How good are 1/3 octave EQ's in practice?

[Amphiprion]

I'm contemplating integrating a subwoofer EQ into one of my projects; for those who have used 1/3 octave EQ's for subwoofers, how did they perform? How much more advantageous would a 1/6th octave EQ perform? The design would be a constant-Q equalizer.

[Kal Rubinson]

A better choice would be a parametric EQ of as many bands you can get. There's no predicting what the frequency or Q needed will be so any GEQ with fixed f and Q is a gamble.

[Amphiprion]

For the hypothetical of this discussion, let's just assume it has to be a 1/3 or 1/6 octave type.

How useful are they?

[---k---]

Kal and others,

Mark is a real geek that is likely building the amp and eq from almost scratch for this project. Therefore, he is likely a little limited in what he can do.

[Kal Rubinson]

For the hypothetical of this discussion, let's just assume it has to be a 1/3 or 1/6 octave type.

How useful are they?

IMHO, they are sophisticated tone controls. How many bands can you get into the sub-80Hz range?

Kal

[Johnloudb]

For the hypothetical of this discussion, let's just assume it has to be a 1/3 or 1/6 octave type.

How useful are they?

I'd think a 1/3 or 1/6 octave would be better than most people have available to them. I've usually used no equalization with subs and got excellent results. I just worked with subwoofer placement when possible.

While equalization is usually helpful, trying to defeat bad room modes just using EQ is not always a good option. In one direction of our listening room requires about 20dB of boost at 25Hz. So we listen with the speakers on the long wall. The bass is pretty flat there.

Anyway, for me 1/6 octave would be more than adequate.

[Johnloudb]

IMHO, they are sophisticated tone controls. How many bands can you get into the sub-80Hz range?

Kal

Maybe I'm missing something, but I think a 1/6 octave equalizer would be 12 bands from 20Hz to 80Hz. That's a lot IMO. I haven't used EQ enough to know how much is really enough though.

[Kal Rubinson]

Maybe I'm missing something, but I think a 1/6 octave equalizer would be 12 bands from 20Hz to 80Hz. That's a lot IMO.

You calculate what I calculate. The problem is the spacing which is fixed (as is the Q). Now, although one can make real improvements with such a device, it cannot make the specific corrections required. If you run REW or similar and look at the filters necessary for correction, even if keeping only to the cutting of peaks, it is unlikely that the fixed parameters of the GEQ will be on point. I did just this in reviewing a parametric sub EQ and found that I needed filters high Q filters (Q=2.8 to 8.4) at frequencies like 23, 29 and 41Hz.

Of course, room acoustics and careful placement will minimize this and should always be attended to first.

Kal

[Johnloudb]

Kal, I see your point. How about a six band equalizer with variable frequency and Q for each band? That's more adjustment than I would want to mess with myself.

Mark, I don't know what you're working on but you can build a simple bandpass circuit with variable Q and frequency using a single op-amp. So using say 6 op-amps for the filters and some buffers and a summing circuit, and some resistors and caps, you could build this fairly easy.

But, maybe this doesn't fit in with your project?

[Kal Rubinson]

Kal, I see your point. How about a six band equalizer with variable frequency and Q for each band? That's more adjustment than I would want to mess with myself.

Should be more than adequate. The Rives sub-PARC, for example, only uses 3 bands.

Kal

[chasw98]

I have built hi pass filters for low frequencies (sub 10 Hz) and used both digital and analog 1/3 and 1/6 octave graphic (constant Q) and parametric equalizers. IMHO a lot more can be done with 3 bands of parametric than with 6 or 12 bands of graphic. You might want to go to the Rane site and look at the schematic for a PE17. http://www.rane.com/pe17.html It would be farly simple to build 3 bands with variable gain and Q, but use plugin resistor or capacitors for the frequency to keep the clutter down.

Chuck

[Amphiprion]

The idea I was contemplating was a microcontroller interfaced to a computer with either RS-232 (easy) or USB (hard) which would allow control of the 1/X octave filter through a visual basic program. The micro would control low distortion (0.00x %) digital pots (one per band). It would also allow me to try and make (this is the real hard part) an automated room correction algorithm into the microcontroller alone, which is much easier to do with fixed centers and Q's (compare with LspCAD trying to do a crossover optimization - it screws it up a LOT of the time vs. starting off by hand).

FWIW I got my degree in electrical engineering and actually have work experience in most of the above. There are some issues in going to 1/6 octave filters (R values get big) and capacitors are big enough at low frequencies that I am tempted to use PET (mylar) instead of polypropylene.

