Should we correct to Schroder, or full range?

[sarumbear]

This is up to you. My understanding of this thread was that is was about room correction (as Schroeder was the topic), not speaker correction.

But will there be room anomalies above Schroeder range? Doesn’t higher frequencies act as rays whereas low-frequency sound energy propagates in waves? As the latter is modal, room affects them. Does room affects the higher frequencies?

Hence, EQ can be used for both room and speaker anomalies.

 

[Tom C]

My current thinking:
Best case scenario is to have speakers with flat anechoic FR, and well behaved off axis. If this is the case for you, correct only the bass region.

Next best is a speaker that is well behaved off axis, but maybe not ideal FR. For this, you can correct for FR errors based on anechoic data, and correct the bass region, and you should be in good shape. ASR member @Maiky76 has provided computer derived corrections, based on anechoic measurements, for many of the speakers measured here by Amir. I have found these to be excellent, and use them when I can. BTW, once or twice I had a pair of speakers I wanted to correct that had been been measured here, but the filters had not been previously calculated. I asked Mr. Maiky if he could run his algorithm, which he graciously provided.

Unfortunatley, many of us have speakers for which a full set of spins is not available, or which have never been measured at all. Many of these have problems in the FR, and have less than ideal off axis behavior. For these, best approach may well be to correct the bass, and sit close to the speaker when listening, to remove as much of the problematic room reflections as possible. If they still don’t sound good to you, you really do need to consider replacing them, if possible.

 

[abdo123]

If your speaker is fairly close to a boundary then the radiation pattern and "anechoic" response will differ signficantly below 1KHz but should be identical above it.

So take as many (but relevant) measurements as possible and try to match the response below 1KHz with the response above 1KHz if you're certain you got good speakers.

If your speaker is far away from the walls i would not even EQ above 100 Hz.

 

[tmtomh]

If it doesn't sound right, of course it's OK to correct by-ear. And if EQ "correction" makes it sound worse to you, don't do it just because it "measures better". Music is for you to enjoy!

You can measure to get a more-clear idea of what the problems might be, but if you try to perfectly-flatten the response it will only work if never move your head!

I think this a nice, concise summary of what is probably the best practical advice.

What I see developing out of this discussion is a very interesting difference of opinion, and it's not actually about "correct to Schroeder" vs "correct full-range."

Instead, it seems like effective full-range correction is possible, but it requires certain techniques or cautions, and not everyone agrees on which are the best and how effective they are: frequency-dependent windowing; MMM measurement, and so on.

So I think the real difference of opinion here is about how far folks feel like going in getting their room response as correct as possible: some folks are keen to investigate and try all available methods for making full-range correction work as properly as possible, while others - myself included, at least for now - feel like the effort and the chances of not getting it quite right are sufficiently great that as a practical matter it just makes more sense to auto-correct lower frequencies and apply manual PEQ as-needed to higher frequencies, based on the measurements, or listening, or both.

One other note and small correction to something I posted above: Genelec Ones have a rear DIP switch for bass tilt, which applies a -2dB low shelf from 800Hz downwards. The Genelec manual says that room reinforcement can boost response up to about 800Hz. Similarly, it appears that depending on the measurements it comes up with, GLM will apply correction/filters up to about 800Hz too. So it does apply correction as far up as 1-1/2 to 2 octaves above Shroeder, depending on what your room's frequency is and depending on how linear or nonlinear the GLM measurements are in the 250-800Hz range.

 

[sigbergaudio]

But will there be room anomalies above Schroeder range? Doesn’t higher frequencies act as rays whereas low-frequency sound energy propagates in waves? As the latter is modal, room affects them. Does room affects the higher frequencies?

Hence, EQ can be used for both room and speaker anomalies.

I have not said EQ cannot be used for both room and speaker anomalies.

 

[sarumbear]

I have not said EQ cannot be used for both room and speaker anomalies.

I must have miss-understood what you said earlier, though I don't know how reading your quote below.

Rule of thumb: No EQ above 500hz, rarely above 250-300hz.

Don't use DSP to force a different target curve than what the speaker would naturally reproduce in your room.

 

[sigbergaudio]

I must have miss-understood what you said earlier, though I don't know how reading your quote below.

Again, this is with regards to room EQ. This was in reply to the original question in this thread, which goes as follows:

"I would like to gather opinions from ASR members as to whether we should apply room correction to the Schroder frequency, or whether we should do full range correction from 20Hz - 20kHz."

 

[goat76]

Unfortunatley, many of us have speakers for which a full set of spins is not available, or which have never been measured at all. Many of these have problems in the FR, and have less than ideal off axis behavior. For these, best approach may well be to correct the bass, and sit close to the speaker when listening, to remove as much of the problematic room reflections as possible. If they still don’t sound good to you, you really do need to consider replacing them, if possible.

As I see it, if room reflections are strong enough vs the direct sound to seriously affect the tonality or the overall perceived sound, they are problematic no matter if the off-axis response of the loudspeakers is ideal or not, at least from a standpoint of a listener who wants to hear the recorded details over what their own room adds to the equation.

