Rooms at the show using DSP

[Keith_W]

9 hours ago, RoHo said:

A point that was brought up earlier but not really expanded on- how effective can the typical domestic DSP be in this very specialised situation where a large percentage of the room is “listening position”?  Do you set the LP to be the middle few seats?

 

Yes. There is some debate in the DSP community over whether we should correct over a larger listening position vs. a smaller one. There are three ways you can take measurements - (1) single point microphone sweep, (2) moving mic method, (3) multi-point averaged. There are advantages and disadvantages to each approach. 

 

Single point measurement: set up the mic on a tripod and take a single sweep. Advantage: easy to do. Gives you timing and phase information. Disadvantages: only corrects for one point, poor reproducibility. No less an authority than Toole has said that doing a single point sweep is wrong. However, this approach is recommended by Uli Brueggemann (author of Acourate), Bernt (author of Audiolense), Mitch Barnett, and others. 

 

Moving mic method: hold the mic in your hand, play pink noise, and wave the mic around the listening position. Advantages: easy to do, in fact it is faster than a single point sweep. Excellent reproducibility. Disadvantages: no timing or phase information, so you will need to do a SPS anyway. 

 

Multi-point averaged: set your mic on a tripod and take multiple single point sweeps across the listening position. Average the sweeps and correct for that. Advantages: corrects for a wider area and gives each listener better correction (this is setup dependent though). Disadvantages: time consuming, no timing or phase information, may correct for unwanted artefacts. 

 

Among the people who DSP in Melbourne, I am aware that @Ittaku and @andyr do multi-point averaged sweeps. @frednork does single point. Me personally, I have tried all 3 but prefer single point. I have even tried using single point to do the time alignment, and then multi-point or MMM to adjust the target curve. It sounds slightly different to single point, but in my setup the difference is not much because the sofa is not wide compared to the angle of the speakers. Both @andyr and @Ittaku have large wide sofas where you could be sitting on one extreme of the speaker to another, so I am sure that the difference would be more dramatic with their setups. 

 

Basically: some experimentation is required, and the type of measurement and correction you choose is situation and setup dependent. I don't think what they are doing is wrong, it is the right approach for their situation which is different to mine. To answer your question over whether you should correct over a larger listening area: it may or may not yield better results; and it depends on who you are going to plonk in unfavourable positions in your system. If it's your family and they don't care about sound, I would optimize over a smaller area - one or a few listeners will get a better result at the expense of everyone else. If it's for an SNA GTG, or the SNA hifi show, I would optimize for as many seats as possible. 

[sir sanders zingmore]

22 hours ago, frednork said:

or changing the natural character of the speakers or other things?

This is a broader question that does my head in a little bit! 
What can you say about the “sound” of your speakers if you can change it almost at will via DSP?

The way I rationalise it is that it’s not at all controversial to adjust the sound by other means such as room treatment, amps, dacs, cables etc. DSP is just another method. 
 

With that in mind, I don’t think it’s “cheating” at all to use DSP at a show. I’m firmly of the opinion that it’s impossible at a show to say what an individual component sounds like. You are listening to a combination of many unfamiliar components in an unfamiliar room so I think all you can comment on is how the system sounds. DSP is a single part of that system just like any other. 

 

[acg]

4 minutes ago, sir sanders zingmore said:

What can you say about the “sound” of your speakers if you can change it almost at will via DSP?

 

There are limits, of course, to how DSP can be used to change the sound of your speakers and the type of changes posssible, but in my opinion the primary use of DSP in this situation is as a room integration tool.  The same speaker will sound different in various rooms because of room modal behavior and the acoustic properties of the space and also according to speaker and listener positions within each room.  DSP is a convenient way of ameliorating some of these variances...not all...but some...also with varying degrees of effectiveness in each space and position. 

 

 

[POV]

1 hour ago, acg said:

 

There are limits, of course, to how DSP can be used to change the sound of your speakers and the type of changes posssible, but in my opinion the primary use of DSP in this situation is as a room integration tool.  The same speaker will sound different in various rooms because of room modal behavior and the acoustic properties of the space and also according to speaker and listener positions within each room.  DSP is a convenient way of ameliorating some of these variances...not all...but some...also with varying degrees of effectiveness in each space and position. 

