Skip to main content

Topic: 31 band Graphic EQ: Q factor values (Read 5770 times) previous topic - next topic

0 Members and 1 Guest are viewing this topic.
  • Juha
  • [*][*][*][*][*]
31 band Graphic EQ: Q factor values
I'm preparing a 31 band Graphic EQ for Windows (Vista/7/8/?). It's gonna be based on biquad (peak) filters.

As being new (noob) with EQ design, I was wondering if the Q factor for each band "needs" to be equally set (lets say 4.318 for all 31 bands) and constant ?

There are lots of reading but, looks like I'm a bit lost there.

http://en.wikipedia.org/wiki/Equalization_(audio)
http://en.wikipedia.org/wiki/Filter_design#Digital_filters
http://www.rane.com/note101.html
http://www.sengpielaudio.com/calculator-bandwidth.htm
http://www.dummies.com/how-to/content/how-...or-an-audi.html
http://www.dafx.ca/proceedings/papers/p_037.pdf


Juha
  • Last Edit: 28 September, 2013, 03:07:13 PM by Juha

  • saratoga
  • [*][*][*][*][*]
31 band Graphic EQ: Q factor values
Reply #1
Generally the Q factor is a parameter that the user sets.  Its not usually constant per band, although it can be if the user wants each band to have the same fractional width.

31 band Graphic EQ: Q factor values
Reply #2
I'm preparing a 31 band Graphic EQ for Windows (Vista/7/8/?). It's gonna be based on biquad (peak) filters.

As being new (noob) with EQ design, I was wondering if the Q factor for each band "needs" to be equally set (lets say 4.318 for all 31 bands) and constant ?

There are lots of reading but, looks like I'm a bit lost there.

http://en.wikipedia.org/wiki/Equalization_(audio)
http://en.wikipedia.org/wiki/Filter_design#Digital_filters
http://www.rane.com/note101.html
http://www.sengpielaudio.com/calculator-bandwidth.htm
http://www.dummies.com/how-to/content/how-...or-an-audi.html
http://www.dafx.ca/proceedings/papers/p_037.pdf


Juha


Traditional Graphic equalizers with 30 +/- bands are always fixed constant Q.  Working with real world coils, capacitors, and resistors fixed parameters was the only practical way. This was almost always carried over into digital implementations of the same.

The basic idea was to approximate critical bands, which were not so well understood when the traditions started. The traditions carried over even when we understood critical bands better.

Graphic equalizers with smaller numbers of bands were mostly based on constant Q and constant bandwidth as well but there were some exceptions.

You can check this out with real world pro audio products. Leaders in this field include Behringer, Rane, Ashley, DBX, etc,

http://www.sweetwater.com/c468--Graphic_EQ

  • Artcore87
  • [*]
Re: 31 band Graphic EQ: Q factor values
Reply #3
Traditional Graphic equalizers with 30 +/- bands are always fixed constant Q.  Working with real world coils, capacitors, and resistors fixed parameters was the only practical way. This was almost always carried over into digital implementations of the same.

I know this is an old thread, but I was searching for a constant Q dsp and found this thread.  The above information is incorrect, and is a misunderstanding of what a "constant-q" equalizer is.  I myself was looking for just such an equalizer dsp, although I have now a superior and ultimate solution, a 5 band parametric eq (thus, the q is user-adjustable for all 5 bands).  While a traditional EQ's "Q" is not adjustable, it is only effectively constant when at full boost or full cut... key word effective.  A typical 1/3 octave EQ only has a 1/3 octave Q in practice when the adjustment is at full boost or full cut, as the boost or cut is lowered the actual Q becomes wider, this is simply a consequence of traditional designs, whether passive or active, or digital (most of which simply digitally mimic analog filters).  The potential benefit of a constant Q EQ is that their Q factor actually stays the same regardless of the level of boost or cut.  If the purpose for the EQ is general shaping of the sound for personal preference reasons, a constant Q EQ is not needed.  But if significant corrections have to be made either for speaker or room response, a constant Q EQ  (or a parametric where the Q is adjustable) is absolutely essential.  If you need to correct for, lets say a 3 or 6 decibel response peak or trough, then a traditional 1/3 octave (31 band usually) EQ will not help you much, because with that low of an adjustment the actual Q at that point will be much wider than 1/3 octave.... you're no longer merely correcting a peak/trough in a very narrow frequency range, you're also affecting frequencies perhaps a whole octave out that may not need any correcting at all.  Personally I have some great speakers that have some inherent flaws that require significant equalization, after which they sound fantastic.  I absolutely needed either constant Q or parametric equalization, as a traditional 31 band would never allow me to make the necessary corrections in some very limited frequency bands that need to be made.  If a traditional EQ is all you had, it is possible (though impractical) to effectively narrow the Q of a particular frequency band adjustment.  For instance, if you want to cut 1khz by 3db, you would have to adjust the adjacent bands to compensate for the wider effective Q of a slight or modest adjustment.  This would be done by setting your 1khz band to perhaps -6db, and the directly adjacent bands to +3db, and the bands adjacent to that by -1.5db, and the ones adjacent to that by +.75db... this is just an approximation to get the general idea across and not necessarily the actual values you may have to use, although I suspect it's probably close to realistic.  The boost right next to the heavier-than-desired final cut effectively narrows the Q for that cut... but then that compensation needs to be compensated for itself on the other end, since on the other side of those bands you're actually producing a boost that you don't want (rather than simply reducing the cut closer to 1khz).  That's why the boost/cut values keep going down as you move away from the center frequency in theory.  The complexity of doing this, especially if you actually do want to make other minor adjustments elsewhere in the frequency range (which would be exponentially more complex, unless your DSP actually showed you the resultant curve of your settings, then you could do it visually) highlights the benefit of a constant Q equalizer.  The 5 band parametric I have indeed has a visual representation of the actual resultant curve after all adjustments, which is extremely useful.  If all you're looking at is the 31 bands of the equalizer itself, and not the actual resultant curve (where boosts and cuts both add and subtract from one another), then you may be able to tune it by ear, but you really have no idea what your actual curve looks like, because I promise it looks nothing like the setting of the 31 bands if you merely connected them with a line, even assuming you gave that line some natural curve and didn't use straight lines.

