[GUB]

Just discovered Iso invariance - typical slow learner - it was flavour of the day a couple of years ago. Anyway is anyone utilizing it in their workflow? - it certainly appears to have its uses, especially highlight control and a fluid work flow in a constantly changing low light environment like perhaps a rock concert with its changing lights.
Anyway I was wondering how Iso invariant my K01 was so I have been experimenting. Here is a test I did last night with my K01.
The squares are a 100x100 pixel crop of an image choosing a range of tones.
The idea is to see if I can match higher Iso by just pushing the exposure in the Pentax utility.
Silkypix only has +3EV exposure adjustment so in the more pushed ones I had to top up with the curves tool so results are approximate.
As you can see the correctly exposed Iso 12800 one (bottom right) is inferior to the Iso 200 one with the same shutter speed(top right) and pushed +6EV !!
So it is fair to say the K01 is Iso invariant.
Interesting an earlier attempt to push the exposure using Darktable was only effective up to about 3 stops but it may have just needed better management.
So is anyone using it?
And for those who have no idea what I am talking about here is a link: ISO Invariance: What it is, and which cameras are ISO-less
I think I might find it useful for night sky photography to get better control of the exposure of the stars.
Hopefully you can understand this chart - it was the best way I could think of to lay the samples out.
The longest diagonal is the correct incamera exposure for that setting. The degree of exp pushing increases on the diagonals to the right.

Attached Images

 

[starbase218]

I don't know if it's the same thing, but in shooting high-contrast scenes (landscapes come to mind) I often choose to preserve highlights, and one way of doing this is dropping ISO. But as for my workflow, I have found that reviewing a significantly underexposed image on either the back screen of my camera or on a computer is not ideal for deciding whether to discard or keep the image. If camera manufacturers or application companies were to automatically balance the exposure for playback it would make things a lot easier.

I'm curious to how other people are doing this.

[Na Horuk]

So it is fair to say the K01 is Iso invariant.

Its a really complicated issue. For example, some manufacturers have some NR and sharpening built into the photo, even if you shoot raw. So if you shoot at a lower ISO and add exposure in post, that part will look different. Sensors are different, as well. And even the different Raw software might use different algorithms, different tricks under the hood, when "adding" brightness. So I don't think we can take really clinical tests with all variables accounted for, though you certainly did a good job in your test. The little variables I mentioned might be really small, but if there are enough of them and the image is pushed to extremes, they can make a difference.

Another important issue is clipping. If you have a dark photo and add exposure in post, it will not look the same as a properly exposed photo, because the dark photo simply does not contain the same amount of recorded information. Now sure, for most uses adding a little brightness in post will be fine, because our exposures are good enough to begin with. Camera metering is pretty good and captures a wide dynamic range, so if the light is halfway decent we get useable photos, or at least salvageable photos.
Anyway, I don't really put much faith in ISO (in)variance. I prefer to expose the photo correctly, or in some situations ETTR (as bright as possible, but avoid highlight clipping). Then you can play with the brightness/exposure a little with minimal IQ deterioration.

[JensE]

So is anyone using it?

Yes, routinely/extensively - ever since I got the K-5. Many of my landscapes, available light macros, indoor sports shots and - yes - night sky pictures are nominally 'underexposed' and lifted up by typically 1EV to 3EV in post-processing and/or tone-mapped. See e.g. this thread for macro examples with up to +3.5EV processed in darktable - you have to enter that numerically as it's out of the exposure tool slider range

I have found that reviewing a significantly underexposed image on either the back screen of my camera or on a computer is not ideal for deciding whether to discard or keep the image.

In-camera, it is what it is and I only rarely review pictures on-screen to determine what to delete. Before importing files from the card, under Windows, I like to use Fast Raw Viewer for culling. It is specifically designed for that purpose and allows to adjust exposure and/or selectively check shadow and highlight detail using keyboard shortcuts very quickly. It also allows to define custom thresholds to highlight raw underexposed areas, very useful.

[GUB]

I have found that reviewing a significantly underexposed image on either the back screen of my camera or on a computer is not ideal for deciding whether to discard or keep the image

Yes it is hard to head off home with a bunch of black images on your camera!!

If you have a dark photo and add exposure in post, it will not look the same as a properly exposed photo, because the dark photo simply does not contain the same amount of recorded information

Remember that the only difference is the Iso setting - the sensor is getting the same amount of light as it would with high Iso. So I would expect the same amount of information to be there.

[GUB]

up to +3.5EV processed in darktable - you have to enter that numerically as it's out of the exposure tool slider range

How do you do that in Darktable? The only way I could see was utilising the tone curve. Which of course is where you should be doing it anyway if it wasn't just a test shot because this is the moment you can protect the highlights.

