Active crossovers measurably better?

[dallasjustice]

I have 4 cheap subs, two on each side.

The mains actually go lower.

I run them all down low as a reduction of distortion experiment. Three drivers on each side can run 9dB lower than a single for the same SPL.

My pending path-of-least-resistance experiment is to reverse two for a slightly lopsided dipole simulation.

You are in deep Ray; On depth row.

 

[stunta]

View attachment 10307
The SPL sweep smoothed to 1/12 per octave is a good way to show the overall frequency balance of a speaker. It’s not the ultimate measurement but it’s a very helpful measurement to answer your question. It’s best to wait for me to overlay the active and passive measurements. Once you see that, I think it will be easy for you to get a feel for how they may sound different.

The other plot is a step response. It is probably the most useful time domain characterization of a loudspeaker. You can see there’s 2 spikes separated by about half a millisecond. That’s the tweeter followed by the midwoofer. The ideal step response should look like the one I posted above. In reality no speaker has this step response. But some active speakers come scary close. You can see that the JBL 4367 has a very good step response but not ideal. When I later overlay the M2, I think you’ll see how much better time domain is with a well designed active speaker.

That makes it very clear. Thanks. With the SPL sweep, why is there a dip at 35 Hz from 25 Hz? Is that due to your room characteristics?

 

[dallasjustice]

That makes it very clear. Thanks. With the SPL sweep, why is there a dip at 35 Hz from 25 Hz? Is that due to your room characteristics?

It’s either the room or the phase reversal below the port tuning frequency. Almost everything below roughly 300hz is determined by the room. Those frequencies are much less important to understand what a speaker is about. Of course, there are speaker gimmicks like the cardiod bass speakers whose manufacturers imply perfect bass due to cardiod low frequency polar patterns. IMO, those are okay engineering goals for a speaker. But the room will still determine bass no matter the marketing claims. There’s never been any convincing data for me to believe otherwise and there likely won’t be.

Low frequency can be made close to perfect in only 2 ways:
1. Without subwoofers, in a gigantic or unusually well shaped room.
2. Smart subwoofery, DSP and crossovers.

In a normal size rectangular room, there’s no such thing as a 2 channel system with great bass. It doesn’t matter how much one spends or how big/small or cool the 2 speakers are.

 

[Rodney Gold]

Have you looked at a Trinnov Dallas?
Its everything you need in a single box

 

[Cosmik]

It's not so noticeable to the ear, but, there it is, right where I sit, nonetheless.

This is what everyone says: acoustic delayed reflection artefacts simply aren't as audible as they 'should be'; 'you hear through the room'. But if you EQ'ed a big hole in your signal and played it over the speakers you would hear it. And if you EQ'ed a reverse 'correction' into your signal you would, too, even though it then looked perfect in the measurement.

By smoothing away the undulations in the FR at higher frequencies you are in denial that the reflections have any effect higher up, but of course they are there. You can say "They are so complex that the ear smooths them over, too". But if your hearing isn't just smoothing them over in the frequency domain, but is 'hearing through' them using a different mechanism than a FFT, then it may well be capable of doing the same thing at lower frequencies too. And indeed, to quote you above, it seems to do just that.

For sure, as you move around the room you hear bass dips and peaks if you consciously listen for them, but does that mean that they sound wrong in music? Accompanying the dips and peaks are time domain-related clues and I suspect you hear them and thereby at some level understand the FR undulations and ignore them - at least partially. A dumb EQ 'correction' would not change those time domain-related clues, thus leading you to the conclusion that the source sounds wrong.

The DSP in active speakers should be restricted to making the speaker's output neutral, and that is all!

 

[svart-hvitt]

This is what everyone says: acoustic delayed reflection artefacts simply aren't as audible as they 'should be'; 'you hear through the room'. But if you EQ'ed a big hole in your signal and played it over the speakers you would hear it. And if you EQ'ed a reverse 'correction' into your signal you would, too, even though it then looked perfect in the measurement.

