Jump to content

Dirac Live Users Discussion Thread


Recommended Posts

A couple of months ago I started exploring room equalisation using Dirac Live.  I had done some work with REW and corrections exported to Roon previously, but nothing as integrated as the Dirac Live concept of target curves.

Going in I found theory on different target curves and thoughts on personal preferences, but not a lot of discussion I found practical.  So I set about listening to different curves and, as a broad generalisation, liked how it tightened the bottom end but not how it added a hint of brightness to the top end – almost independent of curve.  So I went digging.

I had tested quite a few target curves but here I will focus on 2 for illustration – the Dirac Standard and the Harman 6db curves. 

The first two graphs show these in Dirac.  The target curves are the yellow dotted lines on the graphs.  The Dirac Standard is a straight line (on log scale 20Hz to 20kHz) from +2db to -2db.  The Harman Curve is more explicitly 6db boosted at the bottom end but then flat.  

The light green lines are the right speaker raw without any equalisation, and the darker green line (mostly hiding behind yellow) is Dirac’s projection after its adjustments.  Note the difference in the bottom end of the raw curves is different sub integration.  First look at the Dirac projections you can see the ‘taming’ of the bass and mid-range, but the brighter top end is less immediately obvious.

DriacSml.jpg.6ef430bfb38df4b7606e930ab6649d2f.jpg

Dirac Standard Curve (Above)

HarmanSml.jpg.2b6a8310e33ec4d9098e99355c32d6ad.jpg

Harman 6db Curve (Above)

I then took some measurements through REW.  These were using both speakers on, Pink PN across about 50 averages with microphone moved through about the same area as covered by the Dirac 9‑point sampling.

The first graph below shows the raw measurement in grey, with the Dirac Standard adjusted curve in red.  I found this a little more revealing.  The taming of the bottom and mid-range is clear, along with an overall lowering in SPL – but relative to that you can see the contribution of the top end actually increasing.  Strikes me that the top end of the raw curve falls away because the tweeter is slightly off axis given speakers are toe-in about half-way between straight ahead and to listening position.  That felt like best raw listening across balance and sound stage.  But Dirac sees that fall off and, in its standard curve, pushes the top end back up again.  It is more visible in the REW vs Dirac displays, but I sense explains what I was hearing.

RvsDSml.png.33546ebd84313116eab607b0e15c7819.png

Now the Harman Curve, from the Dirac view, doesn’t look like it should present the same way.  But it does.  The final graph shows the REW view of Dirac Standard curve (actual) in red against the Harman 6db curve (actual) in Green.  A further lift in the top end relative to mid-range.

DvsHSml.png.c3b75cb07509d8878abfd8cb7aba4286.png

I sense this explains why some recommend / prefer limiting Dirac optimisation to maximum 1 or 2 kHz so it leave the top end alone.  Will need to try that as the Dirac display typically shows a  discontinuity at the transition point, and would like to tale a look at that through REW.  I have also created a linear +4db to -3db target curve that sounds pretty good.

Not quite what I expected, but maybe some thoughts of interest to others using Dirac or similar target curve based systems.

  • Like 4
Link to comment
Share on other sites



22 minutes ago, Snoopy8 said:

You are not alone in dropping the curtain down

Thanks Snoopy.  Useful thread (and like your top tips) - wish I had found it earlier.  That NAD standard curve looks quite aggressive and different to others.

Link to comment
Share on other sites

41 minutes ago, Bunno77 said:

What hardware/software are you using to run dirac? Thanks

I am running Dirac Live 3 on a miniDSP Studio SHD.  Fits well into my system as also acts as Roon end point and cross-over for sub, and speakers are active so accept digital in.

Edited by gibbo9000
  • Like 4
  • Love 1
Link to comment
Share on other sites



2 hours ago, gibbo9000 said:

Thanks Snoopy.  Useful thread (and like your top tips) - wish I had found it earlier.  That NAD standard curve looks quite aggressive and different to others.

Welcome to the joys of Dirac.  Actually, Dirac is very good and my recommendation for most people.  And most people do not need to tune it like you have done.

