When you say "shelf" what do you mean?

[Gold]

I think that "shelf" when describing an EQ is an ill defined term. Is there one correct definition?

Some seem to be a fixed amount of gain on one side of the selected frequency. Some seem to follow a filter slope. xdB/oct.

 

[SSLtech]

Well a shelf response flattens off at an elevated (boost) or reduced (cut) plateau.

You can't specify any variable boost/variable cut EQ in db/8ve, if the slope changes with amount of boost (which most undoubtably will).

xdB/8ve is a cutoff filter measurement to my way of thinking...

Keith

 

[JohnRoberts]

A shelving EQ starts from flat, rises or falls for some range then flattens out again. This can be boost or cut and can be located at the bottom or top of the passband. This response curve looks like a "shelf" when graphed.

The popular example of this is the bass and treble "Baxandal" tone control.

JR

 

[Gold]

[quote author="SSLtech"]xdB/8ve is a cutoff filter measurement to my way of thinking...
[/quote]

[quote author="JohnRoberts"]A shelving EQ starts from flat, rises or falls for some range then flattens out again. [/quote]


Here is where the worlds collide. Basically I agree with Keith but some (most?) shelves don't rise instantaneously. The rising portion of the shelf happens at some slope. Can the rising portion of the Baxandall circuit be described as a filter slope (xdB/8va)?

 

[SSLtech]

Paul, you're describing a 'step' more than a 'shelf' -correct?

The 'slope' of almost all shelf designs varies with boost, since the corner point where the response deviates from flat, and the corner point at which the response levels off remain fixed (on the frequency axis) as boost varies... thus the slope joining those two points varies in steepness, hence it cannot be referred to in terms of db/8ve...

I've built a few parametric EQs where the steepness of the slope into shelf has been variable... a sort of 'slope' control (you can't call it bandwidth, because a shelf by definition has infinite bandwidth!) and this does more to affect the 'character' of the HF/LF bands in shelf mode than anything else.

I'm baffled that more designers don't offer variable slope... you could vary it all the way from gentle slope to 'step'.

But there's a mighty big difference between 'step' which you seem to be describing (which has theoretically infinite slope or 'Q') and general 'shelf' designs.

The vital thing defining a shelf is NOT the steepness of the slope, it's the flatness of the plateau.

Can the rising portion of the Baxandall circuit be described as a filter slope (xdB/8va)?

No, if for no other reason that the slope varies with boost.

Keith

 

[Gold]

[quote author="SSLtech"]The 'slope' of almost all shelf designs varies with boost, since the corner point where the response deviates from flat, and the corner point at which the response levels off remain fixed (on the frequency axis) as boost varies... thus the slope joining those two points varies in steepness, hence it cannot be referred to in terms of db/8ve...
[/quote]

That's a very good description. That really clears it up for me.

On Igor's Sontecish there is a switch for variable slope on the HF band. Somewhat confusingly labeled 6/12 which immediately brings to mind a filter slope. I guess I should ask him what the 6 & 12 refers to.

 

[bcarso]

[quote author="SSLtech"]I'm baffled that more designers don't offer variable slope... you could vary it all the way from gentle slope to 'step'.
Keith[/quote]
It's not a simple synthesis!

 

[beatpoet]

[quote author="bcarso"][quote author="SSLtech"]I'm baffled that more designers don't offer variable slope... you could vary it all the way from gentle slope to 'step'.
Keith[/quote]
It's not a simple synthesis![/quote]

Perhaps to offer some variation in approximation or rough translation thereof may not be quite as difficult.

 

[Crusty2]

I think he's talking about varying the dampening of the filter. Like varying the Q control on a very wideband bandpass filter. The steeper it is, the bigger the bump before the dropoff...

 

[Viitalahde]

Shelf, as in "She couldn't help touching her shelf".

