Info

ollywood88

ollywood88

Well-known member
Hey all,

Can someone explain the difference between latitude and dynamic range.
How does it work and affect the film\camera

I have only a basic knowledge of it

Thanks
 
I may be wrong, but I've always looked at the two as being fairly similar, just different ways to describe the same thing. DR is basically how many stops of information you have light wise. For example, or eyes would be considered to have 20+ stops of DR, where film/digital normally maxes out at 13-14 stops. Of course, your eyes don't take in information in a linear fashion like cameras do.

The best way to see the difference in DR and what it actually is, is to take a cheap camera and then a really nice one like a RED/Alexa and point it at a scene with high DR. A good test is blasting sun with something dark in a shadow. Stick a camera in a dim room with bright sunlight outside. If you try to expose for the inside, the outside will blow out even on a nice camera. However, a camera with more DR will take less light to equalize the difference (this is highly over simplifying it). You normally want as much DR as you can get in your recorded image because you can always throw the info away (which is done more often than not in a color session), but you can't get back what's not there.

So simply put, the more DR you have, the more latitude you have in your image to push it around in post, and the more natural it will look because it will be closer looking to what your eyes see.

Matthew
 
Thanks man, Thats exactly what i thought. But throws me when people talk about both and when you look at it it means the same. I no film is all about latitude about how you can still blow something out and still see the detail. UnLIKE MANY DIGITAL CAMERAS.

Im guessing to you have to shoot pretty flat, and dont use to much metadata as this would affect/make harder in post.

Thanks
 
ollywood88 said:
Hey all,

Can someone explain the difference between latitude and dynamic range.
How does it work and affect the film\camera

I have only a basic knowledge of it

Thanks
Click to expand...

Since I was not in 'film film' for years, but 'still' film, I only heard of 'latitude' for films. Don't know if movie people in the film film days used dynamic range or not.

But for me the use of the word dynamic range has come in via 'video', along with such terms as 'gain'. That's an 'electronic' term, as is 'dynamic range', at least for me.

In terms of film film, and latitude, it refers to the ability of the material to record a range of values in a scene. Those who followed the Zone System would think in terms of
eventual 'print' values, and then interpret the current scene for light values that would yield those 'output values'.

All of this was based on a the HD graph of the film's response to light, and as a graph one could speak of 'wide latitude', which means the 'shoulder' of the HD graph was 'farther' to
the right than a 'narrow' latitude film.

The most significant difference between digital and film, is that film responded to light with an "S" curve, and in the olden days, giving more light would actually cause an inversion
of the negative, giving a 'solarization' effect, that is light intensity after some point would yield less density than lower intensities.

Not so with digital sensors, at least ones that use single cells to capture the photons... they respond linearly until saturation... Film had a 'shoulder' and so increasing amounts of light
would yield less sensity change, until the above mentioned 'solarization' point, or in more modern materials, just block up with no additional change.

In terms of film, the wider the latitude the more 'grain' the film typically had. So, the cost of being able to shoot such a wide range was 'grain' in the resulting image. Push or Pull processing
would allow the photographer to 'adjust' for light ranges that exceed the 'printable' range of values, which was less than the range of the negative.

There are no real equivalents to this in the digital world at the moment, at least for the 'lowend' camera user... don't know what the big boys do... But I'm sure there some DSP involved to
try to match film characteristics better in this regard...

Fujifilm made (makes still perhaps...) a sensor which had differnt sized cell sites, such that they could get better highlight detail because the sensors had differential saturation points, and
so would mimic some elements of how film responded to light. Don't know if other manufacturers have followed that path or not.
 
Dynamic range is a term used for bit depth in digital acquisition of a signal. Latitude is how many stops of light can be resolved by film. When either are used to define discrete steps in a scale from light to dark in motion capture, they can be used interchangeably.

Just remember that neither of these are as important as you think. A lot of work is done to reduce the amount of dynamic range in scenes (fill light, etc) but increase the perceived contrast. It's this perceived contrast that makes things pop and look good, not the total dynamic range.

However, as mentioned before, the response knee of film and digital are different and contribute to the main difference in how each look.
 
