On Mon, 20 Jan 2003 03:00:55 +1100, "Hughy" <antispam@spamkiller.net> wrote:

 

>The electronics theory I'm fairly confident of - but when it comes to camera

>optics I'm on more shaky ground, happy to defer to more expert comment.

>

>The dB figure refers to video amplifier voltage gain.  0 dB is normally

>taken as a "reference level" and a cameras video amplifiers could be

>considered to be running "normally" when operating at this point.

>

>If a video gain of +6 dB is then selected, the voltage gain of the video

>amplifiers doubles.  From that point, if you increase the gain another 6 dB,

>the voltage gain doubles again.  Mathematically dB=20 log (Voltage Gain

>2/Voltage Gain 1).  At 12 dB, the voltage gain has reached four times its

>original value.  At 18 dB the voltage gain has reached eight times its

>original value.

>

>Looking now at f stops, each time the lens is opened further by an f stop,

>it roughly doubles the light received by the CCD's. All other things being

>equal, this should produce double the voltage from a CCD (ignoring any non

>linearitys).  This is equivalent (AFAIK) to a 6dB increase in voltage gain.

>So two f stops additional lens opening would represent 12 dB - making the

>TRV-900 in theory about 2 f stops "worse" than our XL1.

>

>But it's not quite as simple as this in the performance stakes.  CCD's and

>video amplifiers introduce noise and some have *much* better noise numbers

>than others. A guy called Boltzmann managed to put it all into perspective

>with a formulae: for noise =4KTBR, which is a bit esoteric but almost all

>radio & electronics guys will recognise it instantly. The K refers to

>degrees Kelvin, which is why they cool radio telescope preamps with liquid

>helium (or nitrogen or whatever) - down to close to absolute zero. To reduce

>the noise.  But I diverge (way off).

>

>The TRV-900 is a truly great little cam for the outlay, but in practice, it

>is *significantly* noisier than an XL1 in a given low light situation.  I

>have no figures (maybe David Ruether's website looks at this) but for a

>similar noise performance, my own "guesstimate" would be around 3 f stops

>"worse" (maybe even a little more), not the kinda vague "2 f stops"

>suggested by gain figures in earlier post.  I will gladly defer to those who

>have made careful measurement of the performance of these two cameras under

>better controlled test conditions!!

>

>Looking now at the VX2000 we were running at the same concert, our rough

>estimate (and taking noise into account) would be that the VX2K is around 2

>f stops better than the XL1.  The VX2K noise performance is truly superb!

>And so it should be, the CCD's were developed much more recently than the

>XL1.  When the VX3K arrives, I'll consider being one of the first to buy

>..... a DSR-275 (which on recent history, should electronically be a

>VX3K/PD175 thrown into a bigger box).  But only if 16:9 is available in a

>better form than currently.  These numbers *not* real, but meant to indicate

>later models than those currently on sale!

>

>The substantial difference in bit rate mentioned earlier was caused IMO by

>the much noisier image produced by the TRV-900 at the low light levels

>present during that particular section of the concert.  It reinforced a

>basic fact known to all broadcast/ENG types, but not necessarily to many

>guys using prosumer cameras - the higher the video gain, the more noisy the

>image, the more the bit budget suffers (or it gets softer as noise reduction

>is employed to reduce the noise).

>

>Finally, maybe a bit of a guess about MPEG2 algorithms.  To me personally,

>the tones in very dark video don't *seem* to have as much detail in them as

>lighter shades.  So although this is irrelevent to the above (the two images

>were pretty much equall in brightness), maybe darker images do compress

>better - perhaps due to the varying sensitivity of the human eye??

>Somebody's sure to have good info on this.

>

>That's it.  For those wanting to flame me over some point I've missed or got

>wrong, kindly do it constructively and without nitpicking too hard.

 

Interesting....;-)

In practice, given an "ultimate low-light test" (where the

light is too low for either camera to produce a good image,

but high enough to produce "some" image, the TRV900 (with

lens wide-open and at +18db gain) and XL-1 (not "s" - and

with the lens also wide-open and the gain at its lower

maximum of +12) show the following differences that could

account for the differences in compressed file sizes: the

TRV900 image, being at a higher gain, shows more noise;

the TRV900 fails to provide a good black (unlike the XL-1,

but at +12db rather than +18), making the noise more evident

(see samples at "daylight from small window, low light", at:

www.David-Ruether-Photography.com/camcorder-comparison.htm). Other

samples there indicate that the TRV900 is more than one stop

more sensitive than the XL-1 (perhaps different from the

"s"?), possibly indicating that the predominant lighting

was lower for the TRV900 in the example in the post above

than for the XL-1(?). I suspect that if both cameras were

shooting the same thing nearly exactly, with brightnesses

equalized, and with both operating with gain near "0", that

compression differences (though smaller...) would *still*

show and still favor the XL-1 (!) - in many circumstances,

the XL-1 picture is noticeably less sharp than that of the

TRV900, resulting in less data to compress...;-)