By smoothing away the undulations in the FR at higher frequencies you are in denial that the reflections have any effect higher up, but of course they are there. You can say "They are so complex that the ear smooths them over, too". But if your hearing isn't just smoothing them over in the frequency domain, but is 'hearing through' them using a different mechanism than a FFT, then it may well be capable of doing the same thing at lower frequencies too. And indeed, to quote you above, it seems to do just that.

For sure, as you move around the room you hear bass dips and peaks if you consciously listen for them, but does that mean that they sound wrong in music? Accompanying the dips and peaks are time domain-related clues and I suspect you hear them and thereby at some level understand the FR undulations and ignore them - at least partially. A dumb EQ 'correction' would not change those time domain-related clues, thus leading you to the conclusion that the source sounds wrong.

The DSP in active speakers should be restricted to making the speaker's output neutral, and that is all!

Again @Cosmik , you are hypothesizing. Though that may be an enjoyable pastime, can you show us that room compensation is harming the perceived sound? (For example peer-reviewed research, extensive listening tests). The only guy with some credentials that I can think of who is in total agreement with you is John Watkinson (SEE picture Velle), AES Fellow and co-owner of boutique loudspeaker designer Legend.

 

[Cosmik]

Again @Cosmik , you are hypothesizing. Though that may be an enjoyable pastime, can you show us that room compensation is harming the perceived sound? (For example peer-reviewed research, extensive listening tests).

You are absolutely correct. But there is one bit of actual science I can cite. It is the paper that Amir mentions in this article:
http://www.madronadigital.com/perceptual-effects-of-room-reflecti

The most surprising was scenario #4 where the outcome was “greatly degrading effect.” Let me repeat: the same distortion created electronically and sent out of the speaker was a very negative thing...

...So what seemed like an open and shut case of eliminating wall reflections due to anomalies in frequency response of the room becomes much more complex when one considers how we hear sounds in our home listening spaces. It shatters “gut feelings” one might have about the problem and solution thereof.

The apparent mystery at the end of the piece (there is no real explanation or conclusion) would be explained entirely by my hypothesis!

We could spend forever debating whether 'room correction' is a good thing, and is preferred by users. But a far simpler and more elemental experiment leads us to conclude that while a certain acoustic 'artefact' is not noticed, or is regarded as a positive thing, the EQ 'equivalent' (but only equivalent if you think that the ear only hears the frequency domain) is clearly heard as a negative, annoying thing. This has direct implications for the idea of 'correction' by DSP.

As in any science, though, you can look at it and say, "Subjectively, I don't like that experiment and its results. I prefer this explanation..."

But I just keep noticing people saying "Strangely, this huge dip isn't really audible", and so I just keeping adding 2 + 2 and getting 4.

 

[Rodney Gold]

I can correct for amplitude and phase , group delay ,, impulse response and turn on and off amplitude and phase ,, level alignment and delay individually , I can further implement an eq independent of target curve atop this , all controllable and heard in real time
ANY time based parameter has a marked effect when turned on and off , much more noticeable than messing with the eq or amplitude curve

 

[dallasjustice]

Have you looked at a Trinnov Dallas?
Its everything you need in a single box

I’m not interested DSP in a box. The software solutions are more powerful, more flexible and make use of better components. So I would not be interested in the Trinnov solutions.

 

[svart-hvitt]

I’m not interested DSP in a box. The software solutions are more powerful, more flexible and make use of better components. So I would not be interested in the Trinnov solutions.

Yes, this may be an example of software making another piece of hardware obsolete (because you can re-use your existing, more powerful PC hardware).

The next thing is to do all the calcultions in the cloud to generate filters (Genelec is already there).

