Better to underexpose or overexpose?

Bill Ravens · December 13th, 2007, 08:10 AM

OVEREXPOSE doesn't mean blow the highlights! Blown highlights are gone forever. That's why it's so critical to use a 3 channel histogram. A lightmeter will tell you you're OK when one color channel is blown.

David W. Jones · December 13th, 2007, 01:09 PM

Sometimes the correct answer is not what your were looking for.

You asked if it was better to over expose SD video or under expose SD video and fix in post.
The correct answer is neither. It is better to expose SD video properly when shooting by using industry standard measurement tools, like a light meter, waveform monitor, vector scope, or simply using the built-in zebra bars on your camera.

Nothing was said in your first post about how much time or money you had for a shoot.

It's kind of like rebuilding an engine and asking if it is best to under tighten the bolts, or over tighten the bolts. The correct answer is neither. It's best to tighten the bolts to the recommended torque setting by using industry standard measurement tools, like a torque wrench.

Bill Ravens · December 13th, 2007, 02:17 PM

Sorry, David, I beg to differ.
Built in light meters do a fair job of determining proper exposure, however, it's not necessarily optimum exposure. In the context of getting the most image information recorded on your record media, in most cases it is better to expose on the high side of the bandwidth than on the low side. The built in lightmeter won't advise you on the subtleties of how to do this, unless you dial some exposure bias into the reading. This is tricky, however, because the amount of "overexposure" that is truly optimal changes according to the scene characteristics. No averaging meter sees well enough to predict the best, ie optimum exposure for every scene. The meter only shows you what the optimum exposure is for an 18% grey scene. How many perfectly defined 18% gray scenes do you run across on every shoot?

That all being said, one is better off, in digital imaging, to err on the high side, that is to overexpose, as long as the highlights aren't blown. How does one tell if the highlights are blown? A 3 channel histogram is the most accurate way. This is commonly accepted knowledge, not my invention or opinion.

BTW, "standard" torque wrenches can be grossly inaccurate if not used properly. They measure stiction, not true torque as seen by the head of the bolt. There are precision instruments available when the sledgehammer approach isn't close enough.

Ervin Farkas · December 13th, 2007, 04:08 PM

Quote:

Originally Posted by Roy Beazley

As far as zebras go I like to see them only on the brightest highlights and maybe spread over some to the next level up... but it depends on your camera.

Where do you set your zebras? 80%? 90%? or 100%?

Bill Ravens · December 13th, 2007, 07:47 PM

100% means it's clipping info. Not much clearer decision point than that.

Glenn Chan · December 13th, 2007, 08:00 PM

Most cameras usually don't clip at 100%. They will record a little headroom above that. On particular cameras you can set that clipping level.

2- I highly suggest you do your own tests.

If you expose very bright, the detail will first hit the camera's knee circuit and then it will clip. The behaviour of the knee differs between cameras and depends on menu settings.

If you expose too dark, bringing up exposure in post will increase noise.

Color correction/grading also plays a role in this.

3- Learn your camera. Learn how the zebras + what you see in the viewfinder and monitor correspond to your exposure level. Some zebras can be set between 70~100... this makes a difference obviously.

Roy Beazley · December 13th, 2007, 08:16 PM

Quote:

Originally Posted by Bill Ravens

OVEREXPOSE doesn't mean blow the highlights! Blown highlights are gone forever. That's why it's so critical to use a 3 channel histogram. A lightmeter will tell you you're OK when one color channel is blown.

So where would you get one of these and could it connect to the H1? That would be cool :)

Thanks
Tony

Bill Ravens · December 13th, 2007, 09:06 PM

Only place I know of would be to use Adobe Audition(aka Serious Magic's HD Rack). Unfortunately, Adobe wants an arm and a leg for it. And that means you have to lug a laptop around to make it work. Some of the newer cameras, like the EX1, have built in histograms, but, the video camera makers still haven't wigged to the fact that a monochrome histogram is not the whole story, you need a 3 channel histogram, because you can blow one channel and the mono histogram won't show it.