Constant Q EQ's are great tools, far more useful than regular EQ's for making corrections (as opposed to mere wide-bandwidth tonal preferences... bass boost, mid cut, treble boost, stuff like that).  Parametric EQ's (as many bands as are needed) are the ultimate.  Nothing else matters but either the quality of the circuits and components, or for DSPs, the quality of the processing... bit depth/resolution/sample rate... the digital audio variables) in an equalizer.  Linear-phase EQ is mere marketing hype and offers no benefit, though i've played around with them.  Reading actual data however shows them to be worthless.  "minimum phase shift" is also a meaningless marketing term, since it applies to literally all equalizers by design.  complementary phase shift is another marketing term out there that's virtually meaningless.  For DSP, you want a high FFT... 65553 samples is what i use.  This determines the accuracy/resolution of the applied curve... how closely the digitally implemented curve matches up with the chosen settings.  Lower the FFT and you'll see the curve eventually become straight lines... it's literally just the resolution of the curve, it doesn't directly affect the fundamental quality of the audio signal, just the accuracy of the EQ effect.

Maybe someone will find this useful, cheers.
  • Last Edit: 04 January, 2017, 02:50:59 AM by Artcore87

  • saratoga
  • [*][*][*][*][*]
Re: 31 band Graphic EQ: Q factor values
Reply #4
Quote
While a traditional EQ's "Q" is not adjustable, it is only effectively constant when at full boost or full cut... key word effective.  A typical 1/3 octave EQ only has a 1/3 octave Q in practice when the adjustment is at full boost or full cut, as the boost or cut is lowered the actual Q becomes wider, this is simply a consequence of traditional designs, whether passive or active, or digital (most of which simply digitally mimic analog filters).

What?  Q is just the ratio the center frequency to the  bandwidth of an oscillator. It does not depend on amplitude.

Re: 31 band Graphic EQ: Q factor values
Reply #5
Traditional Graphic equalizers with 30 +/- bands are always fixed constant Q.  Working with real world coils, capacitors, and resistors fixed parameters was the only practical way. This was almost always carried over into digital implementations of the same.

I know this is an old thread, but I was searching for a constant Q dsp and found this thread.  The above information is incorrect,

What is incorrect is your interpretation of what I said. When I mentioned  "real world coils, capacitors, and resistors " I was obviously discussing the bandwidth of the actual filters that are inside the equalizer, not its external measured performance.  You must not know that when adjustments of height and depth of the peak or notch to be adjusted and allow low peaks and/or shallow dips, the effective bandwidth of the filter is naturally widened, and thus the Q is naturally broadened.

You further compounded this error by apparently suggesting that parametric eqs are somehow immune to this effect, which they are not.

For example, consider an equalizer of any kind that is set to create a 1 dB peak. The definition of Q is based on the points where the response fall 3 dB from the peak value, but this never happens since the peak is only 1 dB high.

The lesson is that comments like mine have a natural domain of validity which I stated in way that was relevant to the original context. Taking a comment out of its original context (in this case the actual individual filters) and putting it into a different context  (measured performance) to falsify it  is one of Schoepnhauer's 38 false tricks for winning every argument. This information is over 100 years old (originally written in German by someone who died in 1860 and translated to English in 1896) and is well known among well-educated people who are serious about the discussions that they are involved with.

Schopenhauer's 38 Stategems

Just to ensure that Schopenhauer's actual meaning is not missed, please let me point out that he was not trying to train people in methods for winning every argument by adding additional fallacies to the discussion.  He was trying to point out how to avoid getting hung up on ideas that may seem valid when looked at superficially, but fail to be reasonable when properly understood. Context is very important!

In the process of trying to make my correct statement appear false you made additional errors, one of which I just showed above. But the biggest error was taking my comment out of its natural context.

  • andy_c
  • [*][*]
Re: 31 band Graphic EQ: Q factor values
Reply #6
Though I'm hesitant to dive into this thread, I thought I''d chime in anyway, just to try to clarify some things.  Unlike, say, bandpass filters, there is no single, commonly agreed upon definition for Q of PEQ filters.  I've summarized the mathematical relationships between a few of them here.
  • Last Edit: Today at 03:53:40 AM by andy_c