Compensating for mains and sub delay

[dualazmak]

If all frequencies are delayed equally, the waveform is reproduced correctly despite the fact that the delay introduced a major phase shift. The requirement for is that the system have both flat frequency response and flat time delay. That is, all frequency components must arrive with the correct amplitude and time alignment.

Exactly! We should not mix-up the total signal delay (or total latency) and the relative delay between XO-ed Fq zones.

I shared the same at the beginning of my post here on my project thread.

The relative delay among the SP drivers should be measured and tuned/adjusted which is the "time alignments" all over the SP drivers.

 

[dualazmak]

I've never tried tone burst method. I'd like to see them please.

OK, I will soon PM you sharing the tone burst signals, hopefully within 12 hours.

 

[ernestcarl]

I feel like there’s some confusion and oversimplification going on here in places. Maybe it’s time rewatch/listen to this discussion again. If anything, it can also double-duty as a sleeping aide for my insomnia.

BTW, everyone should really try the headphones/IEM and speaker test I suggested.

 

[Keith_W]

I feel like there’s some confusion and oversimplification going on here in places. Maybe it’s time rewatch/listen to this discussion again. If anything, it can also double-duty as a sleeping aide for my insomnia.

BTW, everyone should really try the headphones/IEM and speaker test I suggested.

Whilst I am sure that it would be a really instructive video, I don't have 4 hours to watch it. I would much rather read. Do you have a good resource for that?

 

[OCA]

OK, I will soon PM you sharing the tone burst signals, hopefully within 12 hours.

I checked your method but I first have to say you have a very interesting system. You have both beryllium and horn tweeters running in parallel! The alignment methods: pulse wave, sine wave and energy peak matching all seem to work great. It's also interesting an active crossover system requires 2nd order crossover phase linearization filters. Do you also have REW mdat measurements at the LP if possible for your system? I would like to check driver impulse peaks after alignment so as to clarify what to align for in the impulse responses.

 

[OCA]

Whilst I am sure that it would be a really instructive video, I don't have 4 hours to watch it. I would much rather read. Do you have a good resource for that?

Here's a favorite of mine on phase and it is only 7 minutes:

 

[ErLan]

Condition of your measurements don't look particularly ideal, so this may be the most excess phase correction that's quite "possibly" psychoacoustically benign (needs to be tested):

View attachment 298676

Also check how the windowing affects the "excess group delay" curves.


Extract the windowed excess phase:

View attachment 298677


If you want to keep using REW for simplicity (rather than rePhase), invert the excess phase:

View attachment 298678


Remove the windowing from the original measurement before convolving it with the generalized phase/time "correction" filter:

View attachment 298679


Before & After:

View attachment 298680 View attachment 298681 View attachment 298682 View attachment 298683

I would not attempt to pull down anything above 120-140 Hz as that would surely result in much more -- quite possibly "severe" -- audible pre-ringing than what this non-optimal correction filter already does.


-------

*Eh, this particular REW generated filter already does not look all that great post export (all pass/phase inverse filter is just sooo steep in some places):

View attachment 298689

View attachment 298687

Hello,
Thanks again for all the information, a lot to catch up.
Taking several steps back.
As a 'warmup' I tried to reproduce steps being mentioned by @ernestcarl and @dasdoing (as I assume my exploration will drive me to do similar with the yet to come measurements).
I managed to do so till the point where I am trying to export impulse as filter. my REW (20.5.13) configuration seems to lack the option to export samples count (or time) and consistently exporting the entire signal (~3sec long).
Any idea how to allow this option?

 

[OCA]

V5.20.14 early access build

V5.20.14 has moved to a new look and feel, FlatLaf, for most components, to modernise the appearance and make it easier to adjust. That involved a lot of changes, as many REW interface components were heavily customised, so I'm making early access builds available to help pick up any interface...

www.avnirvana.com

 

[OCA]

I feel like there’s some confusion and oversimplification going on here in places. Maybe it’s time rewatch/listen to this discussion again. If anything, it can also double-duty as a sleeping aide for my insomnia.

BTW, everyone should really try the headphones/IEM and speaker test I suggested.

I feel like there’s some confusion and oversimplification going on here in places. Maybe it’s time rewatch/listen to this discussion again. If anything, it can also double-duty as a sleeping aide for my insomnia.