Jim Andrada · December 14th, 2007, 12:36 AM

Er- um - I think you meant On Location, now from Adobe, formerly Serious Magic.

Bill Ravens · December 14th, 2007, 08:33 AM

Thanx Jim. Hard to care much about the long fingers of adobe.

Roy Beazley · December 14th, 2007, 06:49 PM

...guess I'll just keep riding the zebras.....lol

Roy Beazley · December 14th, 2007, 07:08 PM

Quote:

Originally Posted by Ervin Farkas

Where do you set your zebras? 80%? 90%? or 100%?

I shot last week on 85%.....

Andy Graham · December 15th, 2007, 03:45 AM

Quote:

Originally Posted by David W. Jones

You asked if it was better to over expose SD video or under expose SD video and fix in post.
The correct answer is neither.

I get your point, you sound like someone who has a strict way of working which is fine , each to their own. Personally i never have the time to set up waveform monitors and i think using a light meter with digital is pointleless because unlike film you can see in the viewfinder or monitor the image you are getting.

To answer the origional question.....I went to a lot of seminars at the Edinburgh film festival this year and one of them was a post digital grading company, they had examples of work and had their grading desk at the front of the auditorium. They said it is best IF ANYTHING to underexpose slightly, now of course as David said its better to expose correctly but once highlights have blasted out they are gone. (there, you heard it from the horses mouth)

The botom line is if you are going to do anything other than expose correctly (the correct exposure being the farthest you can go without anything blasting out) then you should underexpose slightly as this can be fixed unlike overexposure.

Having said that I personally dont even use zebras, i judge the image by eye and iv never had any problems.

Andy.

David W. Jones · December 15th, 2007, 09:58 AM

Quote:

Originally Posted by Andy Graham

I think using a light meter with digital is pointless because unlike film you can see in the viewfinder or monitor the image you are getting.

I guess I'm just an old school fool. Before I ever pull a camera out of it's case, I use my light meter and color temp meter to instantly check exposure level, lighting ratios, and color temperature.

Bill Ravens · December 15th, 2007, 10:37 AM

to rely on what you see in the viewfinder is a fool's errand. the viewfinder has brightness and contrast controls that completely negate any accurate representation of the real image lighting. The only time I would rely on a viewfinder for anything but framing is when it's a reflex image.
time and time again, in this business, I run into people who are so completely stuck in their belief system that they can't accept the latest technological truths. Actually just means more business for me, so have at it.

roy....

there are very valid reasons for shooting at zebra settings other than 100%. For example, skin tends to expose around 70%. Setting the zebra for 70% and exposing to just remove zebra from skin works very well when the subject is primarily a talking head.

Thread Tools	Search this Thread
Show Printable Version Email this Page	Search this Thread: Advanced Search

December 13th, 2007, 08:10 AM	#16
Bill Ravens Join Date: Nov 2001 Location: santa fe, nm Posts: 3,264 Images: 10	OVEREXPOSE doesn't mean blow the highlights! Blown highlights are gone forever. That's why it's so critical to use a 3 channel histogram. A lightmeter will tell you you're OK when one color channel is blown.

December 13th, 2007, 01:09 PM	#17
David W. Jones Trustee Join Date: Dec 2006 Location: Little Rock Posts: 1,383	Sometimes the correct answer is not what your were looking for. You asked if it was better to over expose SD video or under expose SD video and fix in post. The correct answer is neither. It is better to expose SD video properly when shooting by using industry standard measurement tools, like a light meter, waveform monitor, vector scope, or simply using the built-in zebra bars on your camera. Nothing was said in your first post about how much time or money you had for a shoot. It's kind of like rebuilding an engine and asking if it is best to under tighten the bolts, or over tighten the bolts. The correct answer is neither. It's best to tighten the bolts to the recommended torque setting by using industry standard measurement tools, like a torque wrench.

