Dirac Live Users Discussion Thread

[davewantsmoore]

8 hours ago, gibbo9000 said:

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

I don't think this is a great idea..... although as an acedemic excecise (to see what it sounds like), then go for it.

 

Quote

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.

 

I think you might be confused.... now that I read you carefully.

 

Whatever you draw as your target curve.... that is what your speaker response will be.

ie. what you measure with REW after correction.... should look exactly like the target curve you specify.

 

Said another way.   The "target curve" in DL is not the "EQ curve" that it will apply as a correction..... it is the "desired result"

 

I hope that helps.

Quote

raised on how well Dirac Live actually derives filters that deliver the target response curve in practice.  And thirdly, does it sound any better? 

DL corrects the respons of your speaker to match the target.... "perfectly".   (ie. very very close).

 

What GS is talking about.... is that the selection of the target (ie. the result you ask for)  is critical to how you will perceive the sound.   How much lift in the mid/low frequencies vs high frequencies, and where you begin it.... changes the sound immensely.  1dB can be a lot.

 

However, I disagree that starting with a "toole" or "harman" type curve isn't a good idea.   It is.   Start with one of those curves (I'd try toole).   Don't be afraid to modify it some.

 

 

Edited June 13, 2020 by davewantsmoore

[davewantsmoore]

On 12/06/2020 at 10:34 AM, Grant Slack said:

First thought: the bottom chart shows that the Dirac filters are not actually delivering anything very close to the intended curve. 

That isn't what I see.   They look fine.

 

(Note: the scale in the REW charts is very large, making it difficult to see/read.)

 

 

On 12/06/2020 at 10:34 AM, Grant Slack said:

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.

I strongly agree with this in general.

 

However DIrac Live isn't exactly a "dumb equaliser".... and there is more judgement contained within than a simple:  data -> target.    That being said, there is no substitute for good accoustic design.    The paradox being that speaker/room correction is less likely to run into trouble when the setup is very bad (anything but a very hamfisted correction, will be better than uncorrected), or very good (there are less acoustic traps to fall into).

[gillmaverick]

I have been playing around a fair bit with the Dirac Curves in my NAD C-658 and finally have settled down on the one that follows the natural curves of my speakers ( Paradigm Tributes ). I did not like the flat response of the standard dirac curve So all I did was lift and drop the target curve as per the Natural response detected of both my speakers during measurement. After going back and forth comparing all different curves , I have finally settled on the one that follows speakers natural measurement. I have not tinkered with it for a while and It still is my favorite. My 2 cents.

[davewantsmoore]

16 hours ago, gillmaverick said:

 I did not like the flat response of the standard dirac curve

That standard dirac target curve is not flat.    Perhaps you loaded one which is flat above ~400Hz like Gibbo did.

 

Here is the text from the user guide, note red.

 

 

Quote

Even small changes in the target magnitude response give more or less audible results. It is worthwhile to tweak the target in order to address even the smallest nuance in the magnitude response. 31 Avoid sharp peaks in the target curve, as they may result in annoying tonal components in the frequency response. Generally you want to have a smooth target curve. A target curve that is slightly tilting down towards high frequencies (like the auto-target) is often preferable - a flat target often sounds too bright. A loudspeaker with a flat on-axis response will usually have a slightly tilting in-room response. You can experiment with the amount of tilt, especially in the treble region, to get a good timbral balance in your room. If you have strong room-resonances in the low frequency region, completely eliminating these may make the sound thin in comparison. You may want to apply a slightly increasing low-end below 100 Hz. If you have dips in the magnitude response, sometimes it sounds better to not fill these in completely.

 

[gibbo9000]

I guess I may be a typical example of somehow getting lost in the 'which curve' maze, somewhat ill-informed, and creating worse outcomes than I started with.

 

The standard Dirac Live curve in my set up is straight line from +2db to -2db from 20Hz to 20kHz (on log frequency scale), and my testing confirms that Dirac Live pretty much delivers against the target curve.  In my case that meant a significant boost above 500Hz relative to no correction on my speakers - giving a harsh, lightweight sounding outcome.  And when trying other curves looks like I got caught out finding some that turned out to be low range only and flat above 400Hz - giving the same harsh outcome.  Some 'traps' I now understand from the guidance above.

