I’ve been noticing that the Stereo Out level in Cubase seems to be higher than the level of the audio being fed to it. I presumed this was user error, and tried to simplify things to see if I could track down the issue.
What I have discovered is that switching on the high pass filter in the Pre section of any channel increases the level by about 1.5 dB. Obviously this means that any audio that’s been mastered, and had all its headroom removed, will clip, unless you turn off the high pass filter. I’ve been able to repeat this issue under test conditions, without fail - so I think this must be an with the high pass filter, rather than me doing something wrong.
It doesn’t seem to happen with a test tone, but it certainly happens with audio. Just to clarify, I’m not talking about a mix, I’m talking about a single audio track being sent straight to the Stereo Out.
So it seems the filter is active even though the knob is fully counter clockwise. The solution, I suppose, is to re-save my templates with all the filters switched off- and only enable one if I want to use it.
But if you have the filter set to 20Hz, it is filtering! Even if you have a filter at 1Hz, it is filtering (and thus can produce filter artifacts). This is not an issue, it works as expected.
Only solution there, as you already noted, is to switch off the filter.
There are many types of [digital] filter algorithms, each optimizing for different tradeoffs. You’d really have to know the specific algorithm employed to know how it will impact your audio signal. But yes, as pointed out in this thread, output gain will be impacted (esp. if the input signal has energy in critical frequency bands around the filter cutoff point), as will the phase of the audio. The nature and extent of both impacts will, again, depend on the type of filter algorithm employed, and there are many of those out there. Wikipedia can be a jump-off point to learn more about it, there’s a ton of literature out there since this is a really rich topic in [digital] signal processing. FWIW, all of this applies to analog filter circuits as well
This is caused not because of filter overshoot, but because of the phase shift introduced by EQ. It can get specially bad if you’re using very high order filters such as the 48 db/Oct cuts, even if it looks like they’re not touching the body of your signal at all.
For this reason, you should try to use the gentlest filter you can while high passing. Start with 12 db/Oct and use a higher filter if you can still clearly hear low end noise or if you’re trying to avoid affecting the fundamental frequency. Filters of higher order than 24 db/Oct are generally used for sound design or fixing serious issues.
You don’t actually have to worry about this gain boost most of the time though, because you’ll likely regain the lost headroom if the high pass filter is actually removing low end noise or gunk, or through regular EQing.
It depends on how deep the cut made with the shelf is. A deep shelf cut will usually have similar or a bit worse gain boost than a 6 db/Oct high pass filter, while potentially failing to eliminate low end noise. Shelves tend to be used for frequencies that you want to keep, but need to be tamed.
This gain boost is avoided entirely by linear phase EQ bands, since they don’t affect the phase at all. However, they may introduce pre-ringing/transient smearing artifacts, specially when applied to the low end. Linear phase filtering in general is highly specialized and you definitely want to do a little reading on the subject before trying to use it in your mixes.
Would it be fair to say destructive inference is likely taking place at some of the frequencies as well, and so it’s possible there may not be an overall boost at all?
Finally - is it safer to apply several/many 6 dB filters rather than one whopping 24 dB filter or greater?
(Re: linear phase - I was asking mainly to see how this might address the HPF problems being discussed. I’ll use it for parallel comp only, though based on what you are writing I’ll compare it to minimum phase filtering and see what difference that might make).
As @Romantique_Tp pointed out, unless you’re using a linear-phase filter, filter algorithms will impact the phase of audio signals differently at different frequencies. So, yes, if the channel you are applying a filter to happens to have energy in a frequency band that’s being significantly phase-shifted, then those phase-shifted signals can cancel out other signals in your final stereo output/mix.
Here’s an example of the phase response of filters:
And this blog post talks about how phase shifts due to EQ can impact a mix:
In case you missed it in the article, this is a great video going into the details of minimum phase vs. linear phase filter phase responses and associated destructive interference, and how to deal with that in a musical/mix context