 

 

Just to clarify for everyone, which DSP products do you have direct experience with (as in running the measurements, setting up profiles, comparing single position to multi position measurements etc)

 

I’m curious what your point of reference is?

[RoHo]

3 hours ago, Keith_W said:

 

Yes. There is some debate in the DSP community over whether we should correct over a larger listening position vs. a smaller one. There are three ways you can take measurements - (1) single point microphone sweep, (2) moving mic method, (3) multi-point averaged. There are advantages and disadvantages to each approach. 

 

Single point measurement: set up the mic on a tripod and take a single sweep. Advantage: easy to do. Gives you timing and phase information. Disadvantages: only corrects for one point, poor reproducibility. No less an authority than Toole has said that doing a single point sweep is wrong. However, this approach is recommended by Uli Brueggemann (author of Acourate), Bernt (author of Audiolense), Mitch Barnett, and others. 

 

Moving mic method: hold the mic in your hand, play pink noise, and wave the mic around the listening position. Advantages: easy to do, in fact it is faster than a single point sweep. Excellent reproducibility. Disadvantages: no timing or phase information, so you will need to do a SPS anyway. 

 

Multi-point averaged: set your mic on a tripod and take multiple single point sweeps across the listening position. Average the sweeps and correct for that. Advantages: corrects for a wider area and gives each listener better correction (this is setup dependent though). Disadvantages: time consuming, no timing or phase information, may correct for unwanted artefacts. 

 

Among the people who DSP in Melbourne, I am aware that @Ittaku and @andyr do multi-point averaged sweeps. @frednork does single point. Me personally, I have tried all 3 but prefer single point. I have even tried using single point to do the time alignment, and then multi-point or MMM to adjust the target curve. It sounds slightly different to single point, but in my setup the difference is not much because the sofa is not wide compared to the angle of the speakers. Both @andyr and @Ittaku have large wide sofas where you could be sitting on one extreme of the speaker to another, so I am sure that the difference would be more dramatic with their setups. 

 

Basically: some experimentation is required, and the type of measurement and correction you choose is situation and setup dependent. I don't think what they are doing is wrong, it is the right approach for their situation which is different to mine. To answer your question over whether you should correct over a larger listening area: it may or may not yield better results; and it depends on who you are going to plonk in unfavourable positions in your system. If it's your family and they don't care about sound, I would optimize over a smaller area - one or a few listeners will get a better result at the expense of everyone else. If it's for an SNA GTG, or the SNA hifi show, I would optimize for as many seats as possible. 

A very nice summary, thanks.

As I see it, the wider the area you average over the more chance there is that a certain position will have a very different frequency profile to the average.  So a "global" DSP profile is more likely to be sub-optimum.  Will it be better than no DSP? Probably.

On the matter of single or multiple  sweeps -   I went to the trouble of doing the Toole "parallel-piped" set of measurements with my basic REW/MiniDSP/single sub system and the average of this set of sweeps was very close to the single LP sweep.  So I now just do a single LP sweep.  I just need EQ for a single seat LP so this works for me.  Well, it works as well as a single sub set-up can.

[March Audio]

On 26/10/2023 at 3:43 AM, POV said:

I agree it’s the wrong thread for this.

 

The rest of your rant I disagree with and think in parts is just objectively incorrect.

 

He is actually correct.  Levels of bass vary around the room so if you use dsp to counter room modes you can only ever get an "average" correction level.  It will only be correct in one spot.  Other areas too much, others too little.  Hence we used Roon to conservatively dial down the major room modes.  IMO too little correction sounds better than too much.

 

It also can't do anything about other acoustic issues such as reverberation time.

Edited October 26, 2023 by March Audio

[davewantsmoore]

5 minutes ago, RoHo said:

As I see it, the wider the area you average over the more chance there is that a certain position will have a very different frequency profile to the average.  So a "global" DSP profile is more likely to be sub-optimum.  Will it be better than no DSP? Probably.

 

What it depends most on, is the coverage pattern (directivity) of the speakers.    ie. how much seat to seat variation do the speakers give before the EQ is applied.

 

"It" .... being both what range of seats is it possible to effectively optimise over (without introducing too much error to all seats in the optimisation window) .... and, how bad will things get in seats which are not in the optimisation area.

 

There is no escaping the physics of the SPL that the speaker is putting out.

