Once in a year opportunity : ) .... messed up. : (

Started Dec 1, 2013 | Discussions thread
OP GeorgianBay1939 Veteran Member • Posts: 3,993
Re: Once in a year opportunity : ) .... messed up. : (

Photomonkey wrote:

GeorgianBay1939 wrote:

clengman wrote:

GeorgianBay1939 wrote:

clengman wrote:

It blows my mind that more people aren't following the point you've tried to make about exposing for the highlights in this shot rather than for the overall scene.

Mine too. But I am not nearly as experienced a photog as you, and thought that I didn't explain it well enough.

Thanks but I'd hardly call myself an experienced photographer. I think we're coming from a similar place. I've been an enthusiastic (but certainly no better than mediocre) amateur for a couple years. I'm also pretty interested in the technical aspects of photography and I've got enough of a background in math and engineering to follow a lot of the information that the bona fide technical whizzes provide regularly.

Yep, similar experience and background. I am also indebted to the whizzes here for lots of help.

It's clear that, like you're trying to point out, you could have used a fast lens, fast shutter and base ISO to freeze motion AND expose the lights on the train at close to the clipping point.

Right. It took me a little thinking to realize that.

The question for me is whether you would have had enough good noise-free detail in the shadows to provide the context that you want in addition to the bright lights on the train. I'm certain that my e-pl1 wouldn't have done a very good job with it. I know the new sensors are much better. Are they really that much better?

That was my question too. I looked at the converted image in Lightroom with no shadow adjustment and realized that the snow and rails in front of the train had lots of exposure to give good brightness (admittedly at ISO 800, an analogue gain of 2 stops ). The image I posted was virtually unchanged with the shadow slider. I did pull down the White and Highlight sliders to reduce the blowing/clipping of the coloured lights.

Here is a screen shot of the unadulterated Lightroom version:

( I have to use DNG because I haven't yet upgraded LR to 5.3, which I need for GX7 .RW2 files.)

I left the black and highlight clipping indicators on for the above screen shot. A lot of the red, blue and some green lights show red due to the indicators.... and the headlights of course.

Here is the RawDigger rendering :

The clipping indicators are turned off. 45k overexposed, 10M underexposed

So about 6M are salvageable. All that snow/rail in front of the loco is fine, as it is. No boosting is needed.

Your question relates to how many stops can I recover without excessive noise (when I reduce 1)exposure by a few stops and 2) analog gain by a few stops.

Gain first:

Yeah, I reduce gain by two stops by decreasing ISO (from 800 to 200), but (I THINK) that I get ~1.5 back from the DR as a function of ISO curve in DxOMark. This one:

DR, in EV, as a function of measured ISO.

So the above deals with the ISO issue. I think. Can you confirm this, please?

I'm afraid I can't. I understand in general terms that increasing pre-ADU gain can potentially improve SNR relative to a case in which the same increase in brightness is applied to the digitized signal. The improved SNR comes with a decrease in DR, but I really don't know how to quantify that trade-off. I think you're right that since you have ample light in the highlight portions of the image and you want to prevent clipping, using base ISO is the way to go, but I think that also means that you will pay a penalty in increased noise in the shadows even before you factor in a decrease in exposure.

YEah, the trick is to quantify the effects of the trade-off. I tend to be a bit geeky also but can't come up with some simple rules of thumb.

My understanding is that both read noise and photon shot noise are present prior to any analog amplification step.

same here.

Noise introduced after analog amplification but prior to discretization by the ADU is still a form of shot noise, but it is not photon shot noise.

.... to digitization by the ADC ... is what I think you mean.

Isn't that (analogue) amplifier noise part of read noise?

This would be the portion of the total noise whose relative contribution to SNR can be decreased by increasing the analog gain.


Is this electronic shot noise most apparent in better exposed parts of the image like photon shot noise, or is it distributed throughout the image? This is something I haven't gotten my head around

Mine either. I thought I had it withJack Hogan's handholding ... but I don't yet think that I have internalized it. And this old head ain't what it used be when it comes to hanging on to that stuff.


The other issue deals with exposure.

Lets say that I dropped the exposure by (EV) 3 stops from what was taken. Would I be able to recover that (in the snow above.) using the Shadow slider in Lightroom?

I don't know how to calculate this but I suspect, from other stuff that I played with that I could probably recover 3 or 4 stops, maybe more. And it is on this point that I look to you competent folks for some good advice.

I hope to see some additional input as well.

I hope so too. See http://www.dpreview.com/forums/post/52636367

As I write this I can think of a test to make some night.... take repeated shots of illuminated ground/snow at varying EV.... for different ISOs ... and see what I can recover in Lightroom. That way I can confirm my conclusions re the ISO issue above. And I can see how many stops (EV) I can recover in Lightroom using the Shadow slider. While I'm at it I will have to find some sort of specular lights / dark area to see how my metering really works.

Thank you very much for working through my stuff and (hopefully) helping with some critique/suggestions re above.

Wish I could be of more help. I'm trying to learn, myself. Thanks for the thought experiment!


Hi Tom,

I see a ton of responses since my 2cents but I also needed to point out that the time of day is way different than your original image. (I am sure someone has noticed that)

I don't understand the above.  If you're referring to the time of the image, I set my camera at UTC.  5 hours ahead of Eastern Standard, where I live.  So the image was taken around 7 pm abour 2 hours after sunset ... dark night.

As I noted in my earlier response, exposing for the highlights at 800 ISO would give you more than enough shutter speed to stop the train. The problem is assuming that the shadows could be extracted by depending on the DR of the sensor.


IMO the scene has insufficient ambient light to bring the unlit areas to a range that would be retrievable with LR.

Depends how much I want.   I only want the snow/rail ahead of the locomotive.   When I look at that area in RawDigger I see lots of light (admittedly at that over exposure and 2 stops of ISO gain.

You can look at the DxOmark charts all you want but if it ain't within 4-5 stops it ain't going to work.

I am close!

You can say I get x stops of DR but you will be dragging a ton of noise irrespective of the DR of any sensor.

I am not at all sure that I agree.   Clengman and I are trying to understand this better by quantifying it for a particular sensor.

Physics has its limits.

For sure!!!  

We believe in numbers but cannot believe our eyes when we see the facts. Sadly, I have a lot of experience in trying to get detail out of shadows as I am an architectural photographer and I often find my self doing just that with night exposures of buildings.

I understand some of  the issue that you are getting at.... but not enough to argue with any authority.  Perhaps others will join in on  pursuing my discussion with Clengman and provide some (much needed) light(!) on the subject of  brightening with ISO or in Post processing.  At least in the context of that train image.


 GeorgianBay1939's gear list:GeorgianBay1939's gear list
Panasonic FZ1000 Panasonic Lumix DMC-GH2 Panasonic Lumix DMC-GX7 Panasonic Lumix DMC-GH4 Panasonic Lumix G Vario 7-14mm F4 ASPH +10 more
Post (hide subjects) Posted by
(unknown member)
(unknown member)
Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark MMy threads
Color scheme? Blue / Yellow