Film vs Digital - Dynamic Range

<blockquote>

<p>Please reconcile these.</p>

</blockquote>

<p>You see a contradiction there? I don't. In both I refer to the opacity range of film. Where's the contradiction?</p>

<blockquote>

<p>I was going to leave the choice of those values to you. You have already given some ADC data - why not compare the 3-bit ADC with a higher-bit ADC?. I suggest 600 mV/lux s linear to 2100 mV, 3200 mV at full well, 0.2 mV RMS noise for the CCD if you have nothing else that you would prefer to use. Zero dB for the PGA seems sensible.</p>

</blockquote>

<p>Umm, I meant that when I see 0000 (or 0001 or 0010 etc) as a code output from the ADC what should I do with it? You want me play detective, right; and I need this piece of information to continue.</p>

<p>Vijay,</p>

<p>What do you think everyone else has been talking about? Have you realised that everyone else is referring to the exposure range, not the density range? That is the common meaning of 'dynamic range' in photography.</p>

<p>What film that you use has a density range of 4.0 (or 80 db)?</p>

<p>This is rather funny, don't you think?</p>

<p>Best,<br>

Helen</p>

<p>Vijay,</p>

<p><em>"Umm, I meant that when I see 0000 (or 0001 or 0010 etc) as a code output from the ADC what should I do with it? You want me play detective, right; and I need this piece of information to continue."</em></p>

<p>You should work out the maximum amount of data that you can deduce <em>for certain</em> about the exposure that the CCD received,<em> </em> as I suggested originally. No need to go into every step of the ADC output, just the output steps at the ends of the range and those adjacent (eg 000, 001, 110, 111 for the 3-bit ADC), and why not use the ADC data you already quoted. Just my suggestion.</p>

<p>Best,<br>

Helen<em><br /> </em></p>

<blockquote>

<p>You should work out the maximum amount of data that you can deduce <em>for certain</em> about the exposure that the CCD received,<em> </em> as I suggested originally. No need to go into every step of the ADC output, just the output steps at the ends of the range and those adjacent (eg 000, 001, 110, 111 for the 3-bit ADC), and why not use the ADC data you already quoted. Just my suggestion.</p>

</blockquote>

<p>So I should do what a DAC would do, right? After all, I have to reconstruct the image based on the codes I receive, and once reconstructed, I can tell you what exposure the CCD may have received. Yes?</p>

<blockquote>

<p>What do you think everyone else has been talking about? Have you realised that everyone else is referring to the exposure range, not the density range? That is the common meaning of 'dynamic range' in photography.<br>

What film that you use has a density range of 4.0 (or 80 db)?</p>

</blockquote>

<p>I see. So you agree that Tech Pan in Dektol has zero dynamic range. And so if I <em>increase </em> contrast, I <em>decrease </em> dynamic range?</p>

<p>And how do you define the dynamic range of slide film then? Or perhaps a print?</p>

<p>Vijay,</p>

<p>No, not what a DAC would do. Given the CCD and ADC characteristics, what can you tell for certain about the exposure a photosite received that is represented by 000 in the ADC output; what can you tell for certain about the exposure a photosite that is represented by 001 in the ADC output etc... That kind of thing.</p>

<p>Best,<br>

Helen</p>

<p>Vijay,</p>

<p><em>"I see. So you agree that Tech Pan in Dektol has zero dynamic range. And so if I <em>increase </em> contrast, I <em>decrease </em> dynamic range?</em><br>

<em>And how do you define the dynamic range of slide film then? Or perhaps a print?"</em></p>

<p>You sure took me for a ride. I thought that you were being genuine for a while there.</p>

<p>This has to be one of the most surreal threads I have ever seen. So proud to have been a part of it.</p>

<p>Best,<br>

Helen</p>

<p>I don't have your sense of humor, Helen - sorry.</p>

<p>I <em>disagree </em> that <em>increasing </em> contrast <em>decreases </em> dynamic range. I <em>disagree </em> that Tech Pan in Dektol has zero dynamic range.</p>

<p>Look at the film curves Rishi posted: <a href="../photodb/user?user_id=2381463">Rishi Sanyal</a> <a href="../member-status-icons"><img title="Frequent poster" src="http://static.photo.net/v3graphics/member-status-icons/3rolls.gif" alt="" title="Frequent poster" /> </a> , Dec 09, 2008; 11:53 p.m. and please show me how increasing contrast decreases dynamic range.</p>

<p>Please be clear about what you are saying.</p>

<blockquote>

<p>No, not what a DAC would do. Given the CCD and ADC characteristics, what can you tell for certain about the exposure a photosite received that is represented by 000 in the ADC output; what can you tell for certain about the exposure a photosite that is represented by 001 in the ADC output etc... That kind of thing.</p>

</blockquote>

<p>Well then, you can only derive whatever you said. That the photodiodes received exposure the equivalent of at least 127mV to 896 mV worth of output.</p>

<p>Now lets do the same example for a single bit ADC. The threshold is at 512 mV, so a little under this and I get a logic 0 output from the ADC and a little over this and I get 1. What can I deduce from this about the exposure range that the CCD got?</p>

