EQ vs. Room Treatments

[bpape]

Sorry if that was confusing. My point I guess was that using a narrow band absorber only addresses the amplitude and decay of that narrow band by definition and design - and the rest is still left untreated. Yes, that 60hz in the above example would have both changed.

The example I was trying to address was specifically nulls off the rear wall. There are others but this is an easy one to visualize. If we have a reflection off the rear wall that meets a wave coming from the front at point X, there is a cancellation if they arrive at that point out of phase with each other. There is a reinforcement if they meet in phase at that point. If we can minimize the strength of the reflected wave by absorbing some of it, there is still a cancellation, but not as much - thereby creating a less intense cancellation.

I agree that it's nice to be able to have a civil discussion as many times, these go the wrong way and can get very heated.

Bryan

 

[peerlesser]

Agreed on both being needed. I'll also agree that from about 35-40hz down an EQ is a better option due to sheer size and issues with tuned absorbers.

What's important to remember is that ALL peaks and nulls which are room induced are the result of reflection. From the stand point of bass response, ones that come back and mix in phase cause a peak. Those that come back out of phase cause a null. Damping the reflection will not elminiate the issue but can minimize the extent to which they build on each other or cancel each other. Using an EQ will increase or reduce the intensity of both by the same amount - thereby leaving the same relative change in response.

Bryan

In FIR models, peaks can be viewed as two be constructive reflections that have +/90 degree shift compared to the direct path, nulls have destructive 90-180 or -90 to -180 degree phase difference. Complete null on FIR approach (low order reflections) would come with 180 phase shift and amplitude 1. They can best be corrected with damping material, as even gettin the factor to 0.5 will change infinite attenuation to -6dB. For reinforced effect the +6dB is maximal in FIR approach for single early reflection. In small rooms, considering the dimensions of it, low frequency peaks and nulls can not be heritage of FIR approach (1st or 2nd order cancelling reflections) due to the lack of long enough path to inflict more than 30 degree phase difference compared to the direct signal.

Another mechanism is needed to explain this, this is called IIR part of the room. When people talk about time response they usually refer to the mixture of FIR and IIR part of the room. Standing waves are good extreme example of IIR part. The IIR-model simply uses combination of poles and zeros instead of only zeros. It should always be aknowledged that natural systems are mostly minimum-phase, and also that changing the amplitude with any realizable filter will inevitably change the time domain too. Highest peaks are more IIR-caused than FIR, and they can be corrected with proper (nearly) minimum-phase inverse IIR. This will inevitably correct both time and minimum-phase time domain behaviour, if performed correctly.

When again, aforementioned FIR phenomena should not be corrected with DSP in real-time, as it would require massive delay in order to make the system causal. It could be, however, be corrected if the sound file would be preprocessed, where maximum-phase and excessive linear-phase FIR filters could be used (or ultimately no real-time realizable zero-phase filterring). For this reason acoustic damping is better for FIR problems, DSPs for IIR ones. The -20dB late reverberation time is dominated by the worst of these two (EQ&treatment), -60dB time is dominated by the total absorption of the room in Sabines formula (0.161*V[m^3] / A[m^2]).

Indeed, a good conversation.

 

[bpape]

Agreed - for the most part.

Your statement of changing the time domain with EQ is true, however, changing it for the good requires an exact inverse match of the curve to be currected. This is simply not possible for multiple rows of multiple seats. Take the typical home theater with 2 rows of 3 seats. I have yet to see one where I could find anything that matches exactly in all 6 seats to be fully correctable.


In fact, when there is not a perfect match, time domain can actually be shifted in the wrong direction for those parts of the curve which do not match exactly. There was a thread here on Audioholics where some plots were posted showing this effect - wish I could find it.

As for not being able to deal with low frequency abberations in a small space, again, I'll have to respectfully disagree. While you may not have pure full length modal issues, you can certainly have cancellations of frequencies lower than the room length (or any dimension) would suggest that can still be helped via minimizing the intensity of the reflections.

Once you get below about 40Hz, I'll agree that it becomes very difficult to do with velocity type absorbers. Helmholz absorbers present their own issues down that low. Panel type membranes can work well in those cases but in most residential situations, I'll agree that down there (below 40Hz), EQ is likely the best solution most of the time for dealing with peaks in response.

