K5 raw is 14 bit

Started Sep 20, 2010 | Discussions
steelski Senior Member • Posts: 2,555
Re: K5 raw is 14 bit

GordonBGood wrote:

If we are to believe the specifications we have been given, I doubt that the limit is the imaging engine used as it seems it is still Prime II just as for the K-7, K-x, and K-r, all of which use 12 bit processing although there is a 14 bit Analogue to Digital Converter (ADC) available. Although from the Pentax record for this K-5's release, that possibly should have read PRIME III!

Regards, GordonBGood

just to correct you on something. The Pentax 645D also uses Prime II and it outputs 14bit RAW files.

MightyMike Forum Pro • Posts: 38,159
Re: BITs and DR are different things

I can say for sure that the K-7 RAW buffer can give 14 or 15 frames likely depending on the scene and perhaps the card... my K-7 RAW files easily ranged from 12mb to 21mb
--
Mike from Canada

'I like to think so far outside the box that it would require a telephoto lens just to see the box!' ~ 'My Quote :)'

http://www.airliners.net/search/photo.search?sort_order=views%20DESC&first_this_page=0&page_limit=180&&emailsearch=mighty_mike88%40hotmail.com&thumbnails=

GordonBGood Veteran Member • Posts: 6,307
Re: BITs and DR are different things

Kikool wrote:

Many thanks for your explanations

Sorry, my question might appear as trivial, but as my english skills are limited.

So, just to be sure to understand, do you think that the K-5 might be able to shoot about 14-15 raw files in a row under "ideal" shooting conditions, but be limited to 8 files at high ISOs when the files would be at their max. size ?

Yup, with it gradually having less buffer capacity as ISO goes up from about ISO 1600 and perhaps those worst cases not being normally encountered until ISO 25,600 to 51,200.

Regards, GordonBGood

Philip Goh Regular Member • Posts: 447
Re: BITs and DR are different things

Engineering is all about selecting the best compromise to match your product requirement. It would make no sense to reduce the number of frames that the buffer is capable of without actually gaining some performance in another area.

I take it that you've never worked at a large engineering company. In such companies, the engineering department is subservient to management and marketing.

JoepLX3 Regular Member • Posts: 416
Re: BITs and DR are different things

how many bit is the K-r?

GordonBGood Veteran Member • Posts: 6,307
Re: BITs and DR are different things

JoepLX3 wrote:

how many bit is the K-r?

Joe, from DNG's from a Beta version of a K-r, the raw bit depth is 12 bits, which may supports a Dynamic Range (DR) up to a maximum of about 13 stops (the K-x was about 12.3 to 12.4 stops of DR at lowest ISO).

Regards, GordonBGood

CherryO Regular Member • Posts: 332
Re: K5 raw is 14 bit

According to the spec sheet on the Pentax Japan site, an 8Gb card will hold 233 PEF and 232 DNG. This means 34MB files.

Owen
--
http://www.flickr.com/photos/owencherry

KL Matt Veteran Member • Posts: 5,885
Re: K5 raw is 14 bit

Hi Gordon,

Many thanks for all your detailed explanations.

Generally, we need at least about at least 10 "good" linear raw bits in order to convert to a normal eight bit JPEG with reasonable image quality.

Is this perhaps why some photoshop experts recommend editing raw files in 16 bit no matter what your final conversion will be?

Thanks,

Matt
\\

-- hide signature --

... interested in .... photographs? Heh? Know what a mean? Photographs? (He asked him knowingly). Nudge nudge, snap snap, grin grin, wink wink, say no more, say no more, know what a' mean? Know what a' mean?

http://www.pentaxphotogallery.com/home#section=ARTIST&subSection=183820&subSubSection=0&language=EN

GordonBGood Veteran Member • Posts: 6,307
Re: K5 raw is 14 bit

KL Matt wrote:

GordonBGood wrote:

Generally, we need at least about at least 10 "good" linear raw bits in order to convert to a normal eight bit JPEG with reasonable image quality.

Is this perhaps why some photoshop experts recommend editing raw files in 16 bit no matter what your final conversion will be?