 

So I went back to basics and constructed my own 'natural' curve that simply tried to follow my speakers underlying in room response (sub taken out of action for the time being) with the aim that applying it would just iron out a few bumps without changing the overall character of the sound - and it did just that (it crimps a little bit below 100Hz, but nothing major).  The two graphs below show the Dirac view, and the REW view of before and after (before in grey, after in red).

 

 

It tightened bass, while leaving the rest broadly alone - a good outcome.

I then set about some testing of boosting the bass.  Looks like Harman, NAD etc. curves limit bass boost to less than 200Hz so I tried some options between 200 and 500Hz and ended up closer to 200Hz cut off for boosting based on some quick listening tests.  Overall the result is what I missed with prior curves - speakers character remaining intact through mid and top end, but with coherent bass boost.   Below shows the natural curve vs 4db boost curve as they measure in REW.  4db feels a bit too heavy in practice, but 2db feels pretty good.  Now for some more extended listening.

 

Also thinking I may want to try some minor PEQ adjustments to the response profile for each speaker below 100Hz before feeding it to Dirac.  And will also try bringing the top curtain for Dirac optimising back to 1kHz or so -- just not sure how it handles the transition in practice as it looks a bit ugly in the Dirac display.

 

Overall much happier - now the Oscar Peterson Trio are back to sounding how they should - only better.

Gibbo

[davewantsmoore]

21 hours ago, gibbo9000 said:

Also thinking I may want to try some minor PEQ adjustments to the response profile for each speaker below 100Hz before feeding

it to Dirac. 

Dirac is going to transform the speaker response to your target curve.... so its largely going to "undo" your EQ.

 

If you want to adjust the result.... adust  the target curve.

 

[gibbo9000]

3 hours ago, davewantsmoore said:

Dirac is going to transform the speaker response to your target curve.... so its largely going to "undo" your EQ.

Agree that the simplest is to leave the whole lot to Dirac to sort out in one hit.  A couple of questions from things I have picked up elsewhere that I haven't yet calibrated on the fact or myth spectrum:

• The Dirac adjusted responses leave some clear 'holes' around 50, 80 and 120Hz.  From reading I gather there is a school of thought that the human ear is less sensitive to dips than peaks, and fixing dips can do more damage than good.  So not sure whether the dips are deliberately left, or a function of Dirac removing the peaks and having to create a small dip as part of its overall solution.
• With active speakers there also seems to be some preference to move core parts of the PEQ to the speakers DSP rather than in the upstream processor (miniDSP in my case).  The speaker DSPs are already doing there own thing so I guess it may be about integration closer to what they are doing?
• I am leaning towards viewing REW graphs with filters in place - some seem to recommend using the psycho acoustic filter in REW that adjusts smoothing by frequency.
• Are Dirac (or other measurement / corrections) best done with the listening area clear of obstacles (so adjusting for primary room resonances) or with the obstacles there.  In my case it is a large glass topped coffee table I am required not to put a cloth over!

Gibbo

[Grant Slack]

On 12/06/2020 at 10:04 AM, Grant Slack said:

....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....

 

On 13/06/2020 at 7:21 PM, davewantsmoore said:

That isn't what I see.   They look fine. (Note: the scale in the REW charts is very large, making it difficult to see/read.)

Hello Dave, it is what I see. I have overlaid two of David's graphs (and matched scales), to show what I mean (that the measurement is not very close to the prediction).

 

If they were fine, the red line (measured after Dirac) should very closely track the dark green line (predicted after Dirac), which is kind of hiding behind the yellow target line.

Bear in mind, the grid size is 10 dB.

 

I thought that was a big enough deviation from prediction, to comment on.

 

cheers

Grant

[Grant Slack]

48 minutes ago, gibbo9000 said:

• With active speakers there also seems to be some preference to move core parts of the PEQ to the speakers DSP rather than in the upstream processor (miniDSP in my case).  The speaker DSPs are already doing there own thing so I guess it may be about integration closer to what they are doing?

Hi David,

 

tell us a bit about your speakers. If they are active speakers with manufacturer-embedded DSP, and the manufacturer has done a competent job, then I would be inclined to only run Dirac, to a target, below transition frequency.

