Dirac Reviews

[ah-ra]

Hope this helps!

Thanks a lot! That should be really helpful. I will look into this and try if I can achieve better results with your method!

 

[ah-ra]

As far as I know, REW only allows you to create all-pass filters with "falling" phase, which are not suitable for counteracting the crossover-related phase wrap. So I guess rePhase would be the better tool to use for this purpose.

To get a "perfect" looking impulse and step responses you need to correct the excess phase which is introduced by the crossover, which in turn aligns the tweeter and woofer responses.

E.g. here's a comparison of the impulse and step responses of the original measurement, vs the minimum-phase version of the same magnitude response:View attachment 384729View attachment 384730

So if you're able to calculate a filter that counteracts the crossover-related excess phase part of the response you will get the nice-looking impulse/step responses you are after.

The way I tried to do it:

1. Open a measurement in REW and apply some (strong) smoothing to the frequency response. I used "1/1 smoothing" in my test.
2. Click "Actions" > "Estimate IR Delay" and then shift the IR accordingly -> this will remove the phase wraps introduced by the delay of the measurement system, including the delay introduced by propagation of sound through air (from speaker to mic).
3. Click "All SPL" view, then go to "Actions" > "Measurement actions" > "Excess Phase Version" -> this will create a new measurement with flat magnitude response and only the excess phase component of the original measurement
4. Export the previously generated 1/1 smoothed Excess Phase measurement and import into rePhase.
5. In rePhase we're aiming to get the phase close to zero in the mid and high frequencies. It seems to me that aiming for flat phase between 1kHz and 10kHz gives pretty good results for this purpose. Use the "Filters Linearization" view and add a filter to counteract the crossover phase wraps, and then add a few "Paragraphic Phase EQ" filters to tune as needed. My Revel M16 is specified with a crossover at 2100Hz, so I used that with a "LR 24 db/oct" crossover preset in "linearize" mode and then added just 4 paragraphic phase filters with relatively low Q values (0,5 or 1).
   This was the result:View attachment 384852
6. Export the phase correction with the appropriate number of taps and sampling frequency (I used 2048 taps and 48kHz).

In my case, here's how this manual phase correction compares to Dirac (and original measurement without any correction):
Hope this helps!

That was a big step forward! Thanks again....

This is my result, which looks much better than before. I just did a check based on Nearfield measurements. I will still have to do measurements for my MLP after applying the new FIR filter shown here...

1. Basic Nearfield measurement

2. Excess Phase copy of 1. after shifting IR based on IR delay

No. 2 was basis for calculating FIR filter in rePhase

Phase completely aligned to 0-line...

3. Impulse before (red) vs. after FIR (purple)

...and old FIR filter version (green) vs. new FIR filter version (purple)

4. Step before (red) vs. after FIR (purple)

...and old FIR filter version (green) vs. new FIR filter version (purple)

I think these are already pretty good results...

As mentioned I will have to do full measurements on MLP incl. subs and see, if there's anything else to consider.

What I still don't get is the overall picture of the phase. There are still some phase wraps to the right after applying FIR. If I set time offset to -0,300ms it looks like this (black line)... what is the logic behind this?

So overall phase is not on 0-line...

 

[dominikz]

What I still don't get is the overall picture of the phase. There are still some phase wraps to the right after applying FIR. If I set time offset to -0,300ms it looks like this (black line)... what is the logic behind this?

Those phase wraps don't look right, I agree. I assume this is a regular passive two- or three-way loudspeaker?

First thing I'd check is the timing reference configuration in the measurement window - I'd suggest to use acoustical timing reference (with null offset) and to use the same output for measurement and reference signals:

You may also wish to select the "Adjust clock with acoustic ref" option in the "Preferences" > "Analysis" window. From REW manual:
"The Adjust clock with acoustic ref selection controls whether REW compensates for clock rate differences between input and output when using the acoustic timing reference. If this option is selected an additional timing reference signal is played at the end of the sweep and the time between the timing reference signals is used to calculate any clock adjustment required to match the input and output device clock rates."

