Shelving EQ

[Gold]

For a swinging Input version I was thinking of  modifying a White graphic EQ. I could divide it into four “bands” with the top and bottom bands being cut and the middle two bands being boost.

I think I’d just need to remove the inductor from the high frequency sections and the capacitor for the low frequency sections.  The user interface would be rotary switch frequency select and gain.

 

[JohnRoberts]

When Steinberg introduced the 1st version of VST, their EQ algo was based on a digital implementation of the hardware SVF, which resulted in the same underdamped response, whatever the actual Q setting. Devs were convinced thye had done the right thing; audio engineers pulled their hair trying to understand what happened. The brain received several conflicting sensations.

Underdamped HP or LP filters could actually be useful, but too hard to explain to the customers what is going on (I didn't even try).

The classic application for an underdamped HPF is the Bose 901 speaker active EQ... The rise before it rolls off response very cleverly provides useful bass boost for the loudspeakers with plenty of driver/amp*** protection at frequencies below that boosted region. 

JR

***When providing lots of bass boost we need to be careful about clipping the power amps, that would diminish the listening experience. 

 

[L´Andratté]

The high and low shelfs from the Steve Dove Studio Sound EQ are interesting, swinging inputs, low shelf has freq control, high shelf has a slope control...

 

[abbey road d enfer]

high shelf has a slope control...

Actually it's frequency and width control ; it is perceived as a slope control because the center frequency is always above hearing limits.

 

[Gold]

Actually it's frequency and width control ; it is perceived as a slope control because the center frequency is always above hearing limits.

I think the term  slope should be banned in reference shelving EQs. It subtly conflates a shelf with a filter with poles. One thing I dislike about the Maselec EQ is that the bandwidth of the bells are defined in dB/oct . That drives me nuts, especially since it’s an approximation anyway.

The more I look at modifying a White graphic the more I like it.

 

[PRR]

A shelf has poles. Two of them (well, one a zero). At least any EQ-useful shelf I can imagine. EQ is all about poles/zeros.

Here's two shelves. Top is what my old analog R-C filter mind thinks of. A zero at 880 and a pole at 3,520. While each tends toward a 6dB/oct slope, they are close together and interact to make a 4.5dB/oct slope at most. Bottom is what we can now do in digital (and conceptually with infinite numbers of analog parts). Bring the pole down to 950Hz and the slope is 84dB/oct. You can slice with 2-semitone precision. I have done such to separate a woman's choir from a boogie piano (near 250Hz) and invert the balance.

 

[abbey road d enfer]

I think the term  slope should be banned in reference shelving EQs.

I can't agree with that, because slope is an actual fact, even if it's a variable, it can be observed on the graph.

It subtly conflates a shelf with a filter with poles.

Because that's what it is. An EQ IS some kind of filter, so it can be described by its zeros and poles. The basic issue is that all EQ's are not simple filters, they have a numerator and a denominator, with different poles and zeros, that result in interaction in a more or less restricted band.

One thing I dislike about the Maselec EQ is that the bandwidth of the bells are defined in dB/oct .

Technically it is as inappropriate as using "Q". I advocate the use of BW (in octaves) because that's what is close to human perception and solidly implanted by centuries of musical practice. However it is necessary to dissociate bells, shelves and HP/LP filters.

That drives me nuts, especially since it’s an approximation anyway.

Likewise for the use of "Q", which is a mathematical abstraction with no relation to a physical perception. Anyway everything about EQ's is an approximation.

 

[JohnRoberts]

Likewise for the use of "Q", which is a mathematical abstraction with no relation to a physical perception. Anyway everything about EQ's is an approximation.

I gave up trying to get the AES standards committee to offer a concise definition for "Q" in peaking (bell?) type EQ...  All 1/3 or 2/3 octave GEQs are not the same despite claiming the same EQ section bandwidth.

This becomes a major source of errors when loudspeaker manufacturers try to publish preferred EQ curves for their speakers. Different DSP makers can deliver different results when set to the same Q (center frequency, and amount of boost/cut track reasonably well).

JR

 

[abbey road d enfer]

I gave up trying to get the AES standards committee to offer a concise definition for "Q" in peaking (bell?) type EQ...

I was consulted by the AES at the time; basically my answer was that "Q" was utterly improper. The committee should as well have invented a new unit, such as Sherry, in honour of the AES founder. It would not have been more inappropriate than Q.
I think several major players in the industry were reluctant to see a change in their ecosystem.

 

[JohnRoberts]

I was consulted by the AES at the time; basically my answer was that "Q" was utterly improper. The committee should as well have invented a new unit, such as Sherry, in honour of the AES founder. It would not have been more inappropriate than Q.
I think several major players in the industry were reluctant to see a change in their ecosystem.

But the result for people trying to use published speaker manufacture EQ curves encountered a tower of babel where the different brands of DSPs returned different results.

I kind of liked the old school way that fixed EQ was specified with an amplitude vs frequency graph, or look up table of amplitude vs frequency. Of course this is not very useful for characterizing variable EQ.

JR

 

[Gold]

I guess I used “poles and zeros” incorrectly to describe what I mean.  In PRR’s examples I see the zero  and the pole. Is the dot at the center frequencies considered a pole too? I’m probably making a technical argument for an operational perspective.  I realize it’s all just EQ in the end.