BTW, everyone should really try the headphones/IEM and speaker test I suggested.

form your suggested video

 

[ernestcarl]

form your suggested video

View attachment 298961

Did I say that or made such a claim? Honestly, I need to sleep so why not rewatch the rest of it in earnest.

 

[dasdoing]

wow, so much to read, learn and reply.
let me for now just coment on the GD not audible below 100Hz thing.
so many tried to make a valid test for these but we could never come to a consense. I think because if you manipulate phase you manipulate FR, too.
So I have been thinking, we could have an ABX for this without any changes to FR if we use two tone bursts, e.g.:

we create a single burst of them = A
than we can delay the lower one and join them = B

what you guys think? wouldn't this be the perfect GD audibility test?

 

[OCA]

wow, so much to read, learn and reply.
let me for now just coment on the GD not audible below 100Hz thing.
so many tried to make a valid test for these but we could never come to a consense. I think because if you manipulate phase you manipulate FR, too.
So I have been thinking, we could have an ABX for this without any changes to FR if we use two tone bursts, e.g.:

View attachment 298981


we create a single burst of them = A
than we can delay the lower one and join them = B

what you guys think? wouldn't this be the perfect GD audibility test?

Unfortunately audibility involves how the brain perceives or interprets what the ears will hear and this has evolved over thousands of years in many mysterious ways. We are most sensitive to 1000Hz for example and can differentiate the highest SPL range at that frequency and guess why? It's the baby cry frequency.

 

[ErLan]

I had thought "it's possible" and I'm more convinced now after I checked the data a bit more. Here are some supporting facts:

• The flat delay from 70Hz to above 200Hz in your wavelet graph contradicts with a sub induced delay below 100Hz
• The delay up to 300Hz is even harder to explain
• Right speaker has considerably more GD than the left below 100Hz
• Subwoofer filter delays almost never exceed 5ms

Gradually falling GD graph from around 15ms @ 20Hz to around 2ms @200Hz is not only the norm for an idealized minimum phase speaker but also way below "extrapolated" detectability limits. Longer wavelength bass signals will travel slower than the short wavelength high frequency signals in the air and you cannot change that! There were extensive studies about this by Blauert and Laws, Flannigan and Liski back in the 70s:

View attachment 298942View attachment 298943

Intresting points,
1. What i was able to dig from (previously) measuring Rt/Lt without sub is that there is delay in the 200-300Hz that is likely generated by room modes - I am planning to tackle those as well (can send graphs later, not near my pc at the moment).

2. What is am also short to explain is: assuming sub is subjected to delay of 25msec, and mains to 5msec, how come in the spectrograph i do not see a smooth transition from 25msec to 5msec between 100Hz +/- 0.5Octave that reflects the blend of those inputs on the crossover? Can it be a result of potentially another room interaction? And if so, how come it is triggered through the sub connection, which is supposed to be fading out?

 

[dasdoing]

Unfortunately audibility involves how the brain perceives or interprets what the ears will hear and this has evolved over thousands of years in many mysterious ways. We are most sensitive to 1000Hz for example and can differentiate the highest SPL range at that frequency and guess why? It's the baby cry frequency.

you didn't respond, no? is the test valid?

It's the baby cry frequency

also the frequency of my ex complaining, whith whom I have to live in the same apartment together with my 1 year old daughter. so there is a lot cry frequency interferance between the two lol. though I have to say that even same frequency when it is my baby it sounds like music, when it is my ex it sounds like hell hahahha
I actually use industrial ear-muffs when she just wont stop. not kidding.

 

[dasdoing]

whatever, here are 3 files of the type I mentioned in post 71

GD-Kick-0 is a mix of 2 1-cycle burst: 50Hz and 300Hz (did I just create the perfect electronic music kick? just kidding lol)
GD-Kick-20 has the 50Hz one delayed by 20ms
GD-Kick-20-predelay has the 50Hz put 20ms to the front

you @ErLan can actually use these to find out if the delay you messured is real. if the predelay one on your speaker sounds close(r) to the 0 one on headphones it would be confirmed. if it has the obvious predelay you can hear on phones on your speakers, too....you sub is not delayed.
that's my theory, there might be a flaw in this aprouch...don't know

regarding audibility. the 0 and the 20 one are so obviously diferent to me I wont even abx them....too easy

GD-kick.zip

drive.google.com

 

[OCA]

Intresting points,
1. What i was able to dig from (previously) measuring Rt/Lt without sub is that there is delay in the 200-300Hz that is likely generated by room modes - I am planning to tackle those as well (can send graphs later, not near my pc at the moment).