 

[svart-hvitt]

You are absolutely correct. But there is one bit of actual science I can cite. It is the paper that Amir mentions in this article:
http://www.madronadigital.com/perceptual-effects-of-room-reflecti

The apparent mystery at the end of the piece (there is no real explanation or conclusion) would be explained entirely by my hypothesis!

We could spend forever debating whether 'room correction' is a good thing, and is preferred by users. But a far simpler and more elemental experiment leads us to conclude that while a certain acoustic 'artefact' is not noticed, or is regarded as a positive thing, the EQ 'equivalent' (but only equivalent if you think that the ear only hears the frequency domain) is clearly heard as a negative, annoying thing. This has direct implications for the idea of 'correction' by DSP.

As in any science, though, you can look at it and say, "Subjectively, I don't like that experiment and its results. I prefer this explanation..."

But I just keep noticing people saying "Strangely, this huge dip isn't really audible", and so I just keeping adding 2 + 2 and getting 4.

The research piece quoted by Amir (Clark, 1985) is so old and has obviously had little if any influence among audio science peers.

I observe that modern audio research seems to put emphasis on room compensation. So even if I can’t come up with other evidence than pointing to recent research directions, I still think it is the guy who claims the emperor is naked who has to prove that modern audio research is going in the wrong direction.

I think we both agree, however, that we would appreciate peer-reviewed research on the perception of DSP based room compensation.

 

[oivavoi]

AFAIK, there has been published results from three or four controlled listening tests on various room correction products. Maybe I can cough up the links later. They have all, I believe, showed that listeners preferred at least one form of room eq over no correction (curiously, this always seems to be the specific room correction product that the authors have been affiliated with). But as with all experimantal science, the question of "external validity" is vital. How much do these tests tell us about listening conditions that are different from the one in the test? It might be, for example, that the "room correction" product actually functioned as a loudspaker correction.

In order to really test Cosmik's claim, one would need to compare a loudspeaker that was already optimally corrected as much as possible using DSP (for phase, frequency and distortion anomalies), with the same speaker, but now with added room correction (only in the bass, and fullrange). This would at least tell us what the average preference would be in this listening panel under those circumstances.

My own opinion and experience is roughly as follows, as have been stated before in this forum: I believe that room-based eq in the deepest frequencies will usually be beneficial. I think that room-based fullrange eq will also be beneficial with subpar speakers or in rooms/setups with subpar placement and bad acoustics (if the tonal balance in the room is heavily skewed towards either low or high frequencies, for example).

But I have also experienced that fullrange room-based eq may add a slight level of artificiality which detracts from the listening experience, if everything is already optimized - good DSP-based speakers in rooms with good acoustics. This means that the "speaker-correction" is done already, and what the additional room-eq does is to adjust the direct sound so that the total tonal response (speakers + room) adheres more to a perfect flat frequency response than without correction. If one believes that listeners to a certain extent hear through the room, and that the direct sound is of paramount importance, then it's logical that this can lead to a degradation of the total experience.

Summing it up: I think that reseraching room eq makes sense, because speakers or listening setups often are far from ideal. Most applied audio research these days is geared towards non-optimal setups. With regards to the absolute highest fidelity, whether fullrange room eq has a role to play there, is still open to debate, given how I understand these things.

But in any case we can all agree that making speakers with DSP-based crossovers is the rational way forward!

 

[Purité Audio]

Here, speakers such as the Kiis/8Cs with their cardioid response sound much better than any traditional speaker which has subsequently been EQd, which ever EQ software you use.
Keith

 

[svart-hvitt]

AFAIK, there has been published results from three or four controlled listening tests on various room correction products. Maybe I can cough up the links later. They have all, I believe, showed that listeners preferred at least one form of room eq over no correction (curiously, this always seems to be the specific room correction product that the authors have been affiliated with). But as with all experimantal science, the question of "external validity" is vital. How much do these tests tell us about listening conditions that are different from the one in the test? It might be, for example, that the "room correction" product actually functioned as a loudspaker correction.

