Jump to content
IGNORED

Help needed with choosing the optimal Dirac curve


Recommended Posts

I'm new to Dirac. I've got a MiniDSP SHD Studio with Dirac 1.2 for two weeks now and tried several curves. Started with the automatic curve and than added points up to a 30 point curve to follow the measured curve more closely. I like the reduced bass at 55Hz. But I like to know if I can get more from it. So I would like to know what you think is the optimal curve. I have added printscreens and the matching projects. If you've got Dirac 1.2 I think you can even optimise a curve and send ik back with your reply. Thanks!

288491358_automaticcurve.JPG.9dfb26b2a3730c14ff5d6a4e3bac0718.JPG

 

9 points curve.JPG

20 points curve.JPG

30 points curve.JPG

left up to 400Hz.JPG

9 points curve.dprj 20 points curve.dprj 30 points curve.dprj left right up to 400Hz.dprj

right up to 400Hz.JPG

Link to comment

Generally speaking, the target curve should be close to a straight line (like in your first image), usually with about a 1dB/octave slope from bottom to top range.
Trying to copy the bass ambient response with the detail shown in all the other images you posted is totally pointless: human hearing sensitivity is low in this area, while it is much higher in upper midrange (2000-5000 Hz), therefore fine adjustment of the frequency response in this area does make sense. (fine adjustment = + - 2-3 dB, always using "gentle" slopes).

Here are my Martin Logan Electromotion ESLs in my room: the red line is the reference target curve.

 

2017 - Martin_Logan_DIP100_con_tendenza.jpg

Link to comment
40 minutes ago, hellblau said:

Here are my Martin Logan Electromotion ESLs in my room: the red line is the reference target curve.

 

Have you compared with one not sloping downwards for HF?  The recommended curve is based on concert hall acoustics which is the sum of direct plus reverbs. IME, even a slight slope kills the HF. 

Link to comment

I've been using Digital Room Correction for more than ten years now (in the beginning with the TacT digital preamplifiers), so I could try different loudspeakers, different rooms and, yes, different target curves.
I said "generally speaking" and "usually" because your experience, the music you listen to, your sensibility end personal taste may vary, thus leading to different target curves and different ambient responses (including an "almost flat" frequency response).
What I can confirm is that targets with "mountains and valleys" cloning the ambient response are, in the best case, useless, especially in the bass range (20-300 Hz). The rest of the range can be "tailored", with slopes, valleys (the old fashioned "BBC dip" at 2000 Hz being perhaps the most famous example) or gentle peaks copying the ones of the natural response of the system (loudspeakers+room).  The less the target is above the measured response, the better.

Link to comment

I am unable to comment on TacT room correction as I have never used them. However, I did use REWeq extensively before completely abandoning them for no correction except by physical means. I also used another room correction program but I just couldn’t recall the name now. 

 

Just a note on BBC dip. It was introduced to remedy the cone resonance and had some effect of making the sound to be far away. You can have this effect when playing with EQ in the dip region. Unfortunately, the myth continues and some deliberately introduce the dip when there was no need.  

Link to comment

My advice is not to get too hung up on comparing small differences in target curves. Yes you will hear differences but so much depends on how the material you are playing was recorded/mastered in the first place and no curve will be perfect for everything you listen to. Quick comparisons will do your head in, trust me I went through it :). If you are comparing curves then make it more of a long term exercise ie. use one for weeks rather than hours or minutes before trying an alternative.

 

The target curve I have settled on just happened to be in place for a long period of time when my focus was shifted to other things outside my music interest and now I'm happy with it. I believe you will get used to a certain sound as long as it's reasonably close to optimised.

Link to comment
12 hours ago, hellblau said:

The BBC dip was introduced - as far as I know - to counterbalance the excess of upper high frequencies (2000-5000 Hz) due to close miking of - tipically - acoustic instruments (violin, piano, etc.), in comparison to the "natural" sound of the same instrument in an auditorium.

 

That is a myth. From the designer himself ( allegedly).

 

'

We're off walking in the Northumbrian hills today, so I have much time to relive and add to my comments.

