HP & LP Filters and EQ

[moofi]

One thing he doesn´t take into consideration while demonstrating how sounds´ low ends do not interfere sonically is him having the sounds spread across the stereofield, so bascially they have their own space anyway.

I just stumbled across this video and I thought it would be relevant to this discussion. There are no rules in audio, but its good advice to keep the amount of filtering and EQing of sounds to a minimum as there are trade-offs with EQ. Each time you use filters you degrade the sound quality of the source and you lose some headroom. So rather than using a HP on the bass and then next boosting with bell, it's better to use a low shelf to control the low frequencies.

[huggermugger]

The combination of a HP (to eliminate unwanted very low frequencies) and a bell (to boost a specific range of lows) does not equate with a low shelf. A low shelf will boost or cut all frequencies equally below the cutoff. It's not 'better' to use a shelf when the combination of a HPF and a Bell gives the desired result, because a low shelf can't do that.

I just stumbled across this video and I thought it would be relevant to this discussion. There are no rules in audio, but its good advice to keep the amount of filtering and EQing of sounds to a minimum as there are trade-offs with EQ. Each time you use filters you degrade the sound quality of the source and you lose some headroom. So rather than using a HP on the bass and then next boosting with bell, it's better to use a low shelf to control the low frequencies.

[selig]

This comes up from time to time and is good to remember - many processes add gain, and if you follow best practices that include compensating for processing gain changes (up or down), you would likely already have compensated for that change. The idea that you loose “headroom” is nonsensical IMO, since you have literally hundreds dB of dynamic range beyond what is needed - you could theoretically “loose” close to 1400 dB dynamic range and still have enough dynamic range to accurately reproduce a 24 bit signal., so loosing 1-2 dB is not a problem.

Yeah I'm not gonna worry much about the slightly increased peak level, but the quite drastic phase distortion caused by LP/HP filters is something I've become more aware of. From my previous understanding this would not be an audible issue unless I have a parallel channel of the track running as well but in the video he mentions that this could even be a problem for a single track, and that it can "ruin the mids". I don't expect this to be a big issue but maybe I could be a little more mindful of where I use my filters

That’s different - maybe a I missed it, but I don’t think parallel processing was mentioned, and of course parallel EQ will have phase issues since EQ=phase shifts. I thought the entire conversation was about a single channel?
And as I’ve shown, the phase shift of a filter is far from consistent, yes, a ridiculous steep slope like in the video DOES have “drastic phase distortion (shift)”, but a 6 dB/Oct has FAR far less. So one cannot say there is drastic phase shifts with ALL filters.

[selig]

So rather than using a HP on the bass and then next boosting with bell, it's better to use a low shelf to control the low frequencies.

If you want to filter out the sub 40Hz, and also get a bump at 100Hz, I guess the easiest way is to use an EQ that has an HP function that lets you set a Q-value, for a bit of a resonance peak right above the cutoff point. No stock Reason EQ will do this, AFAIK, but you could use a synth-type HPF for this, using a device like Sweeper.

But you can also use a low shelf EQ that has a Q setting - M-Class is one. Use it to CUT your lowest bass frequencies, then crank up the Q value all the way up to get your bump. The M-Class graphic display illustrated this fairly clearly.

That is why I designed an EQ with filters that have a Q control! BUT, that won’t always achieve the same thing since you cannot tune the boost. IF you get lucky and things line up nicely, I LOVE using a HP filter with a little bump. plus, you also cannot get a wide boost with a Q control, so it will only work in some cases.

[Ottostrom]

Yeah I'm not gonna worry much about the slightly increased peak level, but the quite drastic phase distortion caused by LP/HP filters is something I've become more aware of. From my previous understanding this would not be an audible issue unless I have a parallel channel of the track running as well but in the video he mentions that this could even be a problem for a single track, and that it can "ruin the mids". I don't expect this to be a big issue but maybe I could be a little more mindful of where I use my filters

That’s different - maybe a I missed it, but I don’t think parallel processing was mentioned, and of course parallel EQ will have phase issues since EQ=phase shifts. I thought the entire conversation was about a single channel?
And as I’ve shown, the phase shift of a filter is far from consistent, yes, a ridiculous steep slope like in the video DOES have “drastic phase distortion (shift)”, but a 6 dB/Oct has FAR far less. So one cannot say there is drastic phase shifts with ALL filters.

No you're right the conversation was only about single channels. I only brought up the parallel channel thing when talking about what my own previous knowledge was.
I know different slopes will have a big difference on the actual phase shift but doesn't every filter still produce a complete phase inversion at the cutoff point?

