Stage 2

[Ethan Winer]

Chris,

> > at frequencies <70Hz in your own graphs, located at your link, resonances are reduced at all of the measured positions with equalization <

It may look like the ringing was reduced, but only because the overall level was reduced pushing the ringing off the bottom of the graph. Somewhere around Page 15 of that thread on AVS I showed a new composite graph that normalized the levels to better show the real ringing time with EQ. I then added that to the main article on our web site as a new Example 5. If you haven't seen the article since I first posted it, go back again and find that new 5th example in my "Comments and Analysis" section. Look for "Added January 4, 2006" in that section.

> A 200Hz, or even 100Hz, a resonance would not behave sufficiently as minimum phase in any normal size room. <

Agreed completely. And I have stated many times that I'm not opposed to anyone using EQ. But for whatever value EQ may have, so far I have not seen evidence that EQ can reduce ringing in practice. I'm pretty sure you'll agree once you see the new normalized graph I mentioned.

--Ethan

 

[Ethan Winer]

John,

> I'll be happy to put that together <

That would be fabulous!

> The dimensions of the room I'll use are different to the one you used (longer but narrower), but I can make the reference point the same proportions as you used <

Sure, that's fine.

I also have a question about your program. I went to download it, then saw a note about needing a Java runtime engine. I'm very particular about what I put on my computer, and have so far avoided installing new versions of .Net and Media player (with the copy protection nonsense) and DirectX 9 and so forth. Is this runtime engine likely to be present already on an XP Pro computer? Or, if not, is it a huge package that bloats the registry, or is it just an extra interpreter that may take up some disk space but otherwise stays out of the way and does not hook into the system?

Thanks.

--Ethan

 

[Ethan Winer]

sploo,

> Can I stop a potential fight here before one starts. <

Thanks, appreciated. As I'm sure you know, I never attack anyone and I can't understand when people attack me. I prefer to stick to the facts at hand, and keep all discussions focused on the science.

--Ethan

 

[WmAx]

It may look like the ringing was reduced, but only because the overall level was reduced pushing the ringing off the bottom of the graph. Somewhere around Page 15 of that thread on AVS I showed a new composite graph that normalized the levels to better show the real ringing time with EQ. I then added that to the main article on our web site as a new Example 5. If you haven't seen the article since I first posted it, go back again and find that new 5th example in my "Comments and Analysis" section. Look for "Added January 4, 2006" in that section.

I looked at the graphs in that thread. It is now apparent that your study is invalid because the person who set up the parameteric equalizer did not do so properly. The sharp bands seen in bottom residuals of the new graph, make it apparent that this is a demonstration in non-optimal set filters. This is exactly what happens when you do not use a careful set-up methdology. It appears the only frequency which was somewhat closely matched properly was 70Hz. But regardless, the resonances were still reduced appreciably in magnitude by simple mechanism of reducing their power(SPL) relative to the main signal. Remember, when you reduce the level relative to the main signal, it is being masked. Your use of different SPL starting scales for the empty room and EQed room does not lend validity to the graphs. This could be considered misleading. You must normalize to a common non-equalized frequency reference point so that the relative view is the same for comparison for the graph types shown. I did not see this being done. You apparently normalized by some arbitrary reference. It certainly makes one example look worse by artifical mechanism. The proper methdolody of setting up for this experiment should be as I described to Buckle-Meister earlier in this thread. It takes considerable time to fine tune and re-measure with every parameter change with this technique, but it is required for proper performance.

I'm pretty sure you'll agree once you see the new normalized graph I mentioned.

You have convinced me that the person who set up the equalizer did not do so optimally. But now the graphs provided at the link on your website appear to be invalidated from the perspective of correct equalization, unless you change the objective of the study to: comparing bass traps to improper equalization set-up.

-Chris

 

[Buckle-meister]

Stage 2 indefinitely on hold

I was going to post a three-line quote summarising the conclusions of the very informative thread Ethan mentioned (with acknowledgement given of course) but realised that there was no point as I kept on finding other quotes I also wanted to add.

The thread was extremely long to digest in one go, and my thoughts are a little conflicting as I found myself agreeing with many comments from people with contrary thoughts. One thing I now know for certain; at this point I do not consider myself competent enough to properly setup a PEQ. More reading, questioning, wheedling with hopefully the end result of learning is required on my part.

