That is my design policy if FIR is not available or taps is too limited. Exception is TP filters. For example Trifinov Transient Perfect has 3rd order LP, 2nd order HP and Z offset of 1/3 wave length at XO. MF driver could be quite far back with low XO.
Minimum phase is "the natural" version. Linear phase would have severe pre-ringing due to negating of HP slope of the woofer.
...forgot to mention that linear phase down to 5...0 Hz requires close to infinite amount of taps because phase response is equalized from natural slope (or minimum phase) to linear phase i.e. flat 0 deg down to basement. I guess Bodzio's tool makes minimum phase at least at LF. On-axis phase could be flat 0 deg from MF to HF while magnitude response is not perfectly flat though it's not entirely natural. "Linear" just sounds much greater and sells better than "minimum" 🙂
Is it just single driver with some cone resonance at ~3 kHz and lower SPL at LF than HF. Equalized and shown as 300 Hz square wave.
Right, in theory one can disagree with everything.
It's all about how significant something is in practice.
Otherwise it's nothing more than believing in something.
Which sometimes it heavily feels that's what people prefer to do instead of trying to objectively read and dive into some papers that are being referred to.
That excessive group delay will cause issues is already known for something like 30 years at least (if not much longer).
Fyi, I wasn't saying that group delay is never issue, I was saying that those small group delays mentioned won't be a deal breaker by using a simple thought experiments and deductive logic and reasoning.
Those numbers are barely noticeable in an hyper controlled environment with much higher frequencies.
By definition with lower frequencies those same numbers will be a lot less significant.
People can believe everything they want, right so nothing wrong with it, but it would be respectful to also keep calling it believing and not referring to any kind of "science" to find reasons for believing something.
I personally would like to keep my advice and knowledge based on engineering and science.
I am also more than happily change my mind and opinion about certain things if people can show otherwise with proper experiments in combination with well funded theory.
Statistical significance is a communist system developed for average citizen in the mass market. On DIY audio we have full rights to prefer and weight something 98% of random listeners cheap commercial science is using don't hear or whole study is not interested in and arranged for something else 😀
But seriously, reminding that discussion about timing happens e.g. here because shrine of scientific audio called ASR does not recognize significance of it. You say it's known for 30 years, and I say that it's perfectly logical, perceivable and crucial for sound reproduction trying to be more than indifferent.
Attached is measured response and group delay of my diy, passive XO, two way.
Far field 300Hz+, near field bass and port added (minimum phase both), baffle compensated and merged at 300hz with far field. All done by Vituix tools.
Resulting file is then loaded in crossover simulator, one driver response. Is this correct procedure to get measured LF group delay?
If this procedure is correct, should I be satisfied with this numbers for group delay, considering passive X over?
Number is 2.7ms at 100Hz and 11ms at 50hz.
Simulated group delay in enclosure tools was 2ms at 100Hz and 10ms at 50hz.
Far field 300Hz+, near field bass and port added (minimum phase both), baffle compensated and merged at 300hz with far field. All done by Vituix tools.
Resulting file is then loaded in crossover simulator, one driver response. Is this correct procedure to get measured LF group delay?
If this procedure is correct, should I be satisfied with this numbers for group delay, considering passive X over?
Number is 2.7ms at 100Hz and 11ms at 50hz.
Simulated group delay in enclosure tools was 2ms at 100Hz and 10ms at 50hz.
Attachments
Btw, I played with enclosure tool and found that subwoofers simulating smallest group delay.
FAST was first what I thinked about....
FAST was first what I thinked about....
Looks normal. I would look excess group delay only because that's the one we can optimize. Normal group delay depends on HP slope and extension of LF radiatorand may have some small extra delays too.
Last edited:
Single radiator with flat response close to 0 Hz can have 0 ms group delay down to 20 Hz. That fun ends when it's crossed to main speakers for example at 80 Hz with 4th order IIR. That action creates excess group delay of ca. 7 ms. 8th order XO ca. 15 ms and so on.
