Throw Away Those ND Filters Now

[K12beano]

No! I'm not really recommending that!

But was interested/intrigued by Matt Klown-kowski and his latest tip...


http://www.mattk.com/2012/07/25/tips-for-getting-s...

[GetCarter]

It would help if it wasn't a rubbish photo

[Mr Whippy]

Does the 11bit per channel raw format still only cover 4 ish stops of exposure though? I'm fairly sure it does.

So if you over-expose the shot so you can run a longer shutter then you are actually just extrapolating light data out from the top half or third of the intensities collected to cover 4 stops again. Not really the proper use of RAW, and I'm sure it's technically then a wrongly coloured/lit image, and your totally stuffed if you THEN need to colour correct/grade or do anything else with it because you've got no bit-depth left.

I don't get why you wouldn't just use an ND filter, or stop down a bit, or run a lower ISO (if your not at the bottom already)...


I do wonder though why camera manufacturers can't do something in software internally these days though, so they just dump half the intensity or any other amount before saving the actual data. That should do the same as an ND filter, you can run whatever ISO you want, you just dump data in software as you expose but before any formatting is done... then just run an equivalent correction to the EV readout based on the amount of data you will dump so the metering still works nicely.

Software ND on a digital sensor should be really easy?!


Hmmmmm

Dave

[SamHH]

I think you're describing exactly what some cameras do with their extended ISO range. The problem is that for every stop of brightness that you 'dump', you get one stop closer to clipping the highlights.

[Mr Whippy]

There is probably a reason for it, but I still don't get it.

Assuming the sensor is just detecting a stream of hits why can't it just store the lot, then take the statistical 'peak' of intensity and work back from there when it crunches the data down to 'just' 11 bits?


As said, I don't get the real inner workings of cameras and how they manage the raw raw data, because I'm certain there is a format between the sensor and how it's encoded to raw format still... plenty of quality adjustments could happen there before the raw values are hard-coded, data dumped or lost forever or whatever else.
Afterall, 11bit raw is only 11bit raw to avoid the need for gamma encoding and data loss from what I'm reading. You could easily save to 16 bit per channel or even higher?

Collect a bunch of light at known settings, then pick the highlight and it's associated EV level, then work back through the data saving for as many stops as you have available in the data.


As said, maybe I don't get the inner workings, but I'm guessing it's slow SLR world and legacy issues/expectations based more than anything.

Given the processing speeds and FPS speeds of most modern SLR's I can't think that they couldn't manage more from the data they take in... I'd guess 50% more data exists off the bottom of the histogram, and another 50% within the historgram, that is just trimmed out due to the limits of 11bit 4 stop raw.

Dave

[K12beano]

GetCarter said:

It would help if it wasn't a rubbish photo

It is dire, isn't it! What do you do when you have a bad week of photography and you have a blog to write?


Actually, when I first saw the article, i thought he was going to suggest two images blended, so that just the highlights from the overexposed shot would be used to create water movement - wouldn't that be a better dodge?

[baz7175]

It's not exactly rocket science, there is another option, getting your butt out of bed early enough in the morning when there's just a little bit of light, remember your tripod and hey presto...case and point here:-


Glacial Water Flow by www.bazpics.com, on Flickr

Shot at 6.40am in the morning in the North Cascades wilderness up in Washington State, no ND filter needed there either, Exif data:-

Camera Sony DSC-R1
Exposure 0.25 sec (1/4)
Aperture f/16.0
Focal Length 36.4 mm
ISO Speed 160
Exposure Bias 0 EV

[K12beano]

baz7175 said:

.... getting your butt out of bed early enough.....

dbbb, plth, ssslllllck....... er - does not compute!


Nice shot!

[baz7175]

K12beano]pic]http://www.jonrb.com/emoticons/Eyecrazy.gif[/pic said:

dbbb, plth, ssslllllck....... er - does not compute!