 

Not sure whether you have seen this. The author is a guru in room correction

https://audiophilestyle.com/ca/reviews/dirac-live-2-digital-room-correction-software-walkthrough-r884/

2 hours ago, gibbo9000 said:

I am running Dirac Live 3 on a miniDSP Studio SHD.  Fits well into my system as also acts as Roon end point and cross-over for sub, and speakers are active so accept digital in.

I have not used the SHD Studio but like the idea of a box with streamer with Dirac.  

Link to comment
Share on other sites

13 hours ago, gibbo9000 said:

top end

It seem you are thinking about the "top end" being 3Khz and upwards (or there abouts).  I strongly sustpect this isn't the region responsible for the difference.

 

Try something.   Compare your uncorrected response, to the corrected response (either/both, the Harman or Dirac targets).

Match the level at 500hz.   Look at 500hz vs 1kz, 2khz, 4khz.    What do you see?!  ;)

 

 

 

 

PS... What I see... is that in both ocrrections you have adjusted the balance between 100hz vs 1khz  (or 200hz vs 20khz) ..... by a significant number of dB.    This will do exactly what you are describing.

 

The two ocatves above 500hz (ie. up to about 2khz) is much higher in pitch than most people tend to think.... and what happens through these frequency regions is much more important than what is going on in the final two octaves from 4 to 16khz.

  • Like 3
Link to comment
Share on other sites

14 hours ago, gibbo9000 said:

A couple of months ago I started exploring room equalisation using Dirac Live.  I had done some work with REW and corrections exported to Roon previously, but nothing as integrated as the Dirac Live concept of target curves.

 

Going in I found theory on different target curves and thoughts on personal preferences, but not a lot of discussion I found practical.  So I set about listening to different curves and, as a broad generalisation, liked how it tightened the bottom end but not how it added a hint of brightness to the top end – almost independent of curve.  So I went digging.

 

I had tested quite a few target curves but here I will focus on 2 for illustration – the Dirac Standard and the Harman 6db curves. 

 

The first two graphs show these in Dirac.  The target curves are the yellow dotted lines on the graphs.  The Dirac Standard is a straight line (on log scale 20Hz to 20kHz) from +2db to -2db.  The Harman Curve is more explicitly 6db boosted at the bottom end but then flat.  

 

The light green lines are the right speaker raw without any equalisation, and the darker green line (mostly hiding behind yellow) is Dirac’s projection after its adjustments.  Note the difference in the bottom end of the raw curves is different sub integration.  First look at the Dirac projections you can see the ‘taming’ of the bass and mid-range, but the brighter top end is less immediately obvious.

 

DriacSml.jpg.6ef430bfb38df4b7606e930ab6649d2f.jpg

Dirac Standard Curve (Above)

HarmanSml.jpg.2b6a8310e33ec4d9098e99355c32d6ad.jpg

Harman 6db Curve (Above)

I then took some measurements through REW.  These were using both speakers on, Pink PN across about 50 averages with microphone moved through about the same area as covered by the Dirac 9‑point sampling.

 

The first graph below shows the raw measurement in grey, with the Dirac Standard adjusted curve in red.  I found this a little more revealing.  The taming of the bottom and mid-range is clear, along with an overall lowering in SPL – but relative to that you can see the contribution of the top end actually increasing.  Strikes me that the top end of the raw curve falls away because the tweeter is slightly off axis given speakers are toe-in about half-way between straight ahead and to listening position.  That felt like best raw listening across balance and sound stage.  But Dirac sees that fall off and, in its standard curve, pushes the top end back up again.  It is more visible in the REW vs Dirac displays, but I sense explains what I was hearing.

 

RvsDSml.png.33546ebd84313116eab607b0e15c7819.png

Now the Harman Curve, from the Dirac view, doesn’t look like it should present the same way.  But it does.  The final graph shows the REW view of Dirac Standard curve (actual) in red against the Harman 6db curve (actual) in Green.  A further lift in the top end relative to mid-range.

 

 

DvsHSml.png.c3b75cb07509d8878abfd8cb7aba4286.png

I sense this explains why some recommend / prefer limiting Dirac optimisation to maximum 1 or 2 kHz so it leave the top end alone.  Will need to try that as the Dirac display typically shows a  discontinuity at the transition point, and would like to tale a look at that through REW.  I have also created a linear +4db to -3db target curve that sounds pretty good.