Matt, 16 bit editing is important as long as the data is in "linear response" representation as in no "gamma" Tone Response Curve (TRC) has been applied or if the particular operation/filter being applied required conversion to linear space and back. However, as most operations requiring that can be performed by a modern raw conversion program before further editing is required, there is likely less requirement for actual 16 bit editing. For instance, all colour conversions, exposure boosting, and contrast enhancement that most require the full 16-bit numerical range should ideally be done in the raw conversions process.

As Adobe Camera Raw (ACR) is in itself internally a 16 bit utility, and if you do these major adjustments using it, there isn't so much a requirement that you do further editing in Photoshop using 16-bit mode although it is an option if your are a purist.

Regards, GordonBGood

blende8 Senior Member • Posts: 1,494
Re: K5 raw is 14 bit

I'm so ignorant in this matter:
What is the problem to use 16 bit in the camera already?
Why these strange 12 and 14 bit?

-- hide signature --

blende8
PENTAX, mysterium quod absconditum fuit ...

GordonBGood Veteran Member • Posts: 6,307
Re: K5 raw is 14 bit

blende8 wrote:

I'm so ignorant in this matter:
What is the problem to use 16 bit in the camera already?
Why these strange 12 and 14 bit?

Well, a few problems:

1) There is no reason to use 16 bits to the output image file if one only has a 12 or 14 bit Analogue to Digital Converter (ADC), as we can get 16 bit's any time we want just by adding zeros or even better an average random value into the least significant bits. Sixteen bit ADC's are very expensive, especially if they are to have more than 14 Effective Number Of Bits (ENOB) or the number of bits that can actually be distinguished from each other in the noise.

2) There is also no reason to have even 14 bits of output data even if the ENOB of the ADC supports this if the least significant bits are mostly just random sensor/signal noise as is the case of all be the lowest dark read noise cameras at their very lowest ISO's for their very darkest tones something like less than 9 stops below full clipping bright.

Even 14 bit's hardly seems worth it does it, especially considering that those extra two bits will add at least about a quarter of a MByte per MPixel to the average raw image file sizes?

The only use I can see for even a 14 bit file with current cameras is if the images are shot underexposed for the 14 bit range so that most of the extra range is in the "highlight headroom", in which case there is some more "exposure latitude" and the file sizes won't increase as much, but I haven't proven yet that any cameras do this, although perhaps the K-5 and 645D do it.

Regards, GordonBGood

KitTP Regular Member • Posts: 496
Re: K5 raw is 14 bit

I thought I saw an interesting bit in the K-5's specs - the metering goes from 0-22EV, 1 stop higher than the K-7's.

May be related to higher Dynamic Range with the Sensor.

I think it may mean +1 stop in brightness before the metering treats it as "Overexposure". Does that make sense ?

blende8 Senior Member • Posts: 1,494
Re: K5 raw is 14 bit

There is quite a mess regarding the exact specs.
I have found many different values for the K7 and the K5*.

I think that 0-22 EV is correct for both the K7 and K5, because the metering system is the same.

  • for the K5 I have found:

0 - 21 EV
1 - 21,5 EV
0 - 22 EV

-- hide signature --

blende8
PENTAX, mysterium quod absconditum fuit ...

distudio Veteran Member • Posts: 3,913
Re: K5 raw is 14 bit

blende8 wrote:

I'm so ignorant in this matter:
What is the problem to use 16 bit in the camera already?
Why these strange 12 and 14 bit?

Firstly you need to fully appreciate what types of files these bit depths are being attributed to. In the case of the RAW files the bits refers to the bit depth of the final output of the image capture chain in including ADC and RAW processing. In all Pentax cameras to date the RAW files are uninterpolated Bayer filtered images i.e. they are monochrome, they have one discrete filtered colour value recorded per pixel and the image processing chain is designed optimally to provide either 12 or 14 bit per pixel files. There is no point adding data that provides no significant image forming value in the case of RAW files as it just bloats file sizes which increases buffer flushing times and storage times and reduces the number of image files that can be stored on a given card.

In the case of post processed images (either in camera as jpg or via a RAW converter) the bit depth per pixel is actually 24 bits in what is referred to as an 8 bit file or 48 bits in the case of 16 bit files. This is because each pixel in the interpolated images now contains a Red, Green and Blue component value, each at the standard 8 or 16 bits.

