Audibility of group delay at low frequencies

[andreasmaaan]

afaik, there is a tecnical delay in the bass regions of a drivers since the soundwaves start from zero; meaning that the lower the frequency the longer it takes to get to the peak since the soundwaves become bigger. then there is the fact that more mass is beeing moved also.
So the tolerancy you decribe is explained by the fact that drivers were always like this. we are used to it. now that doesn't mean that a perfect group delay doesn't sound better, since it is what occures in nature

Driver mass is not inherently a cause of “slow” bass. In fact, a bass driver of any size/mass may be made to reproduce bass without excess group delay. The reason this does not tend to happen in practice is (a) that bass drivers must roll off, which has corresponding (minimum-phase) effects in the time domain, and (b) that bass drivers tend to be crossed over to midrange/HF drivers: the low-pass filter introduces a delay.

For practical proof of this, look to the step responses of speakers that are linear phase, either as a result of digital processing (in the case of the first example, below) or due to use of 1st-order crossovers (second example, which incidentally uses dual 10" woofers and is a fully passive design):

 

[ernestcarl]

I make a vented subwoofer system with a calculated group delay which peaks at 34 milliseconds at 17 Hz.

Huh... I remember that in one of Merlijn van Veen's videos he stated (typical model) vented subs will have an overall group delay of about 36ms:

I suppose your designs fit the norm; if maybe just a bit with less delay.

In live events indoors and outdoors I wonder what the usual group delay is for separate subs. Perhaps not much different -- but likely influenced by where the mic or listener is at in relation to any of the loudspeakers -- so it depends on how they're spread out or installed in the venue too.

 

[andreasmaaan]

Huh... I remember that in one of Merlijn van Veen's videos he stated (typical model) vented subs will have an overall group delay of about 36ms:

I think both are talking about the same thing but using different words. Group delay is frequency-dependent, so "peak" group delay and "overall" group delay will be the same value (I guess).

 

[ernestcarl]

Driver mass is not inherently a cause of “slow” bass. In fact, a bass driver of any size/mass may be made to reproduce bass without excess group delay. The reason this does not tend to happen in practice is (a) that bass drivers must roll off, which has corresponding (minimum-phase) effects in the time domain, and (b) that bass drivers tend to be crossed over to midrange/HF drivers: the low-pass filter introduces a delay.

For practical proof of this, look to the step responses of speakers that are linear phase, either as a result of digital processing (in the case of the first example, below) or due to use of 1st-order crossovers (second example, which incidentally uses dual 10" woofers and is a fully passive design):

View attachment 98023
View attachment 98024

The above is very similar to @mitchco's own perfect looking step response (although I don't think it's in volts) where he uses FIR filters and convolution for correction. In contrast, mine currently looks quite abnormal due to overlapping sub to mains xo and 'forced' extra delay for the main monitors -- all in an effort to get sound ahead of a node. I could just as easily have changed the delay so the SPL peak occurs after the node. Not that I really know for sure which way sounds more "right" or correct to the ears since the tail-end in the spectrogram in that area would end close to around at the same time either way. Also, this is beyond my own ability to discern with any certainty -- even after AB'ing separate DSP presets via JRiver.

Sceptre S8+SUB ~2m at the listening position

The mains by their own (sceptre s8) should look like this if measured in the anechoic nearfield:

The particular node which I'm trying to get ahead/around of can be easily seen in REW's wavelet spectrogram:

GD in same 2m distance at the MLP

"overall"

Merlijn didn't use that exact word in his description, but just me for emphasis sake.

 

[Kvalsvoll]

@Kvalsvoll, how did you decide to use 20 milliseconds at 50 Hz? Is it representative of the amount of group delay typical of commercial vented box designs?

I used 2 samples, one with larger 100ms delay, and one with less 20ms delay.

Practical in-room installations rarely manage to keep GD within 20ms in the bass range.

 

[Duke]

I used 2 samples, one with larger 100ms delay, and one with less 20ms delay.

Practical in-room installations rarely manage to keep GD within 20ms in the bass range.

How much group delay would we expect a "practical in-room installation" to contribute, in addition to the loudspeaker's inherent group delay?

Like if we had a sealed box, with very low group delay, what might we expect the in-room group delay to look like? (A ballpark educated guesstimate, or guesstimated range of values, is all I'm hoping for.)

 

[Kvalsvoll]

How much group delay would we expect a "practical in-room installation" to contribute, in addition to the loudspeaker's inherent group delay?

Like if we had a sealed box, with very low group delay, what might we expect the in-room group delay to look like? (A ballpark educated guesstimate, or guesstimated range of values, is all I'm hoping for.)

The small-signal gd of the speaker system in itself, is not the problem.

