[biz-engineer]

It's still a huge assumption.

Are you fishing for more internal information? I'm afraid I can't tell you the secrets I know, except the accelerator isn't a gas pedal

---

Last edited by biz-engineer; 01-04-2019 at 01:01 PM.

[Kunzite]

Don't get snarky.

[biz-engineer]

Don't get snarky.

Well, in French , "accelerateur" is the gas pedal in cars. So, every time I read "accelerator" in Pentax camera topics it makes with think of the gas pedal of a car, and when pressing gas pedal on car, it usually makes more noise , not less noise.

[Breakfastographer]

However, I said "the NR detected by Bill Claff"; not NR in general.

I'm afraid I don't understand why Bill Claff would be talking about different NR than anybody else.

The Imaging Resouce interview hinted about a noise processing (among other types of processing) done by the accelerator, and a higher-level noise reduction done in the PRIME.
The Milbeaut processor (PRIME) does have noise reduction capabilities.

Ricoh @ CP+: Rewarding K-1 fans with a major upgrade, plus what comes next after the K-3 II?

Absolutely - some earlier Pentax bodies with the Milbeaut platform had RAW NR starting iirc at ISO 6400, and people complained about that, too. At that time, it was clear that it was done to keep up the frame rate for noisier ISOs, and the market wasn't looking at higher sensitivities as eagerly as it is today. With the Pentax K-1 and K-1 II, I'm not aware of any effect on frame rate, and the file size differences don't suggest there would be any such effect.

[Not a Number]

Another name for an accelerator chip is a co-processor chip.

[biz-engineer]

Absolutely - some earlier Pentax bodies with the Milbeaut platform had RAW NR starting iirc at ISO 6400, and people complained about that, too. At that time, it was clear that it was done to keep up the frame rate for noisier ISOs, and the market wasn't looking at higher sensitivities as eagerly as it is today. With the Pentax K-1 and K-1 II, I'm not aware of any effect on frame rate, and the file size differences don't suggest there would be any such effect.

Either you look at those NR from a photographer standpoint or from a pixel peeper standpoint.
Here, and at DPR, people look at this hardware NR from a pixel peeper standpoint.

If I look at it from photographer standpoint, every time I couldn't work around using a high ISO setting, it was when shooting sports indoors, and the noise was such that acceptable visual quality was when downsizing at around 12Mpixels, and other guys shooting with sport cameras 1DxII and D4 didn't do better, and customers were content to get 12Mpixel, 12 effective MP was good for them. After presetting exposure and WB manually, I shot 12Mp JPEG all day long.

For me, from a photographer's standpoint, it's weird looking a loss of details over a 36Mpixels image and high ISO, because all camera shooting sports have much less than 36Mp. I think I already wrote something like that earlier but no one seem to have understood, so I'm asking myself if all this discussion is not from people who actually never shoot at high ISO. The only case where a photographer must use a high ISO setting is when shooting moving subjects in low light, and I can't think of anything else than indoor sports where 36Mpixel resolution is not needed. For all other shooting cases, a tripod or SR will allow to shoot at base ISO, or the light will be sufficient.

Now, if the discussion is about to look at hardware NR from a pixel peeping and signal processing science standpoint, there is no need to use a Pentax camera for doing a study, there are tons of digital signal processing and image signal processing books that anyone can buy and study.

And if people want to ignore practical shooting condition in order to continue to discuss about hard wired NR, they can still do it. This remembers me when a company offered life warranty over tools for mechanics, and a customer asked "what is put it under a train? do you still replace the tool for free?" .... the sales man replied "if you really want to damage things, you will always find a way". It's the same idea for the Pentax K1 II, you can always set it at ISO256000 and shutter speed 1/8000th and complain that you don't get 36 Mpixel sharp, it's just not the way to use the camera, except maybe it's exciting for forum discussions and dPR reviewers, if it is so then it is a disappointment, and surely a sign that we don't make any progress with those review sites.

[Kunzite]

I'm afraid I don't understand why Bill Claff would be talking about different NR than anybody else.

