Stage 2

[Ethan Winer]

John,

> I didn't say nulls were not audible. I said that the studies I have seen indicate that peaks and long delay are subjectively more objectionable. I am not pro-null! <

Understood, and I've heard others quote that study (or maybe there's more than one?) about the audibility of nulls. My point is those studies are useless unless they take into account the frequency of the nulls and the key of the music. My own personal experience treating a lot of rooms is that nulls are at least as damaging as peaks if not more so.

--Ethan

 

[Ethan Winer]

Chris,

> I may take the effort to produce a careful test scenario in an empty room with a corner loaded subwoofer, using very careful equalization below 70Hz, and provide the results here. <

That would be awesome. Make no mistake, I'm here to learn as much as to teach, and the more data we have to examine, the more we all can learn.

--Ethan

 

[bpape]

OK. Time for me to learn now.

If EQ can help to TAME ringing in a perfectly inverse world, then how can it not make it worse when you're not perfectly inverse? Seems to me we're trying to have it both ways. Is it really that the imperfect one would OVER-deaden a particular band?

 

[JohnPM]

If EQ can help to TAME ringing in a perfectly inverse world, then how can it not make it worse when you're not perfectly inverse? Seems to me we're trying to have it both ways. Is it really that the imperfect one would OVER-deaden a particular band?

I think some confusion arises from thinking that an EQ filter applied to a mode somehow generates an inverse of the mode's response to cancel it. That is not the case. At low frequencies the room's modal behaviour at a given point can be accurately modelled as a 2nd order damped resonant system. The mode will have a certain gain at its centre frequency, and a certain bandwidth which is directly related to its decay time - the narrow the bandwidth, the longer the decay. To counter such an effect requires a similar second order system but which has a loss equal to the mode's gain. A parametric EQ filter can provide that. However, 2nd order systems of a given bandwidth with loss decay faster than 2nd order systems of the same bandwidth with gain.

To explain this for those with a background in sampled systems: the z-transform of the 2nd order system has a pair of complex conjugate poles and a pair of complex conjugate zeroes, all inside the unit circle. Systems with gain (like room modes) have the poles closer to the unit circle than the zeroes, systems with loss (negative gain, like EQ filters to counter modes) have the zeroes closer to the unit circle. The closer the poles are to the unit circle, the longer the decay time. To exactly counter the poles and zeroes of a mode requires a filter which has zeroes where the mode has its poles and poles where the mode has its zeroes - when that happens, the mode's effect on the room's response vanishes completely. This will be the case at the set of points in the room where the mode's gain is exactly countered by the filter's loss and the bandwidths of both are the same, this is a set of points because the standing wave corresponding to the mode will have more than one physical location where those conditions arise, the higher the order of the mode (i.e. the higher its frequency) the more such locations there are. If the filter were applied in the absence of a corresponding mode, it still has faster decay than the mode because its poles are further from the unit circle.

To show that in a more digestible form here is a waterfall plot of a perfect isolated mode at 70Hz with a gain of 6dB and a bandwidth of 5Hz. The plot covers SPL from 37 to 97dB, the time axis covers 600ms and the nominal response is flat at 90dB. The decay of the mode is 1.9dB for each 20ms slice of the waterfall. The last visible slice of the response outside the range of the mode is at 180ms, after that it has dropped off the bottom of the plot.

The next plot is of the corresponding correction filter alone, i.e. applied to a system with a flat response, to show the worst effect that could arise from applying a filter where there is no modal response to correct. The filter is -6dB at 70Hz with the same 5Hz bandwidth (which corresponds to 3/60ths of an octave on a BFD Pro). The filter's response hits the 37dB floor of this plot at 360ms. The level at 70Hz is below the decay of the unfiltered parts of the response up to the 160ms slice, by which time the level has dropped by 20dB. Its rate of decay beyond this point is 4.5dB in each 20ms slice.

Clearly the filter has a resonant tail, as do all such second order systems. However, the decay is fast and the starting level is already below the rest of the frequency range. It is doubtful whether the decay of this filter would be audible, though its effect on the frequency response certainly would be.

