You can do this with Angelo's method too.
Do what ? Farina's proposed the exponentially sweeped sine technique does not randomize phase of external disturbance signal. Of course you can apply averaging to eliminate/reduce disturbance signal having 0 mean value (i.e. random noise). However, if the mean value of the background noise/disturbance sinal is not 0, then averaging of the sweep sine response will not help you reduce the background noise in the sweep sine response.
Why don't you try it and see?However, if the mean value of the background noise/disturbance sinal is not 0, then averaging of the sweep sine response will not help you reduce the background noise in the sweep sine response.
In ALL cases, a Farina sweep measurement of the same length (and power) as your MLS signal, whether it's an extra long sweep, or several sweeps concatenated to give the same length (to be averaged), will give AT LEAST 3dB less noise in the IR than MLS and other methods and usually more.
Works with musical noise, banging, birds singing or systemic buzzing, hum, bla bla too. I can do good measurements when the rain on my tin roof is so loud I have to shout to speak to someone in the same room. You can also do the party trick of playing a Farina sweep at very low level during a musical performance and getting a good response without anyone noticing 😊
For da Signal Processing pedants, Angelo's method is a Matched Filter technique with a Synchronous Detector for Response & HD. Yes hum will show up in da HD measurement but less than with other methods. And averaging WILL reduce it.
Last edited:
Thanks for this TNT. I see my error now.Aha OK. Yes, I have noticed I think that the THD figure dont change depending on how the HDs are viewed... but should they really? Its not like the reality is different between the views
To go from DcibeL's "show harmonics at harmonic frequency" display to traditional HD ...
I need to shift the Red 2nd harm curve DOWN an 8ve and the Orange 3rd DOWN 3x in freq. etc. NOT UP. Then it all makes sense. It's now doing exactly what I was trying to describe in #60
Mea culpa, mea culpa, mea maxima culpa

Dammit! I now have to REALLY UNNERSTAN DIS FSAF S**T. It's gonna hurt my single remaining brain cell ba... ad !!
Got any comments about my post #118 ?
Last edited:
You could go to AI like chatGPT, upload the doc and ask it to summarize it to you like you were homer simpson needing help from the brain 🙂 For all kinds of work I find this opportunity highly valuable, to be able to query documents with natural language and reflect with the AI with it, which all helps to make connections between brain cells much faster than without.
edit. oops the document is quite big, perhaps free versions cannot handle it yet.
edit. yeah chatGPT paid version can read it and of you go. This stuff is expert domain though, so don't have too high hopes with it explaining it exaclty correct, but it might still be worth while to reflect upon, helps yourself think it through. Perhaps spot things you didn't notice yet and of you go.
edit. oops the document is quite big, perhaps free versions cannot handle it yet.
edit. yeah chatGPT paid version can read it and of you go. This stuff is expert domain though, so don't have too high hopes with it explaining it exaclty correct, but it might still be worth while to reflect upon, helps yourself think it through. Perhaps spot things you didn't notice yet and of you go.
Last edited:
should be A Farina sweep even 10x FASTER than your stepped sine will still have 10x or better noise immunity.No. If you use a longer Farina sweep, its noise immunity increases. A Farina sweep even 10x slower than your stepped sine will still have better than 10x noise immunity.
Member
Joined 2003
REW shows THD with respect to the fundamental regardless of whether the individual harmonics are shown at the fundamental frequency or the frequency of the harmonics. Not sure if that's a bug or intentional, but that's what it does. THD looks very normal and correct when harmonic display is the conventional method of showing harmonic levels at the fundamental frequency. Personally, I don't know why this is the conventional display, I find easier interpretation when the harmonics are shown at the frequency they occur.Thanks for this TNT.
But is THD(H2..9) supposed to be the summed power of the 2nd to 9th harmonic? If so, there is something wrong with this on DcibeL's HD curves.
Member
Joined 2003
I'm no mathematician, but perhaps you could get some me insight from John M if you asked over in the official REW support forum.DcibeL, Tranh, Kumar and any other DSP and MATLAB gurus here, I'm trying to figure out what the FSAF residual display is. I know I should be able to do this by inspection of the MATLAB code but it might take me the rest of the Millenium to grok something sensible.
Could you gurus please comment on my probably naive, understanding (guess?) of what Michael is doing.
- Use FSAF with arbitrary stimulus to produce a long IR which models the supposedly LTI transfer function of your DUT (including room responses bla bla) This produces an 'accurate' response which we know corresponds to other methods for IR & response. Loadsa buzzwords and TLAs which I don't unnerstan at present but we know it gets 'good' response.
- Convolve this IR with the stimulus to get what should be the perfect result if the transfer function is LTI (ie has no distortion, compression bla bla)
- Subtract the expected perfect result from the actual result to get the residual.
- Analyse the residual to get the display.
- But what are the analysis parameters to get this display?
So the subband adaptive filtering that Micheal discusses in his work combines two areas/subjects: a) Multirate signal processing and filter banks, b) Adaptive filter theoryIf you know of these papers, or any others which explain subband filtering accurately and simply in the public domain, preferably simple enough for my single brain cell, please post links.