Who's "anybody else"? People talking about the exact same kind of NR Bill Claff is detecting, using Bill Claff's chart?
Why are you twisting my words?

Absolutely - some earlier Pentax bodies with the Milbeaut platform had RAW NR starting iirc at ISO 6400, and people complained about that, too. At that time, it was clear that it was done to keep up the frame rate for noisier ISOs, and the market wasn't looking at higher sensitivities as eagerly as it is today. With the Pentax K-1 and K-1 II, I'm not aware of any effect on frame rate, and the file size differences don't suggest there would be any such effect.

To my knowledge, all CMOS cameras employ various noise suppression techniques; and not just starting at high ISO.
CMOS imaging sensors are noisy by their nature.

[Breakfastographer]

Ifyou have a lens that resolve 50 lppmm or less (typical for cameralenses), and the sensor outresolve that lens, you won't get in yourimages more scene details than the lens can resolve, but you willstill get more noise out of the smaller pixels of the image sensor.So, if in your image data you have pixels with values different fromthe pixel beside, you know that the jump in value is from noise andnot from the image. In that case, when you apply noise reduction atpixel level, you remove more unwanted noise than wanted detail. Andso, the image quality of the overall image is truly increased. Andthe less details are let through by the lens, the more significantwill be the image quality increase from noise reduction.

It should probably be pointed out that if the sensor is outresolved, a very simple method for reducing noise is a downsample-upsample cycle,or just downsample if the high resolution is not required by the desired output medium. Some NR methods, and this seems to some extent true of the accelerator, spread the noise rather than removing it, attenuating its intensity while increasing grain size, and the difference between a version denoised in this way and with noise intact before downsampling may not be that great.

Cutting to the chase here, a low resolving lens is really an argument against (1) a high resolving sensor or (2) a high output resolution, not so much an argument for baked-in noise reduction, imo.

Secondly, it should be pointed out that a 36 megapixel sensor does not resolve 36 million RGB pixels. It only resolves 18 million green, 9 million blue and 9 million red pixels. As a consequence, lenses are not as often outresolved as some may think. If there is one thing to take away from the considerable oeuvre of Ron Brandon, perhaps the Pentax community's (late) grand wizard of high resolution tele photography, it's that using a higher resolution sensor or a teleconverter is almost always worth it to resolve more detail, even if this pushes your lens well beyond the point where it might be said to be sharp at pixel level.

---------- Post added 01-04-19 at 02:16 PM ----------

Who's "anybody else"? People talking about the exact same kind of NR Bill Claff is detecting, using Bill Claff's chart?
Why are you twisting my words?

Not trying to do any such thing. I'm trying to understand why you said:

not NR in general.

What exactly are you excluding by that phrase?

[Kunzite]

What exactly are you excluding by that phrase?

I'm not excluding, but rather I'm specific about what I'm talking about.
People are looking at a spike in a PDR chart, and assume that marks a boundary between "no NR" and "NR" - no noise suppression otherwise. They also assume that the image content is irrelevant for the NR applied. And that they can see "detail smoothing" on "images made of only noise"

[BigMackCam]

I'm going to jump in here and say that, based on significant playing around with other people's test shots from the K-1 and K-1II - specifically, those from "the other site" - in darktable, there are parts of those images where the K-1II's image processing for raw files does seem to reduce extremely fine detail to a tiny - and I do mean tiny - extent. But then, there are other parts of the same images that appear to have (only just) better definition of detail.

I truly believe that for 99.9% of users, any tiny (and I say again - in big, to stress the point - TINY) loss of detail is more than offset by the benefits of the accelerator's rather clever real-time noise reduction.

Like most other folks here, I'd prefer that Ricoh hadn't made the accelerator's NR non-optional. But, really, if it wasn't for K-1 raw files being available as a direct comparison, no-one would have noticed anything but the positive effects of the accelerator. We wouldn't even be having these discussions, IMHO...

I still haven't invested in a Pentax full-frame camera, though it may well happen one day. If that time comes, I'd be quite happy with the K-1 or K-1II. Right now, if price were no object, I'd probably pick the K-1II.

