Time-Domain, FFT, and Octave (and 1/3 Octave) band analysis

Started by David Reid February 19, 2004
Hi,

I'm looking to recreate an audio environment for a flight simulator.  I've
been supplied with a document that contains different flight conditions (eg,
plane at idle, plane during takeoff, climb from 15000 to 35000ft, etc) as
well as the location (in the plane) of the recorders for these conditions.
This document also contains an Octave band (or narrow band, depending on the
condition) analysis of each condition.  I'm somewhat familiar with signal
processing and frequency spectrum analysis, but this is the first time I've
encountered Octave band analysis.  What i would like to do is to create a
wave file based on this data and hopefully be able to play back the sound
this data represents (in any media player).

My question is on how to interpret the data.  The plots  are sound level
(dB) vs time (seconds), and there are plots for various center frequncies
(can you define this term?): 63Hz, 125Hz, 250Hz, 500Hz......up to 8000Hz. If
i were to determine the time domain equvalent of each plot, and generate
their sine waves, mix them and write the output to file, would that
essentially recreate the sound for a particular condition?  My intuition
tells me that there is more to it than that, as these center frequencies
don't seem to represent the entire frequency spectrum of a given condition.
Can I also be pointed to an online resource that may give me a better
understanding of Octave band analysis?  My background is Computer Engineer,
software, but I think i understand the fundamentals of signal processing (ie
a first course in signals and systems), so I don't want anything too basic,
but not too advanced either.  Any help is very welcome.

Thanks in advance.
David Reid


"David Reid" <no_spam@mechtronix.ca> wrote in message
news:ou4Zb.6610$w65.610499@news20.bellglobal.com...
> Hi, > > I'm looking to recreate an audio environment for a flight simulator. I've > been supplied with a document that contains different flight conditions
(eg,
> plane at idle, plane during takeoff, climb from 15000 to 35000ft, etc) as > well as the location (in the plane) of the recorders for these conditions. > This document also contains an Octave band (or narrow band, depending on
the
> condition) analysis of each condition. I'm somewhat familiar with signal > processing and frequency spectrum analysis, but this is the first time
I've
> encountered Octave band analysis. What i would like to do is to create a > wave file based on this data and hopefully be able to play back the sound > this data represents (in any media player).
David, I suggest you use Google Groups on comp.dsp - we've had a few discussions about how to do this sort of thing.
>What i would like to do is to create a > wave file based on this data and hopefully be able to play back the sound > this data represents (in any media player).
Hmmmmm.... the only thing you'd be able to do with octave band levels would be to generate noise at the proper levels in each band - and possibly further smooth the spectral density curve as part of the process. Any tonal information you wouldn't know / have .. in order to go backwards from octave levels. Is that what you want? Fred
Yes, that answers my question.  Thanks, I'll search comp.dsp for more info.

David Reid

"Fred Marshall" <fmarshallx@remove_the_x.acm.org> wrote in message
news:eeGdncAMZ6t6n6jdRVn-hQ@centurytel.net...
> > "David Reid" <no_spam@mechtronix.ca> wrote in message > news:ou4Zb.6610$w65.610499@news20.bellglobal.com... > > Hi, > > > > I'm looking to recreate an audio environment for a flight simulator.
I've
> > been supplied with a document that contains different flight conditions > (eg, > > plane at idle, plane during takeoff, climb from 15000 to 35000ft, etc)
as
> > well as the location (in the plane) of the recorders for these
conditions.
> > This document also contains an Octave band (or narrow band, depending on > the > > condition) analysis of each condition. I'm somewhat familiar with
signal
> > processing and frequency spectrum analysis, but this is the first time > I've > > encountered Octave band analysis. What i would like to do is to create
a
> > wave file based on this data and hopefully be able to play back the
sound
> > this data represents (in any media player). > > David, > > I suggest you use Google Groups on comp.dsp - we've had a few discussions > about how to do this sort of thing. > > >What i would like to do is to create a > > wave file based on this data and hopefully be able to play back the
sound
> > this data represents (in any media player). > > Hmmmmm.... the only thing you'd be able to do with octave band levels
would
> be to generate noise at the proper levels in each band - and possibly > further smooth the spectral density curve as part of the process. Any
tonal
> information you wouldn't know / have .. in order to go backwards from
octave
> levels. Is that what you want? > > Fred > >
To refine my question, based on your reply and new info....

I noticed at the end of the document that there is indeed the full spectrum
but for certain conditions only.  If i understand you correctly, if i took
one of the base conditions with the full spectrum (eg, engines at cruise at
5000ft), i could use the octave band plots (of lets say, climb from 5000ft
to 35000ft over 60 secs) to adjust the full spectrum at the different center
freq's of the octave plots to get the full spectrum of the latter condition?

I hope i made that clear...