So yes, if your speakers must be set up in a non-ideal way, far away enough for the listening room to be a dominant factor of the perceived sound at the listening position, it will probably be much more important to choose a loudspeaker with an ideal off-axis response.
But for people who are interested and strive for a better hi-fi experience, the aim should be to maximize the direct sound ratio vs the reflections from their listening environment to be able to hear more of what’s actually in the recordings. So a shorter listening distance will be ideal for that and “wins” therefore over the importance of perfect off-axis response in this case.

It's like an oxymoron.
An ideal off-axis response is way more important for a non-ideal listening situation where you have way too much interference from your own listening environment. Why not get it right to begin with, just set up your loudspeakers with a smaller listening triangle and the importance of the ideal off-axis response will be less of a factor.

 

[Davide]

On my 2.2 system with DLBC I experimented with the correction up to Schroeder (about 200hz, from calculator), then up to the area with more peaks and nulls (about 500hz) and the clarity of the stereo image was better in the second case, but the music had something lifeless about it, probably due to the lower perceived reflections (to which our ear is actually accustomed for that specific room).
These two corrections were with a slightly tilted target, like 0dB at 20hz to -8dB at 20khz.

Then I did an interesting test. I extended the correction up to 20khz, tracing a target that follows the trend of the measured frequency response.
In my case it consists of a dip around 2khz (crossover) and a bump from 3 to 5 khz, then roll-off very rapidly.
This results in little suck of life from the speakers and improved the definition quite markedly. This is probably because Dirac intervened by improving the phase of/between the speakers.
This was a very pleasant and natural sound, as opposed to the straight target line (tilted) which was unlistenable.

Despite this, however, I think the most satisfying thing for me is correcting up to Schroeder. It is a bit blurred but more engaging.

I specify that Dirac allows you to make an instant AB comparison, therefore the perceived differences are certainly valid and not the result of mental conditioning.

So in relation to the subject, it depends on what the correction system does in practice.

 

[tmtomh]

On my 2.2 system with DLBC I experimented with the correction up to Schroeder (about 200hz, from calculator), then up to the area with more peaks and nulls (about 500hz) and the clarity of the stereo image was better in the second case, but the music had something lifeless about it, probably due to the lower perceived reflections (to which our ear is actually accustomed for that specific room).
These two corrections were with a slightly tilted target, like 0dB at 20hz to -8dB at 20khz.

Then I did an interesting test. I extended the correction up to 20khz, tracing a target that follows the trend of the measured frequency response.
In my case it consists of a dip around 2khz (crossover) and a bump from 3 to 5 khz, then roll-off very rapidly.
This results in little suck of life from the speakers and improved the definition quite markedly. This is probably because Dirac intervened by improving the phase of/between the speakers.
This was a very pleasant and natural sound, as opposed to the straight target line (tilted) which was unlistenable.

Despite this, however, I think the most satisfying thing for me is correcting up to Schroeder. It is a bit blurred but more engaging.

I specify that Dirac allows you to make an instant AB comparison, therefore the perceived differences are certainly valid and not the result of mental conditioning.

So in relation to the subject, it depends on what the correction system does in practice.

Very interesting - thanks!

One question regarding the phase: my understanding is that Dirac always applies impulse response correction full-range, regardless of whether the user applies frequency response correction full-range. So wouldn’t that mean it was/is doing full-range phase correction in all the scenarios?

 

[Davide]

Very interesting - thanks!

One question regarding the phase: my understanding is that Dirac always applies impulse response correction full-range, regardless of whether the user applies frequency response correction full-range. So wouldn’t that mean it was/is doing full-range phase correction in all the scenarios?

I can't find any documentation but I'm pretty sure the correction of Dirac is limited to the threshold set by the user on both the magnitude and time domain.
In fact, with the correction up to 200Hz, the above frequency band is definitely the same by ears, so not phase corrected.
Perhaps the only thing that persists is the delay and gain of the single speaker, which is surely freq independent.

 

[Nvk]

Now I read that most prefer to use dsp only to the Schroder frequency, or at least no higher than 400 a 500hz, because otherwise it might degrade the sound quality..

Now I was wondering if these experiences mainly belong to rather 'intensive' DSP use where you want to pursue a target curve and generate a relatively complex filter? Or does this also apply when you have a small amount of let's say at most 10 PEQ per channel?

 

[daftcombo]

When I do anything, I EQ out the FR peaks due to room modes speaker by speaker, upward starting from 20 Hz to 150 Hz max. One peak at a time until it doesn't sound bloated anymore. Usually between 0 and 2 peaks to tame per speaker.

The best is to place speakers and seat so that you don't even need to do that.

Bad speakers EQing is another story and need full pseudoanechoic measurements.

 

[Steve Dallas]

I have tested this extensively with KEF and Revel speakers in different rooms with Dirac and XT32. My preference is to correct up to somewhere between 800 and 1200Hz if the FR benefits from it. In some cases covering the entire range of the human voice fundamentals is beneficial. Above 1200Hz, my speakers begin to sound "closed in" for lack of a better description.

