ISO 50

[Meh]

Factual consistency does not seem to be one of Mikael's strong suits... But in other ways, he's quite consistent - his use of bold text, his derogatory questioning of others' understanding coupled with urging others to 'read' and 'try to understand', his repetition of the same statements in post after post, in these and similar areas, he seems consistent to the point of boredom.

Michael IS a very rude boy indeed.

 

[TrumpetPower!]

The so-called analog gain is applied prior to readout[. . . .]

Because the readout is the point where the signal logically leaves the sensor, it should be self-apparent that analog gain must be applied by the sensor itself to the signal before (or as) it is read out.

You're of course correct that expanded high ISOs are again digital simulations and not analog processes, but I suspect your cutoff of ISO 1600 only applies to very old cameras. I may be mistraken, but I'm pretty sure that digital ISO boost only happens with expanded ISO settings. With the 5DIII, native (analog) ISO should extend all the way to ISO 25,600...but I wouldn't bet more than a cup of coffee on that exact number.

Cheers,

b&

 

[rpt]

"Halves the amount of light"? Please correct me if I am wrong..... But there is nothing you can do with sensor settings or camera modes that will change the amount of light.... You can play with gain and linearity and mapping color depths, but the amount of light remains unchanged. You can change the amount of light with shutter speed, aperture, or slapping on a neutral density filter, but not with the sensor.

Obviously a number of us are clueless. HTP stands for Half The Photons! Now how that happens in the sensor needs to be discovered...

And oh, just in case somebody takes my post at face value, here is the tag:
<SARCASM/>

 

[Meh]

Because the readout is the point where the signal logically leaves the sensor, it should be self-apparent that analog gain must be applied by the sensor itself to the signal before (or as) it is read out.

Depends when, where, and how that "analog gain" is being applied and on what we think is the beginning of the read out process. My suggestion is that we think of the "sensor" as the photo-sensitive material and the electron wells. Everything after that is electronics and should be considered as part of the read-out circuitry including the analog gain. Applying the analog gain is not part of the exposure per se and, it adds noise which by the usual definition is part of the read-out noise (if we agree on two categories of noise those being photon shot-noise and read-out noise).

 

[Meh]

"Halves the amount of light"? Please correct me if I am wrong..... But there is nothing you can do with sensor settings or camera modes that will change the amount of light.... You can play with gain and linearity and mapping color depths, but the amount of light remains unchanged. You can change the amount of light with shutter speed, aperture, or slapping on a neutral density filter, but not with the sensor.

Obviously a number of us are clueless. HTP stands for Half The Photons! Now how that happens in the sensor needs to be discovered...

And oh, just in case somebody takes my post at face value, here is the tag:
<SARCASM/>

Absolute. Best. Response.

 

[Meh]

I suspect your cutoff of ISO 1600 only applies to very old cameras. I may be mistraken, but I'm pretty sure that digital ISO boost only happens with expanded ISO settings.

The highest ISO setting that is based on analog gain is certainly going to be different among bodies and may have increased with the latest tech but it's not my understanding that digital gain is only with expanded ISO settings. However, I'm no expert... Neuro?

 

[friedmud]

ISO 50 is a complicated subject... and really depends on the nature of the sensor in your particular camera.

The following is something I posted over at Nikon Rumors in response to seeing someone say that ISO 50 was underexposing and then correcting:

(EDIT: I just want to be clear that his was in a discussion about landscape photography, where an aperture has been chosen for DoF and sharpness and the shutter speed is ranging to get the exposure right)

"Firstly, it's not "underexposing then correcting in software"... that's more of what happens with things like "Active-DLighting" and "Highlight Tone Priority" (on the Canon side) or using very high ISO. If anything, ISO less than 100 _overexposes_ and then pulls back in software. If you fix aperture and you expose at ISO 100 and get 1 second exposure... Then you drop ISO to 50 and take the same shot the shutter time will be 2 seconds. If the sensor can't be "less sensitive" than ISO 100 then you are overcooking the sensor (ie, overexposing). The camera can then correct in software to give you the correct looking exposure (possibly at the expense of highlights that might have been blown out and can't be recovered). Because of this there can be a small drop in dynamic range (you might have lost some highlights).

A few thing about this:

1. Physics means that an ISO 50 shot will definitely include the effects of more photons striking the sensor. That will reduce "random shot noise" which can give you a visible reduction in noise. I have done tests with my D600 and there is less noise at ISO 50,

2. At the worst case of a sensor that really can't go below ISO 100 at all, this process is equivalent to overexposing by one stop and pulling back your RAW file in post. Essentially it is equivalent to Exposing To The Right (ETTR http://en.wikipedia.org/wiki/Exposing_to_the_right)... Which, in scenes of lower dynamic range can give you a tangible benefit of having less noise in the shadow areas. So once again we might be shaving off noise.