The upshot of all that is correction filters of course will have an audible effect on the frequency response in locations where the modal response they are aimed at does not exist, but they add little of consequence to the decay time. It is worth noting that this is NOT the case for filters that have positive gain, as some might apply to address a null. A narrow filter (say the 5Hz used in this example) in a region where the null does not exist would produce a response like the ideal mode pictured above and would most definitely contribute ringing. That is why filters with positive gain, if used, should be restricted to wider bandwidths (not less than 1/6th of an octave) as they will then decay rapidly and avoid this problem.

HTH,

John

 

[bpape]

Excellent. Thanks John.

So, based on your explaination and graphs, there IS in fact an introduced additional ringing (though lower in level relatively) when EQ is applied in a non-inverse fashion. If it's not too much trouble, would it be possible to show a graph where you DO have a peak and ringing - say at 65Hz but the cut is applied at 70Hz? I'd also be interested (as would others I think) in seeing a case where say you have a peak at 60Hz and at 70Hz but the EQ applied is a BROAD cut applied across them with a wider Q?

Thanks again John.

Bryan

 

[JohnPM]

Applying a broad filter to correct a pair of modal peaks (such as one might be tempted to apply if looking at a one-third octave frequency response where the individual peaks merge into one), there is no effect at all on the modes' ringing, just a gentle drop in the overall level following the broad filter's response. The image below shows a pair of 6dB peaks at 70Hz and 80Hz, both with bandwidths of approx 5Hz, plus a -9dB correction filter at 75Hz with a bandwidth of 1/3rd of an octave (about 18Hz). The peaks' ringing has not been reduced at all.

Similarly applying a 75Hz correction filter for a 70Hz modal peak has no effect on the peak's ringing - effective correction requires that the centre frequency of the filter is within 1% of the modal frequency, which needs 1/60th octave resolution. The image below shows this for a 70Hz, 6dB, 5Hz wide peak and a 75Hz, -6dB, 5Hz wide filter.

 

[Buckle-meister]

Tentative decision.

It's always nice when you find out you were right. Given the depth of discussion going on, I am definitely not ready for a PEQ. I've more or less decided that even if I struggle to get it to work, I quite fancy having a go at building a Helmholtz resonator anyway. If nothing else, physically building it should deepen my understanding of the influence of the various parameters involved. Just for the record though, could somebody provide an answer to this part of my original question please?:

If the volume stays the same and the resonator is tuned to the exact same frequency, are short dumpy resonators just as good as tall thin resonators? In other words, do I have the freedom to build a resonator of any cylindrical diameter and length and simply use the port to tune it to my target frequency?

 

[JohnPM]

You shouldn't let the discussion put you off Robbie, the process of using EQ is pretty straightforward if you have the right tools to help. You don't need to know the underlying systems theory any more than you do for the resonator. To answer your question, in theory the shape of the resonator does not matter, only its volume, though in practice the shape will affect the stiffness of the structure and hence the Q. The proportions of the tube do affect the Q, with larger diameters giving higher Q figures, but the theoretical Q usually bears little relation to the final result.

 

[Buckle-meister]

You shouldn't let the discussion put you off Robbie, the process of using EQ is pretty straightforward if you have the right tools to help. You don't need to know the underlying systems theory any more than you do for the resonator.

But knowing the underlying theory must be advantageous. It's like math; I'm a firm believer that telling someone who's stuck with a math problem the answer is of only negligible value. Sure, it may get them through the question, but what happens when they get asked something only slightly different? They can't do it. Better to show them the solution. That way they can solve any number of similar problems because they understand the underlying theory.

...in theory the shape of the resonator does not matter, only its volume, though in practice the shape will affect the stiffness of the structure...

Ah! Now we're talkng my language. The stiffness of the structure will not be a problem. Many thanks.

 

[bpape]

Thanks again John. This is an excellent visual representation of the real world issues faced when trying to use EQ to tame ringing and decay times.

Robbie, I'd agree with John. Don't be put off. While treatments IMO are the first solution to room issues, PEQ definitely has it's place in the overall scheme. There are some things - especially at very low frequencies - that treatments just aren't going to deal with. Also, if for instance you have 1 row of seating and it's all in a row and you have remaining issues based on length and height where it would be pretty much the same for all seats, where PEQ can definitely be useful.

As John stated, the trick is to know EXACTLY where the issues are. That's why you hear people harping all the time about not taking measurements with 1/3 octave resolution. It hides where the issues really are and can lead to what you saw in John's second set of graphs. If you measure in 1Hz increments from 20-200Hz, you'll have the information you need to accurately use PEQ to help.

