather than “raw”. 

I will try to continue to be consistent, though being the sole editor of everything I write I do make mistakes from time to time that I don’t catch. I’m sure you’ll remind me when I get it wrong ;~). I’ll use the term raw to refer to the type of file generically, and I’ll use RAF, NEF, CR2, ARW, DNG, etc., when referring to the specific file type (and most of those are acronyms, thus justifying the capitalization). 

*Update: it was pointed out that GNU, PHP, SPOOL, HAM, and a few other of the more geeky sides of Unix take on this recursive form. I should have remembered that, as I was once one of those geeks. 

Looking for gear-specific information? Check out our other Web sites:
DSLRS: | mirrorless: | Z System: | film SLR: all text and original images © 2023 Thom Hogan
portions Copyright 1999-2022 Thom Hogan-- All Rights Reserved