Thanks. I did try and follow that document.
1. I do not have the drivers since I wanted to see if the drivers I have chosen will actually work together before importing them to India from the US (and incurring the customs duties and all the import hassles).
2. My idea to keep things simple was to use a full range and augment the low-frequency response using a woofer (or several woofers) below 200Hz or 300Hz. A mini version of the KEF Blade 2 but with simpler construction (see image below).
2. I used the FRD and ZMA files available via the SPL Trace tool and used the woofer's T/S specs available on the Loudspeaker Database and for the full range I used T/S specs published by Madisound (see links).
http://www.loudspeakerdatabase.com/SEAS/H1224-08_L18RNX-P
https://www.madisoundspeakerstore.c...Magnetic Arc Oxidized Full Range Matched Pair
3. I used WINISD to get the approximate box size and vent dimensions (46-47L). I chose a woofer that would work within my cabinet constraints (external dimensions of 46" H x 9" W x 16" D ) after leaving room for bracing and a smaller internal cabinet of 5L for the full range.
4. I understand that using 4 woofers in push-push (see image below) as used in the Audio Physic Loudspeakers and KEF Blade would mean that the baffle step might be limited if any at all. I am currently designing for a baffle step but am prepared to remove or reduce the baffle step circuit if it is deemed necessary once the speaker is built.
5. Using Vituix I now have a VXP file, a VXE file and a VXB file but the merger tool only accepts TXT, ZMA and FRD files. Does VituixCad have the ability to combine crossover response (VXP), box response (VXE), driver FRD/ZMA files, and diffraction compensation (VXB) into one curve?
I cannot find this option unless there is a different way to execute this.
Thanks.
Navin
I have to read the full paper, but isn't this just GD around crossover?
Because it was already well known that it will cause audible issues at those higher frequencies.
edit; For those who are interested, the paper can be read here (is open document)
https://research.aalto.fi/en/publications/audibility-of-loudspeaker-group-delay-characteristics
Do you want to open a separate topic about it, or leave it here?
I do have some comments about it. Definitely not as straightforward as it seems.
Last edited:
Member
Joined 2003
Hi Navin,
I hope that you have plans to acquire measurement gear, using manufacturer data has its limits in accuracy and reliability, and especially so for drivers aimed 90 degree to the listener. My recommendation for this type of simulation would be for familiarization of the tools, final design work of a speaker that you intent to enjoy for years to come should be completed with real measured data.
WinISD is not needed, VituixCAD includes a perfectly well functioning enclosure modelling tool.
The "export" buttons is what you are looking for in the Diffraction and enclosure tools to save response data. VXE/VXB files contain only the program settings, not the response result.
I've attached a document that I wrote a while back to help you generate response files using manufacturer data. Please read the first page to understand the limitations of following this process.
No.
I think we have to specify what is "low frequencies" or and "higher frequencies" to avoid accidental misunderstanding.
My experience is that big change in excess group delay from HF to LF damages sound within wide range. At least from upper bass to low treble. Sound could be quite balances and pleasant, but relatively bright, transients/punch at upper bass...low mid are weak (piano starts to sound like an organ), sound stage is incoherent splitting high frequencies to drivers despite close to perfect diffraction handling with big wave guide, etc.
Excess group delay changes continuously from LF to HF at least with normal 4-way IIR XO because next XO range starts when previous ends. 3-way IIR has usually wider GD steps and total delay at LF is shorter than with 4-way, but effect is still wider than just crossover ranges.
Particularly funny is to follow piano's hammer from right keys towards left keys. Transients are wide-range signals to both directions from fundamental so basically all keys benefit short excess GD down to bass. All keys are relatively weak if GD variation is high. Right keys are quite okay but left are weak if GD is flat from mid to HF but increases much at low mid. Whole scale down to bass extension is probably strong with minimum-phase response. Not guaranteed because GD is not the only feature affecting to dynamics.
Finnish study above is done with headphones and 100 ms impulse with pink spectrum is not very long so I think it's not entirely valid test and result for speaker reproduction. But it shows that >1-2 ms GD variation is audible at least down to low mid. Not just theory. Personal significance and judgement to good or bad is different story. My preference is significant to me. Not statistical average from random citizens some professor has selected to listening group.
Let's continue with VCAD.
No, this whole experiment doesn't show us anything at all.
The only thing it shows is the audible difference of impulse responses.
As far as I can tell, they just listened to some different impulse "clicks"
(they also added some weighting filter to make it a bit easier)
That's how they at least describe it in the text.
Yes, that is audible. But that has been already known for years.
But it doesn't have much to do at all with the audibility for music and such, let alone inroom listening, so there is no correlation.
This is why it's always EXTREMELY important to read the whole test description of certain experiments.
So I don't know were you draw your conclusions from, but they can't be from this experiment.
Certainly does not have anything to do with "Toole et al"
It also completely beats me why they went trough all the trouble of using simulated and measured loudspeaker responses.
They also could have made some group delay responses without any of that.
GD differences above 500Hz (or so) also have been known for at least a decade.
Just a quick google search will get you plenty of research and papers.
I am fine letting it go, I am not fine drawing conclusion that are not related to things.
There is already to much misinformation and myths around.
Try to decide.
Wide-band transients were listened, and music has wide-band transients (transients are wide-band). I think there will never be total agreement that generated transient and transient from actual instrument would be comparable due to denial, preferences and weighting. I would test with some actual instrument to reduce denial, though simulated transient can be comparable.
Reading alone is not adequate. Understanding is required too.
What conclusion? That effect of group delay is audible with program containing transients?