Bryan

 

[peerlesser]

In fact, when there is not a perfect match, time domain can actually be shifted in the wrong direction for those parts of the curve which do not match exactly. There was a thread here on Audioholics where some plots were posted showing this effect - wish I could find it.

No need for plots, I've seen that many times on my work. Every nothcing filter is also a impulse excited oscillator, so somewhere else the time domain can get worse if only one singular point is used for EQ. Same thing happens with bad EQ when center frequency of tight mode is off by 2Hz. This can happen especially near walls. For large areas, it takes comlicated DSP algorithm to make sure local adjustments are not done with the cost of global results. With acoustic treatment you can't go wrong there, always safe.

As for not being able to deal with low frequency abberations in a small space, again, I'll have to respectfully disagree. While you may not have pure full length modal issues, you can certainly have cancellations of frequencies lower than the room length (or any dimension) would suggest that can still be helped via minimizing the intensity of the reflections.

They can happen of course, but they are high order IIR type or multiple paths early reflection (FIR type), not 1st or 2nd order.With higher order, propability increases that they start to get quite local. It will be bad luck if even both ears can fit to this kind of zone. Dealing efficiently with them will most likely need measurement setup and well placed treatment.

Once you get below about 40Hz, I'll agree that it becomes very difficult to do with velocity type absorbers. Helmholz absorbers present their own issues down that low. Panel type membranes can work well in those cases but in most residential situations, I'll agree that down there (below 40Hz), EQ is likely the best solution most of the time for dealing with peaks in response.

DSPs have great impact on T(-20dB), whereas treatment (and furniture) affects T(-60dB). It should be noted, that T(-20dB) presents the time it takes the energy to decay 99%, whereas T(-60dB) is from there on the time for the rest 1%.

Before the late reverberation time from -20dB to -60dB has more meaning, first 0 to -20dB should be cleaned. Modern DSP such as 8033 can improve T(-20dB) beyond any realistic treatments automatically in few minutes without any input from user. But then, yes there might be sweet spot and countering nulls needs independent access from DSP to multiple sources. Twos subs with good DSP is quite good a deal even when absolutely no treatments or placing the subs can be done.

 

[bpape]

Regardless of WHY the abberations are there, what kind they are, etc. - the fact is that they can be dealt with via treatment as you stated.

Also, while I'm not a statistician by any means, decaying 20db is not 99%. When you have a properly designed space with a noise floor of say 35-40db (or lower) and you have bass impulses pushing 100db in a home theater, saying that I can drop that ringing down to 80db in 1/3 second (and admittedly making it longer in other areas if the EQ isn't exactly correct re center frequency and Q) doesn't make me think that's sufficient - especially when you may have dialog going on that's below 80db and in the range of the harmonics of some of those fundamentals.

Again, not disagreeing that EQ can be a very useful tool - just looking at things practically and trying to get the facts straight for those who are not into the science.

Bryan

 

[peerlesser]

Also, while I'm not a statistician by any means, decaying 20db is not 99%.
Bryan

Yes it is, -20dB is 10^(-2.0) = 0.01 by definition

If the remaining energy is 1%, it has attenuated 99%. The magnitude is on the other hand decayed 90%.

I'm not saying it's sufficient condition for anything, but it has huge weight on frequency masking. Yes, speech can still be below 80dB compared to 100dB initial bass value in your example, but it lies in the different Equivalent Rectangular Bandwidth compared to subwoofer range, as is everything else coming from other speakers, if the cross-over is steep. Subwoofer distortion masks in frequency as you stated. And there will be still post-masking effect, which is why 99% is not enough by any means. Having M.Sc. degree in both Acoustics and Audio Signal Processing, I don't want to take sides in EQ vs treatment, but I can point out they both solve different problems. My current work is actually more related to the acoustics, but the fact is ASP is evolving right now way faster because of DSPs. They will never completely remove the need for treatment, but all the time they give a lot more audible improvements for the effort and money.