2. What is am also short to explain is: assuming sub is subjected to delay of 25msec, and mains to 5msec, how come in the spectrograph i do not see a smooth transition from 25msec to 5msec between 100Hz +/- 0.5Octave that reflects the blend of those inputs on the crossover? Can it be a result of potentially another room interaction? And if so, how come it is triggered through the sub connection, which is supposed to be fading out?

You need to make repeated, 0-100Hz, 100-24000Hz and 0-24000Hz measurements for both left and right. Some of them will surely be distortion free and then we can have a better understanding.

 

[OCA]

you didn't respond, no? is the test valid?



also the frequency of my ex complaining, whith whom I have to live in the same apartment together with my 1 year old daughter. so there is a lot cry frequency interferance between the two lol. though I have to say that even same frequency when it is my baby it sounds like music, when it is my ex it sounds like hell hahahha
I actually use industrial ear-muffs when she just wont stop. not kidding.

It makes sense but standing waves will interact in odd ways with both the original and the delayed tones but I think it's worth giving a try.

In my experience, I don't hear a difference in the sound but I sense a clear lift at the height of center stage when I align mid/tweeter XO phase shifts (typically around 1400Hz) and the explanation for this is that sound is moved towards the higher frequency source when drivers are time aligned.

When I do a port and woofer/mid phase linearization (40Hz/260Hz), it's not audible to me on its own. But it helps greatly to clean up excess phase between 100-1000Hz with soft filters avoiding pre-echo effects which help remove L+R response dips in the frequency response.

 

[dasdoing]

I sense a clear lift at the height of center stage

it's funny you say that, cause I have experienced this a lot while correcting stuff and have never seen anybody mention it.
I still have to study your posts here, and watch more videos on your channel. you have a lot to contribute to us "autodidacts"

 

[ernestcarl]

I've never tried tone burst method. I'd like to see them please.

The burst tone plots @dualazmak shows are more or less similar to REW's filtered wavelets and its envelope which can be generated in the "Filtered IR" window. Advantage of time referenced REW measurements is that effects of any phase filtering and/or time offset can be simulated as well as overlayed over on top of one another. However, the workflow isn't entirely seamless and there's still a lot of switching required between windows.



In a lot of setups in untreated/un-heavily dampened rooms (and with no EQ), I expect these plots to be much messier -- even difficult to make any sense of...



-----


Thinking about this further, @OCA may be correct in that it is enirely possible that the OP's room and sub(s) measured response may already be aligned closely to the mains.

However, the measured room+sub and/or speaker response could be, in fact, just simply be not all as ideal:

FIR EQ could help quite a bit above... but, it's still probably going to be difficult or far from perfect.

 

[OCA]

The burst tone plots @dualazmak shows are more or less similar to REW's filtered wavelets and its envelope which can be generated in the "Filtered IR" window. Advantage of time referenced REW measurements is that effects of any phase filtering and/or time offset can be simulated as well as overlayed over on top of one another. However, the workflow isn't entirely seamless and there's still a lot of switching required between windows.

View attachment 299240 View attachment 299241 View attachment 299242 View attachment 299243 View attachment 299244 View attachment 299245 View attachment 299246

In a lot of setups in untreated/un-heavily dampened rooms (and with no EQ), I expect these plots to be much messier -- even difficult make any sense of...



-----


Thinking about this further, @OCA may be correct in that it is enirely possible that the OP's room and sub(s) measured response may already be aligned closely to the mains.

However, the measured room+sub and/or speaker response could be, in fact, just simply be not all as ideal:

View attachment 299252

View attachment 299253

View attachment 299254

FIR EQ could help quite a bit above... but, it's still probably going to be difficult or far from perfect.

I am trying to produce your graphs. Not being easy! Btw, I am so used to using REW in white background, dark mode looks like an entirely different program to me