| View previous topic :: View next topic |
| Author |
Message |
ChrisCoaster
Guest
|
 Posted: Thu Jul 10, 2008 6:17 pm Post subject: Re: Frequency/Sample rate |
 |
|
On Jul 10, 8:16 am, dpierce.cartchunk....@gmail.com wrote:
[quote]On Jul 10, 7:29 am, ChrisCoaster <ckozi...@snet.net> wrote:
On Jul 10, 1:50 am, Willem <wil...@stack.nl> wrote:
ChrisCoaster wrote:
) On Topic: If someone can explain the 44.x sampling rate of CDs to
) someone whose math skills are limited to adding/subracting whole
) numbers, it would be appreciated. My idea that 44kHZ was space enough
) to contain 22kHZ for two stereo channels was obviously wrong, and
) something I would like clarified/corrected after having believed it
) for 20 years now.
Easy:
If you have a sound with a frequency of 20kHz, that means that the wave has
to go through one complete cycle 20.000 times a second. That is: up *and*
down 20.000 times a second. To record that you need to record both the ups
and the downs, so that>s 20.000 ups plus 20.000 downs, makes 40.000 samples
per second.
Why they went up from 40kHz to 44.1kHz is some weird technical reason,
explained in another part of this thread.
SaSW, Willem
--
Disclaimer: I am in no way responsible for any of the statements
made in the above text. For all I know I might be
drugged or something..
No I>m not paranoid. You all think I>m paranoid, don>t you !
#EOT
_________________
Thanks Willem! You are the only one to finally clarify this in
ENGLISH
Except that the explanation, while seemingly intuitively
correct, is misleading. Among other things, it assumes
that the waveform IS sinsusoidal AND that it has "ups
and downs". It doesn>t have to, and, in fact, the signal
could consist of only differeing "ups".
(remember I can barely add 2+2).
Then if you insist on casting everything into a 2nd grade
mathematical vocabulary, you>re guaranteed to get it
wrong, wven when any number of people try patiently
to explain to you that a 2+2 understanding isn>t going to
cut it.
As for that 44kHz bit, there are some people out there who hear like
dogs; that is they can sense sounds equal to or greater than 20kHz.
The designers of the CD standard probably chose 22kHz because a CD>s
freq response actually drops off like the Cayman Wall at the ends of
it>s specified range, due to its digital nature. Unlike analog, which
"rolls" off gradually at varying rates, subject to source.
Wrong. The "sampling" portion is not in the least bit
"digital", in fact, the sampler can be PURELY analog.
One very practical example is the old analog bucket-
brigade delay lines, which absolutely HAD to have
a brick-wall ANALOG anti-aliasing feature. Another
example is switched-capacitor filters: the signals
are PURELY analog, but discrete-time sampled, and
thus absolutely REQUIRE a brick-wall ANALOG filter
to prevent aliasing.
Yet another example is found in sampling oscilloscopes,
technology which comes from well before 1970. The
fignals are purely ANALOG in a discrete-time analog system.
And guess what? That brick wall ANALOG filter has
to be placed at LESS THAN TWICE THE SAMPLING
RATE to prevent aliasing. The exception is in the case
of the sampling oscilloscope which DELIBERATELY
aliases a very high frequency down to the base band.
So they
figured, move that Cayman Wall slightly above 20kHz to ensure that
99.999999% of musical content gets on the disc. Hence, 22kHz.
So now
you have the high & low portions of soundwaves up to 22kHz = 44.1(to
be exact)kHz!
Did you bother AT ALL, to read some of the posts,
including my own, which addressed QUITE SPECIFICALLY
where the 44.1 kHz sample rate came from?
They could have done 40 kHz, 45 kHz, 50 kHz. Some for the
broadcast industry even do 32 kHz sample rates.
But the ONLY reason 44.1 kHz was chosen was because
the ONLY high-speed storage and transmission medium
that was readily available at the time (late 1970>s) for
digital audio was video tape recorders. And based on the
NTSC frame rate (60 Hz), the number of scan lines per frame
(525), the number of blanked lines per frame (35) leading
to 490 available lines per field or 245 per frame) and the
video bandwidth adequate for the bit rate which lead to putting
3 sample PAIR per scan line leads to:
60 * 245 * 3 = 44100 sample PAIRS per second
On 50 Hz video, a similar calculation based on 625/50 Hz
with 37 blanked lines leads to:
50 * 294 * 3 = 44100 sample PAIRS per second
You can hold on to your "Cayman cliff" understanding if
it feels right for you, but you>d be wrong. 44.1 kHz was
chosen for the reasons given here. Period. It>s not a
debatable point, it>s not subject to what you think you
can or cannot understand, it>s a matter of established
technical and historicl fact.
