## iRoom goes 'public beta'.

Post and discuss acoustic topics, Studio design, construction, and soundproofing here
Jonessy.

Do I understand this correct?
You use the 1/3 octave center frequencies as band boundaries?
If so this is principly wrong.

Octave or fractional octave bands are expressed as CENTER frequencies of the bands they represent.
And it are Nominal frequencies (for convinience emphiric rounded numbers).
The exact frequencies result from a log 10 scale.

Or do I misunderstand you here?
Eric Desart
Moderator

Posts: 1174
Joined: Tue Dec 13, 2005 10:04 am
Location: Antwerp/Belgium

Jonessy,

The mathematical correct manner to assign arbitrary frequencies to a correct corresponding 1/3 octave band, is based on using the scale the official octave band scale is build upon.

+ + + + + + + + + + + + + + + + + + + ++ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +

Some background

In fact the official scale is a 1/3 octave band scale where 10 1/3 octave bands are 1 decade.
Therefore you will notice that the frequencies 0.1, 1, 10, 100, 1000, 10000, 10000 are exact numbers.
All frequency numbers in between have decimals.

It is for practical purposes, and in analogy with music scales, that we speak about octave bands (doubling/halving of frequencies) and assume that a 1/3 octave band is really a 1/3 of that octave band.
Both scales are about, but not complete equal.

Mathematically however based on the log 10 scale this isn't correct.
A 1/3 octave band = a 1/10 decade.
A 1/1 octave band is in fact a 3/10 decade.

Hence here the saying: "If it quacks as a duck and walks as a duck, it is a duck", isn't true at all in this case.

The 1/1 octave band is not a real mathematical unit in the official octave band scale, the decade is (e.g. from 10 to 100, 100 to 1000, and so on). All other official (non-music) frequency scales are mathematically calculated versus this decade, but expressed versus the popular octave.
To avoid confusion and in analogy with music scales, one never or hardly ever hears about a decade in relation to those frequency scales (which, from a practical point of view, is only interesting, but necessary to comprehend, for people being busy with the mathematical side of acoustics).

Hence the musical scale diverts from the octaveband scale since the musical scale, as eg the equal tempered scale is really a log scale based on logarithms with base 2 not base 10.
Even the Official Standards make a chaos from that.
At one hand they define the log base 10 scale as the base to define the exact center frequencies.
At the other hand they define the cutoff frequencies to be defiined on a log base 2 scale.

Professional equipment as B&K makes this simple: they define cutoff frequencies at exact the same principle as they define the center frequencies. This fits the main standard defining frequency scales on a log10 base, and they ignore illogical contradicting info.

Hence the frequency scale in 1/3 octave is calculated from: 10^(n/10) where n is an integer number and the integer 0 always results in 1 Hz (independent of the fractional octave band scale).
Resulting from this, the logical cutoff frequencies between those bands are:
Lower cutoff: 10^((n-0.5)/10)
Upper Cutoff: 10^((n+0.5)/10)
But since the "Upper cutoff "equals the "Lower cutoff "of the next band you only need to calculate one of them.

+ + + + + + + + + + + + + + + + + + + ++ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +

Calculation to see to which band an arbitrary frequency belongs:

To find the EXACT bandnumber of an arbitrary frequency in a 1/3 octave band scale:
Band number exact = 10*LOG(f) ; This gives a fractional result.
To get them in the exact band expressed as an integer:
Band number integer = ROUND(10*LOG(f),0) ; This gives you the band number representing the corresponding nominal 1/3 octave center frequency.
It can't get easier than this. One doesn't need to match frequencies: style: if higher or equal to x and lower than y, then ....
Go to the math base frequency scales are build upon and you get the mathematical correct band number just with a simple standard rounding function.
I use MS Excel notation here hence the LOG is base 10, and the ROUND rounds to the integer.

Then match the band number with the corresponding nominal frequency (INDEX or MATCH function).
One could calculate this mathematically, 10^(ROUND(10*LOG(f),0)/10) but then you get the exact center frequency returned not the nominal frequency, which are pragmatic rounded numbers for convinience.
Normally you want to present the commonly recognized Nominal frequencies.