One thing to be aware of which has direct bearing on your comments about the presence dip on the early Harwood-designed Harbeths (HL1, Mk2 etc.) is this: Dudley Harwood, unbeknown to me at the time I took over Harbeth, was a member of a fundamentalist Christian faith. He lived a simple life, and was disinterested in the trappings of modern consumer life. He was prohibited by the rules of his group to take any interest in entertainment, and that apparently included music of any and all types. Now, you may ask how a world-renowned loudspeaker engineer could evaluate loudspeakers, unable to listen to music on them - it is difficult to comprehend what work-arounds he used. But certainly speech recordings, which were permitted, became critical to his evaluatory process. Without the formal ongoing feedback from BBC users who you mention, it must have been very difficult for Harwood, now ex-BBC and running Harbeth, to be sure about what was a correct speaker response balance bwteen bass, mid and top. If you recall, in one HFC, he was criticised for having set the tweeter level too high (Mk3?) and this was surely a direct consequence of working alone, and not listening to music and natural hearing loss with age.

However, judging, controlling and shaping the energy in the presence region (say, 1-3kHz) and the position of musical performers front-back (z-plane) is much easier in stereo, and with music. Also, again this did come up in conversation with Harwood and is in the transcript, I asked him 'what is the purpose and function of the so-called Gundry-dip where energy is reduced in the (typically) 1-2kHz region'? His answer was unequivocal .... "to mask coloration in the polypropylene cones'. This was the first admission that even though he was the father and patentee of PP cones, they had known deficiencies - quite an admission.

The BC1 had a bextrene cone, where coloration in broadly that region was ameliorated with a thick coating of PVA 'dope'.
Alan A. Shaw
Designer, owner
Harbeth Audio UK
p.s. I was an active member of Harbeth forum then. That's why I remember. :) 

 

Link to comment
On 5/13/2019 at 9:54 AM, STC said:

 

Have you compared with one not sloping downwards for HF?  The recommended curve is based on concert hall acoustics which is the sum of direct plus reverbs. IME, even a slight slope kills the HF. 

 

There's a trend for speaker to exaggerate the treble either by shelving it up a few dB or by producing a rising response from the presence region onwards. This is probably market-driven, since audiophiles have become somewhat obsessed with "spaciousness" and "air".

But I would disagree that this sounds more natural/realistic. It's an effect, one that engineers use in their pop/rock mixes.

 

Perhaps the fact that many audiophiles listen to music at overly loud levels is helping with this trend as well by accelerating hearing loss at higher frequencies...

 

 

Here's the Olive/Toole target curve in red superimposed to the B&K target curve, showing the "ideal" in-room response of a speaker with a flat anechoic frequency response measured from the listening spot:

 

478237d1434900455-room-curve-after-room-

 

"Science draws the wave, poetry fills it with water" Teixeira de Pascoaes

 

HQPlayer Desktop / Mac mini → Intona 7054 → RME ADI-2 DAC FS (DSD256)

Link to comment
On 5/14/2019 at 2:23 AM, STC said:

 

That is a myth. From the designer himself ( allegedly).

 

'

We're off walking in the Northumbrian hills today, so I have much time to relive and add to my comments.

One thing to be aware of which has direct bearing on your comments about the presence dip on the early Harwood-designed Harbeths (HL1, Mk2 etc.) is this: Dudley Harwood, unbeknown to me at the time I took over Harbeth, was a member of a fundamentalist Christian faith. He lived a simple life, and was disinterested in the trappings of modern consumer life. He was prohibited by the rules of his group to take any interest in entertainment, and that apparently included music of any and all types. Now, you may ask how a world-renowned loudspeaker engineer could evaluate loudspeakers, unable to listen to music on them - it is difficult to comprehend what work-arounds he used. But certainly speech recordings, which were permitted, became critical to his evaluatory process. Without the formal ongoing feedback from BBC users who you mention, it must have been very difficult for Harwood, now ex-BBC and running Harbeth, to be sure about what was a correct speaker response balance bwteen bass, mid and top. If you recall, in one HFC, he was criticised for having set the tweeter level too high (Mk3?) and this was surely a direct consequence of working alone, and not listening to music and natural hearing loss with age.

However, judging, controlling and shaping the energy in the presence region (say, 1-3kHz) and the position of musical performers front-back (z-plane) is much easier in stereo, and with music. Also, again this did come up in conversation with Harwood and is in the transcript, I asked him 'what is the purpose and function of the so-called Gundry-dip where energy is reduced in the (typically) 1-2kHz region'? His answer was unequivocal .... "to mask coloration in the polypropylene cones'. This was the first admission that even though he was the father and patentee of PP cones, they had known deficiencies - quite an admission.

The BC1 had a bextrene cone, where coloration in broadly that region was ameliorated with a thick coating of PVA 'dope'.
Alan A. Shaw
Designer, owner
Harbeth Audio UK
p.s. I was an active member of Harbeth forum then. That's why I remember. :) 

 

 

Shaw is a bit of a salesman, I'm affraid, so I'd take his word with a pinch of salt...