You don>t like that explanation? YOu can get your
arms around it? Fine, go rewrite history and change
physics while you>re at it. Good luck with that.- Hide quoted text -
- Show quoted text -
_____________[/quote]
Whoa, easy buddy. I can feel your heart racing through my high-speed
connection.
All I>m saying is that I suffered a significant traumatic head injury
as a child and from that point in my life was never able to keep up or
recover in math. My english/grammar on the other hand earned me not
As, not double-As, but AAA+!! Social Studies/History - fuggedaboudit
- worst grade was a B+!! It>s too bad because good grammar or
history grades do not get you ahead in this world; good math & tech
skills do. We>ve got a guy in the WhiteHouse who can>t even pronounce
NEWKYOOLER yet he>s leader of the free world!
I>m not attempting to rewrite history or change anything, I just need
this explained in terms that someone who has terrible math skills can
understand!
What I meant with my Cayman wall analogy was the actual analog
frequency response curve of the CD medium. That is, if you started to
roll sine wave from 20Hz all the way up to 22kHz, and record it to a
CD, the drop off at the high end of that would be instantaneous, not a
soft gradual roll off as on a record or cassette. |
|
| |
|
 Back to top |
ChrisCoaster
Guest
|
| Posted: Thu Jul 10, 2008 11:31 pm Post subject: Re: Frequency/Sample rate |
|
|
On Jul 10, 6:23 pm, "geoff" <ge...@nospam-paf.co.nz> wrote:
[quote]Willem wrote:
ChrisCoaster wrote:
) On Topic: If someone can explain the 44.x sampling rate of CDs to
) someone whose math skills are limited to adding/subracting whole
) numbers, it would be appreciated. My idea that 44kHZ was space
enough ) to contain 22kHZ for two stereo channels was obviously
wrong, and ) something I would like clarified/corrected after having
believed it ) for 20 years now.
Easy:
If you have a sound with a frequency of 20kHz, that means that the
wave has to go through one complete cycle 20.000 times a second.
That is: up *and* down 20.000 times a second. To record that you
need to record both the ups and the downs, so that>s 20.000 ups plus
20.000 downs, makes 40.000 samples per second.
Why they went up from 40kHz to 44.1kHz is some weird technical reason,
explained in another part of this thread.
You are, I hope, joking. Cos that is total crap.
geoff- Hide quoted text -
- Show quoted text -
_______________[/quote]
Alrighty geoff, then YOU explain it. And remember, easy on the math!
'Cause at this point I>m about hit "Wiki" in my favorites.
-CC |
|
| |
|
| Back to top |
Guest
|
| Posted: Fri Jul 11, 2008 12:56 am Post subject: Re: Frequency/Sample rate |
|
|
On Jul 10, 7:31 pm, ChrisCoaster <ckozi...@snet.net> wrote:
[quote]If you have a sound with a frequency of 20kHz, that means that the
wave has to go through one complete cycle 20.000 times a second.
That is: up *and* down 20.000 times a second. To record that you
need to record both the ups and the downs, so that>s 20.000 ups plus
20.000 downs, makes 40.000 samples per second.
Why they went up from 40kHz to 44.1kHz is some weird
technical reason,
explained in another part of this thread
You are, I hope, joking. Cos that is total crap.
Alrighty geoff, then YOU explain it. And remember,
easy on the math!
[/quote]
Any explanation that>s CORRECT and ACCURATE
involves math beyond addition and subtraction:
assuredly multiplication and division for one.
However, perhaps an explanation by analogy might
help.
Imagine one of them ol' western movies, where the
indians on horses are chasing the settlers in their
covered wagons. Did you ever sometimes notice
that the wheels seem to be turning BACKWARDS?
Even backwards slowly?
That>s due to a phenomenon known as "aliasing."
It>s a direct result of having the wheel spin too fast
for the movie camera. What>s actually happening
is quite simple. Pretend our movie camera is running
at 25 frames per second. And let>s also preten that
ONE of the wagon wheel spokes is painted white and
all the rest are dark wood colored.
If the wagon wheel rotates slowly, sy, one revolution
per second, in one second, we will have captrued 25
images of the wheel, each one with the white spoke
rotated about 14 dgrees further than the previous frame.