Those band numbers in 1/3 octave go from 16 Hz to 20000 Hz from Band n° 12 to Band n° 43.
You can extend this of course up and downwards and those integers can even be negative resulting in the corresponding frequencies > 0Hz and <1 Hz  (0 = 1 Hz).

Hopes this helps

Best Regards
Eric Desart
Last edited by Eric Desart on Mon Apr 24, 2006 7:59 am, edited 17 times in total.
Eric Desart
Moderator

Posts: 1174
Joined: Tue Dec 13, 2005 10:04 am
Location: Antwerp/Belgium

Eric,

You are my KING for the day.  :)
I'm well aware of what you said, and I actually teach every year the common confusion between center and sidebands...
I actually thought of that while typing my previous message, but wanted to wait until I come home and figure out whether I was stupid enough to do that.
I don't know for what reason in the world I implemented a wrong algorithm in my programme.
For that, I am truely ashamed.  :?  :oops:

So THANK-YOU for your promptness. This will be taken care of ASAP.  :)

Regards,

Jonathan.
jonessy

Posts: 580
Joined: Wed Jan 04, 2006 3:39 pm
Location: Israel

J.F.Oros wrote:Excellent ! We're making history here  :D  :mrgreen:

Well, it's time to make history, my friend...

I'm working day & night to improve iRoom, many bugs were already fixed and I'm adding many new and cool features.

So I'd like to work on a module that correlates calculated modes on a LF graph.
Here's the concept:
1. The user produces a LF response graph of the room from a measurement software as instructed (20 to 300 Hz sounds reasonable to me).
2. iRoom imports the JPEG or whatever file format it is.
3. iRoom inserts the modal distribution on the JPEG.
4. The user can look for correlations between actual LF behaviour and the estimated physical analysis - to try and identify whether certain problems are modal.

In order for iRoom to 'integrate' with popular measurement software, I'm going to need some files to play around with.
I am using a programme called EASERA to do measurements, but I will be happy if anyone that has other software will produce LF graphs.
This will allow iRoom to integrate with more software.

So AFAIK these are the most common measurement programmes:
MLSSA
SmaartLive / AcousticTools
Dirac
WinMLS
ETF
SampleChampion

Anything else comes to mind?
Contributions are always welcome.
jonessy

Posts: 580
Joined: Wed Jan 04, 2006 3:39 pm
Location: Israel

Hey, this is great news, I can't wait for the uppgrade  :D

Don't forget about Room EQ wizard, it's the latest "must have" over here, considering it's also free  :D
And you can probably talk directly with John Mulcahy (the designer) to make this integration better (he also posted here).
... studiOTipper ...
J.F.Oros

Posts: 1870
Joined: Mon Feb 09, 2004 11:22 pm
Location: Romania

jonessy wrote:So AFAIK these are the most common measurement programmes:
MLSSA
SmaartLive / AcousticTools
Dirac
WinMLS
ETF
SampleChampion

Anything else comes to mind?
Contributions are always welcome.

http://www.ramsete.com/aurora/homepage.html
If you view life with the knowledge that there are no problems, only opportunities, you are a marketing manager.......this is my personal philosophy
bert stoltenborg

Posts: 4574
Joined: Sun Apr 18, 2004 11:03 am
Location: Achterhood, Netherlands

http://www.ymec.com/products/dssf3e/
Feels as real quality stuff to me, but not very well documented (but lots on the site if searching hard enough).
Eric Desart
Moderator

Posts: 1174
Joined: Tue Dec 13, 2005 10:04 am
Location: Antwerp/Belgium

Funny you mentioned Ramsete\Aurora, by Prof. Farina.
He wrote quite a few fascinating papers on concert-hall acoustics, and was one of the scientists behind Waves' convolution reverb.
I got acquainted with his work, when I wrote a white paper for Waves on implementation of IR measurement techniques.

Eric Desart wrote:http://www.ymec.com/products/dssf3e/
Feels as real quality stuff to me, but not very well documented (but lots on the site if searching hard enough).

It definately looks good. I'll see if I can get a trial version.