 

regards,

Grant

[davewantsmoore]

18 minutes ago, Grant Slack said:

 

If they were fine, the red line (measured after Dirac) should very closely track the dark green line (predicted after Dirac), which is kind of hiding behind the yellow target line.

Bear in mind, the grid size is 10 dB.

 

I thought that was a big enough deviation from prediction, to comment on.

Ahh.  I see what you mean.

 

What I was comparing was the green vs red in the last chart.... which is showig that the two different target responses are working.   ie. the diffrnce between green and red, is the difference between the two target responses.

 

 

I had ignored the whole issue of what you're commenting on.   It's clear that there is some amount of microphone movement (or something) happening.... as we can see that the measured response into Dirac Live is diffrent in both cases (the first two charts in the opening post).

 

 

IME (and everyone esle I've encountered) when done carefully.... the measured results (eg. using REW) will match almost idetnically to the predicted response (ie. the one hiding behind the target curve line).

 

[davewantsmoore]

19 minutes ago, Grant Slack said:

then I would be inclined to only run Dirac, to a target, below transition frequency.

I'm conflicted by this sort of advice.... but in the end, I think I begrudingly agree for most people.

 

As I've mention a few times.... a small number of dB in the target curve.... has  huge effect on the sound.

 

This means that the "toe in" of the speakers.... and the target curve are crtical, and have a profound effect when they are "wrong".

 

 

... and so often better left alone, as opposed to "worse".   

[gibbo9000]

14 minutes ago, davewantsmoore said:

It's clear that there is some amount of microphone movement (or something) happening.... as we can see that the measured response into Dirac Live is different in both cases (the first two charts in the opening post).

That is my bad - looks like I have grabbed one from right channel and one from left channel as Dirac separates the two.  Beginning to recognise the footprint of each now!

10 minutes ago, davewantsmoore said:

As I've mention a few times.... a small number of dB in the target curve.... has  huge effect on the sound.

Agreed - now I am not inadvertently messing with the frequencies >400Hz differences of +/- 1db below that are very significant in practice. 

36 minutes ago, Grant Slack said:

tell us a bit about your speakers. If they are active speakers with manufacturer-embedded DSP, and the manufacturer has done a competent job, then I would be inclined to only run Dirac, to a target, below transition frequency.

Speakers are Kii Three's - so sophisticated DSPs managing cardioid output down to 50 Hz.  If anything they are slightly limited in bass 20 - 40Hz at high SPL where their inbuilt electronics will act to limit power.  So short of mega $ for their BXT extensions a few have been working on sub woofers at 30 or 50 Hz cross over (I am using a Rhythmik sub that is solid down that low) to help unload the Kii's very low down while not messing with their cardioid distribution.  Some of the Dirac learning has been a detour from that goal (which I am finding very valuable) but I am rapidly heading back to where you guys seem to be aligned - Dirac intervention low down to create/control boosted response, with sub supporting the added low down load. 

 

Gibbo

[Grant Slack]

Hi David,

 

yes, those speakers I would consider sorted above the transition area 200-300 Hz.

 

All you will achieve, with Dirac above transition range, is worsen the on-axis response (#1 priority) in exchange for  smoother total sound response (#2 priority). Not a good idea IMHO.

 

cheers

Grant

[Satanica]

On 15/06/2020 at 8:12 PM, Grant Slack said:

All you will achieve, with Dirac above transition range, is worsen the on-axis response (#1 priority) in exchange for  smoother total sound response (#2 priority). Not a good idea IMHO.

Hi, regarding Dirac Live do you know how much "weight" on-axis measurements versus off axis measurements go towards equalisation?

Taking into consideration you can choose from three options (Tightly Focused Imaging, Focused Imaging and Wide Imaging).

Also, with Dirac you can choose to do a single measurement and although it will complain that you should do more it can then provide equalisation based on a single measurement which presumably the user would do on-axis and at the main listening position.

[Grant Slack]

I’m certainly interested in your question, but don’t know the answer, sorry.