 

[Ethan Winer]

Robbie,

> I quite fancy having a go at building a Helmholtz resonator anyway. <

Are you sure you need to even bother? I have only a few more traps than you in my living room HT and my response is far from perfect. But it sounds great and I have no interest in spending the effort to make things better on paper when the sound is already excellent. I think sometimes people obsess too much over things that don't matter as much as it may seem. I'm not trying to talk you out of improving your system! Rather, I'm just asking if you're doing this because it still doesn't sound as good as you'd like, or if you're just put off by the deviation from a flat line.

--Ethan

 

[Buckle-meister]

I have only a few more traps than you in my living room HT and my response is far from perfect. But it sounds great and I have no interest in spending the effort to make things better on paper when the sound is already excellent. I think sometimes people obsess too much over things that don't matter as much as it may seem. I'm not trying to talk you out of improving your system! Rather, I'm just asking if you're doing this because it still doesn't sound as good as you'd like, or if you're just put off by the deviation from a flat line.

I understand what you are saying and you've got me pegged pretty well; I am a perfectionist. However, I am also, at least to a degree, a realist, and am well aware that I will never achieve a flat response from 20Hz to 20kHz, nor would I wish to for the upper frequencies it seems, according to previous studies WmAx has posted.

There must be an end somewhere, but I do not feel I've reached it yet. Remember, I thought my sound was pretty good before room treatment. It improved no end when it was treated. Logically (and in fact proved by plots), there is yet room for improvement (though I am also aware of the law of diminishing returns).

I believe I can improve upon the sound that I currently enjoy. Whether by constructing a Helmotz resonator, purchasing additional room treatment or a PEQ, or a combination of them all, I intend to do just that.

 

[Ethan Winer]

Robbie,

> I believe I can improve upon the sound that I currently enjoy. Whether by constructing a Helmotz resonator, purchasing additional room treatment or a PEQ, or a combination of them all, I intend to do just that. <

Absolutely. Try all of those, and please report what happens.

--Ethan

 

[Buckle-meister]

Stage 1a complete!

Man, my system just gets better and better sounding all the time!

Before starting with the Helmholtz, I thought I'd integrate my sub for two-channel music. After a bit of puttering about, I finally settled for the results linked to below, and which I'll get to in a moment.

I've included some of the results from my recent review for comparison and have adjusted the time on the waterfall plot to make this easier. However, because I've had to move the traps around a bit to get the best results with the sub integrated (I now have one MondoTrap, 2'x4' MiniTrap and 2'x2' Minitrap in each of the room's four corners; the side and ceiling MicroTraps remain in the same position as before, because the couch, front left and right speakers are also in the same position as before), the two sets of plots can't be compared explicitly. That said, they may still be compared qualitatively, and that is what I will do.

Ok, this was my previous low-frequency time-slice. Apart from the room's first axial mode centred at approximately 40Hz, the rest of the graph is reasonably smooth with an average variation of say ±3dB. Great! Apart from the fact that overall, the response falls with frequency. I'd assumed that this was perhaps due to my front two towers increasingly struggling to put out deeper bass, but as can be seen from the plot with the sub integrated (crossed at 80Hz) with the towers, I guess this wasn't true because from approximately 65Hz and above (though the towers are only reproducing 80Hz and above in principle), the response is fairly level. I'll come back to the hump around 200Hz later.

Comparing the previous waterfall plot to that with the sub integrated shows just how flat my response is from approximately 65Hz to 180Hz. The massive blade of the room's first axial mode at approximately 40Hz is very apparent.

Again, the previous one-third octave plot for higher frequencies (which is a fair indicator of the response the ear actually hears) is nowhere near as good as it is now. By joining up the low-frequency time-slice plot with the one-third octave response plot for the integrated sub, you can see that the hump around 200Hz actually extends from about 180Hz to 250Hz with a peak level approximately 5dB above the average of say 82dB. Whilst this isn't ideal, I wouldn't say it's a disaster either.

The resulting sound is, I must say, quite spectacular. I had friends around last night and we were going through tunes which sounded so rich in bass (from a flat-ish response) yet at the same time so incredibly clean (from a quick decay due to absorption) at the same time.