Currently, the KEF R3s in my office are corrected to 900Hz with Dirac, and the Revel F206s in my media room are corrected to 450Hz with Dirac and 1100Hz with XT32.

Schroeder is 200 and change in both rooms.

 

[ernestcarl]

It should be okay to EQ “full-range” if you take the extra steps to check and verify how EQ might affect the direct sound or anechoic/quasi-anechoic on- and off-axis response. I does help greatly, however, if the directivity (and room acoustics) is already good. Moving microphone measurements (MMM) can help lessen the likelihood of over-correction. Post EQ, normal swept sine measurements and the accompanying spectral reflection-decay pattern should also exhibit an improvement. Boosting specific areas too much might cause additional distortion especially with louder volume listening. Again, normal swept sine measurements with the EQ turned on (while gradually increasing SPL in 5dB steps, for example) and examining the change in distortion will easily reveal this.

 

[daverosenthal]

I'll vote for full-range correction is best. BUT... You really do have to be careful when playing with FR above the room modes. A few cases where I feel there is a benefit to full range:

1. Adjusting for a different warm/cool tilt. That is, applying EQ that is essentially a straight line, with a few dB tilt either up or down to your own taste (i.e. house curve).
2. Fixing a tendency to exaggerate/reduce a frequency range. But be careful! The trick here is that you have to look at several versions of the frequency response to make sure you aren't making things better in one way and worse in another. For example, look at 1) the FR of direct sound, 2) the FR of steady-state room sound, 3) the perceptual FR. (And, yes, these measurements need to be well spatially-averaged.) If all of these measurements have the same bump at, say, at 2-4 kHz, then it's going to sound great to fix it. But, for example, if steady-state has the bump but the direct sound doesn't, you're not going to help intervening with EQ.

I haven't personally done a detailed analysis of all of the full-range room correction stuff out there but I have to say that a lot of the "before and after" graphs make me very suspicious. Every little wrinkle in the full range seemingly totally flattened. I suspect that most of these system heavily over-correct.

 

[Deleted member 48726]

I'll vote for full-range correction is best. BUT... You really do have to be careful when playing with FR above the room modes. A few cases where feel there is a benefit to full range:

1. Adjusting for a different warm/cool tilt. That is, applying EQ that is essentially a straight line, with a few dB tilt either up or down to your own taste (i.e. house curve).
2. Fixing a tendency to exaggerate/reduce a frequency range. But be careful! The trick here is that you have to look at several versions of the frequency response to make sure you aren't making things better in one way and worse in another. For example, look at 1) the FR of direct sound, 2) the FR of steady-state room sound, 3) the perceptual FR. (And, yes, these measurements need to be well spatially-averaged.) If all of these measurements have the same bump at, say, at 2-4 kHz, then it's going to sound great to fix it. But, for example, if steady-state has the bump but the direct sound doesn't, you're not going to help intervening with EQ.

I haven't personally done a detailed analysis of all of the full-range room correction stuff out there but I have to say that a lot of the "before and after" graphs make me very suspicious. Every little wrinkle in the full range seemingly totally flattened. I suspect that most of these system heavily over-correct.

So? -What does that even mean?

If it sounds good, it sounds good. Then it is not "over-corrected., but "corrected".

 

[daverosenthal]

So? -What does that even mean?

If it sounds good, it sounds good. Then it is not "over-corrected., but "corrected".

Sorry if I was unclear. By "over corrected" I'm not saying corrected too well, I'm saying corrected too aggressively. For example, if you do a single measurement of the frequency response at the listening position and just try to "fix" every little wiggle in the curve that shows up with EQ, it will sound terrible (and be "over corrected" in my terminology). I'm further saying that I suspect (based on their marketing literature showing super-flat "after" curves) that some of the room correction systems fall into this trap.

 

[Deleted member 48726]

Sorry if I was unclear. By "over corrected" I'm not saying corrected too well, I'm saying corrected too aggressively. For example, if you do a single measurement of the frequency response at the listening position and just try to "fix" every little wiggle in the curve that shows up with EQ, it will sound terrible (and be "over corrected" in my terminology). I'm further saying that I suspect (based on their marketing literature showing super-flat "after" curves) that some of the room correction systems fall into this trap.

Of course you shouldn't base you correction on a single measurement. Which marketing are you referring to?

The early version of Yamaha YPAO did correct from a single measurement as I recall. I use Dirac 13-point myself and corrects full range with great succes.

 

[goat76]

Of course you shouldn't base you correction on a single measurement. Which marketing are you referring to?

The early version of Yamaha YPAO did correct from a single measurement as I recall. I use Dirac 13-point myself and corrects full range with great succes.

Have you ever made a gated measurement from the same toe-in angle as your listening position, to confirm that your non-gated measurements lines up with the direct response from your loudspeakers? Otherwise, it really doesn't matter how many positions you take your measurements from if none of them shows the accurate direct response.