David Reid
"Fred Marshall" <fmarshallx@remove_the_x.acm.org> wrote in message
news:eeGdncAMZ6t6n6jdRVn-hQ@centurytel.net...
> > "David Reid" <no_spam@mechtronix.ca> wrote in message > news:ou4Zb.6610$w65.610499@news20.bellglobal.com... > > Hi, > > > > I'm looking to recreate an audio environment for a flight simulator.
I've
> > been supplied with a document that contains different flight conditions > (eg, > > plane at idle, plane during takeoff, climb from 15000 to 35000ft, etc)
as
> > well as the location (in the plane) of the recorders for these
conditions.
> > This document also contains an Octave band (or narrow band, depending on > the > > condition) analysis of each condition. I'm somewhat familiar with
signal
> > processing and frequency spectrum analysis, but this is the first time > I've > > encountered Octave band analysis. What i would like to do is to create
a
> > wave file based on this data and hopefully be able to play back the
sound
> > this data represents (in any media player). > > David, > > I suggest you use Google Groups on comp.dsp - we've had a few discussions > about how to do this sort of thing. > > >What i would like to do is to create a > > wave file based on this data and hopefully be able to play back the
sound
> > this data represents (in any media player). > > Hmmmmm.... the only thing you'd be able to do with octave band levels
would
> be to generate noise at the proper levels in each band - and possibly > further smooth the spectral density curve as part of the process. Any
tonal
> information you wouldn't know / have .. in order to go backwards from
octave
> levels. Is that what you want? > > Fred > >
"David Reid" <no_spam@mechtronix.ca> wrote in message
news:1E8Zb.6857$w65.627280@news20.bellglobal.com...
> To refine my question, based on your reply and new info.... > > I noticed at the end of the document that there is indeed the full
spectrum
> but for certain conditions only. If i understand you correctly, if i took > one of the base conditions with the full spectrum (eg, engines at cruise
at
> 5000ft), i could use the octave band plots (of lets say, climb from 5000ft > to 35000ft over 60 secs) to adjust the full spectrum at the different
center
> freq's of the octave plots to get the full spectrum of the latter
condition?
> > I hope i made that clear...
David, I think it's a lot more complicated than that. I don't know how you "build" a sound from such little information unless it is almost purely colored (i.e. not flat spectral density) noise. If it's just colored broadband noise then maybe.... However, my experience is that noise often has considerable tonal and harmonic constituents of import. These you don't seem to have a model for. For example, the engines may increase in frequency (rpm) during climb but reduce once more after the climb. At the same time, pumps may turn off and on and off again and some pumps may simply run at the same speed all the time. etc. etc. Now *that's* complicated and probably untenable. I used to construct unending sound "loops" from actual recordings - and *that* was hard. Fred
Well that's more like what i'll be doing, building loops from recordings.  I
was just wondering about the theory behind what I previously mentioned.  The
sample model that you have mentioned is pretty much the kind of thing that
we'll be doing.  The document mentioned contains octave band analysis for
such a condition and others (too many to mention) and our goal is to build a
model based on all the data.  The final implemenation in the simulator (ie,
sounds coming out of all of the speakers) will be measured with a spectrum
analyser and the output has to match what was contained in the document.
But i think our problem is the lack of a dsp expert.  As things become
clearer, I'll post more questions.

Thanks for the replies, they're very helpful.

David Reid

"Fred Marshall" <fmarshallx@remove_the_x.acm.org> wrote in message
news:WZ-dnYrJkbFGP6jd4p2dnA@centurytel.net...
> > "David Reid" <no_spam@mechtronix.ca> wrote in message > news:1E8Zb.6857$w65.627280@news20.bellglobal.com... > > To refine my question, based on your reply and new info.... > > > > I noticed at the end of the document that there is indeed the full > spectrum > > but for certain conditions only. If i understand you correctly, if i
took
> > one of the base conditions with the full spectrum (eg, engines at cruise > at > > 5000ft), i could use the octave band plots (of lets say, climb from
5000ft
> > to 35000ft over 60 secs) to adjust the full spectrum at the different > center > > freq's of the octave plots to get the full spectrum of the latter > condition? > > > > I hope i made that clear... > > David, > > I think it's a lot more complicated than that. I don't know how you
"build"
> a sound from such little information unless it is almost purely colored > (i.e. not flat spectral density) noise. If it's just colored broadband > noise then maybe.... > > However, my experience is that noise often has considerable tonal and > harmonic constituents of import. These you don't seem to have a model
for.
> For example, the engines may increase in frequency (rpm) during climb but > reduce once more after the climb. At the same time, pumps may turn off
and
> on and off again and some pumps may simply run at the same speed all the > time. etc. etc. Now *that's* complicated and probably untenable. > > I used to construct unending sound "loops" from actual recordings - and > *that* was hard. > > Fred > >