Many thanks to all who contributed.

 

[JohnPM]

I also have a question about your program. I went to download it, then saw a note about needing a Java runtime engine. I'm very particular about what I put on my computer, and have so far avoided installing new versions of .Net and Media player (with the copy protection nonsense) and DirectX 9 and so forth. Is this runtime engine likely to be present already on an XP Pro computer? Or, if not, is it a huge package that bloats the registry, or is it just an extra interpreter that may take up some disk space but otherwise stays out of the way and does not hook into the system?

The Java Runtime Environment is the software package that provides the Java bytecode interpreter to allow a Java application or a Java applet (in a web browser) to run, it essentially stays out of the way until you run some Java code or a Java applet. Your XP Pro PC will have a Java interpreter, maybe Microsoft's or Sun's, but my app requires the latest JRE from Sun. You can find out more about Java and the runtime environment at http://www.java.com. It is a large download (typically about 7MByte if using an online installation, according to Sun). If you wait until after the weekend there will be a new update to my app, to save installing twice in quick succession should you decide to go for the JRE.

 

[WmAx]

One thing I now know for certain; at this point I do not consider myself competent enough to properly setup a PEQ. More reading, questioning, wheedling with hopefully the end result of learning is required on my part.

Many thanks to all who contributed.

Just remember what JohnPM noted: a Hemholtz resonantor will be homologous to a parameteric band. If you do not precisely align the frequency and Q, the same effect will occur as with the parametric filter. It is certainly easier/faster to adjust the Q and frequency instantly with a parametric equalizer as compared to physically changing the resonator properties.

-Chris

 

[bpape]

I'll have to agree with Ethan on this one. Even IF you could provide a perfect inverse EQ for a modal issue (which you can't - at least not without an expensive computerized solution or MANY hours of trial and error) it would only be valid at one seat. That same EQ would NOT be the exact inverse of the room response for another seat. Without an exact inverse, the response will be come worse. The ringing that the EQ is supposed to be able to help will also get worse.

In THEORY, EQ should be able to do pretty much whatever you want it to. In practice it's just not feasible. Even if it could do it for peaks, how are you going to use EQ to deal with nulls? Treatments and placement can help with nulls. EQ cannot - pure and simple. In reality, even with careful placement, the biggest issues in a room usually end up being nulls - not peaks.

As for the 'challenge' that took place, there were a lot of things that went wrong. Time was an issue. Ethan was sure that 17 traps would do teh job no matter where they were. I'd love to see the same thing done with 8 traps, 2 in each vertical corner. I'd also love to see what a multi kilo-buck EQ could do - as well as someone given the time to go through the analysis and trial and error.

I will agree that parametric EQ absolutely has a place in an overall acoustical scheme. However, it should be to 'tweak' the last little bit - not to provide the lion's share of the solution. For the average or even above average guy, getting it 'right' at 6-8 seats with EQ as the primary solution is just not going to happen.

 

[JohnPM]

Even IF you could provide a perfect inverse EQ for a modal issue (which you can't - at least not without an expensive computerized solution or MANY hours of trial and error) it would only be valid at one seat. That same EQ would NOT be the exact inverse of the room response for another seat. Without an exact inverse, the response will be come worse. The ringing that the EQ is supposed to be able to help will also get worse.

Whilst PCs are not cheap, many people have them. The other statements are far too definite for me as I haven't seen evidence to support them. There is also a mixture of comment on correction for a mode and correction of an entire room response, which are quite different things. The spatial range over which the EQ of a mode is useful should intuitively depend on the mode's frequency, the lower the frequency the greater the area over which the correction should be worthwhile. I struggle to think of a theoretical justification for the statement that ringing will get worse away from the point at which the EQ was optimised. All a bit academic of course without a set of measurements, so more incentive for me to get on and make some

In THEORY, EQ should be able to do pretty much whatever you want it to. In practice it's just not feasible. Even if it could do it for peaks, how are you going to use EQ to deal with nulls? Treatments and placement can help with nulls. EQ cannot - pure and simple. In reality, even with careful placement, the biggest issues in a room usually end up being nulls - not peaks.

