Home of Professional Cinematography since 1996

style="margin-bottom: 0"> 

class="style5" The Truth of -3 db

>Published : 19th June 2005

>I was told recently by a respected DIT that the -3 DB setting that Sony had incorporated into their cameras was the biggest mistake they ever made, having led many videographers to actually reduce the Dynamic Range of the ideo images they are capturing. He said that it does nothing to help the mage or to reduce noise or anything although it may help the image appear blacker and richer on the monitor. He said that in post it leaved much less room for adjusting levels because of the reduced dynamic range and that the DB setting should always be on 0. Anyone care to chime in here?

>Toby Birney, D.P.
Los Angeles, CA & Vilnius Lithuania


>I guess it's really about how the different camera manufacturer's rate their products, since 0db is a somewhat meaningless term anyway. I have found that on the Panasonic, Ikegami and JVC cameras dropping to -3db reduces noise and helps maintain the blacks. It does not have as much of an effect on the Sony cameras. The Varicam for one noticeably improved both on the monitor and on the waveform when switched to -3db so I only used it at this setting.

Mitch Gross
NYC DP


>Whoever you spoke to was Absolutely Correct.

>Dave Satin
Video Engineer


class="Paragraph">>I was told recently by a respected DIT that the -3 DB setting that Sony >had incorporated into their cameras was the biggest mistake they ever >made

>I've always noticed a reduction in noise and a cleaner picture at -3db. A waveform monitor shows no obvious decrease in dynamic range. Every DIT I work with is a firm believer in -3db.

>0db is an arbitrary setting, and at some point someone figured out a better arbitrary setting and thought, "Hey, I'll market it a little differently. Ours don't go to 0 anymore, they go further: to -3!"

>If -3db reduces dynamic range from 0db then, by your DIT's logic, +6db should give you dramatically increased dynamic range... and it doesn't. The image has much more contrast and noise.

>Sony made such a big mistake that other manufacturers have been incorporating the -3db setting into their cameras for years.

>I'm proud to say I used -3db for a shoot this very day.

>Art Adams, DP [film|hdtv|sdtv]
Mountain View, California - "Silicon Valley"
http://www.artadams.net/


class="Paragraph">>Whoever you spoke to was Absolutely Correct.

>Please explain. I need details!

>Art Adams, DP [film|hdtv|sdtv]
Mountain View, California - "Silicon Valley


>>If -3db reduces dynamic range from 0db then, by your DIT's logic, +6db >should give you dramatically increased dynamic range... and it doesn't.

>I have heard this concept advanced before, but with no reasonable explanation as to why this should be so. I can't think of why reducing amplification by 3db would have any effect on the dynamic range of the captured image, any more than closing down the iris an equivalent amount would.

>I am, however, more than willing to have my mind changed by someone with a proper technical explanation that stands up under scrutiny, so if there is one, let's hear it.

>The other thing is that while reducing the gain of an amplifier by 3db will, in general, decrease its noise, we have gotten to the point in current *expensive* camera (not DVCRAP) technology that the practical aspect of -3db is close to meaningless. If I look VERY hard on an expensive 20" monitor, I can barely see a minuscule reduction in shadow noise while matting off blue and going to -3db. And whatever noise reduction I do see disappears in the noise of the recording medium.

>Instead of participating in this "me too" bullshit, camera manufacturers would be better off dealing with noise issues properly in the first place. If -3db is so useful, why not make that the "new 0db" baseline, and brag that your camera is the quietest?

>Bob Kertesz
BlueScreen LLC
http://www.bluescreen.com


>Bob Kertesz wrote:

class="Paragraph">>...If -3db is so useful, why not make that the "new 0db" baseline, and >brag that your camera is the quietest?

>Because then you'd have to admit that it was less sensitive than Brand B.

>Wade K. Ramsey, DP
Dept. of Cinema & Video Production
Bob Jones University
Greenville, SC 29614


class="Paragraph">>If -3db is so useful, why not make that the "new 0db" baseline, and brag >that your camera is the quietest?

>Vestigial, just like the crosshair setting on the filter wheel. If they did away with the -3db choice then people would complain that a feature had been removed. It's all very "Spinal Tap."

Mitch "But this goes to eleven." Gross
NYC DP


My understanding of this issue has been that negative gain causes the exposure range to be moved away from the middle of the straight line section of what an old sprocket hole like me would call the characteristic curve.

Minus three is a compromise setting that minimises this effect...Any more and the problems described could be quite noticeable.

>Shooting a relatively low contrast scene, this problem shouldn't crop up, but the higher the contrast goes, the more likely you are to stray from the straight-line part of the curve.