Dynamic range is a term used for bit depth in digital acquisition of a signal.
Click to expand...
Actually, dynamic range and bit depth are two different (often independent) properties. Dynamic range is the ratio of the maximum possible signal level relative to its noise level (usually expressed in decibels), while bit depth is simply the number of shades possible from a digital pixel.

Imagine that we have two digital cameras with the same number of possible shades per pixel (the same bit depth), but the range of brightness between "consecutive" shades of each pixel is greater on one camera, compared to the other. So, the camera with the greater range of brightness between shades produces a greater total range of brightness, in spite of the fact that the two cameras have the same bit depth.

Of course, this example works the other way round: we can have two digital cameras with different bit depths (one that is 8-bit and one that is 10-bit, for example) with both cameras giving the same range of "brightness" in their signals. Furthermore, the 8-bit camera could easily produce a greater range of brightness than the 10-bit camera.

Keep in mind that in the decades before digital systems, we had analog systems (audio, video, radio, etc.) that lacked bit depth (incremental levels). In spite of the lack of bit depth, one could measure the dynamic range of these analog systems.

By the way, color depth and bit depth are also two different properties.
 
In the electronics industry, they are used interchangeably for all practical purposes. You are technically correct in your statements however the final outcome is that dynamic range and bit depth are always bound together in practice. ENOB, (effective number of bits, or how many bits of usable data) is the direct correlation of bits to dynamic range when you subtract the noise floor from the total. Generally, without going into the math, you lose about 2 bits to noise in the signal chain and/or in the pre-A/D filtering.

I don't know about these cameras today but in the test and measurement portion of the electronics industry, we generally use the next highest bit A/D so that the bits we lose to noise don't affect actual acquisition. If we expect to need 8 bits, we'll go with a 10 bit A/D and so on. I'd expect the cameras to do the same.
 
ollywood88 said:
Im guessing to you have to shoot pretty flat
Click to expand...

No, you just shoot appropriately for the look you're after -- which you should know before you shoot. If you shoot for the image you want during production, you have less to do in post.

and dont use to much metadata as this would affect/make harder in post.
Click to expand...
Metadata doesn't have anything to do with the image. It's information which is stored with the footage, including (depending on the camera you're using) timecode, frame rate, clip name, take number, pulldown flags, all kinds of things. It actually makes post easier from an organizational standpoint.
 
@Svart

We cannot really use the terms "bit depth" and "dynamic range" interchangeably, because they refer to two very different and independent properties. The two terms are not bound together.

Bit depth only applies to digital systems, and, actually, dynamic range only applies to analog systems.

Digital systems do have a signal-to-noise ratio, but digital signals are largely immune to noise until the noise reaches a certain threshold. Once the noise breeches this threshold, it mostly destroys the digital information, which is dramatically different than the analog noise level that we see in video images (and hear in audio recordings). If noise affects a camera's digital sections, something is terribly wrong.

The analog sections in a digital camera are: the sensor; the analog-to-digital converter; component/composite output stages; connections between digital and analog stages; etc. The noise level inherent in a camera's image is introduced in these analog sections (and primarily in the sensor).

Furthermore, it does not follow that an increase in bit depth yields an increase in dynamic range. The inherent dynamic range of a sensor does not change with the bit depth of the system that it is driving. If we take the output from a Genesis sensor and direct it into a 12-bit analog-to-digital converter, the resulting dynamic range will be identcal to that of the Genesis' current, 10-bit, analog-to-digital converter. The resulting signals will have the same percentage of noise on both systems. For instance, there will be "two bits of noise" with the 12-bit system and "one bit of noise" with the 10-bit system. So, we have two different bit depths, but the same dynamic range.
 
Last edited:
h-munster said:
Furthermore, it does not follow that an increase in bit depth yields an increase in dynamic range.
Click to expand...

Well, yeah... one would hope that if a manufacturer puts out a product with 10 or 12 bits of AD resolution... that it would in fact, be the case that the noise is down in the least significant bits worth of values... and not 4 or 5 bits worth of noise...

But yes, they could just do that for 'specsmanship' purposes...
 
h-munster said:
@Svart

We cannot really use the terms "bit depth" and "dynamic range" interchangeably, because they refer to two very different and independent properties. The two terms are not bound together.