Also from manual:
"If the input and output are on the same device and so share a common clock longer sweeps will provide higher signal-to-noise ratio (S/N) in the measurements. Long sweeps may be problematic when the input and output are on different devices, such as when using a USB mic, as their sample clock rates will differ. Over a long sweep a significant difference in clock rates could cause severe distortions in the shape of the impulse response and affect the phase response. This can be corrected by using the acoustic timing reference and the Analysis option to Adjust clock with acoustic ref or, if a loopback connection is being used as a timing reference, the Analysis option to Adjust clock with loopback."

Also, perhaps double-check if you have the correct microphone and soundcard calibration files loaded.

So overall phase is not on 0-line...

Note that you are doing only excess phase correction - not total phase correction.
So after correction the phase of your measurement should be close to a minimum phase response (corresponding to the magnitude frequency response). I.e. in the resulting (corrected) response phase will not be zero and that is OK! But the phase wraps in the mid and high frequencies should be gone, however.

 

[ah-ra]

Those phase wraps don't look right, I agree. I assume this is a regular passive two- or three-way loudspeaker?

First thing I'd check is the timing reference configuration in the measurement window - I'd suggest to use acoustical timing reference (with null offset) and to use the same output for measurement and reference signals:

View attachment 384950

You may also wish to select the "Adjust clock with acoustic ref" option in the "Preferences" > "Analysis" window. From REW manual:
"The Adjust clock with acoustic ref selection controls whether REW compensates for clock rate differences between input and output when using the acoustic timing reference. If this option is selected an additional timing reference signal is played at the end of the sweep and the time between the timing reference signals is used to calculate any clock adjustment required to match the input and output device clock rates."

Also from manual:
"If the input and output are on the same device and so share a common clock longer sweeps will provide higher signal-to-noise ratio (S/N) in the measurements. Long sweeps may be problematic when the input and output are on different devices, such as when using a USB mic, as their sample clock rates will differ. Over a long sweep a significant difference in clock rates could cause severe distortions in the shape of the impulse response and affect the phase response. This can be corrected by using the acoustic timing reference and the Analysis option to Adjust clock with acoustic ref or, if a loopback connection is being used as a timing reference, the Analysis option to Adjust clock with loopback."

Also, perhaps double-check if you have the correct microphone and soundcard calibration files loaded.


Note that you are doing only excess phase correction - not total phase correction.
So after correction the phase of your measurement should be close to a minimum phase response (corresponding to the magnitude frequency response). I.e. in the resulting (corrected) response phase will not be zero and that is OK! But the phase wraps in the mid and high frequencies should be gone, however.

Thanks for the explanation! Yes, the measurements shown above are Nearfield Measurements of my 2-Way speaker only, as these were the only ones I had at hand.

Yes, I will check the settings in the measurement window and change accordingly.

Yes it's a correction of the excess phase measurement. It definitely gives a better Impulse/Step Response than my previous correction which was based on the initial measurement and thus corrected total phase. That was my "mistake".

I will give an update on settings and also on overall measurements incl. subs in the next days hopefully.

 

[ah-ra]

Those phase wraps don't look right, I agree. I assume this is a regular passive two- or three-way loudspeaker?

First thing I'd check is the timing reference configuration in the measurement window - I'd suggest to use acoustical timing reference (with null offset) and to use the same output for measurement and reference signals:

View attachment 384950

You may also wish to select the "Adjust clock with acoustic ref" option in the "Preferences" > "Analysis" window. From REW manual:
"The Adjust clock with acoustic ref selection controls whether REW compensates for clock rate differences between input and output when using the acoustic timing reference. If this option is selected an additional timing reference signal is played at the end of the sweep and the time between the timing reference signals is used to calculate any clock adjustment required to match the input and output device clock rates."

Also from manual:
"If the input and output are on the same device and so share a common clock longer sweeps will provide higher signal-to-noise ratio (S/N) in the measurements. Long sweeps may be problematic when the input and output are on different devices, such as when using a USB mic, as their sample clock rates will differ. Over a long sweep a significant difference in clock rates could cause severe distortions in the shape of the impulse response and affect the phase response. This can be corrected by using the acoustic timing reference and the Analysis option to Adjust clock with acoustic ref or, if a loopback connection is being used as a timing reference, the Analysis option to Adjust clock with loopback."