<p>That the photodiodes received exposure the equivalent of at least 512 mV to 512 mV of output, and assuming 0 noise for the sake of theory, a 0 mV input caused a change of output.</p>

<p>Which is fine; but what does this have to do with reality? If a DAC assumed this, then you'd never be able to hear any music on CDs at all.</p>

<p>Hey guys,</p>

<p>OK I have a <strong>really</strong> naive question: A CMOS sensor, or CCD, or what have you, has an inherent 'sensitivity', correct? As in: X number of electrons (or charge buildup) for Y number of photons. ISO speeds just adjust the sensitivity of the amplifiers, correct?</p>

<p>Now, 'blown-out highlights' result when pixels reach their saturation in charge, correct? But when you hike up the ISO, you lower your exposure. So, then, for the same image taken at the same EV but at ISO 100 vs. ISO 400 -- shouldn't the ISO 400 image, by definition, have less blown-out highlights than the ISO 100 image?</p>

<p>-Rishi</p>

<p>Rishi, (I misread the author's name first time)<br>

<em>"So, then, for the same image taken at the same EV but at ISO 100 vs. ISO 400 -- shouldn't the ISO 400 image, by definition, have less blown-out highlights than the ISO 100 image?"</em><br>

<em></em> <br /> First, if the EV is not changed between two exposures (and the lighting is not changed) the CCD gets the same exposure, so turning up the ISO (gain) from 100 to 400 will increase the likelihood of blown highlights, not decrease them. But I think you made a typo?<br>

<br /> If, however, the exposure is reduced and the gain is increased you can work out what happens from the CCD information I gave earlier.<br>

<br /> Here is the information again. It is genuine stuff.<br>

<br /> 600 mV/lux s linear to 2100 mV, 3200 mV at full well, 0.2 mV RMS noise.<br>

<br /> The analogue out from the CCD goes to the analogue in of the PGA (programmable gain amplifier), which is typically packaged with the ADC. The PGA could be 0-36 dB, with the gain set by a 10-bit controller. The analogue out from the PGA goes to the ADC as you might imagine.<br>

<br /> That should be enough information for you to figure out that as you turn up the gain, the dynamic range (the exposure range over which detail is recorded, ie the meaning used in this thread) falls. You can also work out what the dynamic range is likely to be...</p>

<p>The noise floor of the CCD is amplified, but the upper limit of the ADC input stays at 2000 mV or just under. So the ADC ouput maxes out from the amplified signal even though the sensor may not be saturated.</p>

<p>Sorry, I'm rushing off for the weekend - hope this isn't too brief, no doubt someone else will help!</p>

<p>Best,<br /> Helen</p>

<p>Um, Helen, you realize that Vijay didn't pose that question. I (Rishi) did. I'm a molecular biologist who likes to dabble in engineering, computers, photography, whatnot.</p>

<p>My question didn't pertain to anything anyone had written above. It was a <em>de novo</em> question in regards to this thread.</p>

<p>Also, by keeping the EV the same, I meant keeping the exposure taking into account the ISO the same. That is if you double the sensitivity, halve the shutter speed. So the sensor would get less exposure.</p>

<p>I'm just generally asking: as long as the amps don't blow out any signals (which I would think they'd be designed to never do, but I may be entirely wrong), shouldn't the same shot at ISO 400 vs ISO 100 show less blown-out highlights?</p>

<p>I'm sure someone gets my question.</p>

<p>Yep, reply edited. In a big rush.</p>

<p>Re EV. The classic use of Ev or EV is that EV represents combinations of shutter speed and aperture that gives the same exposure. Logarithmically</p>

<p>Ev = Tv + Av = Sv + Bv</p>

<p>Oh, wow, but then that means that if you could actually capture the RAW data before being amplified, there could be more dynamic range, because at higher ISOs you'd be less likely to blow out the exposure?</p>

<p>Oh wait, isn't this how highlight tone priority works.</p>

<p>So that means that sometimes it's actually the AMP that blows out the signal? Thereby limiting the dynamic range? Isn't there some way around this? Like at higher ISOs, apply a non-linear gain? A VERY non-linear gain (in my non-technical terms)?</p>

<p>Rishi</p>

<p>The maximum dynamic range is set by the noise floor and saturation capacity of the CCD. The ADC should then be designed to handle that. Raw will make the most of it, at the exposure that represents the 'native' ISO. At all other ISOs (or EIs), higher or lower, the dynamic range is likely to be less. In simple, rushed terms.</p>

<p>I didn't think anybody else would be reading this thread all the way down here. I'm amazed.</p>

<p>(I really have to leave for the weekend, much as I would love to continue this pleasant discussion)<br>

Best,<br /> Helen</p>

<blockquote>

<p>So the ADC ouput maxes out from the amplified signal even though the sensor may not be saturated.</p>

</blockquote>

<p>But isn't that poor amplifier/algorithmic design?</p>

<p>Haha. Like an old audio optocompressor on the output. You could get that real 60's sound in the highlights.</p>

<p>must go.</p>

<blockquote>

<p>The maximum dynamic range is set by the noise floor and saturation capacity of the CCD.</p>