 

cheers

Grant

[davewantsmoore]

On 19/10/2020 at 1:06 PM, Satanica said:

Hi, regarding Dirac Live do you know how much "weight" on-axis measurements versus off axis measurements go towards equalisation?

It isn't a simple answer to this question.    The measurement points you feed it are optimised in a group.... the "off axis" sound is a component of the later arriving sound (ie. the reflected/room sound) and this is considered in the optmisation.

On 19/10/2020 at 1:06 PM, Satanica said:

Also, with Dirac you can choose to do a single measurement and although it will complain that you should do more it can then provide equalisation based on a single measurement which presumably the user would do on-axis and at the main listening position.

The "off axis" sound is still recorded in this one position.....  in so far as the reflected sound from the room (which is comproised of the "off axis" sound) also enters the microphone.

[Satanica]

2 hours ago, davewantsmoore said:

It isn't a simple answer to this question.    The measurement points you feed it are optimised in a group.... the "off axis" sound is a component of the later arriving sound (ie. the reflected/room sound) and this is considered in the optmisation.

The "off axis" sound is still recorded in this one position.....  in so far as the reflected sound from the room (which is comproised of the "off axis" sound) also enters the microphone.

 

OK, thinking some more I hope it is fair to say room measurements are neither on-axis or of-axis as they include content from both as opposed to actual speaker measurements. 

[davewantsmoore]

2 hours ago, Satanica said:

OK, thinking some more I hope it is fair to say room measurements are neither on-axis or of-axis as they include content from both as opposed to actual speaker measurements. 

Sort of.   The initial part of the response is almost all speaker.... and the end is all room.     How loud they are relative to each other, and what the delay is, depends on where you measure.

 

You can look at which ever part of the response (the start, or the end, or both, or whatever) you want to.

[Midget]

Newb questions about MiniDSP SHD with Dirac Live to improve the sound in my living room as I am considering getting one. One speaker is kind of stuck in the corner so I would like to even out the bass response , plus extract any additional bass from the mini floorstanders that I am using as my subwoofer is currently out of the system due to space constraints.

 

1. Does it require a substantially powerful power amp to boost the bass frequencies, or does the strain happen in the pre amp side of the path? I currently have a Purifi which is 425w at 4ohms. 

 

2. Is there a limit on how many dB's you can or should boost the bass signal ie. +10 dB? What determines this and can a more powerful power amp improve this? ie. Hypex NC1200, or is it more about the speaker or pre amp section?

 

3. Do speaker ports play any role? Since the ports are tuned to a frequency can they be boosted? Would a sealed speaker be a better candidate since they actually go lower but suffer from SPL output which the DSP makes up for? I noticed some actives are sealed designs.

[Snoopy8]

10 hours ago, Midget said:

Newb questions about MiniDSP SHD with Dirac Live to improve the sound in my living room as I am considering getting one. One speaker is kind of stuck in the corner so I would like to even out the bass response , plus extract any additional bass from the mini floorstanders that I am using as my subwoofer is currently out of the system due to space constraints.

The MiniDSP SHD with Dirac Live is not the silver bullet that you are looking for. 

 

Uneven bass is due to the ROOM and how your speakers interact with the room.  If you have a bass trough in the room, it cannot be corrected by boosting bass!  If you try boosting the bass, it will lead to distortion and may damage your equipment.  Please forget about boosting bass with more power amps, extracting additional bass from the mini floor standers, playing around with speaker ports.

 

Getting the Anti Mode 2.0 will help reduce some of the problems.  Some of the reviews do explain in more detail how it works.

http://www.dspeaker.com/en/products/20-dual-core.shtml

 

Yes, you have space restrictions, but adding a small sub such as the SVS 3000 Micro will go a long way to getting even bass. 

[Midget]

So DSP can not increase the natural roll off of a speaker? I thought that was one of the things that separated non active and active versions of the same speaker design?

 

Where does the distortion come from? The preamp that is adding the boost, the power amp, or the speaker driver?

 

In regards to the uneven bass, I thought the speaker in the corner might have too much bass. I already have the port plugged but thought there might be a way to reduce bass using DSP to make it more even.

 

Thanks for the link Snoop, I'll look into it!

