Denon 5805 - primi dati disponibili

[Michele Spinolo]

audyssey ha scritto:
Hi Michele,

The advantage of longer FIR filters is that they allow better frequency resolution. So they can correct even very narrow peaks or dips in the frequency domain. You are correct, that with a smaller filter it is not normally possible correct very fine features.

However, with our method, the filter is not applied with uniform resolution in the frequency domain. This means it can correct better in the low frequencies than a "regular" 512 tap filter. This is an important part of our patent because, as you know, most problems in small rooms are in the low frequencies.

I would appreciate if you could translate in Italian for the other forum users. Unfortunately, although I speak Italian, I don't know all of the technical terms in Italian.

Regards,
Chris
CTO Audyssey Laboratories

Chris,

I was wondering if I could link here the PDF file you sent me with frequency response graphs before and after MultEQ correction measured in 6 different positions.
I think it could be very interesting!

Regarding my doubts they concern, as I wrote, phase and frequency response.
As you wrote on AVSForum, and I completly agree, the IIR filters problem is you have not any control on time domain (i.e. on phase), and if IIR filters are not properly designed frequency domain correction could leads to worse results dued to phase problems given by them.

I do not know if what I am going to write is correct, as I said I am only a digital room correction systems enthusiast: due to Gabor inequality you have df*dt>1/2, so practically speaking strong frequency corrections (i.e. trying to correct even very narrow dips or peaks) will leads to poor phase correction if FIR filter is not long enough.
So I was wondering how deep a 512taps non-uniform filter could correct both in phase and frequency domain, simply speaking could you give and approximation, in a standard case, for the taps relation between a uniform resolution FIR filters performance and a non-uniform FIR filters performance? (hope it is understandable, I know it's a quite confused period, but my english is quite poor!).

Thank you very much for the time you are spending here.

[CalabrOne]

L'ho trovata su avsforum.com. Per i pigri, la riporto qui

[audyssey]

Michele,

Yes, it will be fine to post the plots. I also think that the forum members will find them interesting.

In the near future we will be putting out some white papers with more technical details about our approach and I will send you a copy once we can release them.

Chris
CTO Audyssey Laboratories

[Michele Spinolo]

audyssey ha scritto:
Michele,

Yes, it will be fine to post the plots. I also think that the forum members will find them interesting.

In the near future we will be putting out some white papers with more technical details about our approach and I will send you a copy once we can release them.

Chris
CTO Audyssey Laboratories

Thank you Chris, so here is the link to the plots:

PLOTS

link ai grafici della risposta in frequenza, prima e dopo l'equalizzazione, misurata in 6 diverse posizioni:

GRAFICI

[Ninja]

se la misurazione interviene in luoghi più vicini tra loro, al limite distanti un metro da un ipotetico centro, si ha ancora miglioramento nell'eq?

Michele, cosa pensi dei grafici?

traduco per audissey

in the case your measurments are taken in closer places, possibly just far a few feet from an hypothetical center listening poing, will it improve eq?

[AlbertoPN]

CalabrOne ha scritto:
L'ho trovata su avsforum.com. Per i pigri, la riporto qui

Grazie mille !!!

Porca paletta quanta roba da collegare! Per fortuna i connettori casse accettano le banane ....

Grazie ancora!

Mandi!

Alberto

[AlbertoPN]

Michele Spinolo ha scritto:
Thank you Chris, so here is the link to the plots:

PLOTS

link ai grafici della risposta in frequenza, prima e dopo l'equalizzazione, misurata in 6 diverse posizioni:

GRAFICI

Sorry Michele,

but are they the same file ?

Will it be possible to know how far are these six positions from the "typical" listening point ?

Thanks. I don't think that it will be necessary to translate also this question, don't you?

Mandi!

Alberto

[Michele Spinolo]

AlbertoPN ha scritto:
Sorry Michele,

but are they the same file ?

Will it be possible to know how far are these six positions from the "typical" listening point ?

Thanks. I don't think that it will be necessary to translate also this question, don't you?

Mandi!

Alberto

yes Alberto, they are the same file but the link is translated

Chris asked me to translate what he wrote, he is very kind to spend his time here, so I try to be kind as well

For the other question I can't answer, let's wait for Chris!

[Michele Spinolo]

Ninja ha scritto:
se la misurazione interviene in luoghi più vicini tra loro, al limite distanti un metro da un ipotetico centro, si ha ancora miglioramento nell'eq?

Michele, cosa pensi dei grafici?

traduco per audissey

in the case your measurments are taken in closer places, possibly just far a few feet from an hypothetical center listening poing, will it improve eq?

As Chris wrote MultEQ uses a psyco-acoustical model to create FIR filters, do not really display what you hear.
For sure these approach is a lot better than YPAO or MCACC.

Let's say the frequency resolution is something you can easy obtain with a 1/3oct band, well designed, parametric filter.

[gian de bit]

Perche' invece di fare gli sboroni in inglese non postate in Italiano? cosi anche gli "gnurant" di lingue riescono a seguire l'interessante 3D? ...

Mica siamo su AVFSForum! :o

sorry e saluti a tutti

[luny74]

AlbertoPN ha scritto:

Non tutti possiamo avere il camice bianco e girare per casa con strumentazione varia per compensare le carenze dell'ambiente.

x audyssey

Hey! ben venuto nel bel paese! ...non ti scusare per l'italiano...lo parli meglio di me

X tutti

Ahò ma che state a dì? fate una "translation" altrimenti non ci si capisce un piffero

Salutescion

Luigi

[CalabrOne]

AlbertoPN ha scritto:
Grazie mille !!!

Per così poco...

[chromef]

Michele,
i link non funzionano.
Per tirare un po' le somme, quanti MIPS occorrono per un singolo canale per un trattamento "decente"? Sui prodotti consumer siamo ancora molto lontani da quel livello?

Ciao
Francesco

[Michele Spinolo]

chromef ha scritto:
Michele,
i link non funzionano.
Per tirare un po' le somme, quanti MIPS occorrono per un singolo canale per un trattamento "decente"? Sui prodotti consumer siamo ancora molto lontani da quel livello?

Ciao
Francesco

ciao Francesco,

in effetti ho notato che cliccando sul link non funziona, però se clicchi col destro e scegli "salva con nome" e salvi il documento PDF risolvi.

Comunque per far prima linko qui due immagini:

Prima della cura:



Dopo la cura:



Per quanto riguarda l'altra domanda non ho assolutamente idea delle potenze dei DSP commerciali, ne so quanta ne sia necessaria per effettuare una convoluzione con un filtro fa N-punti.
Ad ogni modo so per certo che dipende molto dall'algoritmo utilizzato, quindi è dura fare stime precise.

A dire il vero non ho neppure capito se la convoluzione nel MultEQ è fatta nel dominio del tempo o in quello della frequenza (molto molto più efficiente).
Mo' lo chiedo a Chris!

[Michele Spinolo]

audyssey ha scritto:
Michele,

Yes, it will be fine to post the plots. I also think that the forum members will find them interesting.

In the near future we will be putting out some white papers with more technical details about our approach and I will send you a copy once we can release them.

Chris
CTO Audyssey Laboratories

Chris,

regarding the latency issue, running at 44.1Khz, with 512taps, you have a about 11ms filter lenght: buffering twice the filter lenght you should have a 22ms audio to video delay, how did you solved the problem?
I assume the FIR convolution is made in frequency domain, becouse, as you know, performing a time domain convolution leads to no I/O delay while it has a really worse efficiency.