 

[bpape]

I think you need to read this...

http://www.hearinglosshelp.com/articles/decibelsvspercent.htm

It will explain why you can't do what you're trying to do with the math.

To follow your math, then if I started at 150db and dropped down to 130db, it has decayed 99%? So dropping to 110db is what 99.99%?

Bryan

 

[peerlesser]

To follow your math, then if I started at 150db and dropped down to 130db, it has decayed 99%?

Yes, of course the energy has dropped 99%. It does not matter what the initial level is, +20 dB is always 100 times the power, and -20dB is always 1/100 times the power. Decibel scale was invented to measure energy in log scale. Man we can have some different experiences and opinions, but you can't argue with math. Decibels measure energy, hearing has other measurement units like sones, phones, barks scale and ERBs. So decibels are not directly related to hearing, it's a unit for devices such as SPL meter and microphones. It correlates still well as a link between the hearing and measurements. Decibels are handy when the ratios are more important than absolute differences. I have given lectures in the university about these.


If you read your link carefully, it is talking about the reference level. In this case, it is the maximum level where your waterfall graphs starts. You can use percents just as decibels if you are talking about the relative difference compared to the initial (reference) level. Only thing that the link says is the simple fact that dB are relative measure, and cannot be used to describe absolute power (in absence of reference). It describes power ratios. In SPL readings, treshold of hearing is considered the reference power level 0dB, or pressure of 0.0002 pascal at 1000Hz.

So dropping to 110db is what 99.99%?

Bryan

Yes. Why don't you simply do the math and check it? Dropping 60 dB is always 99.9999% regardless of where you start. Definition for T(-60dB) was invented from the time it takes power to drop to one millionth. Dropping 110dB to 0dB would be in percents 100 x (1-10^(-11)) =99.999999999% attenuation.

And dropping from 150 to 110dB is 1/10000 time the initial power, so its 99.99%

I know decibels have to be recalled once in a while:

decibels = 20*log10(A/A_ref)
power ratio = 10^(decibels/10)

where A_ref is the reference level I mentioned



edit:
ratios to percents

 

[bpape]

I don't argue with the math. I disagree with the interpretation and presentation of the math. Yes, 20db is 99% reduction in power. But we don't hear power, we hear spl. Let's look at an example that all can understand.

3db is pretty universally agreed to be the smallest amount of volume change most people can hear. A well trained ear in a very quiet room with pure tones can hear 1 db. Most of us, with complex music and tones, 3db is a good agreeable point.

According to your logic of how to apply the math, that 'smallest amount we can hear' (3db) is reducing things by 50%. It is from a power standpoint, but not 50% of the SPL which is what we hear. This is the problem with the presentation and interpretation.

Again, if you reference the link I presented, ANY use of percentage in this context is completely incorrect since in order to properly use percentages, you MUST have a defined top and bottom. In SPL, there is no top - hence my question about 150db starting point. In addition, there is variability as to the bottom of the scale. Where is the bottom? Is it 0db? Is it the ambient noise floor of the room? Either can be argued.

Where's that beating the dead horse emoticon whey you need one...

Bryan

 

[peerlesser]

Yes, 20db is 99% reduction in power.

That is all I claimed, good thing that you finally can finally confirm these mathematical basics. I agree measurements scales are not always illustrative for the viewpoint of auditory system, but reflection factors and pass-trough absorption of acoustic tratments can be calculated best in dB and percentage scales. As you know, cascading acoustic material sums decibels, and parallel configuration of materials sums percentages of absorption. This way it is easier to know how much and where the materials are needed to give the desired result in given situation. It is thus also easier to compare the results to those achieveable by only EQ.

But back to the topic. According to the recent test, some Audysseys, such as the one in Onkyo TX-SR875 and Denon AVR-3808 are worthless in improving the lower range for subwoofer. In the test, they merely dropped the level of the subwoofer to compensate for the room resonances. So with any EQ or DSP, you better know exactly if it's one of those that actually work. These are not even the cheapest A/V receivers having Audysseys, and already they are not doing anything useful. They also over-equalized room effect on mid- and high ranges, which is something that should not be done in the first place with any DSP algorithm. Generally, only low ranges benefit from DRC or EQ.