Here it is. Its 2.3ms at 50hz and 0.6ms at 100hz.
Its a bass reflex, passive speaker. How can we optimize excess group delay?
Attachments
I think you should go back to some physics classes. Absolutely doesn't have anything to do with cheap or being that 2% elite. (Which sounds rather arrogant on itself). Yes you're free to believe and follow anything you want that's the thing I agree with.
Don't appreciate the reference you're making at all. Even if it's meant as a joke.
Let aside the wording, i agree with Kimmo's message that the risk of 'one figure says it all' is indeed a mediocre average where improvements / innovations are seen as threads.
This behaviour is present on asr and in essence nog different from the behaviour of Stereophile followers as An example. Just see f.i. the almost absolute focus on Sinad across posts.
Mind you i fully support the principle of "through measurements to understanding". And yes sound reproduction equipment reviews across the board are in dire need of good and comprehensive and comparable measurements.
This mediocre average (i have no better wording) is also present in the pop-music production. ( I have contacts in the mixing/mastering domain).
As a side note, In my profession as Printing Specialist i have seen the same.
This thread is about the time aspect of music reproduction and is an area where in my view the research has not gone to the level of depth as f.i. off-axis/on-axis frequency response.
I can only hope there will be made progress in this area. With all the effort going on on this forum alone it must lead to progress. Not to forget other forums.
My comment about showing excess GD was generic. Your 2-way has already close to zero excess GD due to XO, and LF is typical for vented so all okay. Getting minimum phase i.e. flat zero excess GD down to 20 Hz would require FIR and quite much taps.
You will lose debate by defending weakness proven by reliable studies. Target in my projects is not higher than typical audibility level of old studies. Excess GD <2 ms at 100 Hz is reasonable and achievable for 3-way with quite low XO so I don't have to make questionable decision by accepting e.g. three times longer, and I won't certainly recommend that for the others.
I'm currently in the process of making Living Voice Auditorium clones. This is a 30 litre, vented 2-way MTM design with 2 x 6.5" Vifa midwoofers and a 1" Scanspeak D2608 tweeter. Tweeter is offset on the baffle.
I've been trying to follow the Vituixcad procedure for doing the measurements, however now I'm starting to doubt I have done this correctly.
The Vituixcad tutorial says to "Measure far field responses of woofer and mid-range driver and tweeter at 1000 mm in horizontal plane around the speaker."
So this is how I have done this;
1. Measured at 1 meter on tweeter axis (0-+/-60 degrees horizontal in 15 degree increments)
2. Measured at 1 meter on upper woofer axis (0-+/-60 degrees horizontal in 15 degree increments)
3. Measured at 1 meter on lower woofer axis (0-+/-60 degrees horizontal in 15 degree increments)
I measured upper and lower woofer separately since they will have different baffle step response (upper woofer having less baffle around it).
One pssible issue with my method is that mic distance to the baffle will vary slightly when adjusting mic height for the different drivers.
But now I'm starting to think that maybe I should have done all measurements with microphone on the tweeter axis?
I see some also using the term 'design axis', I presume this means listening height..
I've been trying to follow the Vituixcad procedure for doing the measurements, however now I'm starting to doubt I have done this correctly.
The Vituixcad tutorial says to "Measure far field responses of woofer and mid-range driver and tweeter at 1000 mm in horizontal plane around the speaker."
So this is how I have done this;
1. Measured at 1 meter on tweeter axis (0-+/-60 degrees horizontal in 15 degree increments)
2. Measured at 1 meter on upper woofer axis (0-+/-60 degrees horizontal in 15 degree increments)
3. Measured at 1 meter on lower woofer axis (0-+/-60 degrees horizontal in 15 degree increments)
I measured upper and lower woofer separately since they will have different baffle step response (upper woofer having less baffle around it).
One pssible issue with my method is that mic distance to the baffle will vary slightly when adjusting mic height for the different drivers.
But now I'm starting to think that maybe I should have done all measurements with microphone on the tweeter axis?
I see some also using the term 'design axis', I presume this means listening height..