I'm not sure that the statement re nulls is supported by studies of audibility of response defects, most I've seen would suggest that peaks and long delay times are the most objectionable low frequency characteristics.

I will agree that parametric EQ absolutely has a place in an overall acoustical scheme. However, it should be to 'tweak' the last little bit - not to provide the lion's share of the solution. For the average or even above average guy, getting it 'right' at 6-8 seats with EQ as the primary solution is just not going to happen.

I haven't seen even a hint in this thread of anyone suggesting EQ should provide the majority of the solution to a room's acoustic problems, quite the opposite! Mind you, I'm not sure that the "average guy" needs correction over a range of 6-8 seats, you must have a much bigger room than most (or maybe just a lot more friends )

 

[Ethan Winer]

Chris,

> the resonances were still reduced appreciably in magnitude by simple mechanism of reducing their power(SPL) relative to the main signal. <

Agreed, and I agree with most of your other points too. I still remain to be convinced that EQ can reduce ringing in practice, over an area larger than one cubic inch. I'm hoping John can help settle this because I really would like to know. Since you're such a staunch EQ supporter and user, do you happen to have any waterfall charts showing ringing being reduced by an EQ you set up?

> Your use of different SPL starting scales for the empty room and EQed room does not lend validity to the graphs. <

Sure it does, at least for the purpose of proving that the ringing time was not reduced in that test. Even if the peak level was reduced, the ringing time itself was not. To me this is the defining advantage of bass traps over EQ. Bass traps reduce peaks (and raise nulls), they reduce ringing, and all locations in the room are improved.

By the way, I'd like to point out that I never felt you were being combative here, and my comment above about people who attack was not aimed at you. There's plenty of room for disagreement about audio theory, and disagreeing is not the same as arguing.

--Ethan

 

[Ethan Winer]

John,

> You can find out more about Java and the runtime environment at http://www.java.com. It is a large download <

Thanks. I may put it on my laptop in my HT room. It looks like the download file is about 15 MB but it takes 100 MB of disk space. Yikes! What the heck is in there? (That's a rhetorical question...)

> the lower the frequency the greater the area over which the correction should be worthwhile. I struggle to think of a theoretical justification for the statement that ringing will get worse away from the point at which the EQ was optimised. <

I have to disagree with both of those statements. You might think that low frequencies with their longer wavelengths would have little response change over small distances. But look at the two ETF plots of the room empty, at front center and six inches to the right of front center. There's a big difference for the entire frequency range below 100 Hz. In fact, when I moved the microphone and measured at the right of center, I thought ETF went bonkers because there was so much change. So I moved the mike back to center and the response went back. And when I moved again to right of the center that response returned. So I'm sure the change over a six inch span really is valid. Since there's a really nasty peak at 70 Hz in the center, but not six inches to the right, how could you possibly EQ that? Whatever you do to help what the left ear hears will make the response at the right ear worse.

As for ringing being worse away from the corrected area, I don't see how it could not get worse. Any EQ setting that counters ringing will by definition add the same ringing but out of phase. As soon as the acoustic ringing you're trying to cancel is no longer the same level, the "counter ringing" will then dominate.

--Ethan

 

[Ethan Winer]

John,

Sorry, one more point.

> I'm not sure that the statement re nulls is supported by studies of audibility of response defects, most I've seen would suggest that peaks and long delay times are the most objectionable low frequency characteristics. <

I addressed that in the AVS thread. The notion that bass nulls are not so damaging ignores what happens when the null aligns with the key of the music. If you have a 25 dB null at 110 Hz and a song is in the key of A, I guarantee you it will be very audible.

Bass nulls present one of the biggest problems people have in their untreated home recording studios. The most common complaint I see posted is "My mixes sound great in my room but are way too bassy in the car and everywhere else." This is because they're sitting in a 1/4 or 3/4 wavelength comb filter null caused by reflections off the wall behind them.

--Ethan

 

[ScottMayo]

Just remember what JohnPM noted: a Hemholtz resonantor will be homologous to a parameteric band. If you do not precisely align the frequency and Q, the same effect will occur as with the parametric filter. It is certainly easier/faster to adjust the Q and frequency instantly with a parametric equalizer as compared to physically changing the resonator properties.