 

5 minutes ago, RoHo said:

On the matter of single or multiple  sweeps

There isn't so much of a "right or wrong answer" .... as just a general warning about bad data and gigo.

 

Toole says a single point will be wrong based on the assumption that it won't be representative of the speaker -- it will include too much/many specific diffraction and other errors from the speaker and room.... and that data is not a good candidate for basing a correction from.

 

What he says has a lot of merit ..... but it's a spectrum of right and wrong, that requires investigation for each specific instance.    Is my single point full of problems? ..... is the average I'm creating "better" ? etc.   A single point seems unlikely to be the "right answer in general".

 

More advanced systems actually process the multiple measurements to try to understand what is going on.... rather than just taking a "simplistic" average of them.

 

 

As simple as taking measurements can be.... it is easy to measure things which should not be corrected, at least not as aggressively as "here is my data, and here is my target.... please make data -> target".

 

The paradox being that when you measure the result, you will see (close to) your target, and give yourself a pat on the back for "perfection".    But if you think deeply about what is happening.... you did:

• something (input)
• adjusted to
• target

What if the something wasn't the right thing (ie. data was not representative)

What if target is flawed?

... and even if "something" and "target" are good.... should you have really adjusted it that much.

 

Measuring the result and seeing the target doesn't actually answer any of those questions.

 

5 minutes ago, RoHo said:

it works

Back to the original premise of the thread .... or at least, re: is it needed / cheating / etc.

 

EQ isn't necessarily a "bandaid" ... but it could be, depending on specifics.

 

 

 

[acg]

1 hour ago, POV said:

Just to clarify for everyone, which DSP products do you have direct experience with (as in running the measurements, setting up profiles, comparing single position to multi position measurements etc)

 

I’m curious what your point of reference is?

 

I'm not sure what you are fishing for here...perhaps you can elaborate?

 

The usual simple DSP such as Audessy and ARC, I've messed with MiniDSP and filters in REW and Roon.  I do not yet use DSP in my own system but have not ruled it out for bass as I may need it to clear the final hurdle using multiple subs.  Am familiar with a number of other stereo systems using DSP including one that @frednorksetup that now sounds quite excellent. 
 

In my business daily I use DSP at laser frequencies in remote sensing applications and was university educated in the mathematics and applications of signal processing.

[davewantsmoore]

3 hours ago, March Audio said:

He is actually correct.

Yes.

 

That being said, there is also a lot of merit to the idea that "room treatment" cannot effectively "fix" low frequency problems..... that is without making the room approach an anechoic chamber.

 

It's all to some degree trying to "put the genie back into the bottle" after the speaker "radiation pattern" has done it's thing and released the SPL into the room.

 

At <100Hz the solution is using many SPL sources, and placing them in the right places

At <1000Hz it becomes hard to manage as most speakers are small compared to the wavelength (and so the sound tends towards omni directional)

Edited October 27, 2023 by davewantsmoore

[March Audio]

@sir sanders zingmore can we split this out into a new thread.?  It's a good subject.

[POV]

2 hours ago, acg said:

 

I'm not sure what you are fishing for here...perhaps you can elaborate?

 

The usual simple DSP such as Audessy and ARC, I've messed with MiniDSP and filters in REW and Roon.  I do not yet use DSP in my own system but have not ruled it out for bass as I may need it to clear the final hurdle using multiple subs.  Am familiar with a number of other stereo systems using DSP including one that @frednorksetup that now sounds quite excellent. 
 

In my business daily I use DSP at laser frequencies in remote sensing applications and was university educated in the mathematics and applications of signal processing.

I’m not angling at anything, just you are making statements that indicated to me that you may not have direct experience with contemporary products like DIRAC Live and RoomPerfect is all.  I’m honestly not trying to argue with you, I just disagree with some of the points you are making and I think some of that is coming from what some describe as theoretical position vs practical experience.

 

As a topical example both products allow for multi position measurement or single position measurement or both and you can create user profiles for both and relatively easily switch between so you can hear the difference between the 2.  In my heavily treated room as an example with RoomPerfect engaged in multi position profile I actually moderately prefer the bass (despite it being measurably less linear) at the main listening position but there is a reduction in focus and clarity, vs moving to single position measurement where focus, clarity, imaging is improved at the main listening position but bass I find takes on a very mild unnatural sound (to me).  Both sound materially better than with DSP off, and which one I prefer depends on source material.

