Dirac and frequency shifting?

[tranz]

Just a quick call out to see whether anyone else has experienced a kind of metallic tinge to sound, especially noticable with voices, when using a Dirac filter.

 

The closest I can describe it as is akin to frequency shifting impact of sound.

 

I have just started mucking about with Dirac within Amarra IRC, and am still learning, but if anyone has had this and remembers show they fixed it, I would love to know.

 

Does the Dirac filter perform frequency shifts?

 

Does the integration between Amarra and Dirac perform frequency shifts?

 

Thanks!

[flak]

Hello Tranz,

 

neither Dirac or Amarra perform frequency shifts, what you hear could probably be fixed by fine tuning the target curve...

 

according to a study which has been done in Italy by Giussani to correlate frequency response and perceived effect for example a decrease at 2.000 Hz will reduce the "presence effect" while an increase will make the sound "pungent", at 3.150 Hz it determines the characterization and the articulation of the solo instruments and female voices... and so on.

 

This is the link: http://www.tangerinetech.net/downloads/freq_rg.pdf

(page 2, unfortunately it is in italian)

 

I would wait before definitively adopting a modified target curve... but experimenting will certainly be interesting and eventually beneficial.

 

Ciao, Flavio

[Boris75]

(page 2, unfortunately it is in italian)

 

Italian is a most beautiful, very precise language; in my opinion there is no need to make excuses for using it in scientifical work. I for one am pleased to see that there are still people who write technical papers in other languages than English.

[tranz]

Hello Tranz,

 

neither Dirac or Amarra perform frequency shifts, what you hear could probably be fixed by fine tuning the target curve...

 

according to a study which has been done in Italy by Giussani to correlate frequency response and perceived effect for example a decrease at 2.000 Hz will reduce the "presence effect" while an increase will make the sound "pungent", at 3.150 Hz it determines the characterization and the articulation of the solo instruments and female voices... and so on.

 

This is the link: http://www.tangerinetech.net/downloads/freq_rg.pdf

(page 2, unfortunately it is in italian)

 

I would wait before definitively adopting a modified target curve... but experimenting will certainly be interesting and eventually beneficial.

 

Ciao, Flavio

 

Thanks Flavio!

 

My italian is non-existent except for ordering delicious food, but I will check it out.

 

Glad to read confirmation on the absence of frequency shifts.

 

I tried another set yesterday using a wider box (~60cm radius) around the sweetspot.

 

1. Metallic shimmer especially noticable on voices still there.

 

2. Strong, warm female voices like Diana Krall and Fiona Apple that are solid and right in front of the soundstage suddenly become weak, move backwards on stage, with more sibilance and hiss around their voices with the filter on.

 

I am likely not doing a great job at explaining the result, but it is a bit like listening to a voice in CD quality, and when filter is on it becomes MP3, and a bit out of phase.

 

There is no way for me to only turn on the impulse response only, and not the EQ portion?

 

Thanks for your help.

[tranz]

And a picture...

 

[Boris75]

I tried another set yesterday using a wider box (~60cm radius) around the sweetspot.

 

I have no experience with Dirac as I just read the manual and then realised that there was too much noise in my environment to do measurements right now. However, I remember reading that the measurements should be made in an area with a radius of at more than 1m around the sweet spot.

 

I went to the manual; this is what it says:

it is important to spread out the microphone positions in a sphere of atleast 1 meter of diameter. A too small space will result in over-

compensation that will sound very dry and dull.

 

The description of the outcome seems to correspond roughly to your experience.

[flak]

And a picture...

 

[ATTACH=CONFIG]11720[/ATTACH]

 

Hello Tranz,

please kindly start by moving the right handle (the curtain one which defines the intervention area) to the left so that you will have correction only up to the center point in the graph (700 Hz appx.), also please carefully adjust the dsp gain control to a low value (which does not anyhow incur into clipping) following the manual instructions and then let me know what happens

 

Flavio

[tranz]

I have no experience with Dirac as I just read the manual and then realised that there was too much noise in my environment to do measurements right now. However, I remember reading that the measurements should be made in an area with a radius of at more than 1m around the sweet spot.

 

I went to the manual; this is what it says:

 

 

The description of the outcome seems to correspond roughly to your experience.