3. How much highlight you lose is very sensor dependent. The D600 has a lot of headroom in this area and if you look at DXOMark ( http://www.dxomark.com/index.php/Cameras/Camera-Sensor-Database/Nikon/D600 ) it doesn't show a loss in Dynamic Range when going to ISO 50.... just that it doesn't gain much either.


In summary: be careful about spreading dogma like this. Every sensor / camera is different. In particular, I have done my own testing on the D600 and have found the tradeoff for lower noise to dynamic range to be more than acceptable. In extreme dynamic range scenes I may choose 100 to just be safe... but in all of my shooting with the D600 so far ISO 50 has produced better (less noise with the same / similar dynamic range) photos.

Everyone: don't just take my word for it (or anyone else on the internet) there is no reason not to do these types of tests with your own gear. One of the first things I do when I receive any new piece of gear is test it's limits. Not to be dissappointed but to know where they are so I can make correct decisions about tradeoffs in the field..."

 

[Meh]

@friedmud... what you are describing is the effect of increasing the length of the exposure (from 1 second to 2 seconds in your example) and it's that real increase in the number of collected photos that improves the SNR not any magical effect of using ISO50.

But you're statement that "ISO50 is not underexposing then correcting" is absolutely correct because ISO50 is overexposing (relative to ISO100) then correcting.

 

[friedmud]

Once more unto the breach....

The following "Sunny f/16" exposures all result in the exact same amount of analog gain applied to the readout of the sensor, and therefore the exact same raw file:

1/400s @ f/8 @ ISO 50
1/400s @ f/8 @ ISO 100
1/400s @ f/8 @ ISO 200 + HTP

Hey Trumpet, I don't want to get pedantic (because you are mostly right in what you are saying, and you're trying to provide clarity to someone who is definitely wrong) BUT... I feel like since we've had a good technical discussion here I should point out that you have to be careful about that "exact" word you used there.

Your statement is only "exactly" right if a sensor's "base ISO" is exactly 100. For many modern sensors it is close, but even a 5DMk3 is closer to 80 and it appears as though a Nikon D800 is closer to 75. What this means is that there is some analog gain applied, even at ISO 100... and there will be different analog gain applied at 50.

What this means is that shooting at 50 might not lose you a full stop of highlights. Indeed, shooting at ISO 75 on a D800 won't lose you any highlights at all (compared to ISO 100) and you get the added benefits of more photons (ie, less noise).

Sorry to be pedantic... I don't want to derail what you're saying.... I just want people to realize that not everything about a camera sensor is set in stone by a god somewhere... and each and every sensor type has it's own unique aspects...

 

[friedmud]

@friedmud... what you are describing is the effect of increasing the length of the exposure (from 1 second to 2 seconds in your example) and it's that real increase in the number of collected photos that improves the SNR not any magical effect of using ISO50.

Oh, I agree (with my quibble above about understanding the effect of "base ISO"). I apologize for reposting that here out of context... it was simply more convenient than retyping it

The context was landscape photography... and I was trying to explain why I shoot at ISO 50 sometimes (because of less noise from the longer exposure... well, along with any other reasons why a longer exposure might be wanted, like blurring a waterfall).

 

[Aglet]

clip..
1... I have done tests with my D600 and there is less noise at ISO 50,..

Whoa, stop the presses.
your D600 has noise?!?...
That's a distinctly Canon issue.

BTW, sounds like most of the ISO-arguing around 50-100-HTP and D-L know enough about how to USE it.

@neuro, @trumpet, @ etc...
don't forget @michael seems to speak english as a 2nd language so I often find his sentence syntax confusing but after wasting time reading nearly 3 pages of arguments from you all it seems to me you're perfectly aware how this technology works but can't seem to see that for the scrap that's broken out over semantics and typos.

I initially thot I'd make more use of iso 50 on my 5D2 at some point for making cleaner low key shots. Turns out I rarely shoot anything like that to advantage.

I DID get out for an hour of photo-therapy around sunset on Saturday, a nice way to deal with too much stress from work and family.
Shot with 5d2, d800, d5100s and a Q. The first 2 were equipped with 70-200mm zooms and I was shooting the same high key scene in dim overcast skies and snow everywhere, including inside my car cuz it was too cold to step outside of it.
I liked shooting the 5d2 w 70-200/2.8L is2
I loved shooting the d800w 70-200/4 vr3, crisp handling, spot-on WB, better AF in blowing snow; a condition that often had my Canon cameras failing to AF properly all too often.
D5100s AF'd well but gave me some odd WB, often a bit green but easy to fix in post.