This is what happens when I give up plasma work

[Kal Rubinson]

Automated RC is slick but you still want to measure performance and set up targets. I'd sooner have an interactive, hands-on facility with the ability to A/B filter sets.

Kal

[Dennis H]

I did just this in reviewing a parametric sub EQ and found that I needed filters high Q filters (Q=2.8 to 8.4) at frequencies like 23, 29 and 41Hz.

Q=8.4 is very close to 1/6 octave (actually 1/5.8 ) so, if that's all you need, I think a 1/6 octave GEQ could be pretty useful.

I especially like the ease of use of a constant-Q EQ where you just put the sliders at the inverse of the measured curve and what you see is what you get. With the older GEQs, you have to compensate for the way the filters add. It should be a snap for an automatic algorithm to set and have resolution at least as good as Audyssey at the lowest frequencies. And of course, if Mark's building it, it will have full manual control.

10.0
11.2
12.6
14.1
15.9
17.8
20.0
22.4
25.2
28.3
31.7
35.6
40.0

[Amphiprion]

Yes, if I build this with the microcontroller it will definitely have manual control for everything. That's actually the really easy to do part, just VB programming and using the EUSART on a PIC microcontroller for a serial port hookup.

Although I thought Q = 1 / (width) + 1? So 1/6th has a Q of 7 and so forth? I've been looking at the formulas at Rod Elliot's site and that's what he has.

[Dennis H]

I just plugged it into a little spreadsheet I downloaded from the Rane site. They use constant-Q filters for their EQs so it should be close. I haven't looked at Rod's calcs but he's probably using old-style filters.

[Dennis H]

Okay, I think I see what Rod is doing. As described in a Rane whitepaper, using constant-Q filters but setting the bandwidth a bit wider gives you what they call an 'interpolating constant-Q' filter which still lets you do the old trick of using two filters to hit an Fc in between.

And I was wrong about the WYSIWYG thing although they're close at low gains/cuts. Two adjacent 12dB filters will give you about 15dB with an Fc halfway in between. Lowering one and raising the other will shift Fc one way or the other. Sooooo... a bit more math for your automatic thingy.

http://www.rane.com/note101.html

[Kal Rubinson]

Q=8.4 is very close to 1/6 octave (actually 1/5.8 ) so, if that's all you need, I think a 1/6 octave GEQ could be pretty useful.

Sure but I only needed that Q on one filter; the others were quite different. These would not be possible with a fixed Q EQ.

I especially like the ease of use of a constant-Q EQ where you just put the sliders at the inverse of the measured curve and what you see is what you get.

Ah but that ignores the effect of the Q on the decay.

With the older GEQs, you have to compensate for the way the filters add. It should be a snap for an automatic algorithm to set and have resolution at least as good as Audyssey at the lowest frequencies.

Sure if you want to ignore time-domain issues.

[Saurav]

I thought one of the criticisms of using EQ for room bass correction (as opposed to bass traps etc.) was that they could not fix time domain issues (a.k.a. 'ringing'). How does changing the Q of the filter affect this? In layman's terms if possible

[Dennis H]

I kept digging on the Rane site and I finally found what I was thinking of. Here's the one you want, Mark. 'Perfect-Q' for true WYSIWYG. U.S. patent 7,266,205 oughta give the specifics.

http://www.rane.com/note154.html

http://www.rane.com/deq60.html

[Dennis H]

Sure but I only needed that Q on one filter; the others were quite different. These would not be possible with a fixed Q EQ.

Ah but that ignores the effect of the Q on the decay.

Sure if you want to ignore time-domain issues.

If that's the highest Q you need, you can build a lower Q filter with several sliders. The only thing you can't do is build a higher Q filter.

Decay, time domain, frequency domain, they're all the same thing at low frequencies. Rooms are minimum-phase systems at low frequencies meaning you can calculate the time domain from the frequency domain and vice versa. Fix the frequency response and you fix the phase response at the same time assuming you're using a minimum-phase EQ which most of them are.

Using a linear-phase EQ (cough Audyssey ) is actually wrong for room EQ at low frequencies because it doesn't fix the phase. The DEQX uses linear-phase FIR filters for the crossovers because you don't want phase shifts there but it uses minimum-phase IIR filters for room EQ because you do want to shift (fix) the phase there.

Edit: I shouldn't be so critical of Audyssey. The problem is they use vague technobabble to describe how it works. They imply they are using linear-phase filters but I don't think that can be strictly true. The latency is way less than something like a TacT that for sure uses linear-phase filters.