 

Thanks Boris,

 

Yeah, I had to do the measurements a few times as planes decided to fly above the house just when I started the sweep.

 

Based on some feedback in the Amarra thread I actually went a little larger this time around. The room picture is almost the same. 60cm radius = 120cm diameter. The first time I tried 50cm radius as per the manual.

 

Not sure whether the algorithm changes based on whether a chair or sofa is chosen though.

 

Next I will try a different toe-in angle as well.

 

Cheers

[tranz]

Hello Tranz,

please kindly start by moving the right handle (the curtain one which defines the intervention area) to the left so that you will have correction only up to the center point in the graph (700 Hz appx.), also please carefully adjust the dsp gain control to a low value (which does not anyhow incur into clipping) following the manual instructions and then let me know what happens

 

Flavio

 

Thanks for the suggestions Flavio!

 

If I take it a step further and completely close the 'curtains', does that provide me with the impulse control only correction?

 

I will also try a different toe-in measurement.

 

Cheers

[Melvin]

Tranz .. something else to think about as long as you're remeasuring. Not sure about others but during the measurement process I would sit at the computer which is, of course, right smack in betwixt the speakers. Thinking this might not be a good thing, I set up a screen-share with my other MAC and remeasured from another room. This made a subtle difference, but a difference nonetheless. Everything matters, no? I'm curious if you or other members have done similar.

[tranz]

Tranz .. something else to think about as long as you're remeasuring. Not sure about others but during the measurement process I would sit at the computer which is, of course, right smack in betwixt the speakers. Thinking this might not be a good thing, I set up a screen-share with my other MAC and remeasured from another room. This made a subtle difference, but a difference nonetheless. Everything matters, no? I'm curious if you or other members have done similar.

 

Thanks Melvin,

 

I have done something similar. My gear is all in a closet anyway as I cannot stand the sound of harddrives, fans and transformers whilst listening. I shut down all other electrical devices such as tv, fridge, laptops, and stand in the kitchen with my iPad to trigger the recording. In addition I do it at night when ambient and neighbourhood noise is at its lowest.

[tranz]

Hello Tranz,

according to a study which has been done in Italy by Giussani to correlate frequency response and perceived effect for example a decrease at 2.000 Hz will reduce the "presence effect" while an increase will make the sound "pungent", at 3.150 Hz it determines the characterization and the articulation of the solo instruments and female voices... and so on.

 

This is the link: http://www.tangerinetech.net/downloads/freq_rg.pdf

(page 2, unfortunately it is in italian)

 

Ciao, Flavio

 

H Flavio,

 

Where would we be without Google Translate. I recognize the Møller curve in Dirac's target curve.

[beanbag]

Check to see if dirac is generating a delay on one of the channels. That will mess up your AB comparison.

[flak]

Check to see if dirac is generating a delay on one of the channels. That will mess up your AB comparison.

 

Hello beanbag,

 

I think that the delay should be left in place, it is part of the correction

Flavio

[beanbag]

What's the difference between the delay vs moving your head a few inches to one side?

[flak]

Hello beanbag,

 

when making the correction we have to have a reference point, the central one, and we want the sound of the two speakers to arrive there at the same time and equally loud... while there is no doubt that if (for example) we move one meter sideways the balance will change I see no reason why we should accept as is a speaker which is one meter arrear vs. the other, and not correct it.

In any case if you think that the delay correction makes no difference you will also agree that leaving it on will also make no difference, so that the A/B comparison should be fair.

 

Ciao, Flavio

[tranz]

Well, I tried various more room recordings with different speaker positions, all resulting in similar curves.

 

In addition I tried various curtain positions and curve tweaks for the optimization filter.

 

Close to giving up.

 

In every single trial it sounds better without the Dirac filter, not to mention the pain of the clipping and having to adjust a digital volume ( which does not seem like a good idea to begin with ).

 

It was an interesting experiment, and to be able to visually see the room frequency response to help with physical placement of speakers and other materials is very helpful.

 

At this point I am a bit skeptical and think that if someone cannot change the desired sound through physical room and speaker adjustments this would be an absolute last resort.