Typical situation, before calibration:

A calibrated system (V110+M2):

In one of my rooms, same bass-system (V110), this is in the range as-good-as-it-gets in a practical installation:

As you can see, there will always be narrow peaks reaching quite high, the average gd can be kept fairly flat, if there is output at very low frequnecies, the gd usually increases.

 

[Duke]

The small-signal gd of the speaker system in itself, is not the problem.

Very interesting! If I'm understanding what you are saying, your graphs, and my modelled group delays correctly, it sounds like room interaction is the dominant source of group delay at low frequencies.

I must admit that I don't know what you mean by "calibrated" in the context of your graphs.

Is there a correlation between the in-room frequency response and group delay? Maybe it's obvious and I missed it.

My apologies if these questions are overly simplistic - I have been thinking in terms of in-room frequency response rather than in-room group delay, so this is a new paradigm for me.

 

[Kvalsvoll]

Very interesting! If I'm understanding what you are saying, your graphs, and my modelled group delays correctly, it sounds like room interaction is the dominant source of group delay at low frequencies.

I must admit that I don't know what you mean by "calibrated" in the context of your graphs.

Is there a correlation between the in-room frequency response and group delay? Maybe it's obvious and I missed it.

My apologies if these questions are overly simplistic - I have been thinking in terms of in-room frequency response rather than in-room group delay, so this is a new paradigm for me.

Yes, room and misaligned subwoofers causes gd in magnitudes that completely drowns any time-related issues in a bass-speaker/subwoofer itself.

Calibrated here means that the crossover integration between main and bass-system is adjusted so they are aligned both in time and frequency, and minimum-phase frequency correction applied to correct for room issues.

Keep in mind that a nice gd-graph is never a goal in itself, it is used merely as one source of information to try to see what is going on.

Much of the room interaction at low frequencies are close to minimum phase, and then gd will also improve when the frequency response is equalized. In REW you can see what is minimum phase and what is not, by looking at the excess phase and excess gd graph.

Excess phase may or may not be easy to correct. Acoustic reflections are not so easy, while a misaligned subwoofer is easy to fix by adding appropriate delay on the mains.

 

[ernestcarl]

Is there a correlation between the in-room frequency response and group delay?

One should be able to see these in the measurements:

No delay integration plus all EQ corrections removed -- other than HP and LP filters.

Some of the improvements post correction look subtle, but there definitely is a correlation in the above examples. Of course, one cannot totally remove "errors" in the response caused by speaker-room acoustic interactions/interference.

*forgot to disable stereo upmixing setting in JRiver so surrounds were running and included in the summed response.

 

[bigjacko]

Does room create group delay? A speaker with group delay will delay in time, a speaker with no group delay in a room will have reflections delayed but the first sound will not be delayed. The two scenario will have different sound if we measure, I think the question is whether can human tell the difference or not.

 

[hyperplanar]

Does room create group delay? A speaker with group delay will delay in time, a speaker with no group delay in a room will have reflections delayed but the first sound will not be delayed. The two scenario will have different sound if we measure, I think the question is whether can human tell the difference or not.

Group delay is simply the negative derivative of the system (speaker + room)'s phase response. So yes, a room will basically always create group delay, because of the large frequency response/phase variations caused by room modes. You can split group delay in REW into its minimum phase and excess phase components--minimum phase group delay is improved by room EQ, while excess phase group delay can be reduced with sub/main delay settings, especially if there is a shelving of the excess group delay in the lower frequencies.

 

[bigjacko]

while excess phase group delay can be reduced with sub/main delay settings, especially if there is a shelving of the excess group delay in the lower frequencies.

I know that minimum phase can be EQ but excess group delay cannot be EQ. How can you reduce excess group delay? If we can reduce it to zero, then we can EQ it, but why no one is doing that?

 

[andreasmaaan]

I know that minimum phase can be EQ but excess group delay cannot be EQ. How can you reduce excess group delay? If we can reduce it to zero, then we can EQ it, but why no one is doing that?

It can be reduced with an all-pass filter, which is essentially an EQ for phase. There are DAW plugins (i.e. software plugins) that give you what is essentially a parametric phase equaliser. Or alternatively, you can create a phase correcting FIR filter in software like RePhase and then convolve your audio through it through your software player or a system-wide convolver like Equaliser APO.

People certainly are using these filters to correct phase, although not generally with the aim of eradicating excess group delay entirely, but rather with the aim of cleaning up the in-room phase response to ensure time alignment between different sources (speakers, subs) at the listening position and/or to deal with time-domain issues that are likely to result in audible problems.

We are tolerant of a fairly high degree of excess group delay before it becomes audible in its own right, though, so often this type of processing is implemented more to ensure smooth amplitude response than to eradicate excess group delay per se.

 

[hyperplanar]

I know that minimum phase can be EQ but excess group delay cannot be EQ. How can you reduce excess group delay? If we can reduce it to zero, then we can EQ it, but why no one is doing that?

If you have a setup with subwoofer(s) then optimizing the delay setting would help with that.