Bit depth only applies to digital systems, and, actually, dynamic range only applies to analog systems.

Digital systems do have a signal-to-noise ratio, but digital signals are largely immune to noise until the noise reaches a certain threshold. Once the noise breeches this threshold, it mostly destroys the digital information, which is dramatically different than the analog noise level that we see in video images (and hear in audio recordings). If noise affects a camera's digital sections, something is terribly wrong.

The analog sections in a digital camera are: the sensor; the analog-to-digital converter; component/composite output stages; connections between digital and analog stages; etc. The noise level inherent in a camera's image is introduced in these analog sections (and primarily in the sensor).

Furthermore, it does not follow that an increase in bit depth yields an increase in dynamic range. The inherent dynamic range of a sensor does not change with the bit depth of the system that it is driving. If we take the output from a Genesis sensor and direct it into a 12-bit analog-to-digital converter, the resulting dynamic range will be identcal to that of the Genesis' current, 10-bit, analog-to-digital converter. The resulting signals will have the same percentage of noise on both systems. For instance, there will be "two bits of noise" with the 12-bit system and "one bit of noise" with the 10-bit system. So, we have two different bit depths, but the same dynamic range.
Click to expand...

As you are still misunderstanding what I said then I trust that you are unfamiliar with the term ENOB? Effective Number Of Bits is the number of bits that correspond to the dynamic range of the analog signal being fed into the analog to digital converter. Those two terms, dynamic range and bit depth are used interchangably IN PRACTICE(go back and read my initial post) since the A/D converter is the point where both signals converge. At this point, ENOB is the only thing that matters as it defines the direct relationship between the analog dynamic range and the digital bit depth.. Nobody said anything about gaining increased dynamic range from increase bit depth. I think you need to go back and re-read my post..

You are only partially correct that you throw away bits using different converters. I design systems that scale the analog signals (multiply or divide) to give me the widest dynamic range for the given bit depth. I know all about throwing away bits when using different depth converters, again, I do it everyday at my job, since I design measurement systems using digital converters to measure analog signals.

And I'm not debating what it is, I'm explaining what it is.
 
@svart

Never heard of "ENOB." Not sure how a single value (number of bits) can correspond to a ratio (dynamic range). Now, if "ENOB" is a ratio between the total number of bits and the number of bits lost to noise, then a comparison is understandable.

Nevertheless, dynamic range and ENOB cannot correspond to each other -- they represent different properties. If we change the bit depth, the ENOB ratio will change, but the dynamic range will not change in the slightest. The two ratios are not interchangeable.

Certainly, if my definition of ENOB is correct, dynamic range influences ENOB. But that relationship does not mean that two terms/properties are interchangeable.

In regards to my statement, "it does not follow that an increase in bit depth yields an increase in dynamic range," I am merely pointing out that bit depth and dynamic range are not corresponding, interchangeable properties. You seem to agree this statement.

I am happy that you have experience with electronics, but it would be erroneous for one to conclude that dynamic range and ENOB are interchangeable just because the ENOB increases as one expands the dynamic range.
 
http://en.wikipedia.org/wiki/ENOB

Those of us who design these systems don't care about total bit depth other than making sure that the converter has "enough" bit depth to measure the signal + some amount of headroom.

Only ENOB matters in real design as it defines the dynamic range of the analog signal that is being digitized. It's already assumed that the designer has taken steps to maximize the ENOB for the given signal.

Let me explain.

You can have a small amplitude signal of X dB and your 14 bit A/D bit depth allows you granularity of 6db per bit that yields around 84db of possible dynamic range. You'll lose about 2 bits in system noise due to the A/D architecture and such. So you now have around 72dB of possible measurable dynamic range (ENOB) with around 6db/bit. If X=72 then you have maximum signal to noise and maximum ENOB(but that doesn't take into account headroom that should be designed into the system).

I won't get into supersampling and such as it's not really part of this discussion but it is possible to use a higher precision A/D and then mathematically gain the signal while binning bits to "create" a higher dynamic range signal than was actually sampled. It's not the best choice for most things as it reduces precision and accuracy, but it can be done.

Also, sampling speed means more to us than bit depth does.
 