 

Of course every setup and room is different, and I am open to suggestions of improvement, but in my current situation the room correction filter does more harm than good to the sound.

 

Flavio - as always, thank you for your support on this forum.

 

Cheers

[flak]

Hello Tranz,

 

I understand that you have tried several curtain positions... did you try limiting the correction to the low frequencies only so that just the room modes are corrected?

 

It would seem odd to me if you will not find an improvement 'cause the room certainly has dips and peaks (there is something like a 20 dB excursion) and all the rest would be left untouched with no correction.

(one then may legitimately think that correcting the room modes only is not worth the effort, but that's another story)

Would you try this and let me know?

 

Also did you adjust the DSP gain control at the lowest possible value (I mean as near to zero as possible) without lighting the clipping light in the DAP?

 

Ciao, Flavio

[Miska]

1. Metallic shimmer especially noticable on voices still there.

 

2. Strong, warm female voices like Diana Krall and Fiona Apple that are solid and right in front of the soundstage suddenly become weak, move backwards on stage, with more sibilance and hiss around their voices with the filter on.

 

Can you show how the "after" curve looks like? Especially if you have steep linear phase correction or strong phase adjustment correction around midrange, correction filter pre-ringing may become very obvious.

 

Any steep frequency or phase response corrections around midrange will make sound really strange. I personally prefer to design corrections based on 1/6th octave smoothed response which helps avoiding too aggressive correction filters that would ruin the sound.

[tranz]

Hello Tranz,

 

I understand that you have tried several curtain positions... did you try limiting the correction to the low frequencies only so that just the room modes are corrected?

 

It would seem odd to me if you will not find an improvement 'cause the room certainly has dips and peaks (there is something like a 20 dB excursion) and all the rest would be left untouched with no correction.

(one then may legitimately think that correcting the room modes only is not worth the effort, but that's another story)

Would you try this and let me know?

 

Also did you adjust the DSP gain control at the lowest possible value (I mean as near to zero as possible) without lighting the clipping light in the DAP?

 

Ciao, Flavio

 

Hi Flavio,

 

Thanks for the suggestions. I did try the 700Hz as the highest frequency curtain point with the left curtain untouched, another at 70Hz, and another trying to close the curtains as much as was possible. The 700Hz one sounded better than the others but still not natural.

 

Regarding digital volume gain, it would clip at different levels for every song, so I tried between -6 to -1; 0 would always clip at some point in a song.

 

I will try a >30 Hz and <70 Hz smoothing as well, but I do like a bit of chest slam in my electronica.

 

Cheers

[tranz]

Can you show how the "after" curve looks like? Especially if you have steep linear phase correction or strong phase adjustment correction around midrange, correction filter pre-ringing may become very obvious.

 

Any steep frequency or phase response corrections around midrange will make sound really strange. I personally prefer to design corrections based on 1/6th octave smoothed response which helps avoiding too aggressive correction filters that would ruin the sound.

 

Hi Miska,

 

Thanks for your input. I often read your threads with much interest.

 

Here is the pre-tweaking after, which essentially follows the target line. When I get back home I will load the < 700Hz curtain one as well. Looked similar from what I remember.

 

 

Cheers

[tranz]

One handy chart and link on effects of certain frequencies:

 

Magic Frequencies

[flak]

 

I will try a >30 Hz and <70 Hz smoothing as well, but I do like a bit of chest slam in my electronica.

 

Cheers

 

I would suggest going a little higher as you not only have a 15 DBs peak at 50 Hz but also a 10 dBs dip at 100 Hz, in my opinion you should find no correction if not for the room modes a definite improvement in any case.

But I will not insist anymore as there is obviously room for different preferences (the bit of chest slam that you mentioned is an example) and that's a good reason for the availability of a free two weeks Dirac trial.

 

I thank you anyhow for your time in testing Dirac Live

Flavio

[tranz]

Thank you Flavio!

 

I will try that option as well.

[Miska]

Here is the pre-tweaking after, which essentially follows the target line. When I get back home I will load the < 700Hz curtain one as well. Looked similar from what I remember.

 

Looking at the plot, I would suggest to limit corrections to 30 - 150 Hz range and that should minimize amount of side effects you get. The response is otherwise pretty smooth.