When I play this movie back, and measure the wheel,
I will measure its rotation at 1 per second
If I speed it up to 2, 3, 4, 5 revolutions per second,
it>s ALWAYS the case that on playback, the image
of the wheel is spinning at the right rate in the right
direction.
But now let>s see what happens if it>s spinning
EXACTLY 12.5 revolutions per second (gee,
that>s HALF the frame rate. At that point, the wheel
spins just fast enough to make it halfway around.
So in the first frame it>s pointing up. in the second
it;'s pointing down, the third up, and so on.
On playback, how fast and in what direct is the wheel
turing? Well, it>s impossible to tell, because all you
see is two white spokes pointing up and down, not
moving at all.
And that happend at 1/2 the frame rate of the camera.
But slow it down JUST a little, say 10 revolutions per
second. in the first frame, say, it>s pointing up. By
the second frame, it>s moved 144 degrees, so it>s
NOT quite poiting down. Third frame, it>s moved a
total of 288 degrees and has NOT quite returned to
it>s start.
Play this back, and you will see the wheel turning
at the right speed and in the right direction.
Now speed it up a little, say to 15 revolutions per
second. Now the wheel has moved just past the
bottom, 216 degrees, then 72 dgrees beyond the
start point, and so on.
And what you>ll find is that NOW the wheel, instead
of spinning dorwards at 15 RPM, looks like it>s
spinning BACKWARDS at 10 RPM.
As an extreme case, consider the wheel spinning
forwards at 24 RPM: because your movie camera
is not sampling fast enough, the wheel will actually
look like it>s moving BACKWARDS at a measily
1 revolution per second.
What>s the FAsTEST the wheel can turn before the
playback does NOT accurately depict it? It HAS to
be spinning LESS THAN 1/1 the frame rate of the
camera.
Conversely, how fast does you camera have to
run if it want>s to accurately capture a wheel spinning,
say 20 revolutions per second? Well, it HAS to be
MORE THAN twice that, or FASTER than 60 frames
per second.
Now, the interesting thins about aliasing like this
is that ALL possible aliases get folded down. With
your frame rate of 25 per second you;'' get the same
picture back again with the wheel spinning at 1 rpm,
26 rpm, 51 rpm, 76 rpm, 101 rpm and so on on,
and at 24 RPM, 49 RPM, 74 RPM 99 RPM, the wheel
will look like it>s spinning BACKWARDS at 1 RPM.
Notice also that aliases are occuring at MULTIPLES
of the original sample (or frame) rate +- 1 RPS, e.g.,
25*2 = 50+-1, 25*3+-1, and so on, up to infinity.
Do the experiment yourself in your head: What>s the
fastest the wheel can spin without seeming to stop,
spin backwards, or spin at a rate different on playback?
The answer is ALWAYS, less than 1.2 the sample rate.
ALWAYS (assuming it>s the baseband you want).
And, thank you for your concern, but your ability to
"feel my heart racing through your high-speed
connection" needs some serious reevaluation.
Your ability to sense my stree, kind sir, is quite
defective.
It has been said, however, that I>m not first in line
for the Nobel prize in suffering fools gladly. Especially
when this is easily a topic that>s been discussed
HUNDREDS of times before. |
|
| |
|
| Back to top |
geoff
Guest
|
| Posted: Fri Jul 11, 2008 2:39 am Post subject: Re: Frequency/Sample rate |
|
|
Willem wrote:
[quote]ChrisCoaster wrote:
) On Topic: If someone can explain the 44.x sampling rate of CDs to
) someone whose math skills are limited to adding/subracting whole
) numbers, it would be appreciated. My idea that 44kHZ was space
enough ) to contain 22kHZ for two stereo channels was obviously
wrong, and ) something I would like clarified/corrected after having
believed it ) for 20 years now.
Easy:
If you have a sound with a frequency of 20kHz, that means that the
wave has to go through one complete cycle 20.000 times a second.
That is: up *and* down 20.000 times a second. To record that you
need to record both the ups and the downs, so that>s 20.000 ups plus
20.000 downs, makes 40.000 samples per second.
Why they went up from 40kHz to 44.1kHz is some weird technical reason,
explained in another part of this thread.