[Dennis H]

I thought one of the criticisms of using EQ for room bass correction (as opposed to bass traps etc.) was that they could not fix time domain issues (a.k.a. 'ringing'). How does changing the Q of the filter affect this? In layman's terms if possible

Regular old EQs like all the analog models and most of the digital models fix 'time-domain' issues just fine. See the post to Kal above. The problem is they can only fix the response at one spot in your room so bass traps are a better solution for making your whole room sound better.

Either that or build speakers that don't interact with the room so much, e.g. dipole mains positioned according to SL's recommendations and DBA subs.

[Saurav]

The problem is they can only fix the response at one spot in your room so bass traps are a better solution for making your whole room sound better.

I understand that, ideally, put the sub(s) in the right spot(s), then use room treatments, then use EQ. But not everyone has that luxury, so in my case, I put my sub where it sounds OK and looks OK, use EQ on the peaks, and leave the rest.

I just remember reading a statement about "EQ can't fix ringing" as a criticism, in addition to the "only works for one spot".

I do have dipole mains now, though the room positioning is less than ideal, and am constantly amazed at how much more articulate and 'textured' the bass is. I'm hearing more notes, and hearing more differences between recordings.

What's a DBA sub?

[Dennis H]

I just remember reading a statement about "EQ can't fix ringing" as a criticism,

Yeah, probably by Ethan Winer. He just doesn't 'get' the concept of mimimum phase and doesn't understand the math behind and the limitations of those pretty waterfall plots he likes to post. Short version, waterfalls lie.

[Dennis H]

Double bass array. Put an array of subs on the front wall to minimize horizontal and vertical modes. Cancel longitudinal modes with an active bass trap -- an identical array on the back wall, out of phase and delayed by the time for sound to travel the length of the room.

http://www.avsforum.com/avs-vb/showthread.php?t=837744

Pic with no EQ!

[Saurav]

So this is the same concept I see discussed on Zaph's website then - why a long ridge in a waterfall graph that's outside the driver's passband isn't a problem because the crossover will take care of it?

Thanks for the link. That's an interesting read.

[Amphiprion]

Hi Dennis,

Thank you for looking all that up for me! I was going to start taking the laplace transform of Ron's circuit and do the calculations myself. You saved me a lot of pen and paper engineering

I'll have a look at the links, but it sounds like you've found the reason for the discrepancy. Thanks!

[Amphiprion]

Attached is something I did when I worked for my first employer out of college. I whipped up the following in 5 days with about 70 hours into it from start to 'OMG it actually works' when their old box (built by a long-gone previous employee with no PCB files, documentation, micro code, etc to be found). It's quite simple but this is the type of stuff I used to do with microcontrollers and Visual Basic coding.

http://www.hayenga.com/mark/NON-WEBPAGE%20DATA/work/NaBH4%20Based%20Hydrogen%20Generator%20Control%20Unit.pdf

Apologies for the auto routing I just wanted folks to know I wasn't being pie-in-the-sky about the computer controls for a DIY project like this.

[Kal Rubinson]

I thought one of the criticisms of using EQ for room bass correction (as opposed to bass traps etc.) was that they could not fix time domain issues (a.k.a. 'ringing'). How does changing the Q of the filter affect this? In layman's terms if possible

All filters have time decay, too, even though most of us just think of them in the instantaneous frequency/magnitude domain. So if you have a response peak which decays slowly and you insert a filter at the same frequency with the same/similar decay, you will not only reduce the immediate amplitude but greatly accelerate the decay/ringing. It is not trivial but there are tools that can help, such as REW and the XTZ Room Analyzer.

BTW, a nice, and very conservative, analysis of such an approach is in the Meridian AES paper that led to their EQ system.

Kal

[kgveteran]

My old behringer 1124 could do 24 bands in mono.Thats tough to beat.

Placement of the subs rendered it moot.

[Dennis H]

All filters have time decay, too, even though most of us just think of them in the instantaneous frequency/magnitude domain. So if you have a response peak which decays slowly and you insert a filter at the same frequency with the same/similar decay, you will not only reduce the immediate amplitude but greatly accelerate the decay/ringing. It is not trivial but there are tools that can help, such as REW and the XTZ Room Analyzer.

Kal

If you read what John (REW author) has posted, he shows how a parametric filter that exactly matches the inverse of a room resonance, exactly cancels the 'ringing.' It's just (not so) simple math.

[Kal Rubinson]

If you read what John (REW author) has posted, he shows how a parametric filter that exactly matches the inverse of a room resonance, exactly cancels the 'ringing.' It's just (not so) simple math.

Exactly. :T

The only thing I would add is that one should use multiple measurements, rather than at one position, to determine the response characteristics that need correction.

Kal