Off topic mention here but I've started to use the midi follow function on the ColoringEQ and it's quickly becoming my favorite part about the RE Smart feature to include.

[selig]

This comes up from time to time and is good to remember - many processes add gain, and if you follow best practices that include compensating for processing gain changes (up or down), you would likely already have compensated for that change. The idea that you loose “headroom” is nonsensical IMO, since you have literally hundreds dB of dynamic range beyond what is needed - you could theoretically “loose” close to 1400 dB dynamic range and still have enough dynamic range to accurately reproduce a 24 bit signal., so loosing 1-2 dB is not a problem.

Yeah I'm not gonna worry much about the slightly increased peak level, but the quite drastic phase distortion caused by LP/HP filters is something I've become more aware of. From my previous understanding this would not be an audible issue unless I have a parallel channel of the track running as well but in the video he mentions that this could even be a problem for a single track, and that it can "ruin the mids". I don't expect this to be a big issue but maybe I could be a little more mindful of where I use my filters

(I wrote this earlier and forgot to hit post - sorry for a confusing double reply to your comment)
That's where more gentle slopes come in VERY handy.
Don't know about "drastic phase distortion", but keep in mind that with a filter the highest phase shift occurs at the point of the deepest cut. The attached graph will show this. Note the 6 dB/Oct trace (in "teal") shows only around 60° phase shift down at 10 Hz (the filter is tuned to 40 Hz, and we see four different slopes compared: 6/12/24/48 dB/Oct). Also note the impulse response is almost 100% "intact" after filtering, which demonstrates the filters ability to preserve transients (or not).



And yes indeed the phase shift used to create the filter response continues up the spectrum, but is really quite low at any critical midrange/high end frequencies. By 5kHz it's effect is negligible IMO.

So by all means, only use filters and EQ and compression etc when necessary, but also IMO it pays to learn which settings impart the least degradation to the audio signal - starting with these settings helps to use the least amount of processing necessary to get the job done. And sometimes you realize you don't need the processing in the first place, so doing routine A/B comparisons can also pay dividends in this department.
But bottom line, there is an impact with any processing, but often it's extremely minimal - so don't be afraid to use stuff because someone says it degrades the audio.

[selig]

No you're right the conversation was only about single channels. I only brought up the parallel channel thing when talking about what my own previous knowledge was.
I know different slopes will have a big difference on the actual phase shift but doesn't every filter still produce a complete phase inversion at the cutoff point?

Off topic mention here but I've started to use the midi follow function on the ColoringEQ and it's quickly becoming my favorite part about the RE Smart feature to include.

Only 24 dB/Oct (see my previous graph) is 180° at frequency of interest, at least in the filter I tested above. It would be impossible for all filters to have the same phase response (180° shift at the same frequency) but a different frequency response as they are totally related. Meaning, phase shift = frequency response. Or, more phase shift = more gain change.

[Ottostrom]

Yeah I'm not gonna worry much about the slightly increased peak level, but the quite drastic phase distortion caused by LP/HP filters is something I've become more aware of. From my previous understanding this would not be an audible issue unless I have a parallel channel of the track running as well but in the video he mentions that this could even be a problem for a single track, and that it can "ruin the mids". I don't expect this to be a big issue but maybe I could be a little more mindful of where I use my filters

(I wrote this earlier and forgot to hit post - sorry for a confusing double reply to your comment)
That's where more gentle slopes come in VERY handy.
Don't know about "drastic phase distortion", but keep in mind that with a filter the highest phase shift occurs at the point of the deepest cut. The attached graph will show this. Note the 6 dB/Oct trace (in "teal") shows only around 60° phase shift down at 10 Hz (the filter is tuned to 40 Hz, and we see four different slopes compared: 6/12/24/48 dB/Oct). Also note the impulse response is almost 100% "intact" after filtering, which demonstrates the filters ability to preserve transients (or not).

And yes indeed the phase shift used to create the filter response continues up the spectrum, but is really quite low at any critical midrange/high end frequencies. By 5kHz it's effect is negligible IMO.

So by all means, only use filters and EQ and compression etc when necessary, but also IMO it pays to learn which settings impart the least degradation to the audio signal - starting with these settings helps to use the least amount of processing necessary to get the job done. And sometimes you realize you don't need the processing in the first place, so doing routine A/B comparisons can also pay dividends in this department.
But bottom line, there is an impact with any processing, but often it's extremely minimal - so don't be afraid to use stuff because someone says it degrades the audio.

