"The secret of big speakers"

[Wombat]

FWIW;

https://www.tandfonline.com/doi/pdf/10.1080/19942060.2016.1277166

Group delay guideline, see P 15. http://www.cfuttrup.com/blogspot/Book_05_Futtrup.pdf

Digital Filter Correction Of LF Group Delay: http://www.dafx14.fau.de/papers/dafx14_stephan_herzog_low_frequency_group_delay.pdf

 

[jhaider]

I have been playing with minidsp 4x10HD and 2x4HD for many years now, with 4-way, 3way and 2-way plus sub systems. Also tried to accommodate to passive LR4 xo but switching to LR2 makes all my speakers to sound more natural and "easy to ears". I have eq'd each way of multiways to be similar, but still I and my friends notice easily when XO type is switched - and LR2 is preferred! Same experience with one passive 2-way diy speaker.

Interesting observations. However, I have two questions.

First, when you write LR2 and LR4 are you referring to electrical or acoustic rolloff? In most real world cases an "LR2" electrical filter results in around a third order acoustic rolloff, and an LR4 filter results in a around a 5th order rolloff. The interesting thing about odd-order filters is that they have constant power response.

Second, I do not find it hard to believe that identical speakers with only the filter order changed will sound different if equalized to the same on axis target. That is because the off-axis response will differ. Have you measured the off-axis response of either or both filters?

 

[Juhazi]

^Acoustic slopes. I first equalize each "way", then set xo and delay. The difference in off-axis is relatively small in a 4-way design, biggest difference is vertically. I was really surprised of noticing the difference in sound, I didn't expect it to be so audible - but subtle anyway. I might call LR4 version dry and LR2 wet. LR4 has less distortion when played really loud.

The whole story is here, but naturally I haven't published all my measurements and tests, not even saved everything for myself.
https://www.diyaudio.com/forums/multi-way/231353-aino-gradient-collaborative-speaker-project.html

 

[svart-hvitt]

And that is why I shelled out the big bucks for a Devialet, a purchase that I have regretted in certain ways.

Regret? Curious, why?

 

[Juhazi]

Test failed!
I set 2x4HD box in the line with .bin file made with rePhase. Response was almost the same but step response exactly the same. Something went wrong. 2x4HD accepts only 1024 taps. I left setup connected, I'll try to figure how to do it right.

 

[andreasmaaan]

@oivavoi, for a more practical approach to your question about whether to cross over your speakers with a 4th order filter at 120Hz, you could try this phase distortion ABX tester.

Using headphones, load a familiar high quality recording in wave format and set the "mid phase filter" to 120Hz, LR 24, mid filter normal, normal phase (I find it hard to get the slider to exactly the correct position, but you should be able to get within a few Hertz of 120).

Then look in the top right "ABX" section. A = original signal, B = phase distorted signal, and X = unknown. Take some time to try to work out if you can hear a difference between A and B. Once you think you can, begin flicking back and forth between A/X and B/X and see if you can discern whether X is A or B reliably through headphones. If not, it is very unlikely you will hear such a filter acting on your loudspeakers in a room.

You can also approximate the type of group delay caused by a BR port by using the "low filter" section if you're interested. Frequency = port tuning frequency. Choose "phase 24dB/octave" to simulate the effect of a port on the phase response (without actually modifying the amplitude). The type of group delay caused by a closed box can be simulated with the "phase 12dB/octave" setting.

I haven't analysed the output to make sure that it's doing what it says it's doing BTW... Will get round to this if nobody else beats me to it.

EDIT: unfortunately I've now noticed a little glitch whereby there seems to be the tiniest of pauses/discontinuities when flicking between A and B, or between X and whichever of A and B is not X (but not the other way around). So the results are not going to be reliable unfortunately. Interesting tool to play around with though anyway perhaps.

 

[Cosmik]

You can also approximate the type of group delay caused by a BR port by using the "low filter" section if you're interested. Frequency = port tuning frequency. Choose "phase 24dB/octave" to simulate the effect of a port on the phase response (without actually modifying the amplitude).

