• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

Dirac > Audyssey XT32 ... Sure, Always?

QMuse

Major Contributor
Joined
Feb 20, 2020
Messages
3,124
Likes
2,785
About 3ms.

I'm unaware of low latency modes, but as it sits, it's probably low enough for most purposes.

I'm affraid it's much more than that..

Dirac uses mix of IIR anf FIR filters. With FIR filters delay depends on the number of taps, 8192 taps equals app 93ms of delay. Double number of taps and delay doubles. For a decently precise 44.1kHz filter you need 16384 taps so 186ms would be minimum I would count with.
 

SIY

Grand Contributor
Technical Expert
Joined
Apr 6, 2018
Messages
10,383
Likes
24,749
Location
Alfred, NY
I'm confused.
It seems too low to correct with FIR issues below 300hz (about 3ms between pulses/waves)… even with signal processing tricks...

This was what the measurements show for impulses between Dirac "on" and Dirac "off." Now things could be complicated by latency in the hardware (I don't have DL as stand-alone software), but presumably that's constant between the two modes.
 

QMuse

Major Contributor
Joined
Feb 20, 2020
Messages
3,124
Likes
2,785
This was what the measurements show for impulses between Dirac "on" and Dirac "off." Now things could be complicated by latency in the hardware (I don't have DL as stand-alone software), but presumably that's constant between the two modes.

Did you measure it so that the speaker which was playing timing reference signal was in both cases with Dirac "off" while the other which you measured was "on" in one measurement and "off" in other?
 
OP
G

Gedeon

Active Member
Forum Donor
Joined
Oct 30, 2018
Messages
191
Likes
195
I'm affraid it's much more than that..

Dirac uses mix of IIR anf FIR filters. With FIR filters delay depends on the number of taps, 8192 taps equals app 93ms of delay. Double number of taps and delay doubles. For a decently precise 44.1kHz filter you need 16384 taps so 186ms would be minimum I would count with.
So 2048 taps should be enough to reach around 24hz. Audyssey XT32 claims to use 512 resolution filter which seems to fall short in order to work around below 90hz (if those 512 were taps). Strange, because subwoofer is usually a strong point in these EQ's.
 

SIY

Grand Contributor
Technical Expert
Joined
Apr 6, 2018
Messages
10,383
Likes
24,749
Location
Alfred, NY
Did you measure it so that the speaker which was playing timing reference signal was in both cases with Dirac "off" while the other which you measured was "on" in one measurement and "off" in other?

Neither. I ran the impulse in stereo to both channels with the mike at listening position, and compared the impulse time with Dirac on and Dirac off. This was done several times with several different EQ settings with the same difference in time delay measured.
 

QMuse

Major Contributor
Joined
Feb 20, 2020
Messages
3,124
Likes
2,785
So 2048 taps should be enough to reach around 24hz. Audyssey XT32 claims to use 512 resolution filter which seems to fall short in order to work around below 90hz (if those 512 were taps). Strange, because subwoofer is usually a strong point in these EQ's.

# of taps is not related to LF reach but to the precision of the filter curve. 512 taps might be enough for a filter that only does the job below 90Hz.

512 taps is app 5.8ms of delay.


Neither. I ran the impulse in stereo to both channels with the mike at listening position, and compared the impulse time with Dirac on and Dirac off. This was done several times with several different EQ settings with the same difference in time delay measured.

That way you don't have a proper timing reference to make comparison. I suggest you do it the way I described as that way your (say) right speaker will always be playing the same timing reference with Dirac off and left speaker would play sweep with Dirac on and off.
 

SIY

Grand Contributor
Technical Expert
Joined
Apr 6, 2018
Messages
10,383
Likes
24,749
Location
Alfred, NY
That way you don't have a proper timing reference to make comparison. I suggest you do it the way I described as that way your (say) right speaker will always be playing the same timing reference with Dirac off and left speaker would play sweep with Dirac on and off.