December 13th, 2007, 02:17 PM	#18
Bill Ravens Join Date: Nov 2001 Location: santa fe, nm Posts: 3,264 Images: 10	Sorry, David, I beg to differ. Built in light meters do a fair job of determining proper exposure, however, it's not necessarily optimum exposure. In the context of getting the most image information recorded on your record media, in most cases it is better to expose on the high side of the bandwidth than on the low side. The built in lightmeter won't advise you on the subtleties of how to do this, unless you dial some exposure bias into the reading. This is tricky, however, because the amount of "overexposure" that is truly optimal changes according to the scene characteristics. No averaging meter sees well enough to predict the best, ie optimum exposure for every scene. The meter only shows you what the optimum exposure is for an 18% grey scene. How many perfectly defined 18% gray scenes do you run across on every shoot? That all being said, one is better off, in digital imaging, to err on the high side, that is to overexpose, as long as the highlights aren't blown. How does one tell if the highlights are blown? A 3 channel histogram is the most accurate way. This is commonly accepted knowledge, not my invention or opinion. BTW, "standard" torque wrenches can be grossly inaccurate if not used properly. They measure stiction, not true torque as seen by the head of the bolt. There are precision instruments available when the sledgehammer approach isn't close enough.

December 13th, 2007, 07:47 PM	#20
Bill Ravens Join Date: Nov 2001 Location: santa fe, nm Posts: 3,264 Images: 10	100% means it's clipping info. Not much clearer decision point than that.

December 13th, 2007, 08:00 PM	#21
Glenn Chan Inner Circle Join Date: Jun 2003 Location: Toronto, Canada Posts: 4,750	Most cameras usually don't clip at 100%. They will record a little headroom above that. On particular cameras you can set that clipping level. 2- I highly suggest you do your own tests. If you expose very bright, the detail will first hit the camera's knee circuit and then it will clip. The behaviour of the knee differs between cameras and depends on menu settings. If you expose too dark, bringing up exposure in post will increase noise. Color correction/grading also plays a role in this. 3- Learn your camera. Learn how the zebras + what you see in the viewfinder and monitor correspond to your exposure level. Some zebras can be set between 70~100... this makes a difference obviously.

December 13th, 2007, 09:06 PM	#23
Bill Ravens Join Date: Nov 2001 Location: santa fe, nm Posts: 3,264 Images: 10	Only place I know of would be to use Adobe Audition(aka Serious Magic's HD Rack). Unfortunately, Adobe wants an arm and a leg for it. And that means you have to lug a laptop around to make it work. Some of the newer cameras, like the EX1, have built in histograms, but, the video camera makers still haven't wigged to the fact that a monochrome histogram is not the whole story, you need a 3 channel histogram, because you can blow one channel and the mono histogram won't show it.

December 14th, 2007, 12:36 AM	#24
Jim Andrada Inner Circle Join Date: Feb 2007 Location: Tucson AZ Posts: 2,211	Er- um - I think you meant On Location, now from Adobe, formerly Serious Magic.

December 14th, 2007, 08:33 AM	#25
Bill Ravens Join Date: Nov 2001 Location: santa fe, nm Posts: 3,264 Images: 10	Thanx Jim. Hard to care much about the long fingers of adobe.

December 14th, 2007, 06:49 PM	#26
Roy Beazley Regular Crew Join Date: May 2004 Location: tampa fl Posts: 92	...guess I'll just keep riding the zebras.....lol

December 15th, 2007, 10:37 AM	#30
Bill Ravens Join Date: Nov 2001 Location: santa fe, nm Posts: 3,264 Images: 10	to rely on what you see in the viewfinder is a fool's errand. the viewfinder has brightness and contrast controls that completely negate any accurate representation of the real image lighting. The only time I would rely on a viewfinder for anything but framing is when it's a reflex image. time and time again, in this business, I run into people who are so completely stuck in their belief system that they can't accept the latest technological truths. Actually just means more business for me, so have at it. roy.... there are very valid reasons for shooting at zebra settings other than 100%. For example, skin tends to expose around 70%. Setting the zebra for 70% and exposing to just remove zebra from skin works very well when the subject is primarily a talking head.