Everybody, get out and shoot more!

 

[sanj]

Nice thread. Learning a lot.

I think this evening I will take two shots with ISO 50 and 100 and post them here for us to see and learn.

 

[Loswr]

That's exactly what you said... "HTP. it is a halving of infaling light" are your exact words. Even this last statement from you is not even internally consistent.

no I said by using HTP it is like underexpose 1 stop, go away from 100iso by an under exposure which well be equal to a 200iso exposure and thereby make a head room= which later can compensates with gain and curves

Really? You didn't say, "HTP. it is a halving of infaling light," and you're sure about that??

Let's check. You said:

HTP, here you have exposed the motive 1 stop shorter, halving the number of photons and you get a High light head room and then another curve is applied with a lift in lower areas/levels and a softer curve at the top / high lights
Here 100iso are exposed as i where 200iso which means 1 stop shorter exposure , the sensor collect less photons who are converted in to a charge/signal.

...and you then said:

HTP. it is a halving of infaling light

Sure, you also talked about HTP underexposing by a stop. But for you to deny claiming the light was halved is stubborn intransigence, and a complete lie...a particularly foolish one, too, given that your previous statement that you deny making is there for all to read, and easily catch you in your transparent lie of denial. The second one, "HTP. it is a halving of infaling light," I could accept as linguistic confusion...but it followed the earlier statement where you talk about halving the number of photons, the sensor collecting less photons, and a 1-stop shorter exposure being used, all as a description about how HTP achieves the highlight recovery. That's not linguistic semantics and typos as Aglet suggests, that a very concise, understandable description where it's quite clear what you mean...it's just WRONG.

But...not only can you not admit being wrong, you deny you even made the statements quoted above.

Moreover, you also stated:

You Neuro and others seems to have a very hard time to admit that you are wrong .

It is manifestly clear that YOU are the one unable to admit when he is wrong.

At least you were right about one thing:

it is hard to discuss HTP etc if you do not know what it means

You are obviously having a hard time discussing HTP, because while you appear to understand the consequences in terms of highlight recovery and increased shadow noise (or at least, you can parrot the statements of others who understand that), you haven't got a clue about the image data manipulations that underlie HTP, as clearly shown by your flat out wrong description of a shorter exposure leading to halving the incident light for the exposure.

 

[Loswr]

Your statement is only "exactly" right if a sensor's "base ISO" is exactly 100. For many modern sensors it is close, but even a 5DMk3 is closer to 80 and it appears as though a Nikon D800 is closer to 75. What this means is that there is some analog gain applied, even at ISO 100... and there will be different analog gain applied at 50.

Sorry to be pedantic... I don't want to derail what you're saying.... I just want people to realize that not everything about a camera sensor is set in stone by a god somewhere... and each and every sensor type has it's own unique aspects...

Sorry, friedmud...you're late to the proverbial pedantic party...my own pedantry preceded yours by about 37 posts...

TrumpetPower is absolutely correct in his statement that, "ISO 50 and highlight tone priority are two special cases. The exact same analog signal amplification (i.e., none) is applied with all three exposures: ISO 50, ISO 100, and ISO 200 w/ HTP. If you use the same shutter speed and aperture, you'll get the exact same raw file in all three cases." (well, almost correct - the base ISO isn't exactly 100 for all sensors, so even at ISO 100, some gain may be applied - but it is the same gain in all three cases).

 

[sanj]

My test.
1. It is unscientific, do not pounce on me. Am learning!
2. Stats: 1dx. 24-70 II. Manual focus, exposure, WB (tungston). f5.6 1/4 and 1/8.

Its all for us to see, but to my eyes:
1. The tint at ISO 50 is truer. (This is next to my desk and I can make a visual comparison)
2. The blown whites between the statue and lamp is blown more on 50 ISO.
3. 50 ISO shows marginal noise and sharpness improvement. (This could of course be just my eyes playing games.) Or is there a shake? I did use a sturdy tripod.

Do you think I should make another test? If yes, how should I improve?

Thank you.

Sanjay

 

[Loswr]

In HTP the sensor has now been hit by less light/photons

Like I said, clueless.