The timing reference is the impulse with Dirac off. I am failing to see why measuring the impulse arrival time with Dirac off then comparing to Dirac on is not proper. I can do the same thing electrically, if you think there's something special about that.
 
OP
G

Gedeon

Active Member
Forum Donor
Joined
Oct 30, 2018
Messages
191
Likes
195
# of taps is not related to LF reach but to the precision of the filter curve. 512 taps might be enough for a filter that only does the job below 90Hz.

512 taps is app 5.8ms of delay.




That way you don't have a proper timing reference to make comparison. I suggest you do it the way I described as that way your (say) right speaker will always be playing the same timing reference with Dirac off and left speaker would play sweep with Dirac on and off.


I understand that 512 could be enough when dealing with signal limited to, let's say 250hz max. But I don't fully understand those 5.8 ms delay for 512 taps unless dealing with 96khz sample rate. Am I wrong ?
Thanks.
 

QMuse

Major Contributor
Joined
Feb 20, 2020
Messages
3,124
Likes
2,785
The timing reference is the impulse with Dirac off. I am failing to see why measuring the impulse arrival time with Dirac off then comparing to Dirac on is not proper. I can do the same thing electrically, if you think there's something special about that.

I am here assuming you were doing this measurement with REW. When I said "timing reference" I was referring to the timing reference signal at the beginning of the measurement sweep. If you want to measure timing difference in a precise manner with REW you have to assure that in both measurements timing refernce signal is the same hence my suggesstion to make one speaker play it while measuring the other speaker.
 

SIY

Grand Contributor
Technical Expert
Joined
Apr 6, 2018
Messages
10,383
Likes
24,749
Location
Alfred, NY
I am here assuming you were doing this measurement with REW.

No, I use AP, specifically an APx525 with an APx1701 acoustic interface.

edit: I should mention that I use a 192kHz sample rate.
 

QMuse

Major Contributor
Joined
Feb 20, 2020
Messages
3,124
Likes
2,785
I understand that 512 could be enough when dealing with signal limited to, let's say 250hz max. But I don't fully understand those 5.8 ms delay for 512 taps unless dealing with 96khz sample rate. Am I wrong ?
Thanks.

Sorry, I forgot to mention that 5.8ms delay with 512 taps is with 44.1kHz rate.

With FIR filters delay is a function of number of taps and rate.
 
OP
G

Gedeon

Active Member
Forum Donor
Joined
Oct 30, 2018
Messages
191
Likes
195
Sorry, I forgot to mention that 5.8ms delay with 512 taps is with 44.1kHz rate.

With FIR filters delay is a function of number of taps and rate.
How do you reach that 5.8ms latency from 44100hz and 512 taps ?

Forget it. I thought that it was taps/fs, for linear it is (taps-1)/(2*fs)

So yes, 512 for 44100 is about 5.8
 
Last edited:

sophie smith

Active Member
Joined
Mar 3, 2018
Messages
132
Likes
65
Not to hijack the technical discussion. Few simple questions. Setting up 5.1 HT for movies. Wanted to try room EQ. Few options.

1. Audyssey XT 32 with Denon AVR. All-in-one simple solution, but not easy to upgrade.
2. Other AVR, such as pioneer or Yamaha with EQ software. Anthem is too pricey for me. So is Bryston and new Emotiva/Monoprice TOTL processors that include Dirac.
3. Dirac live. The only hardware that seems affordable is miniDSP. Would need 8 channel (I think) for the HT. But this option would possibly allow for the latersoftware upgrades with future versions for not very much ($100 now I think).
A. Can miniDSP be used as a preamp if DAC included in it to drive 5 channel amp? Or do I still need a processor?
B. Can EQ software be used with other formats (DTS....) if miniDSP used together with dedicated processor? Should I use anything else for movies if Dirac is used? In other words, can (and should) EQ software be combined with DTS, Dolby, or other sound processing formats used in receivers? Or is it one or the other?
4. Any other cost effective options?