</blockquote>

<p>Exactly right. Today, the primary limiting component for DR is the photodetector itself. You can make amplifiers handle much wider dynamic ranges, and while Helen and I seem to be disagreeing on some aspects of the ADC part, that isn't the limiting component either.</p>

<p>Hey Rishi.... finally something I can talk about instead of all this engineering stuff. By the way, I am typing this reply about 2 hours after loading this page up, so by now you probably have all sorts of answers.</p>

<blockquote>

<p>But when you hike up the ISO, you lower your exposure.</p>

</blockquote>

<p>That doesn't sound right to me. Why do you say this? Your exposure stays the same, but the iso effectively causes you to overexpose, coz 400 is more 'sensitive' than 100. So if you are clipping at iso 100, then you will clip more at iso 400. </p>

<p>Helen, you're probably gone for the weekend, and I have plans too so maybe we should turn to the primary question (we can get back to our other interesting topics later).</p>

<p>You've shown that you understand engineering rather well, so we can make this short:</p>

<p>Say I take the raw data from a 14-bit camera and lop off the lower six bits - just make them zero, say. Are you saying that this would change the input exposure range of the camera? Surely you agree that this case is indistinguishable from the case where I have an 8 bit ADC, take 8 bits from it and concatenate these with six ground tied wires to make up 14 bits.</p>

<p>And surely you agree that the information content of 14 bits where 6 bits are tied to zero is exactly that of 8 bits? (There are only 256 distinct levels).</p>

<p>Doesn't this establish the independence of input exposure range from bit-depth of the ADC?</p>

<p>And, you still haven't provided a formula for deriving dynamic range, (or input range) based on the number of ADC bits. Could you please do so?</p>

<p>(I agree, and have always agreed that noise sets an upper limit on the information content of the ADC input signal; so adding bits doesn't do much after this limit is reached, except sampling noise with higher precision. However I maintain that nothing forbids one to lose information about intermediate levels while keeping the endpoints the same or to sample noise with higher precision while keeping the endpoints the same.)</p>

<p>No, Bernie, when you hike up the ISO, you have to decrease the length of time the shutter is open. That's the whole point of using higher ISOs. So you can use a shorter shutter speed and hold that camera still.</p>

<p>Vijay, I understand that, but it doesn't explain why higher ISOs shouldn't show less highlight blow-out.</p>

<p>Ok, i'm going to go through some stuff in order from when i was last here. By the way, Rishi, I see you cleared up the confusion and the question was answered, so you can ignore my last post.</p>

<blockquote>

<p>What makes you so certain that 4 stops of data are identifiable in the data? An assumption of course that the data is "proportional" to the input. If you read my post wherein I describe the actual mathematics, you see an array of some codes. You should treat these codes for what they are: data that only has meaning to a DAC, not to you.<br>

Now even if you want to analyze the data: you'll see a range from 0 to something. Do you want to imply that something/0 = infinity, so the DR of the data is infinite?<br>

Ultimately, until the 0 and max values map to a reflectance or opacity or luminance, there is no way for you to say what the "dynamic range" of the data is.</p>

</blockquote>

<p>I see what you are saying Vijay. The data itself is basically meaningless to the photographer until it is translated back to print or monitor. Couple of points though: DR won't be infinity. No pixel will register 0 signal. Noise will ensure it has some value. In my 5D the darkframe noise is around level 128 out of 4095. The other point is that your point still doesn't change the scenario that Helen and I are on about. You can't make an 8 bit ADC display more than 8 stops of scene DR. A 10 stop printer could make a print with a physical DR of 10 stops, but it would still only translate 8 stops of scene DR.</p>

<blockquote>

<p>This means that the first discernible tone that is not pure black is 6 stops lighter than black. Obviously there is something wrong here, right? Because in the original scene, the tone that was just slightly lighter than black wasn't separated by 6 stops from black.</p>

</blockquote>

<p>It was seperated by 6 stops. Remember, in the real scene 'black' isn't really 'black'. 'Black' is the noise floor. If the scene is 10 stops from brightest thing to the brightest thing that you can measure (ie. the noise floor of the photodiode), then a 16 bit adc exposed to the right will capture the 10 stops and then hit the noise floor. But that is a simplified version of what would happen. What would really happen I suspect is this: If the photodiodes clip at 10 stops, the ADC could never be filled via expose to the right. It would more be a case of expose to the left as viewed from the ADC (same thing that happens in the 40D et al). 10 stops from the photodiode would fill up the lowest 10 bits of the ADC and then 6 bits would be 'empty'. Your final image, in the absence of a saturation clipping adjustment, would be grey (ie. not clipped white) when white balance and a TRC (tone response curve) was applied.</p>

<blockquote>

<p>Vijay, I understand that, but it doesn't explain why higher ISOs shouldn't show less highlight blow-out.</p>

</blockquote>

<p>Helen's explanation was spot on - I was just augmenting one point of hers. She's right: if you increase gain, you increase noise and you hit the ADC's max input value (usually 2 V) - and you clip. This blows out highlights and compresses the input exposure range.</p>