 

Not quite what I expected, but maybe some thoughts of interest to others using Dirac or similar target curve based systems.

 

Hello David, thanks for opening this interesting analysis. A few things occur to me as I read through it.

 

First thought: the bottom chart shows that the Dirac filters are not actually delivering anything very close to the intended curve. Kind of on the right track, yes, but a detailed comparison to the dark green lines in the top two charts, shows it up a bit.

 

Second thought: the Harman Curve should not be flat from 300 Hz to 20 kHz. I don't know where you got that from: Dirac? (An un-gated, in-room target curve needs to taper down steadily as frequency rises from midrange to 20 kHz.) I could be wrong, if Dirac is gating its measurement above a certain frequency -- but if that is the case, then the Dirac Curve should also be flat above that frequency. I assume the Dirac Curve is correctly tailored for how Dirac does/does not gate during measurement, so the Harman Curve should also taper.

 

The 'aggressive' NAD Curve you mentioned, presumably this one, looks more logical to me, combining Harman allowance for room gain in bass, with Dirac allowance for HF taper. Although I would consider trimming the bass boost to 50% for comparison purposes, as Toole says those frequencies need their overall amplitude adjusted 'to taste', and definitely (as shown for the speakers in the graph I linked) avoid having a target curve to boost the deep bass by 15-30 dB in frequencies so low that the natural speakers have given up.

 

Third thought: the right place for DSP above the transition frequency area, is in the speakers. Why? Because the speaker designers are the ones who know exactly why their speaker is deviating, and the consequences of applying phase and amplitude corrections, so they are the ones who know whether the consequences are acceptable, or not, or partially, and can make a fully informed decision about EQ. A 'dumb equater', that sees a measurement and sees a target, but doesn't know much about the why, is just as likely to make bad decisions as good. The dangers of the component equalizer.

 

cheers

Grant

Edited by Grant Slack
  • Like 2
Link to comment
Share on other sites

Lots of interesting thoughts as you all help me get my head around this. 

8 hours ago, davewantsmoore said:

Try something.   Compare your uncorrected response, to the corrected response (either/both, the Harman or Dirac targets).

Match the level at 500hz.   Look at 500hz vs 1kz, 2khz, 4khz.    What do you see?!

This makes a lot of sense - particularly since I can't hear above 10kHz anyway.  Found how to offset the graphs in REW and what you suggest becomes quite clear.  Aligned the two around 500Hz.

 

RawvsD1Align.png.48a938a231c3091630b13e6f20bd726c.png

 

It makes the hefty gain of the red curve (Dirac Standard adjusted sound) over the grey curve (No Dirac adjustment) very clear.  Strong boosting above 700Hz, up to +5db just over 2kHz and continuing close to that out to 10kHa and beyond is what I am hearing as 'bright' and a little fatiguing to listen to. 

 

The 'theory' without seeing the data is easy to interpret as Dirac taking a 'flat' unadjusted response  and giving it a nice little tilt from +2db to -2db response.  The practice as this highlights is far from that - it has done the opposite and taken a downward sloping 'raw' curve and straightened it.

Thanks for the 'aha' moment trigger - helps explain my experience (and others) of the Dirac house curve.  Now much more confident to experiment and assess other curves.

 

Gibbo

Edited by gibbo9000
Link to comment
Share on other sites



23 minutes ago, gibbo9000 said:

Found how to offset the graphs in REW and what you suggest becomes quite clear.  Aligned the two around 500Hz.

?

 

People get trapped into thinking that what happens at these high frequencies is important (or even easy to hear). ... becuase they see other people (all over audiophile forums) saying it is......    It's 99% audiophile bs.

 

The top octave is not only almost inaudible.... but there's quite little content there on most recordings.

Edited by davewantsmoore
  • Like 1
Link to comment
Share on other sites

8 hours ago, Grant Slack said:

Second thought: the Harman Curve should not be flat from 300 Hz to 20 kHz. I don't know where you got that from: Dirac?