Svart said:
http://en.wikipedia.org/wiki/ENOB
Click to expand...
This Wikipedia link shows that ENOB is very different from dynamic range. According to the Wikipedia page, ENOB is actually a complex combination of three ratios, whereas dynamic range is a straightforward, single ratio.

One of the ratios involved in ENOB is the SINAD ratio, which includes harmonic distortion as a variable. If the level of harmonic distortion is high, it would throw-off the SINAD ratio, which would, in turn, throw-off the ENOD -- while the dynamic range remains the same.

It is interesting to note that, in spite of the complex ENOD formula (which includes a harmonic distortion variable), the Wikipedia page states simply that, "ENOB specifies the number of bits in the digitised signal above the noise floor." That simple definition is probably what most would assume ENOD to be, but the formula proves otherwise.

Indeed, there is controversy surrounding the definition and use of ENOD, as indicated by its Wikipedia "talk page". It seems that the ENOD formula allows for the effective number of bits to exceed the actual number of bits (ANOB). How can that be possible?

Even the ENOD Wikipedia page acknowledges this problem: "Note that the effective number of bits can be larger or smaller than the actual number of bits. When the ENOB is smaller than the ANOB, this means that some of the least significant bits of the result are inaccurate." Not sure I understand the conclusion of the second sentence.

At any rate, if accuracy is important to you and your colleagues, I suggest caution when using ENOD as it is defined in the formula that you linked.


Only ENOB matters in real design as it defines the dynamic range of the analog signal that is being digitized.
Click to expand...
I am afraid that I must continue to disagree. ENOB does not define dynamic range, not in any of the differing definitions on the ENOB Wikipedia pages nor in our discussion. I think that most would probably accept ENOB as the number of usable bits above the noise level -- a definition that is fundamentally different from that of dynamic range, which is the ratio of signal's maximum level to its noise level.

Again, with a given ANOB, dynamic range can influence the number of usable bits above the noise, but that does not mean that ENOB and dynamic range are the same signal property.


I won't get into supersampling and such as it's not really part of this discussion but it is possible to use a higher precision A/D and then mathematically gain the signal while binning bits to "create" a higher dynamic range signal than was actually sampled.
Click to expand...
One doesn't need supersampling nor binning nor even bits to increase dynamic range. Various noise reduction schemes were used in analog systems decades before digital systems.
 
Ok I give up. Your interpretation is much different than mine as you are stuck on concrete definitions and arguing semantics. Nobody said they are the same things, they are just used the same way in real world practice.

I've been telling you that nobody uses ratios in measuring dynamic range in the real world. They only measure signal strength and then noise level. You digitize it and then you have your ENOB which is equivalent to your effective dynamic range. The distortion component is generally a PROBLEM that arises from improper design and is usually countered through good design practice rather than through mathematical fretting.

Since I used to design audio gear before moving to RF/Digital design, I can tell you that most analog noise reduction schemes reduce signal accuracy to a degree matching their noise reduction. Not worth the effort really.


I'm just curious how "ENOB specifies the number of bits in the digitised signal above the noise floor." doesn't sound EXACTLY like the digital version of signal-to-noise to you?

Assuming that you setup a device so that your signal is on the verge of clipping so you have maximum SNR and your analog signal is scaled to match the highest range on your A/D, your SNR is now equal to the maximum ENOB. This is the first step in defining your signal scaling further upstream.
 
Your interpretation is much different than mine as you are stuck on concrete definitions and arguing semantics.
Click to expand...
It's not really semantics. Dynamic range and the number of bits (or ENOB) are completely different properties


Nobody said they are the same things, they are just used the same way in real world practice.
Click to expand...
If they are not the same, then the issue is not semantics, and the terms probably should not be used interchangeably.


I've been telling you that nobody uses ratios in measuring dynamic range in the real world. They only measure signal strength and then noise level.
Click to expand...
The signal strength and the noise level *is* the dynamic range ratio.


You digitize it and then you have your ENOB which is equivalent to your effective dynamic range. The distortion component is generally a PROBLEM that arises from improper design and is usually countered through good design practice rather than through mathematical fretting.
Click to expand...
Nevertheless, according to the Wikipedia page linked in this thread, harmonic distortion is a variable of ENOB, and that variable is no part of dynamic range. So, again, I must disagree -- ENOB is not equivalent to dynamic range.