 

From the designer himself (not allegedly):

 

Some factors in loudspeaker quality

by H. D. Harwood, Wireless World May 1976

 

2wmqfxw.thumb.jpg.0f33669d58ad4b05a42c50be7f54b284.jpg

"Science draws the wave, poetry fills it with water" Teixeira de Pascoaes

 

HQPlayer Desktop / Mac mini → Intona 7054 → RME ADI-2 DAC FS (DSD256)

Link to comment
38 minutes ago, semente said:

 

Shaw is a bit of a salesman, I'm affraid, so I'd take his word with a pinch of salt...

 

From the designer himself (not allegedly):

 

Some factors in loudspeaker quality

by H. D. Harwood, Wireless World May 1976

 

2wmqfxw.thumb.jpg.0f33669d58ad4b05a42c50be7f54b284.jpg

 Do you have the full text? As far as I remember the dip at that region is also where crossover is set. I do not have all the papers with me right now since I crashed three drives recently but I am sure there was another interview about the dip. Anyway, no one would happily admit their invention was flawed. 

 

If the dip is crucial then it would continue to exist even now in modern BBC inspired speakers such as Harbeth and Spendor. 

 

Also so please include the link for Toole room response. IIRC, it was the measurement at ear level. An UMIK mic measurement at LP and one at ear level varies greatly. 

 

I have had many interaction with Alan Shaw including one to one meeting and he is not a salesman not a snake oil peddler. He is straight forward and doesn’t toy with words to misrepresent. 

Link to comment

Here is another one from Gundry’s son. 

 

A new range of speakers was developed, but it is possible that at least for those first ones, the uniformity was considered more important than perfect flatness, and thus the speakers may have shown the "Gundry dip". However it would not have been a design aim but a side-effect, and in any case my father would have had no input to the designs, which were developed at the BBC Research Department (Dudley Harwood, Spencer Hughes et al.)

Kenneth Gundry, San Francisco

Link to comment
3 hours ago, STC said:

 Do you have the full text? As far as I remember the dip at that region is also where crossover is set. I do not have all the papers with me right now since I crashed three drives recently but I am sure there was another interview about the dip. Anyway, no one would happily admit their invention was flawed. 

 

If the dip is crucial then it would continue to exist even now in modern BBC inspired speakers such as Harbeth and Spendor. 

 

Also so please include the link for Toole room response. IIRC, it was the measurement at ear level. An UMIK mic measurement at LP and one at ear level varies greatly. 

 

I have had many interaction with Alan Shaw including one to one meeting and he is not a salesman not a snake oil peddler. He is straight forward and doesn’t toy with words to misrepresent. 

 

Here you go:

 

 

factors_harwood.pdf

"Science draws the wave, poetry fills it with water" Teixeira de Pascoaes

 

HQPlayer Desktop / Mac mini → Intona 7054 → RME ADI-2 DAC FS (DSD256)

Link to comment
3 hours ago, STC said:

 Do you have the full text? As far as I remember the dip at that region is also where crossover is set. I do not have all the papers with me right now since I crashed three drives recently but I am sure there was another interview about the dip. Anyway, no one would happily admit their invention was flawed. 

 

If the dip is crucial then it would continue to exist even now in modern BBC inspired speakers such as Harbeth and Spendor.

 

All BBC designs show a dip in the presence region off-axis but some are flat on-axis (i.e my Stirlings).

 

Harwood says that the dip in the presence, which produces a perceived recession of the soundstage, makes orchestral music sound more natural.

And he, as well as the other researchers at the BBC RD should know because they could perform direct comparisons between the sound of the speakers and live orchestras.

 

3 hours ago, STC said:

Also so please include the link for Toole room response. IIRC, it was the measurement at ear level. An UMIK mic measurement at LP and one at ear level varies greatly.

 

The Olive/Toole curve has apparently been redesigned as per the Mitchco plot above.

 

3 hours ago, STC said:

I have had many interaction with Alan Shaw including one to one meeting and he is not a salesman not a snake oil peddler. He is straight forward and doesn’t toy with words to misrepresent. 

 

He is not a snake oil peddler but he does twist the truth somewhat if it suits his busyness... Just go on the HUG forum.

"Science draws the wave, poetry fills it with water" Teixeira de Pascoaes

 

HQPlayer Desktop / Mac mini → Intona 7054 → RME ADI-2 DAC FS (DSD256)

Link to comment
1 hour ago, semente said:

Harwood says that the dip in the presence, which produces a perceived recession of the soundstage, makes orchestral music sound more natural.