[/quote]
You are, I hope, joking. Cos that is total crap.
geoff |
|
| |
|
| Back to top |
geoff
Guest
|
| Posted: Fri Jul 11, 2008 2:39 am Post subject: Re: Frequency/Sample rate |
|
|
ChrisCoaster wrote:
[quote]If you have a sound with a frequency of 20kHz, that means that the
wave has to go through one complete cycle 20.000 times a second.
That is: up *and* down 20.000 times a second. To record that you
need to record both the ups and the downs, so that>s 20.000 ups plus
20.000 downs, makes 40.000 samples per second.
Why they went up from 40kHz to 44.1kHz is some weird technical
reason, explained in another part of this thread.
You are, I hope, joking. Cos that is total crap.
geoff- Hide quoted text -
- Show quoted text -
_______________
Alrighty geoff, then YOU explain it. And remember, easy on the math!
'Cause at this point I>m about hit "Wiki" in my favorites.
-CC
[/quote]
Simple. To accurately recontruct a waveform, you need two samples of the
highest frequency specificied. Look up Nyquist Theory on you beloved Wiki -
even Wiki has that right !
Nothing to do with 20KHz worth or 'up' and 20KHz worth of 'down'.
geoff |
|
| |
|
| Back to top |
Guest
|
| Posted: Fri Jul 11, 2008 11:41 am Post subject: Re: Frequency/Sample rate |
|
|
On Jul 10, 9:32 pm, "geoff" <ge...@nospam-paf.co.nz> wrote:
[quote]Simple. To accurately recontruct a waveform, you need two samples of the
highest frequency specificied. Look up Nyquist Theory on you beloved Wiki -
even Wiki has that right !
[/quote]
Omigod, this is painful!
Look, TWO samples per cycle WILL NOT DO IT.
You MUST have MORE than two samples.
Let me try it a different way
TWO samples per cycle WILL NOT DO IT.
You MUST have MORE than two samples. |
|
| |
|
| Back to top |
Guest
|
| Posted: Fri Jul 11, 2008 11:50 am Post subject: Re: Frequency/Sample rate |
|
|
On Jul 10, 2:17 pm, ChrisCoaster <ckozi...@snet.net> wrote:
[quote]What I meant with my Cayman wall analogy was the actual analog
frequency response curve of the CD medium. That is, if you started to
roll sine wave from 20Hz all the way up to 22kHz, and record it to a
CD, the drop off at the high end of that would be instantaneous,
not a soft gradual roll off as on a record or cassette.
[/quote]
No, the rolloff is NOT instantaneous, nothing like it.
While maybe steeper than a cassette, it is far from
the hypothetical "brick wall." A theoretically perfect
brick wall filter is not possible to implement in a
practical world.
Rather it merely has to be "good enough," which means
that across the significant audio bandwidth, it has to
affect the audible response in the least possible way,
and as you approach 1/2 the sampling rate, it has to
be far enough down so that any expected images or
aliases are sufficiently attenuated so has to have no
impact.
That means that if you>re criteria is to be flat at 20 kHz,
you have a 2.05 kHz transition band available for your
filter, and that provides sufficient margin for many
competent filter implementations. |
|
| |
|
| Back to top |
Arny Krueger
Guest
|
| Posted: Fri Jul 11, 2008 12:47 pm Post subject: Re: Frequency/Sample rate |
|
|
"ChrisCoaster" <ckozicki@snet.net> wrote in message
news:2445952f-415a-45d6-be94-227187b30442@f63g2000hsf.googlegroups.com
[quote]What I meant with my Cayman wall analogy was the actual
analog frequency response curve of the CD medium. That is, if
you started to roll sine wave from 20Hz all the way up to 22kHz, and
record it to a CD, the drop off at the high end of that would be
instantaneous, not a soft gradual roll off as on a record or cassette.
[/quote]
Sort of. So what?
To me the roll-off between 20 KHz and 22.05 KHz is not *instantaneous*.
There is a discernable, measureable slope.
For example, on a Behringer UCA 202, response is less than 1 dB down at 20
KHz, but its only about 4 dB down at 21 kHz, and 8 dB down at 21.8 kHz.
Measuring response between 21.8 and 22.05 kHz gets a little hairy, but if
you want to do the work, the slope is there.
High frequency response is a little like clipping. The debate over soft
roll-offs and sharp roll-offs is like the debate over soft clipping versus
sharp clipping.
Would you rather have a power amp that rises smoothly to 10% THD at 50 watts
and clips at 100 watts, or would you prefer a power amp that is almost
perfectly clean up to 99 watts, and clips at 100 watts? Well, as long as
you stay below 100 watts, you>d probably want the amp that is clean up to 99
watts. With a reasonable application, avoiding going over even 50 watts is
completely doable.
