basic theory behind what we do

[bjs]

Just wondering...

If we succeed in the impossible and manage to equalize all the peaks and valleys out of our system (making it RULER FLAT) will that automatically mean the system impulse response becomes perfect too?

Or is it possible that we might succeed in flattening the frequency response only to find that it actually became worse in the time domain in some manner?

[brucek]

If we succeed in the impossible and manage to equalize all the peaks and valleys out of our system (making it RULER FLAT) will that automatically mean the system impulse response becomes perfect too?

Nope, not at all.

It takes some careful filtering to control the time domain.

And ruler flat is not so nice to listen too either. We don't hear that way. You don't hear a 30Hz and 100Hz tone played at the same SPL at the same level. The 30hz would seem weaker.


brucek

[JohnM]

If the system is minimum phase then a flat frequency response will correspond to a perfect impulse response. However, systems with excess phase can have flat frequency responses but non-ideal impulse responses. Rooms are mostly minimum phase at subwoofer frequencies and filtering that flattens the frequency response also improves the impulse response, but obtaining perfect inverse filters is more of a theoretical than a practical possibility.

[JohnM]

brucek wrote:

And ruler flat is not so nice to listen too either. We don't hear that way. You don't hear a 30Hz and 100Hz tone played at the same SPL at the same level. The 30hz would seem weaker.

How we hear has no bearing on the correct response of the reproduction chain though, if the frequency response of a system (from original input to final output) is perfectly flat it will reproduce recorded signals exactly as they would have been perceived at the original location they were recorded. Altering the balance may be subjectively preferred but it is no longer what was recorded.

[brucek]

How we hear has no bearing on the correct response of the reproduction chain though

Yes, of course. I was trying to ensure that this wasn't some goal that bjs had, trying to create listening nirvana.

brucek

[bjs]

Thanks guys.

If I understood your response correctly JohnM, this leads to a supplemental question.

Hypothetically, if one saw a peak at say 60hz in REW and equalized that down to flat *but* in reality the peak was due to perhaps a room mode at 55hz and another at 63 hertz interacting...then presumably the flattening would mess up the impulse response (perhaps worse than before)? Correct impulse response would only be maintained if one realized the 60hz peak was actually due to two separate modes and equalized them instead (ie the 55hz and the 63 hz separately). The latter would be the only approach that would preserve minimum phase behaviour of the equalization. Is this a correct?

PS. Brucek...I've got REW, and my BFD, AND the correct cables...I'm already in listening nirvana...and I haven't even turned on the stereo!

[thewire]

JohnM wrote:

How we hear has no bearing on the correct response of the reproduction chain though, if the frequency response of a system (from original input to final output) is perfectly flat it will reproduce recorded signals exactly as they would have been perceived at the original location they were recorded. Altering the balance may be subjectively preferred but it is no longer what was recorded.

My room has more bass 70Hz and up or does the signal? Both?



With receiver and bass management



Without a receiver and bass management (same room setup mostly, same type of sub but more of them, different settings.)

Something seems to be occurring before the signal reaches the subwoofer other than just a +10dB boost. Maybe the LFE crossover? I read that an LFE crossover targets certain frequencies. Is that true?

[JohnM]

bjs wrote:

Hypothetically, if one saw a peak at say 60hz in REW and equalized that down to flat *but* in reality the peak was due to perhaps a room mode at 55hz and another at 63 hertz interacting...then presumably the flattening would mess up the impulse response (perhaps worse than before)? Correct impulse response would only be maintained if one realized the 60hz peak was actually due to two separate modes and equalized them instead (ie the 55hz and the 63 hz separately). The latter would be the only approach that would preserve minimum phase behaviour of the equalization. Is this a correct?

The wrong correction in that case wouldn't mess up the IR but it wouldn't improve it much, and the frequency response would not change as predicted (i.e. the measurement of the equalised response would not match the prediction) if the peak being countered was actually a combined effect of two nearby peaks.

[JohnM]

thewire wrote:

Something seems to be occurring before the signal reaches the subwoofer other than just a +10dB boost. Maybe the LFE crossover? I read that an LFE crossover targets certain frequencies. Is that true?