 

[sanj]

It should be more blown out, 50 iso is 100 iso over exposed 1 stop= 1 stop earlier clipping in highlights

Sir do not understand. Am not disputing, just unclear as to what you saying. Could you please expand?

 

[TexasBadger]

If you stare at the sun (not recommended) with and without sunglasses it will not seem as bright with the sunglasses. However, the amount of light reaching you has not changed. The effect the sunglasses have on the light reaching your retina is similar to using a neutral density filter. The intensity of the light (measured in photons) has not changed anywhere except for the amount of light hitting your retina (sensor). Each pixel on the sensor collects fotons like a bucket. When the bucket is full it cannot hold any more photons. This is why we get clipping. Please keep in mind that all of this is not linear but logarithmic. In film we called it reciprocity failure. The loss in efficiency was caused by the physical properties of the emulsion. In digital, the sensor creates heat which has a negative effect in the efficiency. Any time you increase the ISO, you amplify the entire signal which causes the noise as well as the light values to be increased. Since a sensor will always have a noise component, the digital processing of the analog scene can be addresses by the digital processing of the information comming from the sensor. That is why apparent noise can be reduced by software/firmware. The 5D3 has much better noise reduction than the 5D2, even though the sensor is essentially the same. This was achieved by better firmware processing the information coming from the sensor. You can also reduce the apparent noise by improving the sensor by reducing the signal noise going to the digital processor. Another caveat is that the smaller the pixel is, the more noise with all other factors remaining equal.

The bottom line is that the ISO is changed for one of two reasons. The first is to change the aperature/shutter speed combination for the image capture. The second is for desired effect that noise will have on the resulting captured image.

If you have ever pulled film you overexposed the image and underdeveloped the film. However, if you pulled Tri-X, you might expose the image for ISO 125 ~2 stop pull and then decreased the development time by 20%. Remember this is logarithmic not linear. The purpose of pulling film is to increase the grey scale in the image for more depth and richness in the final print.

ISO 50 is similar to pull processing. You are overexposing and digitally underdeveloping. Pulled film will have less dynamic range and or contrast than film not pulled. You only pull film because you are wanting to increase the grey scale in the resulting image. You shoot at ISO 50 for the same reason. Shooting to the right is another way to manipulate the resulting image.

The reality is that you use different ISO values based on the desired resulting image capture. This is similar to using different types of film in the analog world. Increasing the ISO changes the amplification of the signal from the sensor. It also increases the contrast and the noise. Increasing the speed of your film increases contrast and grain. Grain is the analog equivalent of noise in digital. The reason that faster film was more grainy was that the silver crystals in the emulsion were larger which allowed for shorter exposures with increased grain being the trade-off.

Remember that there are still only three components to a proper exposure. Shutter speed, aperature and ISO. The art is in the composition and exposure decisions we make.

Now the equivalent of pushing film in the digital arena ...

 

[sanj]

50iso is no 50 iso, it is equivalent to 100 iso over exposed 1 stop, what you gain is for example slower exposure time which can be handy , running water etc but at the same time that the sensor is richer exposed you lose 1 stop of DR and that is in the highest levels= earlier clipping

Thank you for the explanation. To me, when I re examine the photos based on what you saying, it seems like ONLY the whites are getting overexposed, NOT the entire frame. Could I conclude that not the entire frame but only the whites get over exposed?

 

[Loswr]

My 5dmk2 starts at 200 iso if I enable HTP= 1 stop under 100iso= let fewer photons hit the sensor

Why are you perseverating on this? Please explain: when the HTP is enabled and ISO 200 is set, the camera is actually exposing at ISO 100...how does that in any way alter the number of photons hitting the sensor? You seem completely unable to comprehend the simple facts.

Try to read and study carefully all those other pages you linked earlier. Let's take it by steps:

• Enabling HTP does not change the selected shutter speed.
• Enabling HTP does not change the selected aperture.
• Therefore, enabling HTP does not change the number of photons collected.
• (Pause for a moment, and study the above three points carefully.)
• Enabling HTP results in the camera using a 1-stop lower ISO than selected (although the selected ISO is 'incorrectly' recorded in the metadata).
• The 1-stop lower ISO means less analog gain applied to the readout from the photosites.
• The decreased analog gain ('underexposure') reduces blown highlights in the RAW file
• The RAW converter (in-camera or computer software that recognizes the HTP flag in the metadata) applies a selective tone curve to boost exposure in the low- and mid-tones by a stop, sparing the highlights

What part of the above do you not understand? In particular, you seem unable to grasp the 3rd bullet point.

It's not complex, yet you fail to understand or acknowledge your errors in comprehension.