Everyone -
If any of you have access to the so-far-discussed programmes and are willing to generate LF response graphs, this would be very helpful.
jonessy

Posts: 580
Joined: Wed Jan 04, 2006 3:39 pm
Location: Israel

Is my impression that a lot of these new measurement programs use uncorrelated stimuli like noise correct?
If you view life with the knowledge that there are no problems, only opportunities, you are a marketing manager.......this is my personal philosophy
bert stoltenborg

Posts: 4574
Joined: Sun Apr 18, 2004 11:03 am
Location: Achterhood, Netherlands

Bert -

It seems that every developer has his own reasons for implementing a certain excitation signal.
About a year ago, a certain scientist argued that log-sweeps are the B-E-S-T.
I asked him why, and as a response I got a 2.5 hours lecture in post-doctorate level mathematics and astro-physics, explaining reliability of cross correlation functions.
Still, many respectful acousticians like to implement pseudo-random signals (MLS) or random noise.

The measurement software I use, for example, allows you to excite the room with just about ANY stimulus heard-of, including the option to import your own stimulae.
One of its cool features, is the "Excitation Wizard" that helps you choose the correct signal for the job. The user characterizes the measurements (type, room size, target T60, background noise, etc...) and the programme recommends the most suitable signal for the job.
jonessy

Posts: 580
Joined: Wed Jan 04, 2006 3:39 pm
Location: Israel

jonessy wrote:Everyone - If any of you have access to the so-far-discussed programmes and are willing to generate LF response graphs, this would be very helpful.

LF plots for EASERA, ETF, RoomEQWizard and WinMLS.

All the measurements are made for the same spot (the mic and speaker were not moved between changing programs), but they are not all level calibrated, so the Y axis values are not the same. But still it can be observed a similarity in the frequency responses.

OBS:
-ETF doesn't alow setting 20 to 300 Hz, only 20 to 200 or 20 to 2K, so I chosed the former.
-At this range (20 to 300) WinMLS by default doesn't show the frequency grids (only one grid at 100 Hz), so I had to alter the chart settings (hit F6) to show 20 Hz divisions.

I hope this helps you.
Attachments
LFplots.rar
LF plots
... studiOTipper ...
J.F.Oros

Posts: 1870
Joined: Mon Feb 09, 2004 11:22 pm
Location: Romania

jonessy wrote:Bert -

It seems that every developer has his own reasons for implementing a certain excitation signal.
About a year ago, a certain scientist argued that log-sweeps are the B-E-S-T.
I asked him why, and as a response I got a 2.5 hours lecture in post-doctorate level mathematics and astro-physics, explaining reliability of cross correlation functions.
Still, many respectful acousticians like to implement pseudo-random signals (MLS) or random noise.

The measurement software I use, for example, allows you to excite the room with just about ANY stimulus heard-of, including the option to import your own stimulae.
One of its cool features, is the "Excitation Wizard" that helps you choose the correct signal for the job. The user characterizes the measurements (type, room size, target T60, background noise, etc...) and the programme recommends the most suitable signal for the job.

MLS is a train of pulses encoded by the maximum length sequence.
The computer wants to see this sequence back in the measured response. When there is variation in this response, fe caused by wind or a delay system or something, your measurement sucks.
Sweeps are more immune to that, until the day someone starts moaning about them. In  fact both system are  good as long as used properly.

I cannot see how you derive time information from random noise, that's what startles me.
Must be a lack of imagination, maybe I should take up smoking wiet or something.

bert stoltenborg

Posts: 4574
Joined: Sun Apr 18, 2004 11:03 am
Location: Achterhood, Netherlands

bert stoltenborg wrote:I cannot see how you derive time information from random noise, that's what startles me.
They probably use the random noise (white, pink, brown, rainbowish, etc) for RTA functions, where there is no impulse/time info outputed but realtime level info.
EASERA has a really neat RTA function, I'm gonna try to use it on some TL measurements  :)

Must be a lack of imagination, maybe I should take up smoking wiet or something.
Common, stop saying that, people might get a false impression of you  :mrgreen:
... studiOTipper ...
J.F.Oros

Posts: 1870
Joined: Mon Feb 09, 2004 11:22 pm
Location: Romania

bert stoltenborg wrote:I cannot see how you derive time information from random noise, that's what startles me.
Must be a lack of imagination, maybe I should take up smoking wiet or something.