The bass management in your receiver applies a low pass filter to the signal, the -3dB point of the filter is at the bass management cutoff frequency. That low pass filtered signal gets sent to the subwoofer, whilst a high pass filtered version gets sent to the main speaker for that channel. If you want to see what the signal looks like completely free of room effects connect the subwoofer output signal from your receiver to the line input on your soundcard and measure that.

[thewire]

JohnM wrote:

The bass management in your receiver applies a low pass filter to the signal, the -3dB point of the filter is at the bass management cutoff frequency. That low pass filtered signal gets sent to the subwoofer, whilst a high pass filtered version gets sent to the main speaker for that channel. If you want to see what the signal looks like completely free of room effects connect the subwoofer output signal from your receiver to the line input on your soundcard and measure that.

Can you please explain what this quote means. Are they saying that a subwoofer needs to measure louder to be flat?

Home-theater subwoofers require a boost in output in the 40-80Hz range to make effects seem real and solid. (Why this is not encoded on the DVDs themselves is beyond me. If the DVDs simply had the boost encoded where it is needed, one subwoofer could serve for both music and movies.) Subwoofers for music need linear response from around 20Hz up to 160-200Hz. Surprisingly, movies are not that dependent on response below 40Hz or so. When there is sub-40Hz content in the soundtrack, there seems to be some built-in augmentation to make those frequencies heard and felt, even on subwoofers that have a hard time going that low. DVD soundtracks played back on music subwoofers with linear frequency response sound a bit restrained and lack the impact and serious amounts of air, floor, and wall motion that home-theater subwoofers can generate.

[Wayne A. Pflughaupt]

JohnM wrote:

How we hear has no bearing on the correct response of the reproduction chain though, if the frequency response of a system (from original input to final output) is perfectly flat it will reproduce recorded signals exactly as they would have been perceived at the original location they were recorded. Altering the balance may be subjectively preferred but it is no longer what was recorded.

No production house or recording studio uses "perfectly flat response" (the movie industry uses the X-curve, for instance). And, unless you're listening in an identical room with identical speakers and electronics as the production house or studio, you're not going to hear "what was recorded" anyway.

Regards,
Wayne

[thewire]

Wayne A. Pflughaupt wrote:

No production house or recording studio uses "perfectly flat response" (the movie industry uses the X-curve, for instance). And, unless you're listening in an identical room with identical speakers and electronics as the production house or studio, you're not going to hear "what was recorded" anyway.

Regards,
Wayne

I have to agree with this. I look at many graphs studying and even the finest studios have a curve. None of the rooms measure perfectly flat. There are so many other great features in REW and a sweep is just one of them. When a person says here is a flat response... I would rather look at RT60, room size details (air resonance?, room modes?), what diffusion is there and are the other speakers doing well, whats the decay like? What is the reference level sound good at? That kind of stuff.

[JohnM]

Wayne A. Pflughaupt wrote:

No production house or recording studio uses "perfectly flat response" (the movie industry uses the X-curve, for instance).

The X curve is primarily to compensate for the perceptual differences of listening in very large spaces and the correction becomes less and less as the size of the space reduces, in domestic sized rooms (extrapolating the data for the larger spaces in the spec) the X curve correction would be flat or very close to it.

Wayne A. Pflughaupt wrote:

And, unless you're listening in an identical room with identical speakers and electronics as the production house or studio, you're not going to hear "what was recorded" anyway

What was recorded is what the mic picked up. The mastering engineers may well have applied all kinds of EQ to it afterwards so it sounds how they want it in their production facility, or they may not, but I don't believe they have a consistent bias in the balance they apply that warrants targeting a bass reproduction response that is not flat. Many people like a rising low frequency response, just as many people like ketchup on their food, but both are personal taste and not a correction for some underlying imbalance.

[JohnM]

thewire wrote:

Can you please explain what this quote means. Are they saying that a subwoofer needs to measure louder to be flat?

In the opinion of the author of that extract DVD soundtracks require boosting between 40 and 80Hz.

[lcaillo]

brucek wrote:

And ruler flat is not so nice to listen too either. We don't hear that way. You don't hear a 30Hz and 100Hz tone played at the same SPL at the same level. The 30hz would seem weaker.
brucek

This is a point that is correct, but one has to be careful of the context in which it is applied. If we are talking about calibrating a system that is not itself intended to correct for perceptual effects, then it is not relevant. If we are trying to get the system to faithfully reproduce what is input, we want flat response. One of the important points about calibration is that it removes variables from the reproduction chain so that we can better appreciate the production. If we then want to equalize to our preference, either for program specific correction or a general preference, there is nothing wrong with doing so. When calibrating a system, or comparing systems, however, IMO it is best to target flat response to start with.