 

[SSLtech]

[quote author="Crusty2"]I think he's talking about varying the dampening of the filter. Like varying the Q control on a very wideband bandpass filter. The steeper it is, the bigger the bump before the dropoff...[/quote]
Simple it may not be, and perfect the eventual result may not be...

But I have done it, and it's proven to be one of the definitive sonic 'character' controls...

In the digital realm, the Sony DMX-100R consoles and the sony Oxford both offer a similar control. The HF and LF sections are bell/shelf switchable, and in the 'bell' position, the 'slope' of the bell sides (for any given amount of boost or cut) are affected by the 'bandwidth' control. -When 'shelf' is selected, most manufacturers render the bandwidth control inactive, whereas the Sony implementation 'parlays' the bandwidth control's function into a 'slope' control... (which it kinda already was...)

Anyhow, even with 'bump' imperfections (state-variable implementations being a classic example), the ability to change the sonic character is most useful...

Keith

 

[Crusty2]

What's slope? db per octave set by the filter order, yes? This control doesn't change that, but instead changes the filter polar response.

Here's a sim sweep of Q control of a Sontec/GML-type eq while set to set to shelf mode (simply shorting out one side of passive BPF):

It's a cool effect, similar to using cut and boost of Pultec at the same time...

 

[SSLtech]

Yes, and also like the 'resonance' control on analog synthesizer filters.

It may not be text-book, but it's sometimes VERY useful.

Keith

 

[bcarso]

I guess I should clarify by saying that a sharp shelf without peaking ahead of and after is fairly difficult to realize.

 

[Gold]

Hmmm. So the slope can be described in dB/8va in some cases. No wonder I've never had a clear understanding. It's not very clear!

 

[SSLtech]

Well, only for prescribed boost levels.

-Halve the boost: halve the db/8ve of the resultant slope.
-Double the boost... well... -you get the picktcha!

Keith

 

[bcarso]

[quote author="Gold"]Hmmm. So the slope can be described in dB/8va in some cases. No wonder I've never had a clear understanding. It's not very clear![/quote]

For small delta dB on each side, the peak slope doesn't come that close to the limiting value for large delta dB for a really simple first-order circuit. With the second order shelves (various configurations of the so-called generalized biquad) things can be adjusted to get a high slope, but it entails peaking/dipping on each side. These can then be dipped and peaked to compensate/flatten with more filters, if objectionable.

 

[Gold]

When selecting a shelf on te front panel of an EQ you select where the bottom of the slope starts. Does the bottom and top of the shelf move in tandem with frequency like a Q control?

I'm pretty sure the answer is yes but I thought I'd check. I've heard of people shelving up at the top end of the audio band. I never do this myself but it makes me wonder what they are hearing.

 

[SSLtech]

When you switch to shelf, the nomiated frequency -according to convention- is the upper frequency for HF shelf and the lower frequency for LF shelf... -Perhaps better expressed as being the corner frequency at the HEIGHT (or depth) of the boost (or cut).

Picture a Bell EQ. the nominal frequency is NOT either frequency where the response deviated from -or returns to- flat... -it's where it achieves mavimum deviation from flat... it's the same for shelf.

Now... -picture a switchable Bell/Shelf EQ. -Ignoring any bandwidth/Q changes, and taking an HF shelf as an example. -When you boost a bell at 10kHz, the response may deviate from flat at 2.5kHz, peak at 10kHz, and return to flat at 40kHz. -Note that the deviation points are an equal number of octaves above and below by the way. -When you switch it to shelf, the lower deviation frequency remains the same... the frequency where the response achieves peak deviation remains the same... it just never comes back down again. -This explains or demonstrates very clearly how the 'peak' corner frequency is used as the nominal frequency for describing the setting.

Hope that helps...

Keith

 

[Gold]

Let me see if I have this right. The graph (left most trace) that Crusty posted would be described as LF 3k shelf with a 15dB boost and a 10dB bump in response at resonance. Yes?