And he, as well as the other researchers at the BBC RD should know because they could perform direct comparisons between the sound of the speakers and live orchestras.

 

After reading his paper all over again, I am now inclined to believe Alan Shaw’s version. 

 

There were factual errors in his paper. Look at two paragraphs prior to where you highlighted. Since when Omni directional speakers cannot  produce first quality stereo? MBL Radial is a myth?  My own setup thought to be more accurate with MBL omnidirectional stereo speakers. 

 

It is normal for human to perceive distance when the high frequencies are lessen as distance is perceived with loss of HF and reverbs but to justify the speaker designs doesn’t make sense. 

 

Furthermore, in the immediately preceding paragraph he mentioned that stereo cannot produce the wave-front produced in concert hall. But that was an issue that Blumlein mentioned from the very onset of stereophonics. 

 

Thanks for for bringing to my attention of this paper. It is enlightening to read them all over again from a different perspective. 

 

Link to comment
18 hours ago, mitchco said:

Toole and Olive spent a great deal of R&D effort over years that included controlled subjective listening tests. I wrote a summary of that here, which Sean Olive peer reviewed before it was published:

https://audiophilestyle.com/ca/reviews/nad-viso-hp50-with-roomfeel-headphone-review-r720/#science

 

My preference is for a flat, but tilted target response from 20 Hz to -10 dB @ 20 kHz as measured in-room at the listening position. According to Harman’s research, this objectively measured flat, but tilted frequency response, (i.e. spectral balance), is subjectively perceived by our ears as a flat or neutral or accurate response. The deets are in the link above.

 

Another good read is Toole's open access AES paper on The Measurement and Calibration of Sound Reproducing Systems: http://www.aes.org/e-lib/browse.cfm?elib=17839  Check out Fig. 14. Subjectively preferred steady-state room curve targets in a typical domestic listening room and the "trained listeners" preference.

 

Depending on the speakers directivity and how damp/reflective ones room is, there maybe some fine tuning of how much more or less tilt is required by a couple of dB. This is easily accomplished by moving up or down the -10 dB @ 20 kHz  point in the target response and then give it a listen and compare. 

 

I have tried many targets over the years, including the ones @semente linked above from the ASR forum. The straight line, but tilted target response sounds the most neutral to my ears. I think Toole and Olive's R&D in this area is the most credible, and really the only body of work I know of that uses subjective listening tests to correlate their objective measurements. They have repeated these tests over and over again, with different speakers and listening subjects and end up with virtually the same results each time. The first link above points to the body of work at the end of the article section in presentations, which points to over a dozen AES papers on the subject area.

 

Good luck and have fun!

 

 

Do you remember the measurement method by Toole?  IIRC, correctly the measurement was taken at ear level. I have the frequency response of sound from a speaker at 30 degrees to the left have two different frequency response at left and right level by Toole. 

 

Was the measurement taken on axis of the speaker or at ear level?  

Link to comment

I tried a lot of different curves. Also OliveToole and EBU. After a bit of listening EBU sounds good. But when I turn Dirac off the music livens up a lot and I get more air. So for me the EBU curve is to severe. I want to keep more of the caracter of the speakers. 

The other extreme is to use Dirac only to get the bass around 50Hz in check. Suppressing the bass around 50Hz really helps a lot and is the reason I will continue using Dirac. 

But Dirac can do more. In my setup with Dirac I get a better sound scape between the speakers. With Dirac off the music is focussed slightly to the right. So I've been looking for a curve that keeps the bass in check and centers the music but preserves the sense of air. 

I think I found this by a curve that reduces the bass than follows the measured response up to 5kHz and no correction above 5kHz. 

Link to comment
  • 3 weeks later...

Having settled for two weeks with my best curve I began noticing some artifacts in certain songs. For example in the track "The lighthouse and the cabin" from Third coast Percussion there is a lot of resonation at 1:05. At that point I get a sudden distortion. The rest of the track is okay. Used my phone to find it's a distortion on a 100Hz sound. With other tracks I did not hear any distortion. But with Dirac off the distortion disappeared. When I found another track with distortion it was time to take action.

 

First I lowered the total curve by 2dB. That made a difference but did not totally remove the distortion. So I made a dip in the curve at 100Hz where the original response got clearly below the target curve. Now the distortions are completely gone. 

Screen Shot 06-05-19 dip.JPG

Link to comment

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now



×
×
  • Create New...