It may simulate the effect of a port on steady state waves, but it won't simulate the effect on transients i.e. the resonance, the ringing, the delay in 'getting going' or 'stopping'.

 

[KSTR]

It may simulate the effect of a port on steady state waves, but it won't simulate the effect on transients i.e. the resonance, the ringing, the delay in 'getting going' or 'stopping'.

Now, how do you know? 4th order response is 4th order response no matter haw it is physically implemented. If the "steady-state" response is correctly modelled then "transient response" is correct as well.
Did you try the experiment I laid out here (do a convolution of measured IR with a music signal and compare to the direct recording) and if so, what have you found, to what extend does it violate signal theory other than in the aspects I mentioned (difference is large only when system is overloaded, both from distortion and port noise)

 

[andreasmaaan]

It may simulate the effect of a port on steady state waves, but it won't simulate the effect on transients i.e. the resonance, the ringing, the delay in 'getting going' or 'stopping'.

Assuming that the port is well-designed (and for argument’s sake let’s take that to mean that particle velocity never exceeds 7m/sec within the speaker’s operating range), what other than the system’s minimum phase behaviour causes the delay in “getting going or stopping” in your view?

Or in other words what is the cause of the claimed nonlinear resonance and ringing?

 

[andreasmaaan]

FWIW, the ABX tester appears to perform as claimed (notwithstanding the unfortunate glitch mentioned in post #187).

 

[andreasmaaan]

@KSTR, I'm curious about the tests you undertook personally using speakers. Could you please explain your test setup and methods so that I can attempt to replicate them? (later in the year though - our workshop is under construction at present)

I've read your post #137 and all that you've said there is clear.

So my questions would be:

• Which software did you use to create the allpass filters and to convolve the program material?
• What program material did you use (and if music, what type and/or specific examples)?
• What type of loudspeakers and room did you use?
• Did you test for specific group delays at specific frequencies in isolation, or did you do something more like what the Genelec test did and convolve measured or calculated group delays for specific whole-speaker systems?
• How many subjects participated?
• What format did the ABX tests take? In particular, were subjects able to instantaneously flick between stimuli without interrupting the program? Or would a whole stimulus be played, and then another, with a short pause separating them? And were subjects able to listen to the stimuli multiple times before deciding whether X was A or B?
• How were subjects trained?
• Any other information you think may be relevant to replication.

Of course, if you've done this a number of times you may have used different methods each time. If so, please suggest under which circumstances the lowest thresholds were obtained.

I'd also be interested to hear under which circumstances these thresholds were not obtained, and any views you might have as to what distinguished these tests from the most "successful" ones.

Many thanks

Andreas

PS: feel free to PM me if you'd prefer, although I imagine others here might also be interested to hear more on this topic.

 

[KSTR]

PM sent

 

[Wombat]

Yeh totally, if you're not gonna involve some kind of EQ or subs and need to rely on whatever bass response and roll-off the speaker naturally gives you, there's a better chance a sealed box will integrate well with your room. It's also gonna give you more flexibility in terms of sub crossover because a 12dB roll-off is easier to integrate into a typical 24dB woofer to sub filter than tbe 24dB roll-off of a ported box. But in most cases the port tuning frequency will be sufficiently low that this won't be a factor.

IMO apart from these two quite specific cases, the only potential advantage of a sealed box is the remote possibility that the threshold of audibility of phase rotation happens to lie somewhere between 1/2 cycle and 1 cycle at typical port tuning frequencies. But given what we know about sensitivity to phase in general, and given the fact that these frequencies are well below the Schroeder frequency in most rooms anyway, so these frequencies are ringing through the room for many times longer than the initial rise, the chances of this are very very low IMO.



Thanks, will give that a read

And I hope I haven't seemed to imply that I think we can't detect group delay. We absolutely can. It's just the thresholds seem to be higher than what many assume.

But I'm glad to see a new study on this and will read it.

FWIW: https://www.researchgate.net/public...of_Human_Hearing_to_Changes_in_Phase_Spectrum