From a user perspective it helps to draw a distinction between a shelving EQ and a LPF/HPF. The HPF/LPF is always an order that is a multiple of 6dB ( excluding weird cases). The slope also extends to infinity, theoretically unlike a shelf which flattens out at the center frequency.

 

[abbey road d enfer]

Is the dot at the center frequencies considered a pole too?

I don't get it. There is no center frequency in a shelf EQ.

The HPF/LPF is always an order that is a multiple of 6dB ( excluding weird cases). The slope also extends to infinity, theoretically unlike a shelf which flattens out at the center frequency.

Many HF shelf EQ's go asymptotically to -infinity; some LF also. This trait cannot be used for characterization.

 

[Gold]

I don't get it. There is no center frequency in a shelf EQ.

I mean at the top of the shelf which would be the center frequency of it was a bell. In PRR’s example the top graph would be ‘half a bell ‘ with a wide bandwidth. The bottom graph would be ‘half a notch filter’

Many HF shelf EQ's go asymptotically to -infinity; some LF also. This trait cannot be used for characterization.

Isn’t the  Bax is like that? I don’t like shelves where the ‘center frequency’ is out of the audio band or very wide.  I’d rather control the bandwidth until I have the ‘slope’  I want without having to screw around with out of audio band information.

 

[john12ax7]

I mean at the top of the shelf which would be the center frequency of it was a bell. In PRR’s example the top graph would be ‘half a bell ‘ with a wide bandwidth.

This might have been part of our miscommunication earlier. The center crossover frequency I was referring to would be 1780 in PRRs graph.

 

[Gold]

The center crossover frequency I was referring to would be 1780 in PRRs graph.

Im still confused. On PRR’s graph it looks like each Y axis line is a doubling of frequency. The dot furthest to the right is one Y axis line past the 14,080Hz line. Wouldn’t that land the center frequency crossover at 28,160Hz?

I think of a shelf as ‘half a bell’ which lands the ‘center frequency’ at the top of the shelf where it flattens out.

 

[PRR]

> The dot furthest to the right

Is just arbitrary.

Basic flat response: from zero to infinity(*), gain is 0 dB.

Audio-guy flat response: from 20 to 20k, gain is 0 dB.

Shelf: from zero or 20Hz to 880Hz, head for gain=0dB. From infinity or 20kHz to 3.5kHz, tend to +14dB. From 880 to 3.5k, take a smooth curve.

(*) yes, in REAL systems nothing has gain to infinity frequency. But in audio we can usually have gain well past 100kHz, which is usually same-as infinity to the cutter, speaker, or ear.

 

[john12ax7]

I think of a shelf as ‘half a bell’ which lands the ‘center frequency’ at the top of the shelf where it flattens out.

The problem is the flat part theoretically goes on forever,  what is the center of that?

There was a thread a while back discussing what frequency to call a shelf. Some manufacturers will define it as the frequency a little before where it flattens out.  For example at 3dB down from the max boost of 15dB (the flat part).  But there seems to be no consistency to this.  Some are 3dB down,  others are 1.5dB.

It might have been Ian who suggested using the term crossover frequency , the point halfway up the slope, 1760 in the graph.  This is a good consistent technical definition, but the downside is you might not hear it that way.  I think it would have  been more clear if I originally said crossover frequency instead of center crossover frequency .

 

[Gold]

The problem is the flat part theoretically goes on forever,  what is the center of that?

I picture a shelving EQ as half of the equivalent bell. With the mirror image missing or cleaved off. I realize this is not technically correct in all cases but in many bandpass filter designs if you remove one of the tuning L or C you get a shelving response.

There was a thread a while back discussing what frequency to call a shelf.


Some manufacturers will define it as the frequency a little before where it flattens out.  For example at 3dB down from the max boost of 15dB

Snip

It might have been Ian who suggested using the term crossover frequency , the point halfway up the slope, 1760 in the graph.  This is a good consistent technical definition

I think for a low shelf it’s easy. You call it at its ‘center frequency ‘.  That’s the top of the skirt and where the ear hears the lift.

A high shelf is a mess. Sontec calls a high shelf that flattens out at 10K and lifts from about 1K a 10K shelf. Maselec calls a shelf that sounds like it lifts from 1K a 1K shelf. It’s unknown if it flattens out at the same frequency.

I’d call the Sontec technically correct but the Maselec more useful to the operator.

 

[ruffrecords]

I picture a shelving EQ as half of the equivalent bell. With the mirror image missing or cleaved off. I realize this is not technically correct in all cases but in many bandpass filter designs if you remove one of the tuning L or C you get a shelving response.

However, the shape of the curves is different because one is due to a pole and a zero separated in frequency and the other is due to a coincident pair. I do not like shelving curves because they continue to boost outside the audio band which can introduce unnecessary noise necessitating the use of additional HPF or LPF. Instead I now use low Q (sorry) bell curves which have the advantage of attenuating out of band signals.

Cheers

Ian

 

[scott2000]

I do not like shelving curves because they continue to boost outside the audio band which can introduce unnecessary noise

Cheers

Ian

Looks like NYD also mentioned something to this effect below in one of the related topics

https://groupdiy.com/index.php?topic=3200.0