 

One of the divides in this thread though is some thinking DSP as only applicable/useful for the modal zone, vs those that have experienced full frequency range DSP, or banded/shelved DSP which is I think probably the best solution for most partially treated rooms (dealing with specific issues in specific ranges) and Dirac Live makes this really easy to implement.

 

My next move is to 4 subs with DIRAC Live Bass control which I expect will be a game changer for me in terms of bass linearity both at listening position and more broadly across the room.

[POV]

4 hours ago, March Audio said:

 

He is actually correct.  Levels of bass vary around the room so if you use dsp to counter room modes you can only ever get an "average" correction level.  It will only be correct in one spot.  Other areas too much, others too little.  Hence we used Roon to conservatively dial down the major room modes.  IMO too little correction sounds better than too much.

 

It’s true that you get different results depending on the correction curve applied, as all things different folks will prefer different things (see my previous post) but that’s not the only point he was making.  I’m yet to see a room where passive treatment is able ‘fix’ all issues with room modes and bass linearity, if someone can point me to one I’d be most keen to check it out.  So I disagree that he is correct.

 

4 hours ago, March Audio said:

It also can't do anything about other acoustic issues such as reverberation time.

What can’t?  Adding parametric eq can’t but both RoomPerfect and Dirac Live are able to reduce the audible impact of reverberation at the listening position.  So when we say it can’t do anything, it’s true that the fundamental acoustics of the room aren’t changing, but if the audible impact is reduced at the listening position via DSP then the net outcome is the equivalent of ‘doing something’. 

 

[Guest]

26 minutes ago, POV said:

My next move is to 4 subs with DIRAC Live Bass control which I expect will be a game changer for me in terms of bass linearity both at listening position and more broadly across the room.

Recommend DLBC.  Definitely the current game changer until DLART hopefully makes an appearance next year across various brands.

[POV]

6 minutes ago, Snoopy8 said:

Recommend DLBC.  Definitely the current game changer until DLART hopefully makes an appearance next year across various brands.

Yeah NAD have committed to adding ART via firmware update for the M66 as soon as practically possible.  I’ve emailed them and they said they couldn’t commit to a timeframe but anticipate it will be in the first half of next year.

[Keith_W]

5 hours ago, davewantsmoore said:

At <100Hz the solution is using many SPL sources, and placing them in the right places

 

That's what Jon Herron of Trinnov would call, "a cacophony of bass", or the "flood the room with bass" approach, an example of what he also calls "old fashioned thinking". His "modern" approach would be to "manage" the bass, i.e. cancel reflections or steer it into other acoustic spaces. 

[March Audio]

1 hour ago, POV said:

It’s true that you get different results depending on the correction curve applied, as all things different folks will prefer different things (see my previous post) but that’s not the only point he was making.  I’m yet to see a room where passive treatment is able ‘fix’ all issues with room modes and bass linearity, if someone can point me to one I’d be most keen to check it out.  So I disagree that he is correct.

 

What can’t?  Adding parametric eq can’t but both RoomPerfect and Dirac Live are able to reduce the audible impact of reverberation at the listening position.  So when we say it can’t do anything, it’s true that the fundamental acoustics of the room aren’t changing, but if the audible impact is reduced at the listening position via DSP then the net outcome is the equivalent of ‘doing something’. 

 

 

I will start a new thread as this is way off topic.

[frednork]

9 minutes ago, March Audio said:

 

I will start a new thread as this is way off topic.

It really doesnt concern me if it goes off topic. 

[acg]

1 hour ago, Keith_W said:

 

That's what Jon Herron of Trinnov would call, "a cacophony of bass", or the "flood the room with bass" approach, an example of what he also calls "old fashioned thinking". His "modern" approach would be to "manage" the bass, i.e. cancel reflections or steer it into other acoustic spaces. 

 

Isn't that what Dave means by "in the right places"?

 

Hard not to flood a room with bass considering the wavelengths involved, but I am sure it was meant in a marketing way considering Trinnov are selling something.

[pass the bass]

Fascinating stuff, although a lot of the technicalities is way over my head.

Pardon this naive question:

Is it within the realm of possibility to employ a sensor that detects

- where you sit in the room

- for one or more listeners

and adjust the DSP accordingly?