-- hide signature --

Rob

 distudio's gear list:distudio's gear list
Sony RX100 II Pentax K-3 Pentax K-5 IIs Pentax 645Z Nikon D750 +19 more
borno
borno Senior Member • Posts: 1,191
Re: K5 raw is 14 bit

I do think my 50D seems less likely to get the color bands in the sky than my k20d or k10d. I was thinking the 14bit files might be why, but I'm far from an expert on such things. : )

 borno's gear list:borno's gear list
Canon PowerShot G1 X Canon EOS 7D Mark II Canon EOS-1Ds Mark II
PhotoHawk Contributing Member • Posts: 832
Re: K5 raw is 14 bit

Regarding 12 versus 14 bits I don't think Clarke agrees with you. Moreover I would think Canon engineers wouldn't either (or Nikon or Pentax or Mamiya or Hasselblad).

Regarding the 300s you could be right. However I believe higher quality AD would give better results.

samusan Regular Member • Posts: 146
Re: K5 raw is 14 bit

PhotoHawk wrote:

Regarding 12 versus 14 bits I don't think Clarke agrees with you.

You might have been misleaded by figure 4 plotting the 1DIV at 15 stops : this is 15 "virtual" stops : a model of what would happen with perfect ADC.

But Clarke does agree with Gordon wrt 12 bit files being enough, as you can see at the bottom of http://www.clarkvision.com/articles/evaluation-canon-1div/index.html

"Dynamic range is still limited to a little over 11-stops, apparently by downstream electronics that must process the data extremely fast"

So yeah, supposing camera always chooses proper quantization for each ISO, 12 bit RAW files could always suffice.

Moreover I would think Canon engineers wouldn't either (or Nikon or Pentax or Mamiya or Hasselblad).

bah, 14 bit RAW files ? any marketing team will love them, even if it just means 4 times finer recorded noise. Customers love big numbers and they don't understand the theory anyway.

The brands' choices do not mean much of their engineers' opinions.

-- hide signature --

Samusan

GordonBGood Veteran Member • Posts: 6,307
Re: K5 raw is 14 bit

KitTP wrote:

I thought I saw an interesting bit in the K-5's specs - the metering goes from 0-22EV, 1 stop higher than the K-7's.

May be related to higher Dynamic Range with the Sensor.

I think it may mean +1 stop in brightness before the metering treats it as "Overexposure". Does that make sense ?

I think you are confused between the metering system and the sensor itself, which are two entirely different electronic devices.

The metering sensor is offset from the sensor light path in a similar way to the way that the Auto Focus phase detection sensor is, and is only used to determine the "correct" measurement of that zero to 22 EV range exposure according to the active algorithm based on those 16 metering segments, wither matrix, centre weighted, or spot (center) based.

Once the "correct" exposure has been determined as to aperture and shutter speed, this sets the actual range of light that the sensor is exposed to over the scene which may still be limited at the top end as to clipping and at the lower end by the base black read noise of the sensor. Some might consider that the ISO sensitivity gain comes into this, but actually all that is done at higher sensitivities is to not use more and more of the brighter tones available from the sensor by effectively reducing the upper bright clipping level and expanding the remaining tone range to fill the output range, thus effectively limiting the effective scene Dynamic Range (DR) proportionally to the increase in sensitivity.

One could have a 0 to 40 EV exposure system (if it made any sense to have such), and still have only a 6 EV DR of the sensor; the two are completely unrelated.

Regards, GordonBGood

GordonBGood Veteran Member • Posts: 6,307
Re: K5 raw is 14 bit

PhotoHawk wrote:

Regarding 12 versus 14 bits I don't think Clarke agrees with you. Moreover I would think Canon engineers wouldn't either (or Nikon or Pentax or Mamiya or Hasselblad).

Regarding the 300s you could be right. However I believe higher quality AD would give better results.