In order to really test Cosmik's claim, one would need to compare a loudspeaker that was already optimally corrected as much as possible using DSP (for phase, frequency and distortion anomalies), with the same speaker, but now with added room correction (only in the bass, and fullrange). This would at least tell us what the average preference would be in this listening panel under those circumstances.

My own opinion and experience is roughly as follows, as have been stated before in this forum: I believe that room-based eq in the deepest frequencies will usually be beneficial. I think that room-based fullrange eq will also be beneficial with subpar speakers or in rooms/setups with subpar placement and bad acoustics (if the tonal balance in the room is heavily skewed towards either low or high frequencies, for example).

But I have also experienced that fullrange room-based eq may add a slight level of artificiality which detracts from the listening experience, if everything is already optimized - good DSP-based speakers in rooms with good acoustics. This means that the "speaker-correction" is done already, and what the additional room-eq does is to adjust the direct sound so that the total tonal response (speakers + room) adheres more to a perfect flat frequency response than without correction. If one believes that listeners to a certain extent hear through the room, and that the direct sound is of paramount importance, then it's logical that this can lead to a degradation of the total experience.

Summing it up: I think that reseraching room eq makes sense, because speakers or listening setups often are far from ideal. Most applied audio research these days is geared towards non-optimal setups. With regards to the absolute highest fidelity, whether fullrange room eq has a role to play there, is still open to debate, given how I understand these things.

But in any case we can all agree that making speakers with DSP-based crossovers is the rational way forward!

@oivavoi ,

You wrote:

«In order to really test Cosmik's claim, one would need to compare a loudspeaker that was already optimally corrected as much as possible using DSP (for phase, frequency and distortion anomalies), with the samespeaker, but now with added room correction (only in the bass, and fullrange). This would at least tell us what the average preference would be in this listening panel under those circumstances».

This is exactly what one could do with Genelec’s SAM range monitors. What one would need is an auditorium with people and said speakers to conduct simple tests. I would have guessed Genelec already did this on their own employees. I can Ask them.

 

[Cosmik]

This is exactly what one could do with Genelec’s SAM range monitors. What one would need is an auditorium with people and said speakers to conduct simple tests. I would have guessed Genelec already did this on their own employees. I can Ask them.

But as a control, I would also like you to include on the DSP list 'stereo chorus', 'echo reverb', 'pitch shift down by 1/2 semitone', 'super bass boost', 'super loudness EQ' etc. I am not just being facetious. The claim is that 'correction' may be preferred, but there are infinite, arbitrary possibilities for 'correction' while there is only one 'neutral'. 'Correction' can literally be anything: the signal has been changed. For example, one difficulty is that you can't level match signals with radically different EQs. Just because you call it 'correction' doesn't mean it isn't just 'an effect'.

I put it to you that if you gather a bunch of people together and don't prejudice the test by telling them what they are listening for, and ask them to select their 'preference', then it is quite likely that many people would 'like' the sound of a loudness button - for a few hours every now and again. If a majority say they prefer 'correction' but they also like 'stereo chorus' or 'super loudness EQ', then you have only demonstrated that people 'like' the sound of DSP possibly as a novelty.

If you tell your listeners what they are listening for, and these are trained people who know what 'state-of-the-art correction' [sic] sounds like, then you are just asking them to pick out what they already 'know' to be 'correct'.

Messy isn't it? Unlike clean, streamlined, rational, logical, hypothesising...

 

[svart-hvitt]

But as a control, I would also like you to include on the DSP list 'stereo chorus', 'echo reverb', 'pitch shift down by 1/2 semitone', 'super bass boost', 'super loudness EQ' etc. I am not just being facetious. The claim is that 'correction' may be preferred, but there are infinite, arbitrary possibilities for 'correction' while there is only one 'neutral'. 'Correction' can literally be anything: the signal has been changed. For example, one difficulty is that you can't level match signals with radically different EQs. Just because you call it 'correction' doesn't mean it isn't just 'an effect'.