No worries, and thank you for explaining this confusing subject for me.
I actually tried to find exactly this type of Phase response graph for different slopes when replying to you earlier but didn't know how to word it correctly.
I see now that I indeed did have some misconceptions about how filters behave. Good info

[moofi]

Isn´t there a technical way to compensate for the occuring phase shift correcting the phaserotations after the filter?

[orthodox]

Isn´t there a technical way to compensate for the occuring phase shift correcting the phaserotations after the filter?

There are linear phase or symmetric FIR filters, which produce equally delayed filtered output. The drawbacks are significant latency and CPU load, especially with lower frequency filters. Izotope and Fabfilter make those, I guess.

[moofi]

I was thinking more in direction of a phase comparison before and after the filter and howsoever alter the filtered signal´s phase to match the original after filtering.
Sorry, if that´s what the mentioned filter types do.

Isn´t there a technical way to compensate for the occuring phase shift correcting the phaserotations after the filter?

There are linear phase or symmetric FIR filters, which produce equally delayed filtered output. The drawbacks are significant latency and CPU load, especially with lower frequency filters. Izotope and Fabfilter make those, I guess.

[orthodox]

I was thinking more in direction of a phase comparison before and after the filter and howsoever alter the filtered signal´s phase to match the original after filtering.

If that was possible, everyone would have been doing that. It's not possible mathematically.

[moofi]

How come if you don´t mind me asking?

I was thinking more in direction of a phase comparison before and after the filter and howsoever alter the filtered signal´s phase to match the original after filtering.

If that was possible, everyone would have been doing that. It's not possible mathematically.

[orthodox]

How come if you don´t mind me asking?

If that was possible, everyone would have been doing that. It's not possible mathematically.

If we're talking about plain IIR filters (which are fast and lightweight), the phase shift at some frequency is inextricably linked to the response at the frequency. You cannot correct the phase shift at the ends of the spectrum with another IIR filter without altering the response again and thus modifying the original filter.
The alternative is using FIR filters, which are heavy and comparable to two-way Fourier transform.

[moofi]

Can´t you simply see in comparison where the phase is shifted and adjust it accordingly? Is the only method to correct through a second filter?

How come if you don´t mind me asking?

If we're talking about plain IIR filters (which are fast and lightweight), the phase shift at some frequency is inextricably linked to the response at the frequency. You cannot correct the phase shift at the ends of the spectrum with another IIR filter without altering the response again and thus modifying the original filter.
The alternative is using FIR filters, which are heavy and comparable to two-way Fourier transform.

[orthodox]

Can´t you simply see in comparison where the phase is shifted and adjust it accordingly? Is the only method to correct through a second filter?

Yes, it's the only way. Since the adjustment must have the linearity property, it is a Linear Transformation, by definition. And there are not many types of such transform, all are the well-known filters.

[moofi]

I just felt like you could actually take both waveforms and look at them like you look at two different graphs in a coordinate system , then altering the equasion for the one after the filter to match the phase of the original.

Can´t you simply see in comparison where the phase is shifted and adjust it accordingly? Is the only method to correct through a second filter?

Yes, it's the only way. Since the adjustment must have the linearity property, it is a Linear Transformation, by definition. And there are not many types of such transform, all are the well-known filters.

[jam-s]

I just felt like you could actually take both waveforms and look at them like you look at two different graphs in a coordinate system , then altering the equasion for the one after the filter to match the phase of the original.

You're ignoring the fact that the phase shift of the filter is a function of the frequency. So you cannot compensate by just delaying the signal.

EDIT: See https://www.analog.com/en/analog-dialog ... ers-2.html

[moofi]

That´s a lot of math going on for just having gotten up
Gonna take a closer look eventually, though I doubt I would understand the equations without further education
Then I hoped for an explanation laymen could easily understand.
For example, how does "the phaseshift of the filter is a function of the frequency" relate to the comparison and further transformation of the resulting waveforms I asked about?

I just felt like you could actually take both waveforms and look at them like you look at two different graphs in a coordinate system , then altering the equasion for the one after the filter to match the phase of the original.

You're ignoring the fact that the phase shift of the filter is a function of the frequency. So you cannot compensate by just delaying the signal.

EDIT: See https://www.analog.com/en/analog-dialog ... ers-2.html

[client6]

I think much of this thread has gone over my head but I really appreciate people patiently explaining the details. I have learned a lot.