Hmm - More traps for the average punter.  These are available from various sources when you go searching for Dirac target curves - I think I took my set from here:  https://www.audiosciencereview.com/forum/index.php?threads/harman-target-curve-calculator.9618/ .  I have checked the set and they are all flat from 400Hz (see pic) 

 

Looking deeper he may have got them from here:  https://mehlau.net/audio/dirac-live-2/   I have checked those and they do have a very minor fall off - but tiny - the 6db curve raises bottom end up 6db, coming back to 0.03db at 400Hz, and then 0db at 20kHz.  So negligible.
 

So we need a source of truth on target curves as well !!

The NAD one is making more sense to me - time to do some more testing.

 

1243543052_HarmanCurves.png.20500cd97353f4d4bae739b392862a27.png 

Link to comment
Share on other sites

49 minutes ago, gibbo9000 said:

It makes the hefty gain of the red curve (Dirac Standard adjusted sound) over the grey curve (No Dirac adjustment) very clear.  Strong boosting above 700Hz, up to +5db just over 2kHz and continuing close to that out to 10kHa and beyond is what I am hearing as 'bright' and a little fatiguing to listen to. 

Experiment with more tilt in your curve.

 

eg. keep it flat (or the same target as you have now) above 500Hz.... but make it keep climing below 500Hz.... currently you have about 3dB lift betwen 100hz and 500hz..... try 6db.

 

49 minutes ago, gibbo9000 said:

 

The 'theory' without seeing the data is easy to interpret as Dirac taking a 'flat' unadjusted response  and giving it a nice little tilt from +2db to -2db response.  The practice as this highlights is far from that - it has done the opposite and taken a downward sloping 'raw' curve and straightened it.

No.  Dirac has matched your resposne perfectly to it's target curve.

 

You might like a differnt target.

 

I like about 10dB fall between 100hz and 10khz..... but it does depend.

 

 

49 minutes ago, gibbo9000 said:

Thanks for the 'aha' moment trigger - helps explain my experience (and others) of the Dirac house curve.  Now much more confident to experiment and assess other curves.

1dB is quite large.

Remember 3dB = double the accoutic energy... and 6dB = 4x.......   small changes produce quite profound subjective differences

 

??

  • Like 1
Link to comment
Share on other sites

7 minutes ago, gibbo9000 said:

Hmm - More traps for the average punter.  These are available from various sources when you go searching for Dirac target curves - I think I took my set from here:  https://www.audiosciencereview.com/forum/index.php?threads/harman-target-curve-calculator.9618/ .  I have checked the set and they are all flat from 400Hz (see pic)  

These curves may be for/from version of Dirac live which don't do full frequency range correction.

  • Like 1
Link to comment
Share on other sites

Just now, davewantsmoore said:

These curves may be for/from version of Dirac live which don't do full frequency range correction.

Quite likely in retrospect - all part of the learning curve and traps for the unwary.

 

Now I am getting a better understanding, thanks to these discussions, the NAD Full Range curve is making more sense - with the bottom end boost tailored to my taste.  Question will be whether it sounds better with Dirac trying to adjust above 500 - 1 kHz, or leave that to the raw speakers.  Dirac certainly seems to be helping around some room issues below 500 Hz. 

Link to comment
Share on other sites



4 hours ago, gibbo9000 said:

Quite likely in retrospect - all part of the learning curve and traps for the unwary.

That does make sense. (Your chosen Harman Curve predating full range Dirac)

Quote

Now I am getting a better understanding, thanks to these discussions, the NAD Full Range curve is making more sense - with the bottom end boost tailored to my taste.  Question will be whether it sounds better with Dirac trying to adjust above 500 - 1 kHz, or leave that to the raw speakers.  Dirac certainly seems to be helping around some room issues below 500 Hz. 

I like your way of thinking. The answer to your question might revolve around what are your speakers, and how good is their response, i.e. do they have internal DSP and is it well executed.

 

Regards

Grant

  • Like 1
Link to comment
Share on other sites

12 hours ago, gibbo9000 said:

Hmm - More traps for the average punter.  ...

So we need a source of truth on target curves as well !!

Hello David,

 

These tapering target curves need to be put into context. In a sense, they represent a 'Level 1' understanding and application of equalization, and the more advanced user can put them to one side. They emerged at a time when measurements were almost always taken ungated.