There are several very well-known books/IEEE signal processing magazine papers on both subjects. The popular books that were used during my graduate graduate studies were the following:
1) Multirate signal processing: https://authors.library.caltech.edu/records/rmhds-22q28
2) Adaptive signal processing: https://books.google.co.in/books/about/Adaptive_Filter_Theory.html?id=MdDi_PF7gMsC&redir_esc=y
If someone with some basic signal processing knowledge wants to get a quick summary about the basics of relevant multi-rate signal processing concepts, some important adaptive filter designs, and thereafter understand some basics of subband adaptive filtering, I would suggest doing what Michael suggests in his article, i.e., read Prof M. Moonen's articles or his students' thesis. I particularly liked this one: https://theses.eurasip.org/theses/39/subband-and-frequency-domain-adaptive-filtering/download/
Tjis was not a so smart statement... if one can show individual in two different views, on can of course view the total in the two corresponding views... as such, the REW presentation is strange as in one view, the THD presentation don't match the harmonics presentation.Its not like the reality is different between the views 🙂
But again, one is what is heard, the other is where it is caused.
//
The way how harmonics are displayed as default in REW is the way we are used to look at them. Most likely stated in some DIN/ASA/IAEE standard.
https://en.wikipedia.org/wiki/Total_harmonic_distortion#THD.2BN
Stimulus (fundamental) gets disturbed by harmonics of several orders which modulate the sound we hear or measure (look at waveform of sine getting ripples). This is easy to understand (or measure) for pure steady sine wave signal. The problem is that music contains wide spectrum of frequencies whose amplitude and duration varies all the time (for test use eg. 32-multitone IMD)
Another problem is the unlinearity of a loudspeaker's distortion behaviour through the spectrum. Proportions of different orders of D and their amplitudes can change drastically in fraction of an octave.
So, I propose using more multitone (IMD) tests for loudspeakers. We can see them now only in some magazine reviews and with Klippel NFS tests. eg. Erin's
We also need more time and experience learning to evaluate them.
REW's tone generator has many multitone signals!
https://en.wikipedia.org/wiki/Total_harmonic_distortion#THD.2BN
Stimulus (fundamental) gets disturbed by harmonics of several orders which modulate the sound we hear or measure (look at waveform of sine getting ripples). This is easy to understand (or measure) for pure steady sine wave signal. The problem is that music contains wide spectrum of frequencies whose amplitude and duration varies all the time (for test use eg. 32-multitone IMD)
Another problem is the unlinearity of a loudspeaker's distortion behaviour through the spectrum. Proportions of different orders of D and their amplitudes can change drastically in fraction of an octave.
So, I propose using more multitone (IMD) tests for loudspeakers. We can see them now only in some magazine reviews and with Klippel NFS tests. eg. Erin's
We also need more time and experience learning to evaluate them.
REW's tone generator has many multitone signals!
Yes.
- Convolve this IR with the stimulus to get what should be the perfect result if the transfer function is LTI (ie has no distortion, compression bla bla)
- Subtract the expected perfect result from the actual result to get the residual.
It's just a time domain signal, processing is FFT, more specifically the Hann-windowed Welch power spectrum.But what are the analysis parameters to get this display?
This is how HD is displayed on B&K toilet roll when you use their 2010 + 1902 + 2307, distortion measurement as God intended. 😊THD looks very normal and correct when harmonic display is the conventional method of showing harmonic levels at the fundamental frequency. Personally, I don't know why this is the conventional display,
Thanks to you, DcibeL, Tranh, TNT, Kumar and other illustrious gurus on this august forum, I'm finally beginning to grok Michael's FSAF stuff. My post #118 stuff is some of the remaining detail which I hope will explain Michael's residual and put in its correct place in the pantheon of iconic measuring sticks.
We are nearly there 😊
Thanks for this JohnPM, the father & mother of REW, 😲It's just a time domain signal, processing is FFT, more specifically the Hann-windowed Welch power spectrum.
I must now get to grips with REW properly when I've avoided doing so in the past cos I preferred my own Jurassic software. But my stuff needs a Windows XP machine to run and these are nearly extinct.
May I beg an indulgence that you provide the faciltiy for REW to print stuff on B&K toilet roll (or at least the same size & aspect ratio) so I can compare curves by the time honoured method of holding two pieces of aligned toilet roll to the light 😊
There's loadsa work for me to do on this subject so I will be back shortly.
But we are now into the good weather on the Great Barrier Reef has started and my beach bum commitments call.
You can choose an IEC 263 dB/decade aspect ratio on the image capture dialog.provide the faciltiy for REW to print stuff on B&K toilet roll (or at least the same size & aspect ratio)
🙂👍The easiest to interpret is FSAF residual on real music. You don't need to interpret it, you simply listen to it and subjectively judge the "annoyance" factor vs another driver.
I do not think that Mike is coming back any time soon.Thats unfortunate... come back Mike!
//
He was a angry guy and Teed Off on another member.
My guess is that he was Deleted.
Being able to isolate "distortion" is the stuff of error correction we are seeing in the new generation of amplification.
Thanks DT
- Home
- Loudspeakers
- Multi-Way
- Measuring speakers using music (Michael Tsiroulnikov's FSAF)