>Perhaps in a strictly technical sense, the use of -3db does cause the described deficiencies but in a practical sense, they are so barely noticeable that the benefits outweigh the disadvantages. I use -3, 0 and +6 on my card with +12 on turbo. I've never had a problem with the -3 setting and I do a lot of high contrast, full sun in the middle of the day stuff on farms. Because most people don't seem to use these settings, I ALWAYS do a black after reading the card to the camera.

>I'm keen to be enlightened further on this one.

>Mick Fanning
ABC Brisbane


>Bob Kertesz writes :

class="Paragraph">>I can't think of why reducing amplification by 3db would have any effect >on the dynamic range of the captured image reducing the gain of an >amplifier by 3db will, in general, decrease its noise

>Decreased noise is the same as an increased signal-to-noise ratio, which is the same as increased dynamic range.

>Ultimately, however, this issue all comes down to the sensitivity of your imaging chip. The less amplification the chip's signal requires, the less noisy the result.

>This is directly analogous to the sensitivity of microphones vis-a-vis audio noise.

>Dan Drasin
Producer/DP
Marin County, CA


>The truth about -3db or -6db is that it does not and cannot change contrast or anything about the image except to increase the signal to noise ratio. In other words the picture has less noise.

>As this switching of gain is in the pre-amp stage before any signal processing nothing about the picture changes.

>IT IS EXPECTED THAT YOU WILL OPEN THE IRIS TO COMPENSATE FOR THE LOWER ELECTRICAL GAIN. You restore the picture to the same level as before switching the gain by changing the iris. This also goes for changing the gain to +6db or +12db and you have to lower the exposure to bring the level back to "normal". There will be more video noise but the contrast and other characteristics will be the same.

>(By the way in the Panasonic Varicam and maybe other Panasonic products the exact change of electrical gain when switching between the gain settings is chosen in the menu system. In other words the -3db or +6db labels are just labels and the amount of gain change is set elsewhere.)

>DISCLAIMER : I believe that there are some camera designs that change items like lowering the amount of image enhancement when raising the electrical gain. There maybe other items that change in some specific design areas. But this is a specific camera design and cannot be generalized about.

>Regards,

>Bill Hogan


class="Paragraph">>Decreased noise is the same as an increased signal-to-noise ratio, >which is the same as increased dynamic range.

>Signal to noise ratios have absolutely NOTHING to do with increased dynamic range. It could have a noisy camera that would handle 9 stops of dynamic range without clipping and a very quite camera that could only handle 7 or 8 stops of dynamic range. The two terms and the results are totally unrelated.

>Regards,

>Bill Hogan


class="Paragraph">>Decreased noise is the same as an increased signal-to-noise ratio, >which is the same as increased dynamic range.

>I don't know where you got this information, but the last part is incorrect.

>Signal to noise ratio has nothing to do with dynamic range. If it did, your conclusion would be opposite that of the original poster's assertion that -3db decreases dynamic range, and I believe neither is correct.

>Bob Kertesz
BlueScreen LLC


class="Paragraph">>My understanding of this issue has been that negative gain causes the >exposure range to be moved away from the middle of the straight line >section of what an old sprocket hole like me would call the >characteristic curve.

>I'm not sure that changing the db setting on a video camera is directly analogous to moving exposure up and down a film curve. We're just talking signal to noise changes, right?

class="Paragraph">>If -3db is so useful, why not make that the "new 0db" baseline, and brag >that your camera is the quietest?

>Because coming out with a -3db setting is a marketing gimmick : "Our camera has a negative gain setting! How cool is that! Nobody else has that! Buy! Consume! Be happy!"

>In a battle between common sense technical thinking and marketing, marketing will win every time.

>Art Adams, DP [film|hdtv|sdtv]
Mountain View, California - "Silicon Valley"


class="Paragraph">>Decreased noise is the same as an increased signal-to-noise ratio, >which is the same as increased dynamic range.

>Exactly.

>Denny Lajeunesse
d/d/e
Vancouver, bc


>D.J. Lajeunesse wrote:

class="style7">>Decreased noise is the same as an increased signal-to-noise ratio, >which is the same as increased dynamic range.

class="style7">> Exactly.

>I think that there's a level of confusion here that is leading to all of these generalizations.

>What's dynamic range?

>Simple : Compose a frame with both bright and dark elements. Whatever you can capture through a single exposure is your dynamic range.

>Say that the bright element is a window with sunlight coming through. The dark element is a wastebasket under a desk that is completely obscured. There are trees outside the window and magazines in the wastebasket.

>How much of that scene can you capture without loosing lowlight and highlight information?

>The answer to that question is your dynamic range.