I put it to you that if you gather a bunch of people together and don't prejudice the test by telling them what they are listening for, and ask them to select their 'preference', then it is quite likely that many people would 'like' the sound of a loudness button - for a few hours every now and again. If a majority say they prefer 'correction' but they also like 'stereo chorus' or 'super loudness EQ', then you have only demonstrated that people 'like' the sound of DSP possibly as a novelty.

If you tell your listeners what they are listening for, and these are trained people who know what 'state-of-the-art correction' [sic] sounds like, then you are just asking them to pick out what they already 'know' to be 'correct'.

Messy isn't it? Unlike clean, streamlined, rational, logical, hypothesising...

FWIW,

your talk on «loudness» is nonsense in this case. Genelec’s room compensation shaves off peaks, it doesn’t try to fill nulls. So in this case the DSP is anti-loudness in lack of a better word. So even if the volume is in sum lower, people seem to prefer it. Other room DSPs are different and fill out nulls, so all DSPs are not equal.

When it comes to the fidelity of the signal at the listening, I believe the DSPers have the upper hand. Your case rests on the premise that a heavily distorted signal at the listening position will be heard through by the super computer which is the human brain.

However, I agree that tests of preferences are difficult.

 

[RayDunzl]

This is what everyone says: acoustic delayed reflection artefacts simply aren't as audible as they 'should be'; 'you hear through the room'.

What I meant to say was the dip wasn't bothersome either before or after other corrective measures (reduction of peaks) that were noticeable and annoying, which was what led to adopting a little DSP Correction of the Speaker Output as Heard at the Listening Position.

My measurably flat (per Harman) JBL LSR 308 exhibit essentially the same curve as the Martin Logan dipoles (which may or may not be measurably flat in the echoless chamber). The Bass humps are not objectionable at low levels, but objectionable at higher SPL.

"Uncorrected" JBL LSR 308 (blue) and Martin Logan reQuest (red) measured (1/12 octave smoothing) at the listening position in-room:

I could hear something that didn't sound natural to me at elevated volumes.

I perceived it to be in the bass region.

I had no control of the speaker output other than volume.

Fifteen years passed.

I bought a measurement microphone to provide a way to quantifiably examine what I perceived as a problem area.

I found two humps that (on paper) corresponded to my earlier ignorant dissatisfaction.

I bought digital EQ, and improved the situation some (DEQ2496).

Another year passed.

After further investigation, decided to try a more flexible device (miniDSP OpenDRC-DI) and manual corrections (rePhase).

Finally, decided to try automated filter generation (AcourateDRC). That's where I am now, and I am pleased with the results:

Gold JBL, green MartinLogan:

This is what everyone says: acoustic delayed reflection artefacts simply aren't as audible as they 'should be'; 'you hear through the room'.

The JBL sprays sound all around the room, the ML avoids the floor, ceiling, and sidewalls. this shows up in an impulse response, but also shows up in an un-smoothed frequency response. I don't hear the defects tonally, but I do hear them spatially. The "correction" applied has no effect on this.

See the level of "hash" in the higher frequencies which I interpret as narrow cancellations due to reflections in my largely untreated room. It gives the sound an "out of phase" quality when critically listening, verified one night by feeding the JBL an out of phase signal, and noticing the unexpected lack of difference between in phase/out of phase, particularly on correlated pink noise.

---

Results of "room correction" here:

Elimination of uncomfortable excesses in the bass region, no noticeable degradation of the frequency response higher up, no degradation in the spatial image.

Improvement in the measured Impulse and Step response at the listening position. Audible difference? Maybe on some transients.

No noticeable problems added when casually listening outside the sweet spot. The boominess is gone there, too.