-Chris

EQ is easier. But as has been pointed out, it spits out a solution that's unique to a point in a room.

A resonator is a pain to build, and you end up scooping sand (or some equivalent) in and out to adjust it. But it benefits the whole room.

Use both. Both is good. But realistically, since either solution takes hours to get right anyway, choose what you can afford, based on whether you like fiddling with wood tools or fiddling with electronics better, and whether a wider listening area is worthwhile. Neither is universally "right".

(Well, EQ becomes right if WAF factors prevent the use of large boxes sitting at modal points, but that's a different issue).

 

[JohnPM]

As for ringing being worse away from the corrected area, I don't see how it could not get worse. Any EQ setting that counters ringing will by definition add the same ringing but out of phase. As soon as the acoustic ringing you're trying to cancel is no longer the same level, the "counter ringing" will then dominate.

No, that's a misconception. The EQ does not work by generating a signal of the same amplitude but opposite phase as the original peak, phase doesn't enter into it. The filter starts out with as much attenuation at the mode's frequency as the room has gain. If you look at a waterfall plot of the filter alone, as time advances the dip corresponding to the filter's attenuation gets shallower relative to the frequencies either side of it but it is quite some time into the decay before the filter's contribution rises above the decaying level of the regions either side, by that time the level is already well down. There is then a period where the tail of the filter's response is above the floor of the unfiltered regions, but it is nothing like the ringing associated with a mode. A picture would probably be worth a thousand words here but I'm not able to generate and host one from where I'm typing this, but hopefully the above makes sense. Just remember that we are NOT generating some sort of inverse of the music signal passing through, we are generating an attenuation at a modal frequency that is opposite to the room's modal gain at that frequency.

 

[JohnPM]

The notion that bass nulls are not so damaging ignores what happens when the null aligns with the key of the music. If you have a 25 dB null at 110 Hz and a song is in the key of A, I guarantee you it will be very audible.

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!

 

[WmAx]

No, that's a misconception. The EQ does not work by generating a signal of the same amplitude but opposite phase as the original peak, phase doesn't enter into it.

Amplitude and phase, relative to the reference, are inter-related. I'm not sure what you mean by the above.

Here is an amplitude vs. phase measurement for a standard non-linear phase(as opposed to a digital FIR linear phase) equalizer, applying a negative and positive curve of the same gain, bandwidth and frequency:

http://www.linaeum.com/images/eq_inversedcompare.gif

The amplitude and phase is precisely inversed.

-Chris

 

[JohnPM]

Chris, you're talking about the filter's phase response. Ethan thought Any EQ setting that counters ringing will by definition add the same ringing but out of phase. which is the misconception I addressed in my post, phase has nothing to do with the action of the EQ filter in addressing the room's modal gain.

 

[WmAx]

do you happen to have any waterfall charts showing ringing being reduced by an EQ you set up?

I have not kept such measurements. However, 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.

-Chris

 

[WmAx]

Chris, you're talking about the filter's phase response. Ethan thought Any EQ setting that counters ringing will by definition add the same ringing but out of phase. which is the misconception I addressed in my post, phase has nothing to do with the action of the EQ filter in addressing the room's modal gain.

Thank you for the clarification. I think it should be noted for readers here that he non-linear phase model(as seen in the graph) must be used to equalize modal points, however, to correct them properly. A linear phase equalizer would have undesirable results, since it's phase would not be inter-related to the minimum phase curve it is being used to correct.

-Chris

 

[Ethan Winer]

John,

> No, that's a misconception. <

Thanks for clarifying. I knew that was likely to be wrong technically, and I almost added "or something like that" - which I should have! But I'm pretty sure the mechanism by which EQ can theoretically reduce ringing is still highly position dependent, and will still add artifacts away from the place the EQ was optimized for. I don't see how it couldn't make things worse even a few inches away.

My "expert friend" Bill Eppler writes exactly this sort of DSP code, and he explained to me that the more you try to correct, the more localized the correction becomes. Since you offered to actually test this, I'll wait for your results before putting my foot farther into my own mouth.

> A picture would probably be worth a thousand words here but I'm not able to generate and host one from where I'm typing <

If you'd like to email me an image I'll be glad to host it and send you the link.

--Ethan