Also, perhaps double-check if you have the correct microphone and soundcard calibration files loaded.


Note that you are doing only excess phase correction - not total phase correction.
So after correction the phase of your measurement should be close to a minimum phase response (corresponding to the magnitude frequency response). I.e. in the resulting (corrected) response phase will not be zero and that is OK! But the phase wraps in the mid and high frequencies should be gone, however.

I checked the settings. Indeed I had the measurement config set to 48khz and 256k. Maybe that's the reason....? It's now...

I couldn't change anything for "output" as I haven't had anything connected to the laptop. I'm using an USB microphone with respective calibration file and I'm using my Pro-ject Pre Box DS2 Digital as DAC via USB, but with no dedicated calibration file (don't have one and I was always told it doesn't matter in their case...).

The preferences were already set, I guess...

I need to test the next days. Anything else to consider? Thanks in advance!

 

[dominikz]

I had the measurement config set to 48khz and 256k. Maybe that's the reason....?

I wouldn't expect so, but in the very unlikely case this is related to the DAC low-pass filter using 192kHz sampling rate should push that to much higher frequencies. Make sure that the same sampling rate that you use in REW is selected on the audio device in Windows sound control panel too (for both input and output devices), so that no unnecessary resampling is happening.

It's now...

Note that in this screenshot you are using channel "Default Output" channel "R" while in the reference output you're using channel "L" - this could be causing a timing offset in your measurement. I suggest to use the same channel in both "Output" and "Ref output".

I'm using an USB microphone with respective calibration file and I'm using my Pro-ject Pre Box DS2 Digital as DAC via USB, but with no dedicated calibration file (don't have one and I was always told it doesn't matter in their case...).

Amir's review of Pro-ject Pre Box DS2 Digital indicates it is a well-engineered device with a flat frequency response, so indeed I see no need for soundcard calibration in REW. I see it allows configuration of several filter types, however. It is probably a good idea to use one of the "sharp" filters - especially when using lower sampling rates (e.g. 44,1kHz or 48kHz). Also, I'd personally disable resampling while doing tests - just in case!

Anything else to consider?

The only other thing that comes to my mind is the obvious - make sure that all crossover filters, delays, PEQ, as well as any other processing is disabled in miniDSP when doing the baseline tests (which you will then use to calculate the corrections).
I'd personally also disable "Audio enhancements" for the associated audio device in Windows (see link) to make sure there's no undesired processing happening there. If the Pro-ject Pre Box DS2 Digital comes with any kind of proprietary audio control panel, make sure any processing/effects are disabled there as well.
When measuring the corrected response again make sure only the corrections you need are applied in miniDSP and everything else disabled.
Good luck with your tests!

 

[win]

Dirac Live full range with DLBC, 2 channel, 4 subs

A — Outstanding

What hardware are you running that on?

 

[Ricwa]

What hardware are you running that on?

Mac Mini M2

 

[win]

Mac Mini M2

Thank you. Obviously you need to run Dirac to create the filters for the DLBC set up. My question is, would you then be able to export the filters to some other hardware? Or, do you always need dirac running to utilize DLBC?

 

[Ricwa]

Thank you. Obviously you need to run Dirac to create the filters for the DLBC set up. My question is, would you then be able to export the filters to some other hardware? Or, do you always need dirac running to utilize DLBC?

Think of DLBC as an optional feature of Dirac. Dirac itself (with or without DLBC) utilizes 2 apps; the Dirac Live app creates the filters and the Dirac Live Processor uses the filters. So yes, the Processor program is always running. The Processor sits between your digital audio source and your DAC. And yes, you could export the files to some other Dirac-based hardware, but I don’t see why anyone would want to do that.

 

[win]

Think of DLBC as an optional feature of Dirac. Dirac itself (with or without DLBC) utilizes 2 apps; the Dirac Live app creates the filters and the Dirac Live Processor uses the filters. So yes, the Processor program is always running. The Processor sits between your digital audio source and your DAC. And yes, you could export the files to some other Dirac-based hardware, but I don’t see why anyone would want to do that.