>Notice that I did not talk about gains and shutter settings and filters and lenses or bits-per-pixel, etc. In fact, I didn't even say that this is a video camera or a still camera. Heck, for that matter, I didn't even say that this is DIGITAL!

>Well, folks, the fact is that dynamic range is defined exactly the same way regardless of acquisition methodology or processing. Either you get the picture (on film, CCD, CMOS, whatever) or you don't.

>Now, going digital, the dynamic range of a camera --any camera-- is purely, absolutely and unequivocally determined --as in: set in stone-- by the sensor characteristics. There's nothing you can do in conventional processing[1] to alter this. The best you can hope is to keep as much of it as you can[2].

>With this in hand, it is easy to understand that an increase in signal-to-noise has nothing to do with an increase in dynamic range. The two are only related in that you can waste good dynamic range by having a very noisy signal[3]. But you cannot INCREASE dynamic range with better SNR, all you can hope to do is capture as much of it as the sensor can produce.

>So, signal-to-noise (and, by extension, this "-3 dB" issue) is about carrying your precious cargo from sensor to analog-to-digital converter without mangling what the sensor is able to do. As someone pointed out, in modern circuit design the amplifier performance for gain settings of "-3" vs. "0" would be so close as to not make much of a difference[4].

>Martin Euredjian
eCinema Systems, Inc.
www.ecinemasys.com


class="style7">>Decreased noise is the same as an increased signal-to-noise ratio, >which is the same as increased dynamic range.

>Bill Hogan writes:

class="style7">>Signal to noise ratios have absolutely NOTHING to do with increased >dynamic range.

>There's the total dynamic range of the *system* (which you're referring to) and then there's the dynamic range of the *signal*, which is what I was addressing.

>In a situation were you're not using the full dynamic range of the system to begin with, you might as well open up a stop and then reduce your noise floor by dropping your preamp gain. This increases the dynamic range of the *signal* (which cannot exceed that of the system of course, but can at least be *optimised* within the system's limitations.)

>Dan Drasin
Producer/DP
Marin County, CA


class="style7">>I'm not sure that changing the db setting on a video camera is directly >analogous to moving exposure up and down a film curve. We're just >talking signal to noise changes, right?

>I seem to remember first hearing of this a few yeas ago on the GTC forum - or maybe it was the old CompuServe Broadcast Professionals' forum (that WAS a long time ago). I think it was from an Ikegami bloke. I followed up by asking our local Sony engineer and he confirmed it.

>Still awaiting enlightenment.

>Mick Fanning
ABC TV Brisbane


class="style7">>...If -3db is so useful, why not make that the "new 0db" baseline, and >brag that your camera is the quietest?

>I'm having a "Spinal Tap" moment here.... in reverse!

>Guy Galloway
DP
Little Rock, Arkansas USA


>Martin Euredjian wrote:

class="style7">>...So, signal-to-noise (and, by extension, this "-3 dB" issue) is about >carrying your precious cargo from sensor to analog-to-digital converter >without mangling what the sensor is able to do.

>Would you say, then, that the value of -3 dB is that it may reduce noise slightly and substitute for some ND?

>Wade K. Ramsey, DP
Dept. of Cinema & Video Production
Bob Jones University
Greenville, SC 29614


>Wade Ramsey wrote :

class="style7">>Would you say, then, that the value of -3 dB is that it may reduce noise >slightly and substitute for some ND?

>To tell you the truth that's the only use and care I have for the -3db setting: loosing depth of field in a hurry by allowing me to open the iris . It's not quite one stop so when a ND3 is too much it's just a nice little difference at the flick of a switch.

>I have never noticed, nor really gone out of my way to look, for any flagrant change with the image except that... it gets darker if you don't open up.

>My two and a half cents.

>Daniel Villeneuve, c.s.c.
Directeur-Photo/Director of Photography
Montréal, Canada
demo à / at : http://pages.infinit.net/davil


>Wade Ramsey wrote:

class="style7">> Would you say, then, that the value of -3 dB is that it may reduce noise >slightly and substitute for some ND?

>Yes, I think so.

>Martin Euredjian
eCinema Systems, Inc.


>I use -3 all the time for reducing depth of field. The other use is for those occasions that I'm working with a teleprompter. I engage -3 while lighting, knowing that when the glass finally goes up in front of the lens I can compensate for the 1/2 stop loss of the mirror by going back
to 0db.

>Randy Miller, DP in LA


>Well there doesn't seem to be a definitive answer yet.

>Perhaps I was given a bum steer or perhaps I misunderstood. But I definitely remember being told by those seemingly in the know that there was some sort of issue with the use of negative gain that meant that it was unsafe to go any further than minus three.

>Mick Fanning
ABC Brisbane