Yes, I definitely recommend you take up smoking weed!  But regardless, here's a way to understand it.

Random noise isn't truly random, in the sense that there is always a waveform with some specific time pattern. If you record such noise after a room "processes" it, that pattern will still be present in some form.

How do you find such a pattern? By cross-correlating the original noise signal and the room-processed noise signal. There will be maximum correlation at a certain time delay between the original signal and the room-processed signal. This delay is in fact exactly the time that it takes the signal to go out the speakers and through the air to the microphone!

In practice, such explicit time-domain correlation isn't necessary. If you take the FFTs of the two signals, you can, through some mathematical jiggery-pokery (that's the technical term) automatically extract this information.

Regards,
Terry
Terry Montlick

Posts: 1143
Joined: Mon Apr 12, 2004 9:31 pm
Location: Rhode Island, USA

Terry Montlick wrote:In practice, such explicit time-domain correlation isn't necessary. If you take the FFTs of the two signals, you can, through some mathematical jiggery-pokery (that's the technical term) automatically extract this information.

Deconvolving the signal with it's inverse in the frequency domain? Relying on Fourier's convolution theorem?
And then doing an IDFT?

For MLS and TDS signals we perform time-domain cross-correlation, for sweeps we deconvolve.
But what's the process of extracting the IR from recorded noise?

----
Flaviu -
Thanks for the graphs!
I'll start working on it right away.
jonessy

Posts: 580
Joined: Wed Jan 04, 2006 3:39 pm
Location: Israel

jonessy wrote:For MLS and TDS signals we perform time-domain cross-correlation, for sweeps we deconvolve.
But what's the process of extracting the IR from recorded noise?

Deconvolve, just as for sweeps. You have to capture in stereo using a loopback cable. One channel has the original looped-back signal, whatever it is, and the other channel the room-processed signal. You can't extract the IR from noise unless you've got both the "before" and "after" signals.

Regards,
Terry
Terry Montlick

Posts: 1143
Joined: Mon Apr 12, 2004 9:31 pm
Location: Rhode Island, USA

Thanks Terry.
jonessy

Posts: 580
Joined: Wed Jan 04, 2006 3:39 pm
Location: Israel

***** PROGRESS REPORT ON THE NEW AND EXCITING iROOM *****

Flaviu - thanks for the graphs, they have been very useful.

Attached to this message is a quick screenshot of the new version, including a brand new GUI.

iRoom now has gone beyond a "mode calculator" only, and includes diffuser and absorber calculators (all types).
I promise I will make it available for download as soon as I get past some technical issues.

I'm looking for feedback on the following idea:

I've played around with the demo version of INSUL, which is very nice.
Since INSUL is probably the most expensive calculator ever, I thought of adding similar features to iRoom.
Now, it seems that estimating TL values shouldn't be THAT difficult, since most equations are pretty straight-forward and material data (mass, young's modulus, etc...) is freely available.
I need your feedback from a legal\moral point of view.
Although I'm eager to make a useful and FREEWARE acoustical tool, I have absolutely no intentions to piss anybody off.
From a copyright\fair-use\moral perspective, does anybody think that there would be something wrong with adding features pretty similar to INSUL?
** Given ofcourse that all collected data\formulae is based on available publications. No intention to 'steal' anything...

I'm really having a dillema here, and I really need help making up my mind.
Eric, as you are a copyright expert... Any thoughts on this?
Anybody else?
Attachments
iroomone-screen.jpg
jonessy

Posts: 580
Joined: Wed Jan 04, 2006 3:39 pm
Location: Israel

jonessy,

From tomorrow I will be out for a week or so.
Can I give this some thougt.

That GUI looks impressive.

Eric
Eric Desart
Moderator

Posts: 1174
Joined: Tue Dec 13, 2005 10:04 am
Location: Antwerp/Belgium

Thanks Eric.
I will be eagerly waiting to hear what you think...
jonessy

Posts: 580
Joined: Wed Jan 04, 2006 3:39 pm
Location: Israel

PreviousNext