http://www.fabfilter.com/forum/2268/48db-hi-cut-slope-causes-boost-in-level?replies=1#comment7390
 
I can come close to thinking I can handle it by considering that, to my limited knowledge, at least some, EQs use a sort of parallel processing and then combine signals to get the final output.
 
Then I think, "hey it's only a lo-cut", and the explanation that the peaking is due to comb filtering seems like, well, that's what EQ is anyways... controlled filtering.
 
What do you think?
 
I'm off to work on the road but when I get back I'll try a spectral analysis to see what frequencies are doing the nasty.

Mike with my limited knowledge, I do know that when you use a steep filter to cut, there will be a resonance freq to deal with. The steeper the cut, the higher the resonance freq with raise. Now is that happening with your situation or not I'm not sure.

I also remember reading something about the refreshing of the digital signal in EQ's not catching this resonate freq. Again if this is happening with you not sure. But the article/book I was reading stated this may occur and you wouldn't be able to detect/know about it until it was too late.

In all the reading I did I don't remember if this freq boost was a result of a freq no longer existing not canceling another one out. Or if it was more to do with how eq's work.

All the above may be related to actual hardware eq's and not digital ones! Can't remember right now.

Mike in your question to FabFilter you mention Hi Cut in the heading and Lo Cut in your actual question. Which is it?
 
Also if you mention Q then you cannot be talking about filters. Filters don't have a Q setting usually. So are you referring to a peak style EQ then.
 
Anyway I did some tests and set up a source of pink noise at a very precise level and metered the output with a VU meter. I tried several filters to start with, all at 48 dB/Oct slope at both ends of the spectrum. (I even set up a Cakewalk LP64 EQ and replicated the response and tried moving all the nodes and still got no upward level shift)  In all cases I got no upward level shift here at all anyway no matter what the cutoff freq was. (only a downward shiftt in level as to be expected as you move the cutoff freq further in etc)
 
I also tried this with a peak style EQ as well with a Q of 1.0 and also got no such level shift either. Only a drop in level as to be expected.
 
Sounds a bit strange to me. The fact is if you are taking stuff away the level should go down not up. That is what is happening here for me.  In all the years of mixing I have done and used steep filters at either end of the spectrum I have never seen any upward level shifts. (probably because I don't use any Fab Filter stuff maybe!)

Hi Grem,
I too thought it might be a resonant bump but the reply I was asking about seemed to explain that there is no resonant peak such as we are familiar with.

With a Q of 1 you usually will not see a hump. I thought perhaps the 48dBsuggestion per octave filter might be different.

Sent from android... pardon the mess.

Hi Jeff,
I just saw your message. I am working a gig the next 2 days but look forward to reading the details of the info you have shared.

Thank you!

Mike - enjoy your gig, that gives me time to try to write my explanation in a way that makes sense<G>! In the meantime I found (don't ask me how) a decent explanation from SOS here:
http://www.soundonsound.c...articles/qa-1210-5.htm
 
Jeff - can you elaborate on your statement:
"Also if you mention Q then you cannot be talking about filters. Filters don't have a Q setting usually. So are you referring to a peak style EQ then."
 
The Q factor for a bandpass filter is defined as the center frequency divided by the bandwidth, it is, by definition, a dimensionless ratio.

This can also be applied to high pass and low pass filters, although the definition at that point has more to do with damping.
 
I'm also unclear on why a peak style EQ is not a filter. What am I missing?

Back to Mike - probably my all time favorite EQ topology is parallel filters - my Valley People Maxi-Q used this approach. Something about the way the bands add together is just inherently musical to my poor old ears.

The total amplitude of a signal is the result of the combined frequency parts. If you cut a frequency band with an EQ, especially a strong low end, you can create a peak where there was none, because the wave might have been inhibiting the total at a certain point. Think of a perfect sine of, say, 50 Hz that is part of a rich harmonic signal close to peaking. The 50 Hz wave has parts where it has positive values and parts where it has negative values. If the total signal is close to peaking on a place where the 50 Hz tone has a strong negative value, you can create a peak by cutting it. It occurs with other frequencies too but is usually clearest with strong lows.

Sorry if this is not what you meant, I remember being surprised by this realization when I read it. I think in a manual of a Waves plug.

I am not seeing any boost in level no matter what I try. Except that maybe the pink noise is not the best source.  Perhaps I should try it with a music signal instead.
 
But like I said before in all my mixing experience I have never encountered this so maybe I have just been lucky in that all the various EQ's I have used (in filter mode) have not exhibited this issue.
 
One should not be anywhere near 0 dB FS anyway so if you are working down at a ref level eg -14 or -20 and you did get a level shift then it should not cause any issues with clipping, only that you may have to tweak the level down slightly in your mix.
 
Bill in most of the EQ's I have got here as soon as I go into HPF or LPF mode the Q parameter usually greys out. (But you get a choice of slopes though) I do have the odd EQ where Q can sometimes refer to the slope of the filter.  Yes the peak style EQ is a filter too for sure (When it is cutting though not boosting, filters don't boost remember) But even when I set up a peak style EQ in cut mode I still don't get any boost going on either as I sweep the frequency.

Well, it may not help much but it is (or should be) well known among signal processing engineers that if you have a multiplicity of sinewaves, the phases of them can be chosen to minimize the peak of the entire signal. This is useful to keep a wideband signal within range of a DSP. So, if you move the frequency of any of these sinewaves, or change the phase of any of them, or REMOVE any of the sinewaves, the peak of the entire signal will go up. I think that is what you're seeing. You won't see this effect unless you have "many" sinewaves that happen to fall into certain phase relationships.

AHA! I misunderstood your original statement, thought you were making a general statement about filters and equalizers, not focusing on the filters and equalizers that you own. It is unusual to provide "Q" control on a shelving filter simply because there are other. more conventional ways to describe the effect. Usually the setting is marked as slope, but mathematically it is the same thing, the sharpness of the resulting curve.

As far as Mike's question goes, our ears are sensitive to the energy under the curve, as are some meters. Most meter implementations are peak reading (it is the nature of DSP, reading a maximum value is just easier), and you can end up with some squirrely behavior.

That does not mean, however, that Mike's scenario is not real. There are effects from summing filter outputs that can be... what's the word... less than intuitive?