Or the other end of the scale as the light drops off in the evening, same effect, and it doesn't affect your long lie


Mount Adams Trip - 88 by www.bazpics.com, on Flickr

[markmullen]

All well and good but it takes me all of 3 or 4 seconds to slide an ND into my holder, not sure how much quicker or convenient this is.

[GetCarter]

markmullen said:

All well and good but it takes me all of 3 or 4 seconds to slide an ND into my holder, not sure how much quicker or convenient this is.

...and another 5 to pick it up off the ground 'cos the fixings are loose, then another 30 trying to get it clean.

[markmullen]

GetCarter said:

...and another 5 to pick it up off the ground 'cos the fixings are loose, then another 30 trying to get it clean.

Had a Lee foundation kit spring to pieces on rocks at Bamburgh Castle before dawn, had to rebuild it by headtorch light, not an ideal start.

[dibbly dobbler]

I would have thought that unless you're in bright sunshine you could get a slow enough shutter for some water blur just by whacking up the aperture.

I'll be honest I did find the article helpful though as I didn't realise quite how much could be pulled back from an over-exposed shot

[Mr Whippy]

dibbly dobbler said:

I would have thought that unless you're in bright sunshine you could get a slow enough shutter for some water blur just by whacking up the aperture.

Exactly. I really do despair at some of the junk tips you see these days. And that tip is from an academic teaching this kinda stuff too!

Yes a lot can be pulled back in 11bit raw but that really isn't the purpose of the raw format. It's for adjusting slightly for where you get it a bit wrong, a half a stop here, some white balance there, a bit less contrast etc...
But taking an image so it fills the top half of the histogram then dragging it out to fill the whole one so you can get another 50-100% more shutter time is a bit crackers really.

The end result will be ok but I'd argue you are better spending time doing it right than spending time doing a dodgy fix on purpose in post which takes as much time as using an ND in the first place



I'm all for avoiding the use of ND's the way we do on SLR's though... video cameras with their in-built ND's are much more handy for this kinda job.

I'm surprised that these days we can't throttle the sensitivity down in software elegantly with different grad profiles too, so we get the effect of ND simply by altering the raw data off each part of the sensor.

You can do silly stuff on the A77 right now in-camera. Of course on that it's mostly all junk you could do in post but it would be cool to select the ND's and grads like in Lightroom but in your camera and have it expose as if it had an ND there.


Dave

[dibbly dobbler]

Mr Whippy said:

I'm surprised that these days we can't throttle the sensitivity down in software elegantly with different grad profiles too, so we get the effect of ND simply by altering the raw data off each part of the sensor.

Dave

I agree completely. I raised a similar thing a month or two back about why 100 iso is the typical minimum - why can't we go down to iso 10 (say) and then we really could throw away the NDs, Big Stoppers etc. Given the level of clever processing in the modern DSLR I wouldn't have thought this was beyond the wit of man

[gingerpaul]

dibbly dobbler said:

I agree completely. I raised a similar thing a month or two back about why 100 iso is the typical minimum - why can't we go down to iso 10 (say) and then we really could throw away the NDs, Big Stoppers etc. Given the level of clever processing in the modern DSLR I wouldn't have thought this was beyond the wit of man

ISO 10 would be very useful as an option. I know I'd use it from time to time.

[markmullen]

gingerpaul said:

dibbly dobbler said:

I agree completely. I raised a similar thing a month or two back about why 100 iso is the typical minimum - why can't we go down to iso 10 (say) and then we really could throw away the NDs, Big Stoppers etc. Given the level of clever processing in the modern DSLR I wouldn't have thought this was beyond the wit of man

ISO 10 would be very useful as an option. I know I'd use it from time to time.

Yep, I'd find that handy.

[tim-b]

Mr Whippy said:

I really do despair at some of the junk tips you see these days. And that tip is from an academic teaching this kinda stuff too!

To be fair to the chap, he does state more than once that this isn't a better method than using a filter, so I interpreted the article as showing a way to rescue the situation if you don't have one to hand.

