I've seen a number of post where folks say they set the gain to -3db for almost everything. But I'm not sure why. I've played with it and I don't notice any particular difference between doing that and leaving the gain control on auto.

Maybe you can s'plain it to me, Lucy.

Cleaner footage. If you have enough light to expose the shot, lowering the gain is a way to increase the SNR.

Good question Charles. This matter first came up (to my knowledge) regarding the XL1. Many folks (including moi) remarked that the XL1's standard image was a bit "hot" and that black areas were prone to grayness and grain. Turning down the camera's gain to -3db produced better contrast by producing a better quality of black, so to speak. Some even say it produces a more film-like image with the XL1 on many scenes.

Personally, I use it sparingly with the XL1s which has a slightly different default image than the XL1. If the scene is dim I do find it can seem to smooth-out the image. But in bright scenes I'll generall leave the gain on auto. Use your own judgement. Your eyes are all that count on this subject.

Well, I'm shhoting with an XL-1s so I guess that's why I didn't notice any difference... except in low light, of course.

Thanks

When you turn up the gain, it gives this artifical brightening that makes the image look video-ish. If it is the desired effect, that's fine, but I would explorer keeping the gain down and lighting or unlighting a scene as needed.

This isnt an insult, but you may not have closely observed film and video stock shot at different settings long enough to know the difference right away.

crafting an image and understanding the results for a medium it is delivered to is something that comes with time. If it looks wrong to you, I say crank up that gain dial till it does.

Im sure there is some great stuff whose story justifies it being shot a +12 gain.

The problem with auto gain, is the same as any auto device on video cameras. They can induce more problems than they cure. Auto gain on the audio, auto white balance, auto focus, auto exposure etc. all have their place and it isn't being left on mindlessly. As JeepBastard suggests take some time and learn your camera and it's many settings. But leaving all the settings in auto will yield far less than optimum results.

I'm not anti automatic settings. I use AF and AWB a lot. Each one of the items is a tool to be used. But what separates a pro from an amateur is how well they use their tools. I own a hammer, but believe me, I'm no carpenter.

Jeff

I think this has been a pretty useful post about auto settings. To refresh this point the advice is if lighting conditions permit, using -3dB gain will reduce the noise in the picture.

I guess as long as we are here, what about posterization? Could having -3dB increase this problem when shooting sky shots?

I'm no fan of auto gain as well. Better to choose a gain level
manually (-3, 0, +3 etc.). Just pick one that you like best. Keep
in mind that grain is a lot harder to see on the small LCD viewfinder
on the camera than on a big TV or Projector!! I either use -3
db or 0.

The -3db setting will not increase or decrease the banding (posterization) commonly seen in shots of the sky. Banding is more the result of 8 bit video. There are not enough colors or luminance levels to produce smooth gradations of subtle level changes.

Jeff

Recall that gain amplifies both signal and noise read from the CCD. If the gain is set at -3 dB and the light reching the CCD is increased by +3 dB by virture of a larger aperture (or slower shutter) you have a ~3 dB improvement in signal to noise in the image. That can be valuable in some situations, although most viewers on ther home TV will never notice.

Although gain reduction has its advantages (noise , DOF) when there is sufficient ambient light, there is also a drawback due to the limited "full well" capacity of CCD structures. The reference gain is normally set at allmost full well for peak white. Reducing gain, and allowing more light on the CCD, potentially generates well overflow which induces extra blooming and streaks. In practical terms this means that high contrast scenes tend to show more blooming and streaking in the -3db(or more) gain shift mode.

Does anyone know whether the gain on the XL1 is an analog or digital gain? That is, does the gain knob amplify the analog video signal coming off the CCD prior to AD conversion, or does it simply multiply the brightnesses of the digital pixels?

If it's an analog amplification, turning up the gain will aid in reduction of posterization. Increased noise enhances the overall perception of a quantized signal. This is known as dithering.

Andre, good point about the well overflow. For those that don't know what he's talking about, take your video camera outside and shoot some street lamps at night. You'll see vertical streaks emanating from the flare of the lamps. The CCDs in Sony's CineAlta cameras don't seem to have this issue.

Analog CCD pixel voltages are, after a fixed amount of amplification, digitized a first time (in 10-bit and more resolution) . on this level all kinds of processing are applied (gain, gamma, shading correction...). In a second step, the (DV) 8-bit 720x480 (576) sampling is being applied. So gain-up adds noise before the final 8-bit quantisation and thus reduces quantisation banding visibility by intensity dithering.

In the XL1, pedestal adjustment, AGC, AWB, gamma correction, are applied in the "analog proces IC". The output from the analog process IC is fed to a 9-bit A/D converter for further processing in the DSP chip (e.g., pixel shift, digital zoom, noise reduction, Y/C processing). The analog process IC is fed from the CCD through a sample/hold IC.

Strange...to my knowledge all modern camcorders only have precompressing analog amps (pre knee) and some direct to the CCD related functions like dark current compensation(pedestal). All other processing is done in the digital domain. Analog gain control and gamma compression of video signals is known as unstable and unreliable.

Andre,
Confucious said: "Never argue with man who have technical reference manual for camera and knows what he read." ...which Don has and does. <g>

Actually, isn't this getting awfully thick and irrelevent to Charles' original question?

Well I'm interested in the inner workings of the XL1S.

_____ thick and irrelevant
|
V

If the XL1S gain control is analog, why is the knob designed to be discrete adjust (versus being a continuously variable pot)? That is, why can't I turn the knob between settings to get +4 dB etc.?

I didn't know that confuscious (+/- 500 B.C.) already knew about the existance of technical ref manuals for camera's...Because I was for over 30 years, toroughly involved in circuit and chip design for simular signal processing, and I know about the strip down exercises to be done for so called ref manuals to make them useable for repair people, I am not impressed by the "interconnected blocks" style wiring diagrams which are published through such channels...Sorry Don. Happy Newyear anyway!

Because the unusual 9-bit processing mentioned by Don I searched and found some info from Chris Hurd http://www.dvinfo.net/canon/skinny.php. Now I understand why Jeff Donald allways sees banding in the sky on his DV camera(XL1s I suppose) where it is seldon mentioned (although possible)in other DV articles. The XL1s apparently has only a 6-bit equavalent grayscale resolution... of course resampled on a 8-bit structure for the DV format. This makes me also conclude that gamma and WB are done in the (9-bit) digital domain. These are the two processes by which you "loose bits": 2 for gamma (4xgain at black) and 6 db for WB (=1-bit). Other processes ( pixel shift, noise reduction...) don't reduce grayscale resolution.

That makes a lot of sense. Thanks for doing the research on it. It's very useful information.

Jeff

> ... If the XL1S gain control is analog, why is the knob designed to be discrete adjust ...

The gain control is a switch that selects a specific voltage step on a voltage divider. That voltage is sent to the control/logic circuit. Depending on the voltage level you get the differnt gain settings, or automatic gain. The analog gain is selected in descrete steps, which works for video. If the control directly controlled the gain via a potentiometer (like many stereo volume controls) it could introduce noise, especially as it ages and would be hard preseed to provide logic selections as well.

As for the 6-bit grayscale. Not sure how that computes. Would be easy to check. Someone shoot a gray wedge and read the luma levels in the output. 6-bit implies only 64 possible values for the luma signal. The XL1 DSP shows 9-bit input, 8-bit output.

FYIW, the old L2 and Ai-Digital camcorders Hi-8 system used 6-bit for the color portion of their internals DSP.