[JensE]

Anyway, I don't really put much faith in ISO (in)variance. I prefer to expose the photo correctly, or in some situations ETTR (as bright as possible, but avoid highlight clipping).

One could argue that ETTR is the 'correct' exposure to capture a maximum of information. (side-note: It of course depends on your scene: If I expose a night landscape to capture detail in 'the man in the moon', inevitably the landscape disappears into noise, irrespective of ISO.)

For the use cases which I mentioned, I usually have enough detail captured in the dark areas to lift them up with good detail and little noise (K-5, K-1) even if slightly underexposed. Why would I do this instead of ETTR by incresing the ISO setting? I simply don't have enough time to take several pictures for an optimal ETTR e.g. if bees fly from flower to flower. Turning down the ISO, and thus allowing more of the dynamic range captured by the sensor to make it through the A/D conversion without clipping into RAW data, ensures that I don't have blown-out highlights. It comes at the mere expense of a simple post-processing adjustment, and due to the excellent design of the Pentax processing, as demonstrated above, little to no sacrifice in shadow detail.

[starbase218]

Yes, routinely/extensively - ever since I got the K-5. Many of my landscapes, available light macros, indoor sports shots and - yes - night sky pictures are nominally 'underexposed' and lifted up by typically 1EV to 3EV in post-processing and/or tone-mapped. See e.g. this thread for macro examples with up to +3.5EV processed in darktable - you have to enter that numerically as it's out of the exposure tool slider range

I hadn't considered using it for sports or fast action. Do you do this because you are using one of the auto settings, which may yield blown out highlights?

In-camera, it is what it is and I only rarely review pictures on-screen to determine what to delete. Before importing files from the card, under Windows, I like to use Fast Raw Viewer for culling. It is specifically designed for that purpose and allows to adjust exposure and/or selectively check shadow and highlight detail using keyboard shortcuts very quickly. It also allows to define custom thresholds to highlight raw underexposed areas, very useful.

Actually I don't have a standard workflow yet. But I don't import every photo into Lightroom anymore. I used to run Photosmith on an iPad mini for reviewing photos, but that app has been discontinued. At the same time you can lock photos on your camera, so I might want to use that somehow. And for that, some kind of auto-HDR mode for playback (which can be turned of) would be welcome I think.

Yes it is hard to head off home with a bunch of black images on your camera!!

Yes, especially if you're a perfectionist like me. The thing is, technology can solve it for me. But although I have a wonder of technology with me in my DSLR, it can't do it.

Remember that the only difference is the Iso setting - the sensor is getting the same amount of light as it would with high Iso. So I would expect the same amount of information to be there.

It is getting the same amount of light, but the ISO setting increases gain before A/D conversion. So the image might be cleaner if you increase ISO than if you push your exposure afterwards (with the digitized signal). Which is why I try to only use ISO invariance as far as I need it to preserve highlights. Any further and I'm throwing away information without a good reason.

Sometimes, you can even tell there is a loss of information by looking at the histogram after lifting up the exposure, if it starts looking like a comb.

[JensE]

I hadn't considered using it for sports or fast action. Do you do this because you are using one of the auto settings, which may yield blown out highlights?

Quick answer: Yes.

Fast indoor sports always means you have less than desirable light. On the other hand, in the usual mixed gym lightning here, there are always a few sunny spots with way more light than otherwise available.

So you choose your aperture according to what your lens has to offer or the widest for acceptable depth of field and your shutter speed for the desired level of motion blur, the free decision left being the ISO. Your mid-tone signal will be maybe 3EV above the sensor noise floor and even at intermediate ISO values (say ISO800), that sensor noise floor is amplified enough to be well within the 14bit signal resolution capture ability for each component in the RAW file.

So I do one of the following:

1. If I want to use the out of camera JPEGs, e.g. to quickly share on-site, I use TAv-Mode to let the camera do its best to jusdge how much (analog) amplification to apply by bumping up the ISO. This may blow out highlights at those bright spots - as you suspected. Bright white jerseys/shirts are notorious for that, including parts of the faces lit by reflections.
2. If I plan to do manual post-processing, I either set the ISO to a constant lower value reflecting ETTR in the brightest spots (M mode), or I simply use -2.7EV exposure compensation in TAv-Mode which normally gives enough additional headroom for the hot spots. I'm nowhere near base ISO, so none of this discards any potential shadow detail, the noise floor always determines how much there is in the shadows.