We run into the same issue with video, where I do much more calibration work. People are conditioned to like higher color temperatures. If you calibrate to a higher color temp, however, you bias everything in one direction. Programs that were produced with a bias may look better or worse but those that happen to be off in the same direction are much worse than if you start with a neutral display. There is nothing wrong with changing the settings to one's preference, but given that you have variance in the programming, it is usually useful to start with a system that does as little as possible to alter what is input. Then one can make better judgements about what corrections are appropriate and desired.

The bottom line is, when we talk about calibration of any system, we have to start by understanding what the system is doing to the signal. Only then can you adapt that system to the preferences of the user or if you prefer, the perceptual tendencies of a population.

[brucek]

it is best to target flat response to start with

Yeah, I suppose, but I would think you would need to listen to music at the same SPL level that it was originally mixed at to get the same sound as the engineer heard.

SPL level is quite important in determining the type of emphasis you want to apply to your low frequencies.
The weighting curves certainly have a range they use that changes the emphasis as the level of low and high frequencies change.
See below:

A-weight recommended level use between ~ 20dB - 55dB.
B-weight recommended level use between ~ 55dB - 85dB.
C-weight recommended level use between ~ 85dB - 140dB.

So we can see as the SPL level gets lower, we require more compensation with regard to the filtering needed to make the same frequency 'sound' at the same level to our ears.

Of course we remove any weighting with REW, to produce a flat response, and I sure don't see any other way than that to do it either, as it trys to reproduce the original mixed sound. But I ain't gonna listen to music that way.. especially since I listen at such low levels

brucek

[thewire]

It looks like a Media Director would give us more than a standard X-Curve. Do you see any reasons why this would not work? We would have to start at a flat response? If the director intended a rising response than how does one get that rising response when starting at a flat one? This would require boosting then I would think. If it were flat to 10Hz and the content says there is roll off at 20Hz, this would also mean allot of cutting. I would think then maybe that the best accurate way to reproduce the intended experience would be to get close as possible to the intended frequency response, prior to changes being made. I also wonder if there would be both an intended version for both home, and commercial cinemas. I would think that rather the director would also want this to become a faithful reproduction in both areas, therefore most likely the mix would be intended for commercial theaters, where response is not flat. There would of course be limitations as Marshall discussed with storage and not all may like the idea.

http://www.thx.com/technologies/mediadirector/how.html

[DrWho]

brucek wrote:

Yeah, I suppose, but I would think you would need to listen to music at the same SPL level that it was originally mixed at to get the same sound as the engineer heard.

That is a common misconception.

Take for instance a very simple example of a studio engineer recording a cello...single mic placed about 4 feet away with the cello sitting in a nice full sounding studio. And then assume the goal of this studio engineer is to capture exactly how that cello sounds in that room.

If the engineer is monitoring the mix at say 90dB, then the engineer is going to call upon his acoustic memory of what a cello sounds like at 90dB. If instead, he chooses to listen at 70dB, then he will call upon his acoustic memory of what a cello sounds like at 70dB. His acoustic memory is already compensating for the F-M curves.

Trying to make your system respond with the inverse of the F-M curves is not an approach towards more accurate sound. If anything, it's more an approach of hyper detail - kinda like super bright neon colors. Another result is that we can almost (it's not perfect) make things appear louder than they really are because our acoustic memory gets mixed in with the perceived tonal balance.

But regardless, if you're listening at 50dB, then you're hearing the correct tonal balance of the cello as if it were playing at 50dB. I don't think it makes much sense to listen to a 90dB cello at 50dB (when accuracy is the goal).

EQ...that term makes me shiver. Here are few drive-by comments...

This is A LARGE can of worms that would be well worth discussing and requiring everyone to check off on, is a thorough understanding of the limits of equalization and the recognition of just what can, and cannot be equalized! As this subject is one of the lest well understood, and subsequently, one of the most abused.


First, the frequency domain is not the domain of primacy. That distinction belongs to the time domain. Simply put, I would suggest that for all practical purposes, except to view the room modes, that you stop looking at the system in the frequency domain. (A gross oversimplification perhaps, but one that would have the practical effects of actually resulting in major improvements without the concomitant errors SO COMMONLY made!)

