I don't belong to those ones who think that the system Q alone determines the acoustically relevant transient behaviour due to the following points:
a)
A given Q value alone doesn't define how long some resonance will ring due to the fact that the decay time is also inversely proportional to the resonant frequency.
b)
The relevance of transient behaviour is not the same for subbass frequencies as it is for the spectral content further up. Have a look a the following table and you can see that percussion instruments have content fom 50 Hz upwards.
The Only EQ Chart You'll Ever Need ? Audio Issues
I have enclosed a group -elay simulation for two closed-box speakers. The cyan one is for a speaker with a resonant frequency of 30 Hz and a Q value of 1 and the green one is for a resonant frequency of 60 Hz and a Q value of 0.5 (the often mentioned ideal regarding transient response). I bet the one with Q=1 sounds better on percussive music than the other one ,if we just take the speaker itself into consideration, because it has only half as much group delay distortion as the other one in the frequency range of interest.
Regards
Charles
a)
A given Q value alone doesn't define how long some resonance will ring due to the fact that the decay time is also inversely proportional to the resonant frequency.
b)
The relevance of transient behaviour is not the same for subbass frequencies as it is for the spectral content further up. Have a look a the following table and you can see that percussion instruments have content fom 50 Hz upwards.
The Only EQ Chart You'll Ever Need ? Audio Issues
I have enclosed a group -elay simulation for two closed-box speakers. The cyan one is for a speaker with a resonant frequency of 30 Hz and a Q value of 1 and the green one is for a resonant frequency of 60 Hz and a Q value of 0.5 (the often mentioned ideal regarding transient response). I bet the one with Q=1 sounds better on percussive music than the other one ,if we just take the speaker itself into consideration, because it has only half as much group delay distortion as the other one in the frequency range of interest.
Regards
Charles
My contribution #22 states that EQ applies to a local region.
No - a wavelength is a wavelength and here we are talking about relatively long ones.
No I am not
CABS is a "global" equalisation scheme that attempts to cancel wavefronts and therefore effect equalisation throughout a room. Google will help you find moire information on it.
I never said it was simple. My post clearly states it is relatively simple - and that is relative to applying group delay correction
I have again stated in this thread that this is one case where pre-echo is not audible (under normal listening circumstances). All the references of which I am aware concern pre-echo in different applications. And barring the qualification concerning delays when used with video playback, it can be a very nearly perfect solution. That does not mean some will not prefer minimum phase but that might also be because that is the nature of the world.
yes, it removes something you didn't need to.
Neither am I but it is a distinct and separable component of acoustically relevant factors.
That does not mean that Q is not an acoustically relevant measure for a specific frequency.
No it doesn't but this thread but this thread was concerning low frequencies specifically. It is also worth noting that you need first provide listening conditions devoid of room resonances in which to carry out assessments at low frequencies. When this is done, things get more revealing...
You are not comparing like with like and so you cannot make a conclusion regarding group delay audibility in this case. More importantly my posts have related specifically to compensating for the group delay, that is comparing linear phase to minimum (or ideally close to minimum) phase system roll-offs. It is not the same experiment - real or just imagined...
Yes, of course. I think it is clear that higher order transfer functions (e.g. bass reflex) have more group delay distortion compared to lower order trnsfer functions. But that doies still leve the question open whether the differnece is relevant when used properly. I still stand by my comment that I prefer bass reflex for larger speakers that go low and closed box for smaller ones than don't go very low.
I did intentionally not compare twospeakers with the same resonance frequency in order to show that depending on the application a higher Q transfer function can have its merits.
You do not really want me to make a reflex to closed-box comparison simulation, do you ?
... and my post have always related to the relevance of group-delay distortion caused by reflex tunings .....
Regards
Charles
My research showed it does make a difference and I have tried to make it clear that it can only be considered relevant when all factors are understood and attended to.
An opinion you are welcome to but not one that is commensurate with designing for minimally audible distortion - particularly when group delay is compensated for.
No because I have done the real thing.
But not in isolation and not with respect to delay-compensated speakers.