On what do you base the need for a better Analogue to Digital Converter (ADC) on the Nikon D300s? I state that the D300s already has an excellent ADC based on the following: that at ISO 1600 the "noise floor" limited Dynamic range is about 9 stops and this limit changes by almost exactly one stop per stop change in ISO sensitivity both up and down the range to ISO 200, with a step at ISO 100 only due to that this is not within the normal effective usable range of the normal gain setting of the sensor for this camera . A camera with a poor ADC as to Effective Number Of Bits (ENOB) such as most Canon models, including the 14 bit ADC models, has a distinct flattening of this linear trend line towards the lower ISO sensitivities.

By the Roger Clarke measurement, the D300 sensor has about a 13 stop DR instead of the about 12 stops actually measured, but that is due to using an extrapolated well capacity at ISO 100 of 42,000 electrons, whereas the camera never gets to that due to clipping and gain manipulations in order for the camera to appear to offer this extended ISO 100 range. The real limit to Dynamic Range (DR) is the primarily the base black read noise of the sensor.

OTOH, the Canon 1D Mark IV would indeed have the capability of a 15 stop dynamic range at ISO 100 if it had an excellent 16-bit ADC with an ENOB of at least 15 bits, where it is currently restricted to a low ISO DR of about 11 stops by its limited electronics processing chain.

samusan wrote:

PhotoHawk wrote:

Regarding 12 versus 14 bits I don't think Clarke agrees with you.

You might have been misleaded by figure 4 plotting the 1DIV at 15 stops : this is 15 "virtual" stops : a model of what would happen with perfect ADC.

But Clarke does agree with Gordon wrt 12 bit files being enough, as you can see at the bottom of http://www.clarkvision.com/articles/evaluation-canon-1div/index.html

"Dynamic range is still limited to a little over 11-stops, apparently by downstream electronics that must process the data extremely fast"

So yeah, supposing camera always chooses proper quantization for each ISO, 12 bit RAW files could always suffice.

Moreover I would think Canon engineers wouldn't either (or Nikon or Pentax or Mamiya or Hasselblad).

bah, 14 bit RAW files ? any marketing team will love them, even if it just means 4 times finer recorded noise. Customers love big numbers and they don't understand the theory anyway.

The brands' choices do not mean much of their engineers' opinions.

The coming proliferation of 14 bit raw files is supported by so many users misunderstanding the use of such files, even if one lives with their larger file size and even if the camera using them has both a sensor and data acquisition system that could use the extra bits.

For instance, let's consider the currently top rated Canon 1D Mark IV with an ADC system worthy of its sensor: The ISO 100 electron capacity of 55,600 still means that the camera could only use a true 14 bit output range at light levels below about 12.2 stops below full clipping bright due to the statistical variation of the photon arrivals ("shot" noise), meaning that at ISO 100 one gets a whole extra 3 or 4 discrete levels to play with for a 14 bit ADC and about 6 to 8 extra discreet levels to play with if an excellent 16 bit output electronics were used. Wow! This slight advantage quickly goes away as ISO sensitivity is increased, with the "extra" discreet levels that can be used dropping to zero at just over ISO 200. Thus by forcing an output of 14-bit or 16-bit raw files, one gains these few extra discreet usable levels in the extremely dark tones at lowest ISO sensitivy at the cost of larger raw image files size of over 4 MBytes and 8 MBytes, respectively - for all files at all ISO's and even when glare and flare mean there are none of these darker tones .

You may consider this worth it; I do not and object to Canon (and now Pentax) forcing these bloated file sizes on users without the option of turning it off, which choice Nikon users have. For instance, one might choose to use 14-bits for ISO 200 and under but not for higher ISO shooting, where the extra bits have no use whatsoever.

Regards, GordonBGood

MightyMike Forum Pro • Posts: 38,159
Re: K5 raw is 14 bit

Have you been able to crack the RAW stuff yet? is the 14 bit a waste of time/space and processing power or did they improve something? at least the jpeg output in the one thread at ISO6400 and ISO12800 seem to be a bit better then the K-x, but i don't shoot jpeg anymore
--
Mike from Canada

'I like to think so far outside the box that it would require a telephoto lens just to see the box!' ~ 'My Quote :)'

http://www.airliners.net/search/photo.search?sort_order=views%20DESC&first_this_page=0&page_limit=180&&emailsearch=mighty_mike88%40hotmail.com&thumbnails=

Keyboard shortcuts:
FForum MMy threads