So, would you rather have a recorder that drops off smoothly to 1 dB down at
10 kHz, and is 3 dB down at 15 kHz, or would you rather have a recorder that
is nearly perfectly flat to 20 kHz, and then the response drops off pretty
sharply. I can tell you that if you did an ABX test, the recorder that is 3
dB down at 15 kHz will be pretty easy to pick out, and the one that is
almost perfectly flat to 20 Hz will be very, very tough.
As far as the So what part goes, here>s what I mean. Does the shape of a
system response curve actually matter that much above even 12 kHz? In fact,
all we know for sure is what we hear, and what we hear even as low as 10 KHz
is actually pretty corrupt. I>m speaking acoustically, of course.
It is well-known in the perceptual coder community that with almost all real
world music, you can throw away *everything* above 16 kHz and nobody is the
wiser. Well they won>t be the wiser until they stop listening to the music
and start looking at frequency response curves. They they may have anxiety
attacks. But, its not what they hear that is bugging them. Its what they
think they hear. |
|
| |
|
| Back to top |
GregS
Guest
|
| Posted: Fri Jul 11, 2008 12:47 pm Post subject: Re: Frequency/Sample rate |
|
|
In article <RLednQb_mYs_JevVnZ2dnUVZ_umdnZ2d@giganews.com>, "geoff" <geoff@nospam-paf.co.nz> wrote:
[quote]ChrisCoaster wrote:
If you have a sound with a frequency of 20kHz, that means that the
wave has to go through one complete cycle 20.000 times a second.
That is: up *and* down 20.000 times a second. To record that you
need to record both the ups and the downs, so that>s 20.000 ups plus
20.000 downs, makes 40.000 samples per second.
Why they went up from 40kHz to 44.1kHz is some weird technical
reason, explained in another part of this thread.
You are, I hope, joking. Cos that is total crap.
geoff- Hide quoted text -
- Show quoted text -
_______________
Alrighty geoff, then YOU explain it. And remember, easy on the math!
'Cause at this point I>m about hit "Wiki" in my favorites.
-CC
Simple. To accurately recontruct a waveform, you need two samples of the
highest frequency specificied. Look up Nyquist Theory on you beloved Wiki -
even Wiki has that right !
Nothing to do with 20KHz worth or 'up' and 20KHz worth of 'down'.
[/quote]
Not any waveform. Just a sinewave.
Its interesting aliasing can be seen with nothing to do with time,
but its still a sample, such as with viewing lines through various
filters.
greg |
|
| |
|
| Back to top |
Guest
|
| Posted: Fri Jul 11, 2008 1:19 pm Post subject: Re: Frequency/Sample rate |
|
|
On Jul 11, 8:24 am, zekfr...@zekfrivolous.com (GregS) wrote:
[quote]Not any waveform. Just a sinewave.
[/quote]
No, it>s ANY waveform whose components are within
the Nyquist bandwidth (i.e., < 1/2 sample rate) ANY
waveform whatsoever. |
|
| |
|
| Back to top |
Willem
Guest
|
| Posted: Fri Jul 11, 2008 8:20 pm Post subject: Re: Frequency/Sample rate |
|
|
dpierce.cartchunk.org@gmail.com wrote:
) On Jul 11, 8:24 am, zekfr...@zekfrivolous.com (GregS) wrote:
)> Not any waveform. Just a sinewave.
)
) No, it>s ANY waveform whose components are within
) the Nyquist bandwidth (i.e., < 1/2 sample rate) ANY
) waveform whatsoever.
And what are those components ? Sine waves, right ?
SaSW, Willem
--
Disclaimer: I am in no way responsible for any of the statements
made in the above text. For all I know I might be
drugged or something..
No I>m not paranoid. You all think I>m paranoid, don>t you !
#EOT |
|
| |
|
| Back to top |
Chronic Philharmonic
Guest
|
| Posted: Sat Jul 12, 2008 7:49 am Post subject: Re: Frequency/Sample rate |
|
|
"geoff" <geoff@nospam-paf.co.nz> wrote in message
news:i7-dna1M87pVpujVnZ2dnUVZ_gWdnZ2d@giganews.com...
[quote]Chronic Philharmonic wrote:
"geoff" <geoff@nospam-paf.co.nz> wrote in message
news:jdOdnSNoyt34c-_VnZ2dnUVZ_vadnZ2d@giganews.com...
