S1H Raw Format?

[HockeyFan12]

  On 5/8/2020 at 5:28 AM, Lux Shots said:

You must be mistaking the bit depth with dynamic range. The only thing they have in common is ration numbers. You can have 14 stops of dynamic range with a 10-bit ADC. Now you won't have fine gradation between the darkest and the brightest, but resolution and dynamic range are completely different.

Expand  

Thanks, that directly contradicts the dpreview article I read on the subject (and my understanding of dynamic range in audio):

https://***URL removed***/articles/4653441881/bit-depth-is-about-dynamic-range-not-the-number-of-colors-you-get-to-capture

Is this article wrong? I've been posting it online and will stop posting it if it is. I don't want to spread misinformation.

Granted, I get that the article is about raw bit depth with still cameras (where the raw data is linear, as off the sensor) and this is different with the C200 and Alexa, where the ADC has a greater bit depth and then the raw data is converted to 12 bit log. But C200 and Alexa raw is processed that way to save disk space, and that's different in still cameras. I still thought the ADC's bit depth correlated with maximum dynamic range give-or-take a bit of error for noise reduction or DXOMark normalizing to 8 megapixels.

With those caveats, I always thought ADC bit depth correlated directly with the maximum dynamic range a system could capture and that's verbatim how dpreview explains it. Anyway if I'm wrong I'll stop referencing the article.

[TheBoogieKnight]

  On 5/8/2020 at 6:28 AM, HockeyFan12 said:

Thanks, that directly contradicts the dpreview article I read on the subject (and my understanding of dynamic range in audio):

https://***URL removed***/articles/4653441881/bit-depth-is-about-dynamic-range-not-the-number-of-colors-you-get-to-capture

Is this article wrong? I've been posting it online and will stop posting it if it is. I don't want to spread misinformation.

Expand  

No it's not, as I've already said in a 12-bit system you can't get more than 12-bits of dynamic range, period. In audio it's the same: a 16-bit CD gives you around 96dB of dynamic range, nothing will give you more than that and adding dither isn't increasing the dynamic range.

But, one thing to consider is that noise below the floor will modulate the last bit/bits (this is what dither does anyway) so as others have said you can see see something under this simply due to the fluctuating noise which will have some statistical relevance to the original signal.

 

[TheBoogieKnight]

One thing I don't get is if the sensor outputs 12 bits, why not just store is as 12 bit H.264 (Hi444PP) or H.265? Purely down to CPU/chipset limitations or they don't want to damage sales of higher end products or?

 

[MeanRevert]

  On 5/6/2020 at 9:19 PM, sanveer said:

And perhaps dentistry too ...

 

You sound awfully familiar. Are you from Egypt, by any chance?

🤣

Expand  

I'm asking the same question myself.

They look to be in direct competition and now that the Blackmagic Pocket 6K is only $2K, I think the S1H is a harder sell.

[Neufeldt]

A few things:

 

ADC in this camera is linear, so no funny business with Log encoding or anything like that.

96db of dynamic range is an audio standard. The video/photo equivalent would be 48db (a stop is either 3db or 6db, power or amplitude).

IIRC N-bit ADCs have DR = 6.021N + 1.763dB (audio), or 3.021N + 1.007dB, so a 12-bit ADC = 74dB/37dB or 12.333 stops of dynamic range.

The sensor is 12-bit readout of ~6K photosites in a Bayer array.. When this is tested at 4K we have significantly more than 12 bits of readout per measured pixel. 8-bit 4k IS 10-bit 1080p, etc.

[Lux Shots]

  On 5/8/2020 at 7:39 AM, TheBoogieKnight said:

No it's not, as I've already said in a 12-bit system you can't get more than 12-bits of dynamic range, period. In audio it's the same: a 16-bit CD gives you around 96dB of dynamic range, nothing will give you more than that and adding dither isn't increasing the dynamic range.

But, one thing to consider is that noise below the floor will modulate the last bit/bits (this is what dither does anyway) so as others have said you can see see something under this simply due to the fluctuating noise which will have some statistical relevance to the original signal.

 

Expand  

Bits ain't stops buddy. Why is everyone interchanging bits with stops like they are the same thing, they clearly are not the same thing! Bits allow more graditions between a minimum and maximum luminance. The delta between minimum and maximum luminance is the amount of stops of dynamic range.

[KnightsFan]

  On 5/14/2020 at 12:03 AM, Lux Shots said:

Bits ain't stops buddy. Why is everyone interchanging bits with stops like they are the same thing, they clearly are not the same thing! Bits allow more graditions between a minimum and maximum luminance. The delta between minimum and maximum luminance is the amount of stops of dynamic range.

Expand  

You're right about what bits are and what dynamic range is, but virtually every modern sensor has linear ADC's, so bits equate to dynamic range stops because of the linear relationship. I don't know of any sensors that don't behave this way and suspect that none exist. (That's strictly dealing with the sensor ADC, not any encoding afterwards, which is hardly ever strictly linear.)

[Lux Shots]

  On 5/14/2020 at 1:19 AM, KnightsFan said:

You're right about what bits are and what dynamic range is, but virtually every modern sensor has linear ADC's, so bits equate to dynamic range stops because of the linear relationship. I don't know of any sensors that don't behave this way and suspect that none exist. (That's strictly dealing with the sensor ADC, not any encoding afterwards, which is hardly ever strictly linear.)

Expand  

I can put an 18-bit ADC on a sensor that has 9 stops of dynamic range. The only thing that will change is the delta between steps will be almost indiscernible numerically. Dynamic range is based on the physics of the pixel well, and how many photos it can accept before saturation.  

[TheBoogieKnight]

  On 5/14/2020 at 1:19 AM, KnightsFan said:

You're right about what bits are and what dynamic range is, but virtually every modern sensor has linear ADC's, so bits equate to dynamic range stops because of the linear relationship. I don't know of any sensors that don't behave this way and suspect that none exist. (That's strictly dealing with the sensor ADC, not any encoding afterwards, which is hardly ever strictly linear.)

Expand  

Exactly.

 

I still don't get why all these cameras take a 12-bit linear signal then convert it to a 10-bit format when encoding. Is it just to protect more expensive lines higher up the market?

[KnightsFan]

  On 5/14/2020 at 3:56 AM, Lux Shots said:

I can put an 18-bit ADC on a sensor that has 9 stops of dynamic range. The only thing that will change is the delta between steps will be almost indiscernible numerically. Dynamic range is based on the physics of the pixel well, and how many photos it can accept before saturation.  

Expand  

You can, but what I'm saying is every sensor currently being used in any camera we discuss here uses linear ADCs that double bits for every doubling of light. Theoretical ADC's aside, for every real camera on the market, including the S1H, dynamic range measured at full resolution will never be substantially greater than the bit depth of the ADC.

  On 5/14/2020 at 8:16 AM, TheBoogieKnight said:

Exactly.

 

I still don't get why all these cameras take a 12-bit linear signal then convert it to a 10-bit format when encoding. Is it just to protect more expensive lines higher up the market?

Expand  

Well it's exactly because of what @Lux Shots is saying: bit depth has more to do with gradations than dynamic range (unless you're talking about 1:1 linear ADC's). 10 bits is plenty to store all the usable data from any sensor, when encoded logarithmically. There's really no point to storing a 12 bit image if you sensibly map values into 10 bit.

[deezid]

  On 5/14/2020 at 8:16 AM, TheBoogieKnight said:

Exactly.

 

I still don't get why all these cameras take a 12-bit linear signal then convert it to a 10-bit format when encoding. Is it just to protect more expensive lines higher up the market?

Expand  

L.O.G.