Since I used to design audio gear before moving to RF/Digital design, I can tell you that most analog noise reduction schemes reduce signal accuracy to a degree matching their noise reduction. Not worth the effort really.
Click to expand...
All of the DBX and Dolby B use in the 1970s was worthless?


I'm just curious how "ENOB specifies the number of bits in the digitised signal above the noise floor." doesn't sound EXACTLY like the digital version of signal-to-noise to you?
Click to expand...
I will demonstrate, but let us not introduce signal-to-noise ratio now. It is not exactly the same thing as dynamic range, although the two terms can be used interchangeably, certainly more so than bit depth (ENOB) and dynamic range.

If we ignore the essential differences mentioned earlier in this thread, we can make a mathematical comparison between ENOB and dynamic range.

To make the mathematical comparison, let's set the following definitions:
MA = Maximum analog signal level
NA = Analog noise level
MB = Total number of bits
NB = Number of bits obscured by noise

"MB" is the "digital version" of "MA".
"NB" is the "digital version" of "NA".

If "ENOB specifies the number of bits in the digitized signal above the noise floor," then:
MB - NB = ENOB.

Dynamic range is the ratio of the maximum signal to the noise level, so (without converting to decibels) it would be expressed as:
MA / NA = Dynamic Range.

Do you see the discrepancy?

If ENOB = Dynamic Range, then it must be true that:
MB - NB = MA / NA. Obviously, the ENOB number and the dynamic range ratio will not correspond in most cases.

The only way to make ENOB the "digital version" of dynamic range is to express ENOB as a ratio:
MB / NB = ENOB.

However, we still must disregard the fact that there really cannot be a digital equivalent of dynamic range, because digital bits are fundamentally different from analog levels. Bits are relative "steps" that can be mapped and re-mapped to any point in the range of an analog or digital output.

We must also ignore other instances (given previously in this thread) in which dynamic range and bit-depth (or ENOB) do not correlate.
 
OK since my own words probably aren't good enough(I'm not much of a debater), I give you quotes from other sources:




Maxim IC "defining and testing Dynamic Parameters in High Speed ADCs, part 1:

http://www.maxim-ic.com/app-notes/index.mvp/id/728

For actual (versus ideal) ADCs, a specification often used in place of the SNR or the SINAD is ENOB, which is a global indication of ADC accuracy at a specific input frequency and sampling rate. It is calculated from the converter's digital data record as N - log2 of the ratio of measured and ideal rms error:
ENOB = N × log2 (AMEASURED_ERROR[rms] / AIDEAL_ERROR[rms]),​
where N is the number of digitized bits, AMEASURED_ERROR[rms] is the averaged noise, and AIDEAL_ERROR[rms] is the quantization noise error, expressed as q / √12 = AFS / [2N × √12]. AFS is the converter's full-scale input range as determined by the reference voltage AREF.
ENOB = log2 (AFS / AMEASURED_ERROR[rms] × √12)​
Click to expand...


Dynamic Range Questions and answers by Sage technologies:

http://www.sagetechnologies.com/library_documents/dave_qa/q&amay9.pdf


The useful dynamic range of an analyzer
digitization process is a function of
the analog signal conditioning, the
sample-hold circuit and the stability of
the time base used. All these factors are
involved when evaluating the dynamic
range of an analyzer. The effective bits of
an ADC will be less than the number of
physical bits in the ADC. This effective
number of bits defines the useable dynamic
range of the system.
Click to expand...

National Semiconductor ABCs of ADCs:

http://www.scribd.com/doc/16740475/15/ENOB-Effective-Number-Of-Bits

Effective Bits (also called Effective Number Of Bits, or ENOB) is a specification that helps to quantify
dynamic performance.
Click to expand...

Maxim IC Appnote AN641 Glossary of ADC terms:

http://pdfserv.maxim-ic.com/en/an/AN641.pdf

ENOB specifies the dynamic performance of an ADC at a specific input frequency and sampling rate.
Click to expand...
 
You must log in or register to reply here.
Back
Top