Got it, thanks. I wasn't sure if Dirac just had a clever algorithm to create filters that could be ran on regular dsp hardware (for example, minidsp makes an 'open' version and a Dirac version for a lot of their crossovers).

 

[dlaloum]

A / B

Moving from an early version of Audyssey XT32, Bass tightened up (marginal), midrange, vocals, dialogue became distinctly clearer, imaging improved marginally...

The old version of XT32, had no way to turn off the MRC (mid range compensation)... so it is possible that a current D&M implementation would provide the same benefits.

Be that as it may - Dirac has proven itself a worthwhile upgrade, and resulted in my setup sounding better than it ever has (not just in movie mode, but also in stereo, music mode)

 

[ah-ra]

I wouldn't expect so, but in the very unlikely case this is related to the DAC low-pass filter using 192kHz sampling rate should push that to much higher frequencies. Make sure that the same sampling rate that you use in REW is selected on the audio device in Windows sound control panel too (for both input and output devices), so that no unnecessary resampling is happening.

Note that in this screenshot you are using channel "Default Output" channel "R" while in the reference output you're using channel "L" - this could be causing a timing offset in your measurement. I suggest to use the same channel in both "Output" and "Ref output".

Amir's review of Pro-ject Pre Box DS2 Digital indicates it is a well-engineered device with a flat frequency response, so indeed I see no need for soundcard calibration in REW. I see it allows configuration of several filter types, however. It is probably a good idea to use one of the "sharp" filters - especially when using lower sampling rates (e.g. 44,1kHz or 48kHz). Also, I'd personally disable resampling while doing tests - just in case!

The only other thing that comes to my mind is the obvious - make sure that all crossover filters, delays, PEQ, as well as any other processing is disabled in miniDSP when doing the baseline tests (which you will then use to calculate the corrections).
I'd personally also disable "Audio enhancements" for the associated audio device in Windows (see link) to make sure there's no undesired processing happening there. If the Pro-ject Pre Box DS2 Digital comes with any kind of proprietary audio control panel, make sure any processing/effects are disabled there as well.
When measuring the corrected response again make sure only the corrections you need are applied in miniDSP and everything else disabled.
Good luck with your tests!

Hi, I did some measurements again today with your recommended settings in REW, Windows etc... Somehow I can't get rid of the phase-wraps in the initial measurements...

I need to use t=0 offset and in this case set it to -0,238ms to get a proper picture of the phase

Based on an Excess Phase version of the measurement above I can generate a correction file in rePhase which I uploaded in the miniDSP. The resulting measurement looks like this:

The correction also results in a better impulse and step measurement

1. Impulse before (green) and after correction (yellow)

2. Step before (green) and after correction (yellow)

What do you think? Any more ideas regarding the phase wraps? I remember that some videos on Youtube (from Obsessive Compulsive Audiophile) showed the same and it was handled the same way as I did. Nevertheless it would be nice to have an explanation for it.

 

[ernestcarl]

Hi, I did some measurements again today with your recommended settings in REW, Windows etc... Somehow I can't get rid of the phase-wraps in the initial measurements...
View attachment 386310
I need to use t=0 offset and in this case set it to -0,238ms to get a proper picture of the phase
View attachment 386309

Based on an Excess Phase version of the measurement above I can generate a correction file in rePhase which I uploaded in the miniDSP. The resulting measurement looks like this:
View attachment 386311
The correction also results in a better impulse and step measurement

1. Impulse before (green) and after correction (yellow)
View attachment 386314

2. Step before (green) and after correction (yellow)
View attachment 386315


What do you think? Any more ideas regarding the phase wraps? I remember that some videos on Youtube (from Obsessive Compulsive Audiophile) showed the same and it was handled the same way as I did. Nevertheless it would be nice to have an explanation for it.

Check your acoustic reference. Exact positioning between output speaker and reference speaker is not perfectly aligned -- even a tiny distance difference in HF can look extreme. Alternatively, use the same output speaker as your acoustic reference.

 

[ah-ra]

Check your acoustic reference. Exact positioning between output speaker and reference speaker is not perfectly aligned -- even a tiny distance difference in HF can look extreme. Alternatively, use the same output speaker as your acoustic reference.