---

My Limited Conclusion:

Flat or potentially un-flat speakers can exhibit similar obnoxious traits in-room, and a little EQ can improve the perception, particularly in the bass region. I'm not bothered by the changes in the higher frequencies (at least within the range that I can hear).

The dip in the bass is caused by an out-of-phase condition (nulling) that the correction as applied doesn't address. It is not particularly noticeable with musical content. Narrowband omission is much more benign compared to addition.

 

[svart-hvitt]

What I meant to say was the dip wasn't bothersome either before or after other corrective measures (reduction of peaks) that were noticeable and annoying, which was what led to adopting a little DSP Correction of the Speaker Output as Heard at the Listening Position.

My measurably flat (per Harman) JBL LSR 308 exhibit essentially the same curve as the Martin Logan dipoles (which may or may not be measurably flat in the echoless chamber). The Bass humps are not objectionable at low levels, but objectionable at higher SPL.

"Uncorrected" JBL LSR 308 (blue) and Martin Logan reQuest (red) measured (1/12 octave smoothing) at the listening position in-room:

View attachment 10309

I could hear something that didn't sound natural to me at elevated volumes.

I perceived it to be in the bass region.

I had no control of the speaker output other than volume.

Fifteen years passed.

I bought a measurement microphone to provide a way to quantifiably examine what I perceived as a problem area.

I found two humps that (on paper) corresponded to my earlier ignorant dissatisfaction.

I bought digital EQ, and improved the situation some (DEQ2496).

Another year passed.

After further investigation, decided to try a more flexible device (miniDSP OpenDRC-DI) and manual corrections (rePhase).

Finally, decided to try automated filter generation (AcourateDRC). That's where I am now, and I am pleased with the results:

Gold JBL, green MartinLogan:

View attachment 10310



The JBL sprays sound all around the room, the ML avoids the floor, ceiling, and sidewalls. this shows up in an impulse response, but also shows up in an un-smoothed frequency response. I don't hear the defects tonally, but I do hear them spatially. The "correction" applied has no effect on this.

See the level of "hash" in the higher frequencies which I interpret as narrow cancellations due to reflections in my largely untreated room. It gives the sound an "out of phase" quality when critically listening, verified one night by feeding the JBL an out of phase signal, and noticing the unexpected lack of difference between in phase/out of phase, particularly on correlated pink noise.


View attachment 10311

---

Results of "room correction" here:

Elimination of uncomfortable excesses in the bass region, no noticeable degradation of the frequency response higher up, no degradation in the spatial image.

Improvement in the measured Impulse and Step response at the listening position. Audible difference? Maybe on some transients.

No noticeable problems added when casually listening outside the sweet spot. The boominess is gone there, too.

---

My Limited Conclusion:

Flat or potentially un-flat speakers can exhibit similar obnoxious traits in-room, and a little EQ can improve the perception, particularly in the bass region. I'm not bothered by the changes in the higher frequencies (at least within the range that I can hear).

The dip in the bass is caused by an out-of-phase condition (nulling) that the correction as applied doesn't address. It is not particularly noticeable with musical content. Narrowband omission is much more benign compared to addition.

Wow, that was one marathon. I’ll reread for comprehension

 

[Palladium]

Which, BTW, explains why so many producers pay the audiophile press to focus on everything but performance.

One has to nothing fear from actives if they aren't charlatans.

 

[stunta]

I'm selling my JBL 4367 speakers which uses the same drivers and same waveguide technology as the active JBL M2. I hope to have a pair of JBL M2 setup in my room with FIR crossovers very soon. For now, I want to post a 1/12 per octave smoothed SPL and the Step response without using any DSP. This is just the raw speaker response using the passive crossovers. Later on, I'll post the same measurements from the same speaker/mic positions from the JBL M2 using FIR crossovers and driver time alignment. It should be an interesting comparison. All sweeps are calibrated to 83db in REW:

View attachment 10301 View attachment 10302

@dallasjustice when you get the chance, please post comparisons.