[Breakfastographer]

I'm going to jump in here and say that, based on significant playing around with other people's test shots from the K-1 and K-1II - specifically, those from "the other site" - in darktable, there are parts of those images where the K-1II's image processing for raw files does seem to reduce extremely fine detail to a tiny - and I do mean tiny - extent. But then, there are other parts of the same images that appear to have (only just) better definition of detail.

I truly believe that for 99.9% of users, any tiny (and I say again - in big, to stress the point - TINY) loss of detail is more than offset by the benefits of the accelerator's rather clever real-time noise reduction.

Like most other folks here, I'd prefer that Ricoh hadn't made the accelerator's NR non-optional. But, really, if it wasn't for K-1 raw files being available as a direct comparison, no-one would have noticed anything but the positive effects of the accelerator. We wouldn't even be having these discussions, IMHO...

I still haven't invested in a Pentax full-frame camera, though it may well happen one day. If that time comes, I'd be quite happy with the K-1 or K-1II. Right now, if price were no object, I'd probably pick the K-1II.

I thought that maybe this is something you would be interested in as well:

Image comparison: Digital Photography Review

[BigMackCam]

I thought that maybe this is something you would be interested in as well:

Image comparison: Digital Photography Review

Ha ha Thanks I've been over that numerous times before now. Many times, in fact... and downloaded the images (at every ISO).. and processed them in darktable using a variety of NR, detail extraction and sharpening techniques. It's mostly those raw images and the results I've obtained in darktable that led to my previous statements

I will say, obtaining the best results from K-1 and K-1II images respectively seems to require a subtly different approach in processing at times. Perhaps this isn't so surprising, given the results you obtained from each camera's files when using DxO's PRIME noise reduction...

[Kunzite]

I'm going to jump in here and say that, based on significant playing around with other people's test shots from the K-1 and K-1II - specifically, those from "the other site" - in darktable, there are parts of those images where the K-1II's image processing for raw files does seem to reduce extremely fine detail to a tiny - and I do mean tiny - extent. But then, there are other parts of the same images that appear to have (only just) better definition of detail.

I truly believe that for 99.9% of users, any tiny (and I say again - in big, to stress the point - TINY) loss of detail is more than offset by the benefits of the accelerator's rather clever real-time noise reduction.

That's the thing - I pixel peeped until my eyes bled, and couldn't see any conclusive detail loss. Yet we're making a lot of noise about K-1 II's "smoothing", effectively scaring people away from an excellent camera (and driving them towards cameras with significantly less detail).
The other site's studio scene doesn't stay exactly the same from test to test, by the way.

[BigMackCam]

That's the thing - I pixel peeped until my eyes bled, and couldn't see any conclusive detail loss. Yet we're making a lot of noise about K-1 II's "smoothing", effectively scaring people away from an excellent camera (and driving them towards cameras with significantly less detail).

I'd generally agree. However, one area in DPR's test image to look at is the maroon jacket of the left-most chap in the group that looks like "The Beatles". For me, that shows the most obvious difference between the K-1 and K-1II at high ISO settings. Yet it's quite possible to pull back the detail that appears to have been lost in the K-1II. I showed this in another thread some months ago... I can't remember which one, but the posts will be there.

The other site's studio scene doesn't stay exactly the same from test to test, by the way.

For sure. In fact, the green feathery stuff and brown hair samples clearly move between tests, so they don't provide an opportunity for accurate comparison.

[BigMackCam]

I thought that maybe this is something you would be interested in as well:

Image comparison: Digital Photography Review

My apologies - I just noticed that your link was actually to the dynamic range comparison, which I haven't previously looked at.

I have now, though.

I can imagine there might be a very small number of users who are interested in camera-to-camera differences in +6 EV of shadow recovery, but really not that many at all. This is why many of Canon's cameras - especially slightly older models, with frankly shocking dynamic range compared to the competition (if we're to listen to reviews) - were and still are used to take award-winning landscape and other photos.

Seriously, this K-1II accelerator noise reduction issue is - for the most part, if not entirely - for folks obsessed with test photos and review numbers, and truly not an issue for most people taking photos... Again, that's just my opinion.