Pity you couldn't since it is not comparing like with like and will therefore never show any effect in isolation. Furthermore Q has no meaning in a linear phase system and the results are not then relevant to the main thrust of the work I have described here.
I've been scanning your (soundbloke's) posts with unremitted scepticism. So I'd like to ask if this "research" is by ear or by objective testing and can you describe it or post your data?
B.
My research stems from ongoing work that began decades ago with a PhD study into low frequency sound reproduction and the application of DSP. Some of this work is in the public domain but some is work carried out commercially and has yet to be published.
The work has involved subjective testing but has not encompassed the large number of test subjects that would render it scientifically acceptable. Hence I indicate wherever I publish in this forum my learned impression and urge readers to experiment for themselves.
The subjects I have contributed to in this forum are always where available references are lacking - motional feedback, loudspeaker and room equalisation are just such subjects. I welcome your scepticism very much and hope that you will welcome the invitation to investigate the work I choose to make public here.
With the delay you can compensate in the same way as you would change the position of the respective speakers closer or farther. The change in the (delayed) runtime of the sound wave you can acommodate for a better phase transition at the crossover point between the different speakers. It's not that easy thoug because the phase change isn't linear and often diverts between different kind of speakers greatly, which means, the phase may be more or less correct over the complete crossover/overlap range or only in a very narrow part of the range. To compensate for that, you often have to cross over the different ways asymmetrically and/or with different crossover frequencies, i.e. 100Hz, 24dB/oct L/R and 130Hz, 18dB/oct Butterworth with a delay of some ms. Most of the times such a setup can't be done automatically, you mostly have to do it yourself.
With a DSP you can delay an output channel (sub/mid/high etc.) or a part of the frequency band. The problem is, with each filter and filter order you add further to the group delay, the steeper, the more. That can be avoided by FIR filters but these have other disadvantages, i.e. the pre-ringing, which distorts the signal form at the very important first impulse. Not all DSPs can manage FIR filters, most can't, only the more expensive ones do. And, it delays the whole signal and it takes a lot of processing power. If you want to equalize the phase and group delay perfectly, you'll therefore need a lot of filters and that means your DSP runs out of processing power or becomes quickly very expensive. Besides that, it takes a lot of work to adjust the group delay within one channel, most DSPs can't do a multi-frequency delay on each frequency on one channel.
You don't seem to understand different room dimensions create different modes/resonances. With more different dimensions (i.e. L-shaped room) you get more different resonances which can also create interferences. With some of these you can get peaks and dips in very close proximity. And you can't cancel that out at all because it's inherent in the room geometry.
This one demonstrates you don't understand the magnitude and importance of group delay, which is audible and what not. Well, I will save my time to reply to the other things, I'll just say one more thing: You can't compensate everything with a DSP. I've done my fair part on DSPs, I use one myself but you seem to have to learn that there are things you can't achieve with a DSP. You can do a lot and it is really amazing but there are things that can't be done in a DSP.
TYVM!
Gee, look at the link, just as I said, it already reads in the title that only works in rectangular listening rooms - just as I said.
Well, please try to understand under which circumstances these researches really apply. That is by far the most important factor in the usability of any research and calculations. It might be working in your room, in room xy but it does not apply to any room.
No, I understand the subject very well. You are misunderstanding the nature of the global-like equalisation systems and the behaviour of discrete room modes at low frequencies re local equalisation strategies. The phase changes you report are apparent higher up the frequency range where excess phase behaviour precludes simple local equalisation strategies - but certainly does not prevent more complex "matrix filter" solutions (although it returns an ever-decreasing area of equalisation where usefulness becomes questionable).
Bizarrely not a response posted in reply to my post, but this is the exact same conclusion I have already posted in this thread
But correcting group delay at the system roll-off is something that can be done and is one of the few areas where a linear phase solution afforded by DSP offers something over and above non-DSP audio system engineering.
CABS was given as an example, hence the words "for example". The CABS system is for rectangular rooms but active absorber solutions (of which CABS can be envisaged and even configured as a specific example), whether arrayed or operating as singular units, are not restricted to rectangular volumes.
Simply incorrect. Please try to understand the issues before commenting further.