 

What Toole et al have shown is that the desired frequency response is smooth, flat and extended, for the direct-arrival sound. What these target curves are showing, is roughly what the total ungated room response, in a broad and averaged sense, looks like unequalized when you have perfect loudspeakers with a smooth, flat and extended direct-arrival sound and ideal off-axis behaviour, when measured in a room of average sonic characteristics. So, in that sense, these tapered target curves represent rough-and-ready back-engineering. And the question of 'how much tilt', that the above comment by @rand129678 poses, has no right answer. It also causes significant risk of making good speakers sound worse, if they already have a smooth, flat and extended direct-arrival sound and that is made worse by the back-engineering kludge of equalizing to a target for total sound.

 

The right answer to the question of what shape of Dirac target curve to use, for your room and your speakers, is complex and requires sound knowledge of your room and your speakers. Good room treatment will also be important to good results. Adopting a generic curve as target is something to do when you don't have that sound knowledge, and just want to do something rough-and-ready -- with risks.

 

Regards,

Grant

  • Like 3
Link to comment
Share on other sites

4 minutes ago, cazzesman said:

Did you notice any difference in SQ between DL2.0 and 3.0?

Hi Cazzesman,
Not significant that I can tell as yet - but I must admit I haven't really been listening for that.  They have reduced sample points back to 9 for tightly defined listening and fixed the 'Dirac Corrected' display bug they had late in v2 (where the corrected line would sit 2db or so away from the target - at least on the SHD).
My underlying 'exploration' is still around boosting the SPL through moving load to sub below 50Hz but, as above, I am discovering a few gaping holes in my knowledge / experience / assumptions of Dirac Live curves that the contributors to this thread are helping me through.  Am getting closer to locking in some benefits in bass, but now clear I need to tame the midrange relative to the curves I have used to date.

Also waiting on an OpenDRC-DA8 so I can get the delay and D/A for the sub in one place as the ~100ms delay needed is beyond the SHD's capabilities and would rather not have the current separate DAC followed by Bluestream 'Delay' (which is another A/D, delay, D/A on way to sub).
Who ever said moving to active speakers took the fun out of the hobby!
Gibbo

  • Haha 1
Link to comment
Share on other sites



Excellent.    You do all the research, leg work and hair pulling.    
 

Once you have it all down-pat let me know and lunch is on at my place, so you can set up my Ki3’s with subs and DL.  ?

 

Regards Cazzesman

Edited by cazzesman
Link to comment
Share on other sites

1 minute ago, cazzesman said:

Once you have it all down-pat let me know and lunch is on at my place, so you can set up my Ki3’s with subs and DL.

I look forward to it - still love the idea of integrating subs into stands!

David

Link to comment
Share on other sites

Back on the learning track.  Thank you all for your inputs.  You are putting me on a whole new learning curve (a good thing!).

I am now going back to basics and will try to create my own curve that is simply a smoothed version of my speakers natural response (no sub) across the full range and see what impact that has smoothing a few basic peaks and troughs.  WAF limits me fixing issues like one speaker nearer corner and both having hydronic heating radiators on wall behind. 
That seems to have three challenges.   First is drawing the curve (not too hard), and second is applying it in Dirac Live and adjusting it so the output at the speakers, measured in REW, mirrors the natural response less bumps measured the same way.  The second navigates the issue @Grant Slack raised on how well Dirac Live actually derives filters that deliver the target response curve in practice.  And thirdly, does it sound any better? 

David

Link to comment
Share on other sites

10 hours ago, Grant Slack said:

It also causes significant risk of making good speakers sound worse, if they already have a smooth, flat and extended direct-arrival sound and that is made worse by the back-engineering kludge of equalizing to a target for total sound.

Yes, I’ve made that experience myself. Having these marvellous tools at one’s disposal makes it very tempting to correct what shouldn’t be corrected. I promptly ran into the “fixing the graphs but killing the music” trap. I now apply EQ for room correction very sparingly, basically only to rein in a couple of room modes under 100Hz. Assuming the speakers are good performers, I don’t think EQ should be applied to anything outside the bass range, ever.

  • Like 1
Link to comment
Share on other sites



  • Recently Browsing   0 members

    • No registered users viewing this page.




×
×
  • Create New...
To Top