It's not a technique I'd use, but it does raise the interesting point about whether to expose for highlights or shadows anymore;

comment below article said:

For years we’ve been told to watch our histograms and expose to the left because if a highlight was clipped we couldn’t get it back (better to deal with the noise in the shadows than lost highlights). Recently I’m seeing more and more suggestions to expose to the other end and let LR get the highlight details back.

[gingerpaul]

markmullen said:

gingerpaul said:

dibbly dobbler said:

I agree completely. I raised a similar thing a month or two back about why 100 iso is the typical minimum - why can't we go down to iso 10 (say) and then we really could throw away the NDs, Big Stoppers etc. Given the level of clever processing in the modern DSLR I wouldn't have thought this was beyond the wit of man

ISO 10 would be very useful as an option. I know I'd use it from time to time.

Yep, I'd find that handy.

As another option my old Canon G9 used to have an ND filter internally that you could flip in and out with the press of a button. Could there be some scope for something like that perhaps?

[Mr Whippy]

tim-b said:

Mr Whippy said:

I really do despair at some of the junk tips you see these days. And that tip is from an academic teaching this kinda stuff too!

To be fair to the chap, he does state more than once that this isn't a better method than using a filter, so I interpreted the article as showing a way to rescue the situation if you don't have one to hand.

It's not a technique I'd use, but it does raise the interesting point about whether to expose for highlights or shadows anymore;

comment below article said:

For years we’ve been told to watch our histograms and expose to the left because if a highlight was clipped we couldn’t get it back (better to deal with the noise in the shadows than lost highlights). Recently I’m seeing more and more suggestions to expose to the other end and let LR get the highlight details back.

Yeah I agree he does mention not to do this really hehe... obviously a bit bored but I'd not blog it personally as it's not really useful at all imo. Two exposures one after the other would be more useful, then blend in post... well that is how I would do it anyhoo.


As for the shooting left/right, surely a blown out highlight is blown out?

I think this is where camera gear these days needs to change as it's basically old hat and dumb to put it bluntly. It's not clear at all what EV range the digital camera catches.

For example, I'm fairly sure the digital camera catches from the EV of the brightest point set (2 EV over the metered mid-grey?) and back down forever... ie, the EV width drops way back to zero signal on the sensor, it doesn't suddenly stop after 4 stops of intensity are covered.

What DOES happen is the camera trims back 4 stops form the bright spot (or 2EV up/down from the mid-grey point) and then dumps the data outside those ranges.

So it'd be interesting to know if the camera thus takes the mid-grey metered 'ideal' conditions it thinks you want, and then crops each end.

Or today are they getting smart and taking the middle 3EV linearly, and then taking another 1 EV at each end and squishing it into the 0.5EV range to cover the 4 stops total (like video knee on digital, or the film falloffs on old SLR's etc)


Perhaps there is some 'extra' at each end on modern digital gear, and that is realised in 11bit per channel as extra capability to go out of the legacy 4 stop range when shooting and not get blow-outs? Is that seen as a safety feature? Does it even exist? It must do if there is tangibly more to come after you think you have blown a highlight out.

Or is it just the benefit of 11bit per channel coming to the fore vs 8bit?




What would be nice on a modern DSLR is to just pick the shutter speed YOU want, and the aperture YOU want, and then let the camera manage everything else. Set a max ISO for noise concerns, and then let the camera meter the scene, splodge the brightest point in scene at the time of exposure at 2048, and then start dumping data back from that for as many stops as you want it to cover... less stops, more range for adjusting in post, more stops and you get a nice HDR (or just wider dynamic range is probably better terminology these days)

Indeed why not just move to something like EXR format, the processors on cameras these days is ample to have an EXR option even if it's not at mega FPS and all that jazz.



I think half the problem is legacy and perception. Things seem slow to change in the DSLR world and it still feels like they hold digital back so it acts like film.

Maybe in 5yrs it'll make a lot more sense and digital will finally 'arrive' because it's still just digital pretending to be film in a lot of ways.

Dave