I don't see an incompatibility, as long as you consider that the frequency for the low shelf is lower than the turnover and higher for the high shelf.
The families of curves are pretty symmetric.
Shelving equalizer and frequencies
- Thread starter RuudNL
- Start date
Help Support GroupDIY Audio Forum:
The families of curves are pretty symmetric.
..agree with @Madman - shelves are best described as the frequency where half it's total range (in dB) occurs.
This is the best approximation to what users report back as expectancies, and seems to make sense to most end-users
..this approach also allows for labeling less-than-3dB transfer functions
/Jakob E.
This is the best approximation to what users report back as expectancies, and seems to make sense to most end-users
..this approach also allows for labeling less-than-3dB transfer functions
/Jakob E.
That's certainly not what used to be usual in HiFi (when it existed) and not what I see in the response graphs of many mixers.
The example in post #9 would suggest frequencies of 120, 200, 400, 800 and 1600 Hz, when Ruud post #11 clearly mentions markings at 33, 56, 100, 180 and 330 Hz .
They have chosen to label frequencies 2dB below max boost/cut, which is 1/10th of it. I call it a "significant fraction".
Another example is the K+H UE100, where the 60Hz shelf corresponds to a point that's so close to the max that it is undistinguishable.
Actually, for the so-called "6dB/octave" position, the "nominal" frequency is about double (half for the high shelf) the frequency of max B/C.
Of course these graphs are heavily doctored but they are indicative of the average performance.
I respectfully disagree.
Most of the SE's I know tend to favour 63Hz for the kick drum and 125 for the bass. They expect it to be the point where or near where the boost is the most important.
They would be disoriented with the definition in post #9, where their favourite positions would be labelled 200 and 800.
That's why shelving EQs typically get labelled as "bass" and "treble". To repeat AFAIK there is not a standard definition for how quantify them.
===
FWIW I wasted a bunch of time trying to get the AES standards committee to define "Q" for peaking boost/cut type EQ sections. I gave up tilting at that windmill.
JR
===
FWIW I wasted a bunch of time trying to get the AES standards committee to define "Q" for peaking boost/cut type EQ sections. I gave up tilting at that windmill.
JR
ruffrecords
Well-known member
The slope should be pretty much the same in all cases. All shelving EQs that I am aware of use a pair of RC time constants - one is a pole and the other a zero. By definition, a single pole or zero alters the frequency response at 6dB/octave or 20dB per decade. So for a 20dB 100Hz boost you would expect to see the frequency changing significantly at 1KHz, probably as much a by 3dB at that point.
On the curves posted by RuudNL, the top one was quoted as being a 330Hz boost (by the manufacturer) to you would certainly expect to see maybe 3dB deviation at 3KHz (it looks like 5dB to me) so it is not a million miles out.
It is worth remembering that the RC circuits cause response that are asymptotic to 6dB/octave. So it you have a pair and expect a 20dB difference between them, then you are going to need significantly more than a decade to achieve this in practice.
Probably worth having a play with some simple passive networks with LTspice to get a feel for what actually happens.
Cheers
ian
Actually, most Baxendall type shelving EQ's exhibit more like a 4dB/octave slope, due in part to the fact that the max boost/cut is volontarily restricted and the interaction between the input filter and the NFB filter. Another way of looking at it is that the asymptotic region is never achieved.
The K+H UE series are a notable exception, since they use a different structure, with separate filters for boost or cut, not to mention the so-called "12dBoctave" position.
Last edited:
youré right a shelfing EQ has no Q but an slope which defines the order of the EQ, in this formular is it declare as Q because the comparism to peaking or Bell EQs.
When you are changing the slope you will see the turnover point is half the gain
this is the only constant point of a shelfing EQ and this because the definition is here.
In Hardware based, the EQ i know and i have designed performes like this (sorry for the bad photo i´m a little in hurry)
you see the -3dB point of a Bell is compared with half the gain (apart from little deviations) of the shelf.
Yes i will be mark it as 120, 200, 400, 800 and 1600 Hz
The EQ curves above are are too high for my liking, my designs are a decade lower in comparison, but this is another thing, thes questions was the definition of the frequenca by shelfing EQs.
How do you propose to make such responses in practice?
Do you have a reference text with this definition?
Last edited:
Well, it's clear that a definition that doesn't relate to user's experience has little value.
Are they useful? A decade lower means your EQ would be most active under 20Hz, where there is notably very little musical information, unless you're in specila FX.
Definitions and expectations should converge somewhat. You can't separate them.
Last edited:
I am starting to see that often in this industry the wheel that gets greased has nothing to do with the wheel that is doing the actual work...
Down that road...I recently picked up a Klark Teknik Square One graphical EQ...mostly used in live settings...but reading the manual it occurs to me it is actually a proportional EQ and not all that graphic and a bag of chips...I tend to gravitate to the proportional EQ's and the API 550 is my go-to workhorse for tracking with (when not using a console)...
But I was intrigued if this device was actually a proportional EQ like the literature suggested and so I threw it into hardware mode in REW and measured some of the bands...and sure enough this device functions pretty much the same as the API 550 with the EQ "Q" getting pointy like a Wizards hat when pushed...
Other than not being able to select specific frequencies, and the always dashing and polished sound of an NE 5532 driving the gain (with no transformers) it appears it does a quick and dirty API proportional EQ for the mere $30 I paid for it on the limited "30" bands in a nice 80's Prince Purple Rain colored rack mount unit...
I can see why it would work well for live, but there's no reason to ignore it in a studio simply because it doesn't come in studio-speak.
There has always been a certain amount of snobbery in the industry that measurements should have silenced but have failed to do so because of money I suppose.
No one wants to admit their expensive plugin or hardware can be copied by a lesser god.
As you probably have figured out, I always have been opposed to using Q to define the BW characteristics of EQ's, mainly because Q is only valid for basic quadratic equations, and EQ's are not basic quadratic equations.
I rather use the notion of BW, which has more subjective and operational value than a number. Ask anybody what's the BW of a Q of 3, you'll have mostly embarassed answers. Particularly if the boost or cut is low (under 3 dB).
Now I understand the need for a simple index to define that, but using Q is improper technically.
I understand why the AES would not embark on this adventure, but still, I reckon something should have been done.
The result of inaction is that the notion is completely distorted, so there seems to be as many interpretations as there are manufacturers.
An authoritative declaration from the AES or anybody, that BW is the way to quantify boost/cut EQ sections would work for me, but instead we get babbel.
===
In connection with my old day job I was responsible for merchandising many GEQ not to mention peaking type EQs inside tens of mixer/console SKUs. At least the GEQs were described as 1/3 octave or 2/3 octave, but not all 1/3 octave GEQ behave the same.
JR
PS: Another problem with this lack of clear definition is publishing manufacturer's loudspeaker crossover and corrective equalization curves. Lacking a concise definition for Q/BW these EQ curves cannot be depended upon.
Even witin the same brand, The inductor-based KT DN27 had narrower "bands" than the "gyrator"-based DN360.
Oh, do I know it... I had to translate our whole library of settings from the Linea Research platform to the Powersoft. Talking with the programmers was useless, basically because they used a textbook approach, which simply does not give repeatable results from one textbook to the other, I had to measure both and devised a simple table translator.