If you resolve issues in the time domain, you resolve much of what is displayed as an anomaly in the frequency domain.

A thorough understanding of what can, and cannot, be equalized is required, and EQ is among the last thing to be utilized. Simply put, one cannot equalize a non-minimum phase environment. Without getting into the math or discussing systems where the poles exist only in the negative (left) half of the S plane, or as a system that can release its potential energy in minimum time, (is everyone confused yet? ;-) ) one might think (a bit over-simplistically) of minimum phase as a system (or regions of a system response) that exhibit no destructive superposition of real and virtual signals.

{Edit - add: After addressing room modes...} All the more reason to run for the ETC diagram with displays and details of the discrete reflections within the envelope as a 'first defense' for general acoustical analysis and remediation.

[thewire]

thewire wrote:

It looks like a Media Director would give us more than a standard X-Curve. Do you see any reasons why this would not work? We would have to start at a flat response? If the director intended a rising response than how does one get that rising response when starting at a flat one? This would require boosting then I would think. If it were flat to 10Hz and the content says there is roll off at 20Hz, this would also mean allot of cutting. I would think then maybe that the best accurate way to reproduce the intended experience would be to get close as possible to the intended frequency response, prior to changes being made. I also wonder if there would be both an intended version for both home, and commercial cinemas. I would think that rather the director would also want this to become a faithful reproduction in both areas, therefore most likely the mix would be intended for commercial theaters, where response is not flat. There would of course be limitations as Marshall discussed with storage and not all may like the idea.

http://www.thx.com/technologies/mediadirector/how.html

To correct myself after speaking with my instuctor John Dahl, the Media Director does have a PEQ and everthing else. It will setup the system exactly like the studio had it unless settings are overiden. The reponse we recommend to be flat and within +-2dB of 0dB to start with. The settings come on the disk themselves.

[Wayne A. Pflughaupt]

Mark,

What is the extent of your experience using equalizers?

Regards,
Wayne

Wayne A. Pflughaupt wrote:

What is the extent of your experience using equalizers?

Quite extensive over a near 35 year period.
They work great in a minimum phase environment - notably on source material in recording and to effect the direct signal from a speaker - assuming the various acoustic origins are in alignment (which is usually limited to one physical axis as they are normally stacked and not coincident).

But they are not effective in correcting for the effects of superposition of multiple sources, real and/or virtual. And unfortunately, this is exactly the case that too many try to resolve when one tries to EQ a system in a room - (as only the direct signal Ld can be effectively EQ'd - and the rare regions (passbands) of the response at a particular location where the passband signal may be minimum phase - where a parametric eq can be of use.)

But I will also stick my neck out and argue vehemently against the all too common use of simply measuring the room's frequency response - either at a location or averaged over the room - and simply inverting the signal and reapplying it as feedback in the thought that we can address room anomalies by tampering with, and flattening the frequency response in the frequency domain. If only acoustics were so simple...;-)

Traditional use of EQ simply imparts small LC induced changes in phase of the direct signal(s) which when recombined with multiple direct signals as well as any virtual sources result in small shifts in the comb filtering and polar anomalies. It was always instructive to understand that the 'walking spectrum analyzers' (he said affectionately) I know who can identify a signal within 3-5 Hz in ringing out a system merely caused the nulls at the FOH mix position caused by overlapping coverage of the various arrays in a large hall to be shifted 6 feet onto the paying customers' seats... and since no one really cared what happened there (and as so few folks have ever walked across a hall and have a clue as to the extent of the affect of the polar anomalies), the problem was "solved".

This case was further 'brought home' at the SAC seminars where two Auratone speakers were stacked vertically and fed with identical signals and the comb filtering was displayed (at frequencies above the quarter wavelength limit below which they would effectively sum) and to watch and listen as the top speaker was slowly physically shifted back, rendering both the measured phase and resultant frequency domain comb filtering readily apparent in the measured display while the listening experience was akin to that of listening to a high Q air raid siren or Leslie being rotated about the room as the polar anomalies became more greatly exaggerated as the inter-driver spacing varied.