Earl Kiosterud wrote:
I notice that Audacity displays sample points connected with
straight lines. I don>t think that>s meaningful. Other editors
show what a post-filter, presumably at 20 KHz, would show, along
with the sample points.
That would be very presumptive of editing software. SOundForge shows
straight lines between samples - I would be most peeved if SF made an
arbitrary decision about a filter.
If they>re drawing lines between samples, they>re making arbitrary
decisions about filters. They should either be properly smoothing the
waveform (they know the sample rate don>t they?), or showing a
histogram, in my opinion.
Naa , a line is important in case you miss a sample dot out of field of
veiw. Histogram just plain messy, and how thick would you make the columns
?
[/quote]
The columns on the editor I use are as thick as the dots that represent the
samples. There are a multitude of ways to represent off-screen data in the
UI that does not imply something that is misleading and mathematically
incorrect. I still maintain, if they are going to connect the samples with
lines, the line should reflect the smoothing filter that the sample rate
demands. |
|
| |
|
| Back to top |
Industrial One
Guest
|
| Posted: Sat Jul 12, 2008 2:58 pm Post subject: Re: Frequency/Sample rate |
|
|
On Jul 10, 6:56 pm, dpierce.cartchunk....@gmail.com wrote:
[SNIP]
Very eloquent explanation. However, anything spinning over 24 fps will
appear spinning backwards to the human eye anyway, cuz we don>t
perceive movement any more precise than that, just like no
neurotypical dude can hear over 20 KHz. |
|
| |
|
| Back to top |
Mark Nelson
Guest
|
| Posted: Sat Jul 12, 2008 4:09 pm Post subject: Re: Frequency/Sample rate |
|
|
On Jul 11, 3:20 pm, Willem <wil...@stack.nl> wrote:
[quote]
And what are those components ? Sine waves, right ?
[/quote]
You>re stretching it a bit there. A waveform is a waveform. It happens
that we can take an arbitrary waveform and decompose it into a set of
sine waves. But we can also decompose it into other types of
components. It>s not like sine waves are some atomic part of any
waveform. It just means that we have built up a body of analysis that
works very well on sine waves, and as a result, decomposing it that
way is convenient.
You can also decompose an arbitrary waveform into wavelets, and then
use different analysis tools. But if this leads to arguments about
whether waveforms are made up of wavelets or sine waves, well, it
shows that we>re missing the point.
- Mark |
|
| |
|
| Back to top |
Guest
|
| Posted: Sun Jul 13, 2008 9:05 pm Post subject: Re: Frequency/Sample rate |
|
|
On Jul 12, 10:58 am, Industrial One <industrial_...@hotmail.com>
wrote:
[quote]On Jul 10, 6:56 pm, dpierce.cartchunk....@gmail.com wrote:
[SNIP]
Very eloquent explanation. However, anything spinning over 24 fps will
appear spinning backwards to the human eye anyway, cuz we don>t
perceive movement any more precise than that, just like no
neurotypical dude can hear over 20 KHz.
[/quote]
No, we do NOT perceive them as "spinning backwards,"
we perceive them as blurred.
There>s another phenomenon called "aperture error" or
"aperture filtering" which, in the case of the eye, is due to
the so-called "persistance of vision." It, in effect, "smears"
the continuous input of images.
Now, interestingly enough, as a complete analog
of my wagon-wheel gadanken, how might we prevent
the aliasing of the wagon wheel motion, given that we>re
limited to, oh, 25 frames per second?
Well, you "low-pass filter" the wagon wheel, i.e.,
blur its motion, before sampling, so that nothing
makes it through to the "sampler (the shutter"
any faster than 1/2 the shutter or frame rate
That>s a little tough to actually accomplish in the
real world. "Aperture filtering" is one way to get
part of the way there: you make your shutter speed
as long as possible. IN the simplest case, it can>t
be any longer than 1/25 of a second, which would
have the effect of blurring things enough such that
things goign faster that 1/25 of a second are blurred
wnough that there>s nothuing left to alias, but doesn>t
solve the problem of what happens between < 12.5
and 25/second.
One might imaging a somewhat more sophisticated'
movie camera that exposes each frame for, say 1/12
second, but interleaves frames such that halway
through frame 1, it starts exposing frames to, and
frame 1 continues unitl halway through frame 2, at
which point frame 3 starts getting esposed, etc. |
|
| |
|
| Back to top |
|