That's not the reason... I used the same speaker for reference (measured speaker "left" > reference "left")....

 

[ernestcarl]

That's not the reason... I used the same speaker for reference (measured speaker "left" > reference "left")....

I would also check repeatability of results (e.g. x6 sweeps, short vs long sweep length, and buffer settings.

 

[ah-ra]

View attachment 386354

View attachment 386355

I would also check repeatability of results (e.g. x6 sweeps, short vs long sweep length, and buffer settings.

Thanks a lot, I will look into that. Are these your settings? Timing offset in the measurement window is set to 0 in my case... What do you mean by buffer settings?

 

[fcracer]

Hi, I did some measurements again today with your recommended settings in REW, Windows etc... Somehow I can't get rid of the phase-wraps in the initial measurements...
View attachment 386310
I need to use t=0 offset and in this case set it to -0,238ms to get a proper picture of the phase
View attachment 386309

Based on an Excess Phase version of the measurement above I can generate a correction file in rePhase which I uploaded in the miniDSP. The resulting measurement looks like this:
View attachment 386311
The correction also results in a better impulse and step measurement

1. Impulse before (green) and after correction (yellow)
View attachment 386314

2. Step before (green) and after correction (yellow)
View attachment 386315


What do you think? Any more ideas regarding the phase wraps? I remember that some videos on Youtube (from Obsessive Compulsive Audiophile) showed the same and it was handled the same way as I did. Nevertheless it would be nice to have an explanation for it.

I wonder if the moderators can fork this over to a new thread. Seems like an interesting discussion that deserves its own thread. Keeping it here may confuse future Dirac users looking for info.

 

[ernestcarl]

Thanks a lot, I will look into that. Are these your settings? Timing offset in the measurement window is set to 0 in my case... What do you mean by buffer settings?

I believe my REW settings is near identical to what @dominikz has already previously shown -- other than the sample rate set at default 48 kHz.

Issues exist when using a USB microphone (although, there are workarounds e.g. properly configured/set up timing reference).

UMIK-1 or any USB MIC unsuitable for timing measurements?

For acoustic reference, the impulse of the acoustic reference determines t=0, so the location of measurement impulse is the difference in time of flight between the acoustic reference and DUT, not the distance of DUT to mic. For electrical loopback, impulse of the electrical loopback is same...

www.audiosciencereview.com

More than one or more configurable setting could exist in your system that is influencing the final measurement results depending in one's particular source to output setup chain. Some of these can make the system more prone to glitches or have clock sync issues during the REW sweep measurement procedure.

 

[Snarfie]

It has everything to do with the speakers and the room so which other variables are there? -Other than taste. I don't call peoples gear bad. They have bought it because they like it so I can see why they would be put off by Dirac if it changes a lot of what they liked in the first place. It can easily be that the corrected response is much better, objectively, but that they just don't like it that way.

Dirac should have a recommended burn-in time in their manual..

I got the impression that most people always like what the have (including me ). They think that their home perceived sound is quite good. 99.9% never enterd a professional studio controll room ( besides probably some proffesional ASR members) from at least let say 1 million euro or dollar with a investment ratio of 70/30 which relate to 70% room treathment an 30% gear. On top of that most could probably not distinguished the difference between instruments (so how they must sound on a recording) like a Bösendorfer or Steinway piano's a Telecaster or a Stratocaster an so on.

Than we have seperation between instrument an voices in a well recorded session including the reverb time from the venu. In suche quality recording using the needed an correct placed quality mic's it contain the information for staging imaging for instance. I noticed (in my current mancave) that most of this is revealed with (a well measured) Room correction (DSP) if your room like mine has a horrible acoustic. I guess that 90 % has the same problem. So get acainted with the correct sound which means if you listen to a bad recording the DSP will reveal that without mercy (which means DSP does it job) if it is well recorded you could be in heaven with DSP. If you have a well treathed listening room you don't need DSP so atleast do a Propper measurment with REW for instance it reveal a lot. I use Mathaudio room EQ did compare it with Lyngdorf an Dirac quite similar sound/result all of them i noticed. DSP for me did let me rediscover my music collection.