Welcome to Part 4 of my series about FFT analyzers. If you haven't seen parts 1, 2, and 3
already go check those out first. There are links in the description
below. In Part 3 of this series I mentioned that an FFT analyzer will
perform and display many FFTs per second, in order to give us a simulated real
time measurement. Let's look at what that means again. Here's a digitally sampled
audio waveform flowing into our analyzer.
Let's imagine this waveform is one
second long. If our sample rate is 48 kilohertz then this waveform contains
48 thousand samples. If we set the FFT size to "16k", then each measurement performed
uses about 16,000 of those samples, which span about a third of a second. You'll
recall from Part 3 of the series that the portion of the waveform used by
the FFT is called the time window, and its length is called the time constant.
So if our time constant is a third of a second, let's break this 1-second
waveform into three time windows - one third of a second each, right? But if the
analyzer only showed us a new measurement every one third of a second
it would look choppy and it would be difficult to read. Rather than perform a
measurement three times per second, most analyzers will perform a measurement at
a rate closer to the frame rate of film. Rational Acoustics' Smaart, for instance,
works at 24 measurements per second. 24 frames per second is fast enough to look
continuous to the human eye, rather than looking like a bunch of still images
shown in sequence. Even if our analyzer performs a new measurement every
twenty-fourth of a second it can still use a time constant longer than the time
between measurements. Each consecutive time window will just have overlapping
data. Now let's talk about averaging. In the scenario I've just described, we
measure a 16,000-sample time window every twenty-fourth of a second, and as
time passes we only view data from the most recent measurement. When we use
averaging, we're choosing to view the average of several recent measurements,
rather than only the single most recent one. In Smaart there are four different
averaging modes: [OFF] or [No Averaging], [FIFO], or "first-in first-out",
[Integrated Averaging], and [Infinite Averaging]. "First-in first-out" uses a buffer. When
the buffer is full each new measurement replaces the oldest one in the buffer.
In this example, [4 FIFO] is selected. As the fifth or newest measurement enters
the buffer, the first one (or the oldest one) is pushed out. Smaart has options to
do this with 2, 4, 8, or 16 measurements in the buffer. When viewing
a FIFO averager in Smaart, you can see that the motion of the RTA is kind of
jerky, as old data leaves the averaging buffer all at once.
By comparison, integrated averaging looks like this. Old data decays, meaning that
as time passes, the oldest data in the buffer is given less weight than the new
data. As a result, the bars in the RTA fall more gradually, instead of dropping
suddenly like they do in FIFO mode. Infinite averaging has an infinite
buffer. Old data does not decay, and it stays in the buffer forever. In Smaart
there are also [fast] and [slow] averaging options. These are integrated averaging
methods, just like the one-second through ten-second options, but they use the fast
and slow decay times that are found in SPL meters. Fast mode has a buffer of
slightly less than one second. That's averaging in a nutshell. It can be used
in spectrum measurements and transfer function measurements and it has
benefits in both cases. I'll explain that in detail when I discuss transfer
functions and coherence in later videos. One last note: when we choose to run a
16k FFT, it's not literally 16,000 samples. It's actually 16,384 samples. I
just rounded that to 16,000 to simplify the explanation. If that doesn't make
sense, you can find an explanation in Part 3 of this video series.
Thanks for watching! Please leave your suggestions, thoughts, and questions, in
the comment section below. If you learned something, hit the like button and
subscribe if you want to learn more! See you in the next one.
For more infomation >> John's Vision of the Great Harlot - Duration: 1:15:40. 
For more infomation >> Premio doble para el Barcelona si empata ante la Juventus - Duration: 2:46. 
For more infomation >> GetResponse tutoriel en français : Autorépondeur et Newsletter - Duration: 22:37. 
For more infomation >> CO2 Gun vs Cola bottles Full with water (Is it powerful enough?) - Duration: 4:07. 
For more infomation >> Precisamos falar de igualdade, diz Miss Brasil no Dia da Consciência Negra - Duration: 2:35. 
For more infomation >> Caged mostra abertura de 1600 novas vagas de emprego no Piauí em Outubro - Duration: 3:59. 
For more infomation >> Voici comment se débarrasser des oignons de pieds naturellement|LSF TV - Duration: 6:41. 
For more infomation >> Featuring Musician Megan Ni...
For more infomation >> Woman With 'F*ck Trump' Sticker Attacks Texas Sheriff With Sick New Decal Days Later - Duration: 4:08. 
For more infomation >> Seat Arosa 1.4 44KW Stella - Duration: 0:56. 
No comments:
Post a Comment