[Snoopy8]

I think you are misunderstanding how bass interacts with the room.  The troughs are caused by sound waves interacting with the walls, ceilings, and cancelling each other in different parts of the room.  Boosting a bass signal will not remove the interaction.  If you go down the path of over boosting a bass signal, you may damage your amp or speakers.

 

Please have a read of these reviews; it explains things better than I can.

https://www.theabsolutesound.com/articles/dspeaker-anti-mode-20-dualcore-digital-signal-processor

https://www.avforums.com/reviews/dspeaker-anti-mode-2-0-dual-core-review.355

The second review also covers a bit about waterfall diagrams, which cover why bass can get muddied by long decay times.

[Midget]

Thanks @Snoopy8

 

I have read both articles and it sounds like I was on the right track. It did fill in the blanks and also confirmed some of my preconceived thoughts. I'll summarise to see if I understood it correctly and post up a few follow up questions that I am still curious about.

 

• Boosting bass frequencies puts stress on the both the power amp and the speaker
• Evening out the response is more safely handled by decreasing the spikes
• Increasing big divots can not be corrected by boosting as it is more about room interactions and speaker placement
• Repositioning a speaker is one solution to fixing these big divots in frequency response
• Fixing natural roll off of speakers can be corrected for by increasing the lower frequencies or lowering the higher frequencies
• This is what is called a 'house curve' in which the bass is attempted to be corrected to be flat and the overall response tilted downwards
• DSP works best for 300Hz and below as you are mostly correcting for the room. Correcting too much at higher frequencies can result in a negative effect

A few follow ups:

 

1. Power amp choice - Am I correct that if any boosting is done, instead of lowering everything to create the desired curve, that class D would be a good choice due to its lower thermal properties? Again, does overall power give more headroom?

 

2. Only reducing - Am I correct that you could create the desired response by only using reductions. The tradeoff would be that the overall output of the system might be reduced but neither the amp or speaker would be driven beyond its capabilities?

 

 3. Using one speaker to correct the other - My understanding is that the overall bass response is an accumulated effect from both speakers. The Dual Core seems to be limited by applying the same adjustments to both speakers. If one speaker had a null but the other didn't, would a seperate adjustment to each not allow the speaker that doesn't have a null to correct for the one that does? (To a certain degree of course within the limits of the boost and overall reductions)

 

4. Volume dependant correction - Active speakers tend to lose their DSP correction when they get closer to their limits. I take it this is to protect both the amp and the speaker. It doesn't look like the Dual Core does this, but does the MiniDSP SHD? I would also like the functionality of even more bass boost at very low volumes akin to 'Loudness' but only in the bass region. If it's not automatic then maybe three presets is the answer.

 

5. Recommended limit of boosting - The Dual Core appears to have pre sets for how much it corrects by boosting, "normal/typical" and "maximum". How many dB is maximum do you think? I take it this is what they believe is the most you can do safely.

[Snoopy8]

I do not boost any frequencies and solve the bass problems in my living room with 2 subs and a DSP.

 

Yes, you have limited space, but you can improve things by changing the speaker positions,  by as little as 0.5m, and also your listening position.  Bass is uneven in the room, because of the room. No amount of boosting of existing sources will fix bass problems. 

 

Room treatment and adding more subs, with DSP are potential solutions.  The key is to get even bass, not loud bass.

 

You will need to find someone else to discuss boosting the bass...

 

[Midget]

46 minutes ago, Snoopy8 said:

Room treatment and adding more subs, with DSP are potential solutions.  The key is to get even bass, not loud bass.

 

You will need to find someone else to discuss boosting the bass...

 

 

No worries Snoop I will wait to see if anyone else has advice, however the articles you referenced clearly discuss bass boosting and the inadequacies of typical room treatment for these lower frequencies. I am just trying to understand what options I have if the other methods for correction are not an option, and how to safely do it. (I have a sub with built in DSP room correction but I do not want to use it due to space constraints). 

 

Of course my goal is to get even bass as previously stated, but also to correct for bass roll off which dips below that even frequency response target we all would like to achieve. The only way to do it is for the frequencies below this target to get louder, which is different to "loud bass" which I would describe as a mismatch in frequency response heavily favouring the lower regions. 

 

I appreciate your time though, thank you.