- Thread starter Canon Rumors Guy
- Start date

You are using an out of date browser. It may not display this or other websites correctly.

You should upgrade or use an alternative browser.

You should upgrade or use an alternative browser.

Except that you don't. And it actually takes a little work to get a HDR image to look natural which is why I prefer blending. And before you say "ah ha! blending is not artistic!" blending is an advanced form of dodging and burning.

To me neither blending nor HDR is artistic. It's mostly mechanical process. HDR merge in lightroom, for example, is almost fully automatic, you only choose the amount of deghosting.

So all those years Kodak and Fuji were lying on their data sheets?

What data sheets? On how film respond to exposure?

You realize those are not mutually exclusive, right?

I realise logarithmic function is not linear so yes those are mutually exclusive.

Density or digital value, depending on medium.

If it's a digital value, the value itself is a linear function of number of photons (which in turn gets multiplied by quantum efficiency). Double the amount of photons and you double the value. Then the exposure is the logarithm of that value, and the resulting curve is logarithmic and not linear.

This was an abrupt jump:...then the value of 16384 will actually represent a value of 32768...

There's not one big leap of missing tones at the end of the scale.

Well if 0 maps to 0 and 16384 maps to 32768, it can be a smooth curve in between. Why does it have to be an abrupt jump? But anyway it doesn't work like that in ADC converters, it just doesn't make sense. Again the voltage gets converted to a digital value. It may get amplified before conversion, and there's also distortions and errors in the signal, but in general, 1 Ev = 2 times difference between the values after ADC. Therefore the difference between 0 and 16384 is 14 stops, not 15 and not 14.5.

To me neither blending nor HDR is artistic.

The sum total of your use of them does not determine if they are artistic or not. Plenty of people do make artistic use of them.

What data sheets? On how film respond to exposure?

I realise logarithmic function is not linear so yes those are mutually exclusive.

Stop being obtuse. You know exactly what data sheets and graphs I'm referring to and you know a characteristic curve has a linear portion. You also properly interpreted my statement which is why you argue about the details in every post after arguing about the definitions of words.

Sensors with the highest DR compress a nearly 15ev space into a 14-bit encoding. This is observable fact not subject to your theorizing. Enroll in an EE course if you want to learn more about ADCs and when/why this happens.

Stop being obtuse. You know exactly what data sheets and graphs I'm referring to and you know a characteristic curve has a linear portion.

No, I don't know. If you put the digital value from a pixel on one axis and EV number on another, you get a logarithmic curve, full stop. With no linear parts in it. If you have a link, please show it to me.

Sensors with the highest DR compress a nearly 15ev space into a 14-bit encoding. This is observable fact not subject to your theorizing. Enroll in an EE course if you want to learn more about ADCs and when/why this happens.

It's not possible to encode 15ev space into 14 bits, unless ADC is non-linear. Just physically not possible. And as far as I'm aware, even Sony sensors use linear ADCs.

Having just on/off would be one stop of DR.

Zero? That would be an array where all elements were set to the same number, and could not be changed from that number

And if your pixel is as small as a photon, that’s all the DR you’ll ever need

It's not possible to encode 15ev space into 14 bits, unless ADC is non-linear. Just physically not possible. And as far as I'm aware, even Sony sensors use linear ADCs.

If they’re linear, do the Sony ADCs output at 14 bits? Serious question, not trolling - if they’re linear, but output at a higher bit depth than the final output, it’s trivial to compress at the point it’s written to file.

Where do you get your data/specs on the Sony ADC implementations? Surprises me that they’re linear, I’d heard otherwise, but have no solid data at all, just heard speculation.

I've only seen linearity mentioned somewhere several times but no i don't have technical specs. However Sony raw output is known as 14-bit and TIFF-based, that implies linearity. If ADC squeezes 15-bit range into 14 bits, the output will need to be converted back to 15 bits using some inverse-to-ADC function but it's not happening. So it's linear and 14 bits all the way through.If they’re linear, do the Sony ADCs output at 14 bits? Serious question, not trolling - if they’re linear, but output at a higher bit depth than the final output, it’s trivial to compress at the point it’s written to file.