 

[March Audio]

5 hours ago, POV said:

One of the divides in this thread though is some thinking DSP as only applicable/useful for the modal zone, vs those that have experienced full frequency range DSP, or banded/shelved DSP which is I think probably the best solution for most partially treated rooms (dealing with specific issues in specific ranges) and Dirac Live makes this really easy to implement.

 

Yes, I think it would be useful to discuss the different reasons behind the different thoughts on full range V just modal correction.  Or for that matter what those two things actually mean.  Many readers may not know.

 

Instead of me regurgitating, a good explanation is here

 

https://www.prosoundtraining.com/2021/10/14/divide-and-conquer-the-schroeder-frequency/

 

Essentially the sound in room behaves in different ways dependant upon frequency.  The point at which the behaviour changes from resonant (modal) to diffuse is called the Schroeder frequency, typically around 200Hz but dependant on room size and characteristics.  It is true to say that rooms the size of our typical listening rooms are too small to become truly diffuse, but that's another story which we don't need to go into.

 

Tbc

[March Audio]

"Room modes" are the fundamental modal frequencies the listening room will resonate at, dictated by the physical dimensions length, width and height. .  I have lifted all the following form various other web sources for expediency.

 

There are 3 different types of room modes for room acoustics

 

 

Room Mode Formula:

F = c/2 * sqrt(p^2/L^2 + q^2/W^2 + r^2/H^2)

F= Frequency
c = speed of sound (1130 feet per second or 344 meters per second)
sqrt = Square Root
^2 = squared
L = Length of Room
W = Width of Room
H = Height of Room

p, q and r represent the mode we’re solving for in the room mode equation. If you want to know the axial mode for the room length, p=1, q=0, r=0. If you want to know the 2nd axial mode, p=2, q=0, r=0. To find a tangential mode, use 1 in 2 of the variables from the room mode calculator. So, if you want the first tangential of the length and width, p=1, q=1, r=0.

 

There are a myriad of calculators on the net which make it far easier to calculate your particular rooms behavior's by simply inputting its dimensions.

 

You can also measure your room with a microphone.  I measured a room in our office with the mic at a fairly central position followed by about a meter to the left.

 

 

You can see from the above how diiferent the bass response will be at the two listening positions.  In one position a 36Hz mode dominates, whilst in another a 51Hz mode dominates.

 

You can see from the spectogram how the two modes take a very long time to decay which along with the peak in amplitude causes the one note boomy bass.

 

 

 

So these modes are obviously bad for sound quality.  So we can try to fix the problem by using DSP correction to reduce the levels of those peaks.  Filter those specific frequencies to reduce their levels.  This indeed works! It not only reduces the peak, it also reduces the long ringing.

 

However, you can see an obvious problem.  Which of the two measurements do you base the correction on?  The two listening positions have completely different requirements.  If you set it right for one listening position it will be wrong for the other.  Some systems take an average of many listening positions.  This method will create a correction that is wrong in all listening positions.

 

Dont get me wrong, whilst a single and multiple measurement correction methods for modal issues have problems, they can still provide an overall improvement in perceived sound quality.  However you do need to be aware of the limitations.  Also my experience is that over correction, i.e. completely diallling out a mode to flat will end up making the sound perceived as lean.

 

I have used many different correction systems, TMREQ, Audyssey, Acourate, DIRAC and Roon.  Anything that is automatic rarely gives good results and requires manual tweaking.

 

TBC

Edited October 27, 2023 by March Audio

[March Audio]

So what about correction above Shroeder?  Well there a whole different set of issues here.

 

The issues revolve around the dispersion characteristics of the speaker.  The sound we perceive is a combination of the direct on axis sound from the speaker and the sound reflected from walls, floor and ceiling.  A speaker radiates sound in all directions, each one having a different radiation characteristic.

 

The total energy the speaker radiates into the room is termed the "sound power".  For good sound the off axis response must be similar in characteristic to the on axis response.  Often it isnt and people find this objectionable.  Its quite possible (and common) for a speaker to have a flat on axis response (a good thing) and a wonky off axis response.  You typically see 2 way speakers that have good on axis response but at crossover where the tweeter takes over, a sudden widening in dispersion leads to more sound power in that region and a perceived brightness/hardness to the sound.