Thanks
 

laurelkurt

Active Member
Forum Donor
Joined
Dec 7, 2019
Messages
130
Likes
59
Location
Rochester, NY
Audyssey does not aim for flat. It creates a downward slope. Audyssey flat is an optional setting. I've never heard Audyssey sound bass light.
Same here. Certain recordings sound like they may have less mid/upper bass. Thin sounding recordings often sound better with Audyssey turned off. Recordings with ample bass usually sound better with it on.
 
OP
G

Gedeon

Active Member
Forum Donor
Joined
Oct 30, 2018
Messages
191
Likes
195
Not to hijack the technical discussion. Few simple questions. Setting up 5.1 HT for movies. Wanted to try room EQ. Few options.

1. Audyssey XT 32 with Denon AVR. All-in-one simple solution, but not easy to upgrade.
2. Other AVR, such as pioneer or Yamaha with EQ software. Anthem is too pricey for me. So is Bryston and new Emotiva/Monoprice TOTL processors that include Dirac.
3. Dirac live. The only hardware that seems affordable is miniDSP. Would need 8 channel (I think) for the HT. But this option would possibly allow for the latersoftware upgrades with future versions for not very much ($100 now I think).
A. Can miniDSP be used as a preamp if DAC included in it to drive 5 channel amp? Or do I still need a processor?
B. Can EQ software be used with other formats (DTS....) if miniDSP used together with dedicated processor? Should I use anything else for movies if Dirac is used? In other words, can (and should) EQ software be combined with DTS, Dolby, or other sound processing formats used in receivers? Or is it one or the other?
4. Any other cost effective options?

Thanks
Denon/Marantz XT32 is usually the most cost effective. No doubts about it. You also get Atmos and very similar (if not the same) results than Dirac with the MultiEQ App.
 
OP
G

Gedeon

Active Member
Forum Donor
Joined
Oct 30, 2018
Messages
191
Likes
195
No, I use AP, specifically an APx525 with an APx1701 acoustic interface.

edit: I should mention that I use a 192kHz sample rate.

With 3ms delay and a 192khz signal, if I'm not wrong, the number of taps should be 1024… but with such delay frequencies below 300hz couldn't be fixed with linear phase filters... Could you please test the delay with 96khz and 48khz?

Wouldn't it be better a 512 taps for 48khz to correct lower frequencies than 1024 for 192khz?

I apologise, maybe there are things I understand badly. Any help will be very much appreciated.
 

QMuse

Major Contributor
Joined
Feb 20, 2020
Messages
3,124
Likes
2,785
.. but with such delay frequencies below 300hz couldn't be fixed with linear phase filters...

Filter delay is in no way related to the frequency range you are correcting. For narrower frequency range you will need less taps for any chosen rate.


Wouldn't it be better a 512 taps for 48khz to correct lower frequencies than 1024 for 192khz?

I apologise, maybe there are things I understand badly. Any help will be very much appreciated.

Most convolution engines will not allow you to have filters for different sampling rates so everything will be resampled to the rate of your filters. If most of your music material is 44.1kHz than making filter with that sample rate is a logical choice.
 
OP
G

Gedeon

Active Member
Forum Donor
Joined
Oct 30, 2018
Messages
191
Likes
195
Filter delay is in no way related to the frequency range you are correcting. For narrower frequency range you will need less taps for any chosen rate.




Most convolution engines will not allow you to have filters for different sampling rates so everything will be resampled to the rate of your filters. If most of your music material is 44.1kHz than making filter with that sample rate is a logical choice.

My (probably totally wrong) assumption is based upon the fact that there are 25ms delay between "pulses" in a 40hz sound, no matter if is wrapped in a 44… 48... or 96khz .. signal. Hence, a 3ms delay isn't close enough to fully "detect" that frequency incoming in FIR which "memory" is limited to number of taps.

I think I need to read more about it ...
 
Top Bottom