You can use FIR filtering to essentially force the lower frequencies to start playing earlier, but the results would be pretty unnatural sounding, wouldn't recommend doing that.

Basically, I would just set the optimal sub/main delay, do room EQ, and leave it at that. We have a pretty high tolerance for low frequency group delay regardless.

This is where I ended up after only doing those two steps, about halfway under the audibility threshold:

(ignore the treble spike, this was both speakers playing simultaneously which caused comb filtering there)

 

[bigjacko]

Thanks for teaching guys. If we can reduce excess group delay, can we EQ room modes out, if cannot then why? If we rephase the raw group delay to a target group delay, will the rephased group delay be fundamentally the same as the target group delay?

 

[Honken]

You can use FIR filtering to essentially force the lower frequencies to start playing earlier, but the results would be pretty unnatural sounding, wouldn't recommend doing that.

Very interesting. But I'm confused, why would it sound unnatural?

 

[dasdoing]

we should obviously concentrate on the overall room group delay when we talk about listening position.

what is possible in terms of correction at LP in a room can be seen here: http://drc-fir.sourceforge.net/doc/drc.html#sec237

Driver mass is not inherently a cause of “slow” bass. In fact, a bass driver of any size/mass may be made to reproduce bass without excess group delay. The reason this does not tend to happen in practice is (a) that bass drivers must roll off, which has corresponding (minimum-phase) effects in the time domain, and (b) that bass drivers tend to be crossed over to midrange/HF drivers: the low-pass filter introduces a delay.

For practical proof of this, look to the step responses of speakers that are linear phase, either as a result of digital processing (in the case of the first example, below) or due to use of 1st-order crossovers (second example, which incidentally uses dual 10" woofers and is a fully passive design):

View attachment 98023
View attachment 98024

thanks, I always get confused with the fact that roll-offs cause phase shift, as I can't process this visualy in my brain.
It becomes esier to understand if you think about it the other way around: it's not the roll-off causing the phase shift....it's the phase shift causing the roll-off (?)

Call it constant rather than zero if you prefer. Same result, but zero is easier to work with.

that's actualy the reason it is called LINEAR phase, and not zero phase

 

[hyperplanar]

Very interesting. But I'm confused, why would it sound unnatural?

In theory, you can essentially cancel out the excess phase completely. However, in practice, it’s very hard to avoid preringing from rearing its head. The filter parameters need to be carefully considered to avoid preringing in every listening position. Obviously, room EQ is position dependent as well, but in my experience the variations in frequency response for that are much less audible than preringing appearing with position variations and FIR filtering in the low frequencies.

Basically IMO the amount of effort and potential downsides aren’t worth the benefit. Excess phase group delay shouldn’t be at problematic levels with correct delay settings for subs/mains anyway. Usually the biggest contributor to overall group delay is the 24 dB/oct rolloff of ported speakers/subs and any additional high pass filtering, which are minimum phase. These sources of group delay can’t be corrected without preringing with FIR filtering in theory or in practice, except by flattening/extending the actual frequency response by adding subs or using minimum phase EQ.

 

[dasdoing]

In theory, you can essentially cancel out the excess phase completely. However, in practice, it’s very hard to avoid preringing from rearing its head. The filter parameters need to be carefully considered to avoid preringing in every listening position. Obviously, room EQ is position dependent as well, but in my experience the variations in frequency response for that are much less audible than preringing appearing with position variations and FIR filtering in the low frequencies.

Basically IMO the amount of effort and potential downsides aren’t worth the benefit. Excess phase group delay shouldn’t be at problematic levels with correct delay settings for subs/mains anyway. Usually the biggest contributor to overall group delay is the 24 dB/oct rolloff of ported speakers/subs and any additional high pass filtering, which are minimum phase. These sources of group delay can’t be corrected without preringing with FIR filtering in theory or in practice, except by flattening/extending the actual frequency response by adding subs or using minimum phase EQ.

the more you correct with FIR, them more LP dependend the result becomes, yes*.......personaly I can't stand listening outside sweet spot at all anyways, uncorrected or not; center channel is always messed up.
as always with audio, it's a compromise where you have to find the right point in the middle.
same is for pre-ringing and (to put it into perspective I will call it) pos-ringing. the best aprouch is to correct the group delay in steps until you hear the pre-ringing (if you hear it before making it close to linear phase).

personaly when I compare (listening) FIR room correction (actualy mixed phase) filters to filters created with REW in my system there is dozen times bigger impact on center clarity then audible delay. as Bob Katz describes it (he uses Acourate): The diference is like 3D vs 2D. I have seen other describe this as "lost wideness"; this is not the case, the mid signal is more centered (as it was supoused to be), not the side signal

*since you only can correct in the time domain on-axis. off-axis the "problem" and the "solution" will be shifted resulting in "2 problems"