Excel Tip:
Try a scatter plot style chart with logarithmic axes. You will like it.
Last edited:
TYVM!
Gee, look at the link, just as I said, it already reads in the title that only works in rectangular listening rooms - just as I said.
The link: http://vbn.aau.dk/files/62729248/LF_sound_field_control.pdf is a 7 page "Readers Digest version" of Adriano Celestinos original PhD document: http://vbn.aau.dk/files/12831869/AC-phd.pdf (150 pages).
In the PhD document one finds a lot more details and also that the CABS-concept is not restricted to rectangular rooms only. Chapter E (page 95 and onwards) shows his positive findings for the CABS-concept in irregular shaped / L-shaped rooms.
Last edited:
That is a. correct but b. requires multiple speakers while the topic here is not about multiple sound sources rather than group delay of closed vs. vented enclosures. The documentation of the CARBS concept shows it only delivers reasonable results in irregular rooms with 4 or even more speakers (p. 105 i.e.). That deviates a lot of what "phase_accurate" said, he didn't even mention a lot more subwoofers needed, instead he claimed it would be perfectly achievable with a DSP - which is simply not doable in non-rectangular rooms.
Simply correct. Please understand CARBS at all before claiming more 'facts'. And please understand that for multiple room modes (i.e. non-rectangular rooms) need more sound sources to compensate/correct the resonances. For L-shaped rooms you already need 4 subwoofers (or more) and 6 FIR ch. DSPs (or more).
No. At best you have misunderstood this thread. Technically you are at best misguided.
The thread started with a discussion of transient response and progressed to low frequency aspects...
Firstly there was a clear distinction made between the transient behaviour of single loudspeakers and the audibility of different low-frequency alignments due to Qts.
Secondly there was the further addition of the subject of group delay compensation and the application of linear phase roll-offs (where the discussion of Qts is irrelevant).
The latter (and possibly in certain cases the former) likely requires room equalisation to be properly auditioned and assessed - hence the introduction of it. Both (intentionally) localised schemes and globalised schemes were then mentioned.
Typically the frequency range of interest - that is around the system roll-off - occurs in the region where room modes are sparse and possibly in their fundamental manifestation. Here non-minimum phase behaviour is unlikely and room geometry is unimportant. Sufficient coupling to a given mode is also guaranteed by the very nature of a single loudspeaker driving it in the first place. To all intents and purposes the resulting equalisation will be global for such an example.
The discussion did not exclude multiple loudspeaker compensation schemes at any time - although a minimum of two is implied - and the results can be assumed to be improved if for no other reason than relaxing the demand for adequate volume velocity. The CABS scheme employs multiple loudspeakers always regardless of the room geometry. Also global-like equalisation strategies need not employ any FIR (or other) filters at all barring low-pass filters to curtail the operating bandwidth: CABS was just given as an example...
My contribution has also cited the use of matrix filtering that can compensate for non-minimum phase behaviour under certain conditions and therefore extend the usefulness of the equalisation somewhat regardless of room geometry. This requires a minimum of four loudspeakers and 16 FIR filters - although these filters do not impose any great burden as they can be implemented more than adequately by sub-sampling or sparse filtering to great effect.
I request politely once again that you invest some time in reading before making further contributions.
Duh... what's the problem? Do what it takes, isn't that the DIY spirit?
You seem focussed to bring up hurdles, while all we have to seek are solutions. There are solutions, you know. First pick a speaker topology that actually has a shot of working with your room/space. This is DIY. We get to choose whatever we want to build. That means to me: you're in charge of the solution that fits your specific needs and wants.
So who cares if it takes more drivers, beside budget worries. Save up a little longer and go for it.
Why this focus on impossibilities. A lot more is possible once you set your mind to it. Start by reading up on it. Don't stop reading only to start worrying. That will get you nowhere. Think in solutions, not in problems.
One more note: pre-ringing does not have to be a worry when using FIR filters. I've heard it in real life and I've concurred it as a problem. It should not stop you from trying. Put some trust in a measurement setup and go learn.
But I'd urge you to read up what's possible first.
I won't tell you it's easy, but it isn't impossible either.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.