 

Microphones do not "hear" like 2 ears and a brain.  An in room measurement of a flat response speaker will typicaly show a downward tilt due to increasing beaming as frequency increases and the absorptive nature of the room and air.

 

If a speaker, for example, sounds bright it may be due to directivity problems., even though it has a flat on axis response.  You might be tempted to correct this brightnes with DSP and increase the tilt of the measured response.  In this situation it wont lead to a satisfactory result.  By doing this you will mess up the very important flat on axis result which will not sound right.

 

So by trying to correct above Schroder, you are really making a stab in the dark as you dont know what the dispersion characteristics of the speaker are.  Any response curve you might (or an automatic algorithm) might aim for is quite arbitrary.  So beyond some subtle tweaks I would strongly recommend against correction above Schroder.  If it sounds wrong in basic tonality, your speakers are wrong.  DSP, whilst may help a bit, cannot solve fundamental issues with speaker directivity.  If your speakers have a frequency response issue that is reflected in both the on and off axis responses then DSP can help.  Problem is that without comprehensive measurements (such as Klippel)  you dont know how your speaker behaves.

 

A couple of examples.

 

This speaker, whilst having a good flat on axis response , will have a bright sound due to the big increase in sound power and early reflections from 3 to 7kHz.  All caused by directivity issues.  You cannot correct for this with DSP.  You will just mess up the relatively good on axis response.

 

 

 

A good speaker that will successfully respond to DSP correction, but is very unlikely to need it.  It will sound "right" out of the box.

 

 

Again, dont misinterpret what I'm saying.  DSP room EQ can be helpful within limits, but above schroeder you are simply better off buying a well designed speaker in the first place.

 

Room EQ also cannot help with acoustic issues such as excessively high reverberation time.

 

 

Edited October 27, 2023 by March Audio

[Keith_W]

2 hours ago, pass the bass said:

Fascinating stuff, although a lot of the technicalities is way over my head.

Pardon this naive question:

Is it within the realm of possibility to employ a sensor that detects

- where you sit in the room

- for one or more listeners

and adjust the DSP accordingly?

 

 

There are already systems that adjust the DSP based on head tracking, e.g. BACCH and the Smyth Realizer. BACCH uses a webcam to track the listener, and I Smyth works with headphones and detects when you turn your head. What both have in common is that directional cues are dependent on upper frequencies, and it is easier to aim those. But no products to my knowledge are able to adjust bass in real time based on listener position. For such a product to exist, either (1) the listener has a calibrated mic attached whilst moving around the room, or (2) the room needs to be thoroughly mapped out with the mic in different listening positions, with the software calculating the ideal curve to employ based on where it detects the listener to be. It is not impossible to bring to market such a product, but how well it would work and how many people would want it to justify the development costs are open questions! 

[davewantsmoore]

6 hours ago, Keith_W said:

That's what Jon Herron of Trinnov would call, "a cacophony of bass"

This has bad connotations.... and to unpack it a bit more it depends on the wavelengths involved... and what EQ is applied to the sources.

 

What they're doing, and what Dirac ART is doing in this space works (very well) ..... but better, or worse, in different frequency ranges.   It is very very useful in the ~80 to ~300 zone.

 

But these approaches don't really discount what I said, which is "many sources, placed intelligently"...... once you have that, yes - you can employ very sophisticated ways to shape/cancel the SPL.

 

5 hours ago, acg said:

Isn't that what Dave means by "in the right places"?

 

Hard not to flood a room with bass considering the wavelengths involved, but I am sure it was meant in a marketing way considering Trinnov are selling something.

I didn't read this before I wrote my reply above.

 

Very much.

[Keith_W]

Well, Jon Herron does work for Trinnov and I suspect a lot of what he says is marketing speak. After I watched his video about "steering" the bass and dumping it in adjacent rooms, I tried the same experiment with Acourate and failed. I did try a VBA and it works, sort of. It works really well in one listening position, but no better than bass EQ. For off axis listening, it has a measurable but not dramatic effect. In the end, I have done the experiment and it was A LOT OF WORK and it works no better than simple bass EQ. With a VBA you work out the delays by calculating the time of flight from the speed of sound and room dimensions, and then you guess at the attenuation required. Then it is iteration after iteration until it gets dialled in, and that's just for one reflection. If you want to simulate more reflections, it's even more iterations.