And if that was not dramatic enough, the fun part came when a cut and boost capable EQ was put into the signal path - at which point the valiant attempts to use the 'classical' approach (meaning, use whatever tools you have on hand!) resulted not in resolution, but rather in fascinating phase wrap anomalies that are hard to describe in words, but where it is painfully obvious the something 'ain't quite right'! Not to mention resulting feedback issues...

The above demonstration remains one of the simplest and yet most dramatic demonstrations that I have ever encountered in audio.

The issues of which I refer (attempts to EQ non-minimum phase environments) were hot topics during the ~1987-1991 period when principles such as Don Davis - who ironically first introduced the 1/3 octave Equalizer while at Altec and subsequently illustrated this limitation quite convincingly with the assistance of the TEF and folks such as Don Keele and others; and at least in the pro markets this debate has thankfully ceased to be a debate - and the racks of EQ that used to be so common in SR rigs are nor reduced to one or two parametric or third octave 'cut only' units now - with EQ largely limited to the direct mic sources. Unfortunately, the use of EQ seems to continue unabated in the 'audiophile world' where the frequency response perspective unfortunately remains a primary focal point to the exclusion of an effective awareness of the time domain.


Edit2:
For more detailed information regarding the limitations of equalization, see Sound System Engineering, Davis & Patronis, 3rd Ed. , Chapter 14, Signal Processing.

[Wayne A. Pflughaupt]

I have no idea what any of that means, but this Shack member was very pleased with the results he got using a pair of Rane parametrics.

Regards,
Wayne

[JohnM]

mas wrote:

Unfortunately, the use of EQ seems to continue unabated in the 'audiophile world' where the frequency response perspective unfortunately remains a primary focal point to the exclusion of an effective awareness of the time domain.

On what basis do you say that? Just about all inhabitants of the "audiophile world" I have come across are strongly opposed to equalisers in general and full range EQ, as you were discussing, in particular.

JohnM wrote:

On what basis do you say that?

With all due respect, this question genuinely made me laugh.

If only your assumption were the case! To coin a phrase, "Tis a consummation devoutly to be wished"!I sincerely wish that I could agree.

One begins to relate to how Dorothy felt when asking for directions from the Scarecrow in the Wizard of Oz, as the subsequent responses are diametrically opposed. I only wish the response to my post mentioning some of these issues had been met with anything other than a statement that they were not familiar with anything that I said! So, I guess that we could begin there...or simply with the original assumptions posited in this original question for this very thread!

Not only are the majority still living in the flatland of the frequency domain and far too many totally unaware of time domain aware tools such as TEF, EASERA, SMAART, PRAXIS, et al, simply mentioning the fact that EQ is a technique of limited applicability is almost assured to initiate a heated debate.

In fact, what I wrote above regarding the inability to EQ non-minimum phase 'environments' - which BTW is correct, was met with anything but acceptance, and I am sure that far more who have not commented have little idea of that which I refer!

This topic is perhaps the simplest way to ascertain the mindset and awareness of an audience. How many possess a sufficient understanding of modern acoustical models such that they can delineate the limitations of equalization? And the fact is that this debate is limited almost exclusively nowadays to the realm of home based 'audiophilia'.

And while there has been a growing understanding of this fact, primarily due to the substantial efforts of Don Davis and the consortium of prominent figures in acoustics via SynAudCon, all one has to do is visit any forum. In fact, few things would please and impress me more than to have this issue cease to be a prime source of misunderstanding, as it would reflect an increasing understanding of the primacy of first addressing (most) issues subject to superposition via techniques such as signal alignment in electronics as well as acoustically within the time domain.

In fact, when you ask why one might suggest this topic as a litmus test of folks awareness, may I refer one to DrWho for his take on the response generated elsewhere when this topic was breached...


As I have said, I wish MORE were aware of this. And I wish comments regarding this topic were met more with "yeah, we already know", or "of course, you are preaching to the choir" rather than the all too common response of dealing with masses of folks showing up in droves with pick forks and torches to burn the witch!



A good introduction to the issue can be found in John Murray's post located at:

http://www.prosonicsolutions.com/articles/Equalization%20Revisited.pdf
{edit: Note, I do not completely agree with his conclusion regarding the treatment of room modes. Resonant traps and absorption are indeed useful.}



And Davis and Patronis address this issue in a bit more technical depth in Sound System Engineering, Chapter 14.

Take care...