By the way Ethan, in answer to your unasked question, yes, I'll still be going for a Helmholtz. That room mode must die!

Speaking of questions, I have one of my own:

Take a look at this low-frequency time-slice overlay of just the front towers reproducing a full-range signal (Direct Analogue mode) compared to that with the sub integrated (2ch mode). Note that these particular responses can be compared explicitly.

I know crossovers aren't brick-wall filters, but given that the crossover is set at 80Hz, why is there a difference in response above 80Hz? I thought that the crossover would only attenuate the front two towers' signal from 80Hz and below. Can anyone explain what is happening here?

 

[Ethan Winer]

Robbie,

First, I have to say you are doing an excellent job here, and I'm impressed by your thorough and systematic approach.

Whilst this isn't ideal, I wouldn't say it's a disaster either ... The resulting sound is, I must say, quite spectacular. <

The second statement certainly confirms the first one!

> I know crossovers aren't brick-wall filters, but given that the crossover is set at 80Hz, why is there a difference in response above 80Hz? <

You answered your own question. What is the slope of your crossover? At 12 dB per octave there's still a very real interaction (combining) half an octave above and below the crossover frequency.

> I thought that the crossover would only attenuate the front two towers' signal from 80Hz and below. <

Yeah, but it also rolls off the subwoofer above the 80 Hz point at the same slope.

--Ethan

 

[Buckle-meister]

You answered your own question. What is the slope of your crossover? At 12 dB per octave there's still a very real interaction (combining) half an octave above and below the crossover frequency.

> I thought that the crossover would only attenuate the front two towers' signal from 80Hz and below. <

Yeah, but it also rolls off the subwoofer above the 80 Hz point at the same slope.

Thanks Ethan. (refer to attachment) I've come a long way baby!

I believe I understand now. I was under the impression that as the receiver's crossover (the setting of 80Hz is made at the receiver end and the sub's own crossover dialled to maximum so as to take it out of the equation altogether) was decreasing the front towers' signal, it was increasing the sub's own signal, but I see now that this is incorrect. The receiver is actually crossing over two full-range signals; the front towers' signal will be attenuated from 80Hz down, and the sub's signal attenuated from 80Hz up, and this of course explains why the two plots must eventually converge; the sub's contribution lessens with increasing frequency above 80Hz until all that is left is the full-range signal of the towers.

 

[Rip Van Woofer]

Robbie;

When I finally get around to building my room, you and I are going to have some long talks!

Haven't looked at all your links but I'm bookmarking this thread. Great stuff. You have gone where I want to go eventually (cue Star Trek theme...)

And JohnPM, your free software looks like a dream come true for this Mac-using 'phile. I shall have to download and play with it ASAP.

 

[JohnPM]

Yeah, but it also rolls off the subwoofer above the 80 Hz point at the same slope.

A minor technical point, but the bass management low pass for the LFE output is generally 24dB/octave in contrast the the 12dB/octave high pass for the mains, the thinking being that the mains will have an acoustic roll-off of around 12dB/octave at the bottom end anyway so overall the result will then combine nicely.

 

[JohnPM]

Rip, there are some "challenges" in using the software with the Mac as Apple have their own proprietary classes for handling audio and provide quite limited support for the core Java audio classes. Nonetheless, there are Mac users out there who have run the app successfully and reported good results. You will need to set the sample rate to 44.1k and use the Mac's own level tweaking tools, but there's info in the help files about that.

 

[Buckle-meister]

When I finally get around to building my room, you and I are going to have some long talks!

I look forward to it.

A couple more questions for anybody:

1. If I understand things rightly, once I get the helmholtz in one piece and start trying various lengths of port to find the correct tuning frequency to nail that room mode, I'll know I've hit it because the amplitude of the 'blade' should to a lesser extent and the ringing time to a greater extent reduce. Only at that point should I start adding glass-fibre or an equivalent absorbant material to reduce the blade-like appearance to a (hopefully) much more rounded lump; in other words reduce the blade's Q. Is this correct?
2. Why, when a sub has so much greater positional freedom in a room than the front left and right speakers, do people pursue full-range speakers? I should note that I'm not ruling out the possibility of buying a pair of full-range speakers in the future myself, but I just can't imagine being able to fine-tune any room's response to anything like the same extent as can be achieved with non full-range speakers (or even full-range speakers) and a properly integrated sub.