The -2.7EV still allow some level of review on the camera display and have over the years proven to be sufficient on both K-5 and K-1 even in difficult gyms. On overcast days or in good venues, I use less.

[GUB]

but the ISO setting increases gain before A/D conversion. So the image might be cleaner if you increase ISO than if you push your exposure afterwards (with the digitized signal). Which is why I try to only use ISO invariance as far as I need it to preserve highlights

My understanding is there is a type of noise introduced (read noise?) after Iso conversion and before PP adjustment. Therefore PP is amplifying that noise whereas Iso isn't. And I think the definition Iso invariant is based largely around the degree of that noise.

Which is why I try to only use ISO invariance as far as I need it to preserve highlights

So rather than using base Iso (which I suspect is 200 on the K01) you use the nearest suitable Iso and minimize your pp push?
It would be interesting to know if you are getting the best of both worlds or the worst of both worlds. Given my example above I don't think there would be much in it.
Oh and while I am about it you might notice that the Iso 200 one pushed +6EV (top right) is in fact a bit underexposed - the last 3 EV was done by eyeometer on the tone curve. I would expect it to be at least as bad for noise as the 12800.

[JensE]

My understanding is there is a type of noise introduced (read noise?) after Iso conversion and before PP adjustment.

Yes, the A/D conversion introduces some noise. The characteristic of (largely) ISO-invariant cameras however is that this is negligible compared to the (ISO-amplified) sensor noise sources. And of course there is the limited A/D resolution (12bit to 14bit), which I suspect is why the ISO800 shot on the +5EV column looks better than the somewhat smeared-out ISO200.

[dms]

Yes similar to JensE post, for my theatre photographs
-- shutter speed is pretty much at a value where motion blur isn't significant--depend on the action, but say 1/30 s.
-- f stop is set by lens max opening (and at least 1/2-1 stop from max), and need for DOF--lets say f/5.6
-- that leaves iso. Likely a correct value is iso 800 to 1600, but setting it at lower gives similar results (as DR drops 1 stop every doubling iso, and I do pp anyway). I often use iso 400 on K-x and K-5 (and 800 or 1600 on K-20d) because
---- iso 1600 (and possibly iso 800) is likely to/may overexpose highlights
---- and iso 100 and 200 make it hard to look at screen to assess things
---- so iso 400 it usually is for K-x and K-5 (but K20d is not so iso invarient, thus use iso 800 or 1600).**
_____
** this means the dslr that is noisier at high iso is actually the one that is used at higher iso--which may seem strange

[starbase218]

My understanding is there is a type of noise introduced (read noise?) after Iso conversion and before PP adjustment. Therefore PP is amplifying that noise whereas Iso isn't. And I think the definition Iso invariant is based largely around the degree of that noise.

I think this is correct.

So rather than using base Iso (which I suspect is 200 on the K01) you use the nearest suitable Iso and minimize your pp push?
It would be interesting to know if you are getting the best of both worlds or the worst of both worlds. Given my example above I don't think there would be much in it.
Oh and while I am about it you might notice that the Iso 200 one pushed +6EV (top right) is in fact a bit underexposed - the last 3 EV was done by eyeometer on the tone curve. I would expect it to be at least as bad for noise as the 12800.

Maybe there's not too much in it, but it still makes for an easier reviewing/workflow process.

But let's take an example out of the audio world. CD uses 16-bit samples, which means each sample can have one of 65536 values. That's enough for your ears (and mine). But, for example, if you put a digital volume control in your hifi system to attenuate that 16-bit signal, you are throwing away bits, and the speakers may not reproduce all of those 65536 levels anymore, but only 32768 or 16348 (or somewhere in between). I think this is one of the reasons that some digital amplifiers actually work with signals with a bitrate as high as 35 bit.

Having said that, it's possible that the analog signal amplifiers in the camera also introduce their own errors by not being perfect either.

[GUB]

Maybe there's not too much in it, but it still makes for an easier reviewing/workflow process.

yeah not just a black square!!.
I wonder if we will see a generation of cameras that leave the Iso to the raw developing and have an autocorrected JPG for playback . It would still need an indicator somewhere to let you know how underexposed you are but no Iso control in raw. How simple could photography get!! Set your shutter and aperture simply for the reasons above and off you go!!

[starbase218]

yeah not just a black square!!.
I wonder if we will see a generation of cameras that leave the Iso to the raw developing and have an autocorrected JPG for playback . It would still need an indicator somewhere to let you know how underexposed you are but no Iso control in raw. How simple could photography get!! Set your shutter and aperture simply for the reasons above and off you go!!

Maybe one day we'll have 35-bit RAW files. All we'll have to care about is not overexposing.