Read section "4.8.1 The Audibility of Phase Shift and Group Delay" in Toole's book. He uses close to half of space to tell how listeners prefer magnitude responses. That is total off-topic. Then he tells about phase response; it is changed to off-axis, by reflections in the recording environment, and speakers maintaining phase to off-axis are difficult to design and reflections in listening room disturb more. Sure that happens. Finally he reveals that GD from 1.6 to 2 ms is audible, but then denies everything with wrong information that these numbers do not exist in normal hifi and monitor speakers. That's it, so Toole's contribution to this topic drops to zero in his book.
I think they wanted to imitate GD spectrum of few actual (Genelec) speakers to make sure that result correlates with actual (their own) products.
No need to repeat. I know that already.
i am poor with maths, so I have found step response the best way to look at phase/group delay of a loudspeaker (basically impulse response). The audibility of how much and at what frequency is interesting topic. And if change is soft or sharp? It will take some learning to identify and differentiate GD problems. Room will always add it's own major mask.
So, IMO GD characteristics is secondary to SPL response including suppression of resonance peaks and excessive distortion of individual drivers. Everything in a loudspeaker is compromised!
Here my IIR-dsp 4-way with All-LR2 vs all-LR4. Delays set for smoothest SPL response and step continuity. These are indoor REW measurements from different sessions. Floor reflection effect is seen in 200-300Hz range. Simulations from VCAD would look nicer!
So, IMO GD characteristics is secondary to SPL response including suppression of resonance peaks and excessive distortion of individual drivers. Everything in a loudspeaker is compromised!
Here my IIR-dsp 4-way with All-LR2 vs all-LR4. Delays set for smoothest SPL response and step continuity. These are indoor REW measurements from different sessions. Floor reflection effect is seen in 200-300Hz range. Simulations from VCAD would look nicer!
Last edited:
Try Envelope (ETC) some day. Y-scale in dBFS and smoothing 0.1 ms. It shows nicely both acoustical "noise" such as resonances, diffraction and reflections, and electrical delays in the same simple curve which is easy to read.
Fortunately yes, though nowadays we have better tools to control timing than few decades ago. For example, this is basically the same speaker as passive and active. Not exactly the same box but quite alike (passive had better shape acoustically).
Attachments
Yes, and music also has so much else going on that the masking effect is extremely high.
But anyway, we were talking about system group delays, in a very limited portion of the freq resp, not about some transients.
What I wanted to make very clear (to people), is that this paper doesn't answer anyway questions in that regard.
Only the audibility of certain transients, which is a little silly to investigate because that has already be known for a decade or two. edit: Even quoted by Toole himself; "Discrimination of Group Delay in Click-like Signals Presented via Headphones and Loudspeakers"
I don't have that 4.8.1 section in my copy of the book, probably newer edition?
I have to find it to be able to exactly comment on it.
Last edited:
Looks that you have completely lost the point. Hopefully "people" got it better.
I understand, but it helps to to get a grasp of the whole context.
Anyway, I managed to find something
The way it is written is a little open for interpretation.
But they are referring to GD's basically in the mid range (or better above 100Hz)
In that case, yes it's almost practically impossible to get such high relative group delays.
Reason being that a normal lowpass filter for example, will just give a constant GD over the whole pass-band.
So the only other way of getting any GD issues and jumps is by filters with an high Q or with param EQ peaks with an high Q and/or boost.
When the high-section is also just filtered on a normal and nice way, the relative GD difference between low and high is very small. (practical tip; a reason why it can be important to apply bafflestep correction to woofer as well as the tweeter)
So seen from that context, it's true that these high GD's don't really exists in loudspeaker systems.
Unless somebody decided to just implement a huge Q somewhere for some reason?
So I don't see nor read why Toole is wrong here at all?
For lower frequencies, he calls it a "special case".
But he's point of focus is mostly absolute phase issues.
Which I agree with, is that you don't have any control over at all.
What he does omit here is the contribution of the loudspeaker itself for "low frequencies".
So yes, that is left open a little bit, although he does add that there are many other factors to worry about.
You were saying that Toole was making certain claims, but that doesn't show in the text as soon as you understand the context and point of view it is written in.
I would rather like to keep the discussion constructive.
Misinterpretations, misunderstandings, explanations, translations and perspectives can be all part of it.
Also masking can interfere from little to none transients what at least I'm talking about. Hammer sound of piano and percussion comes clearly so it's effortless to evaluate by listener, Of course assuming that material is not fully compressed and limited modern mud.
Right, okay, that's true in a section of the music were only percussive sounds can be heard.
In that case it's still the question how much early reflections as well as late reflections are gonna ruin that whole "transient".
Because of the roll-off of a loudspeaker, there is always a rising group-delay on the low frequencies.
(because that is exactly what a roll-off does)
So we are talking about a very small frequency band were issues might occur.
That section is determined by the Q-factor of the system, with mostly a difference between 0.5 and 0.707.
That difference is quite small.
In that case I think that the difference in energy (dB's) around the Fb as well as the difference in roll-off is more audible that the GD itself. It's quite tricky to even get roughly the same -3dB point with different Q-factors.
In practice I have very rarely seen that between different speaker systems.
One more thing about "Toole et al". Those ex Harman fellows are continuously writing/talking about statistical preference of "people". Loyal followers quote their texts just like it would or should be valid for everyone. Talking about generalized preference is primarily general info and for manufactures trying to make universal and conventional products for majority. Higher goals and advanced - or just different speaker concepts are certainly allowed for any manufacturer and listener. As an individual with own ears and brain I find filtered preference very close to insult. Especially if someone on discussion forum starts to claim that some feature is insignificant or no one cannot hear or feel it. That is arrogant BS. We really do not have to have the same hearing capabilities and preferences neither in DIY nor commercial audio.