Where do you get your data/specs on the Sony ADC implementations? Surprises me that they’re linear, I’d heard otherwise, but have no solid data at all, just heard speculation.

I've only seen linearity mentioned somewhere several times but no i don't have technical specs. However Sony raw output is known as 14-bit and TIFF-based, that implies linearity. If ADC squeezes 15-bit range into 14 bits, the output will need to be converted back to 15 bits using some inverse-to-ADC function but it's not happening. So it's linear and 14 bits all the way through.

Ah, right, so you you’re not entirely sure. I was hoping you could point me to solid data.

Dtaylor is correct though, at least in the general case. In my audio engineering days for telecom, we squeezed 16 bit audio in to 8 bit all the time using non linear mapping. Exactly the same principle.

I finally found the article that stated it’s non linear encoded: dated 2014, when at the time, Sony was doing a non linear encoding from 14 bits to 11 bits. Then further compressing with an adaptive delta down to 8 bits. This is what caused their shitty rep previously in their raws.

Lossy compression of raw data is currently the only option available in Sony cameras of series NEX, SLT, RX, ILCE, ILCA, and the recent DSLR-A. The first part of this article is showing how to detect artifacts caused by this compression. We will be discussing the technical details of this...

www.rawdigger.com

Still haven’t found anything on their newer files though. Who knows, maybe they’re actually linear

Crucial point we were discussing here, does voltage from a sensor pixel get lineary converted in ADC into the digital value, e.g. in the range of 0..16383 ?

Ah, right, so you you’re not entirely sure. I was hoping you could point me to solid data.

Dtaylor is correct though, at least in the general case. In my audio engineering days for telecom, we squeezed 16 bit audio in to 8 bit all the time using non linear mapping. Exactly the same principle.

I finally found the article that stated it’s non linear encoded: dated 2014, when at the time, Sony was doing a non linear encoding from 14 bits to 11 bits. Then further compressing with an adaptive delta down to 8 bits. This is what caused their shitty rep previously in their raws.

## RawDigger: detecting posterization in SONY cRAW/ARW2 files

Lossy compression of raw data is currently the only option available in Sony cameras of series NEX, SLT, RX, ILCE, ILCA, and the recent DSLR-A. The first part of this article is showing how to detect artifacts caused by this compression. We will be discussing the technical details of this...www.rawdigger.com

Still haven’t found anything on their newer files though. Who knows, maybe they’re actually linear

Very interesting paper, thanks. However I think it's not about ADC linearity. It's all about processing and compressionafterADC, because cRAW is a compressed lossy format.

Crucial point we were discussing here, does voltage from a sensor pixel get lineary converted in ADC into the digital value, e.g. in the range of 0..16383 ?

Right, and I think the answer to that is: in the absence of real data, none of us know, and can only speculate. However, it is entirely possible, and actually quite common, to map large ranges of numbers in to smaller ranges using non linear encoding. Whether that is the case specifically with recent Sony cameras, no one seems to be able to point us to a reliable reference.

Yes it's an assumption but very plausible. If the sensor/pixel signal fits (potentially) say 15 bits before ADC and gets compressed into 14 bits in ADC, then it'll need to be decompressed later on back into 15 bits, but Sony raw files are said to be 14 bits.Right, and I think the answer to that is: in the absence of real data, none of us know, and can only speculate. However, it is entirely possible, and actually quite common, to map large ranges of numbers in to smaller ranges using non linear encoding. Whether that is the case specifically with recent Sony cameras, no one seems to be able to point us to a reliable reference.

It is a rule on this forum that you have to be a fanboy and suck Canon's you-know if you want to leave a comment

No - just don't be a whiny, flamebaiting troll about it.

Canon fans cant handle the truth

Your

Maybe just learn to use the cameras properly?

- Replies
- 17

- Views
- 8K

- Contains 1 staff post(s)

- Replies
- 19

- Views
- 8K

- Replies
- 5

- Views
- 7K

- Replies
- 3

- Views
- 4K