Hi Jean-Michal,
Linear phase has several benefits. You'll find them on page 19 and 20 of this paper:
http://www.bodziosoftware.com.au/Attributes_Of_Linear_Phase_Loudspeakers.pdf
If you have time, I would suggest reading the whole paper.
Best Regards,
Bohdan
as shown in post #415 the total energy of the ringing (and intermodulation it can produce) is the same for a linear and a minimal phase filter: it is just a matter of where around the peak it appear.
(red=minimal-phase, green=linear-phase)
(red=minimal-phase, green=linear-phase)
An externally hosted image should be here but it was not working when we last tested it.
Hello Bohdan,
Be sure I am convinced that linearized phase audio system have their benefits and even if it seems so, I don't want to dimiss phase linearization where it provides its benefits.
That's only for the lower frequencies ( let's say less than 150Hz) for which I think that the audible benefits of phase linearisation are questionable...
But I dont want to spoil that thread, so I'll undesrtand if the discussion avoid that point.
Best regards from Paris, France
Jean-Michel Le Cléac'h
Gentlemen please, lets keep this thread going with a positive attitude (if not purely on topic...)
I have no idea where the noise Bohdan spoke about could come from, and I (or him, or you for that instance) would be hard pressed to affirm that it comes from preringing, either directly hear or through some supposed intermodulation...
I can understand that preringing could possibility be heard directly (but probably not at the frequency of a BR, or at the Nyquist frequency...), but by its intermodulation effect?...
If that was the case then postringing effects of a minimal-phase filter (such as a BR) would also equally be heard, using those pulse test signals.
I have never heard of such a thing. A BR is a 4th order filter, the linearization of which provoke very few preringings (very hard to detect on a IR, contrary to the very sharp LP filter shown above).
I have the feeling that this thread is already leaning toward off topic territories when it deals with linear-phase artifacts audibility, as it is a consequence of a certain use of the tool presented here, for which this thread is more a how-to and announcement thread than a discussion thread. But here you are not even speaking about the use of this particular tool...
If you absolutely feel the need to discuss these subjects of phase linearization and the supposed audibility of their consequences here, please do so after having tried them for yourself using rePhase.
It is a very different approach compared to automated inverse-based techniques such as DRC-FIR, and gives you total control of the pre/post ringing that will possibly occur.
I have no idea where the noise Bohdan spoke about could come from, and I (or him, or you for that instance) would be hard pressed to affirm that it comes from preringing, either directly hear or through some supposed intermodulation...
I can understand that preringing could possibility be heard directly (but probably not at the frequency of a BR, or at the Nyquist frequency...), but by its intermodulation effect?...
If that was the case then postringing effects of a minimal-phase filter (such as a BR) would also equally be heard, using those pulse test signals.
I have never heard of such a thing. A BR is a 4th order filter, the linearization of which provoke very few preringings (very hard to detect on a IR, contrary to the very sharp LP filter shown above).
As already said, I would be really interested in hearing about your experiments with manual phase linearization (of crossovers and/or BR) using rePhase.
I have the feeling that this thread is already leaning toward off topic territories when it deals with linear-phase artifacts audibility, as it is a consequence of a certain use of the tool presented here, for which this thread is more a how-to and announcement thread than a discussion thread. But here you are not even speaking about the use of this particular tool...
If you absolutely feel the need to discuss these subjects of phase linearization and the supposed audibility of their consequences here, please do so after having tried them for yourself using rePhase.
It is a very different approach compared to automated inverse-based techniques such as DRC-FIR, and gives you total control of the pre/post ringing that will possibly occur.
Last edited:
I have heard good reports from people linearizing their subwoofer crossover down to 80Hz (already far lower than the Schroeder frequency in any reasonable private room) using rePhase. In fact it was most of the time the most audible crossover to linearize.
The point is: rePhase is just a tool that let you do those things (minimum or linear phase EQ, pure phase EQ, electrical/natural filters linearization, linear-phase filters of arbitrary slopes and shapes, ...), and convolution is just a way of applying any amplitude/phase modification you want (minimum-phase being also possible...).
Nothing fancy really, and certainly no embedded dogma...
Hi Jean-Michael,
Thank you for your comments.
Yes, you are correct, the benefits of linear-phase performance can not be dismissed. In my tests, the linear-phase subwoofer was a clear winner.
What specific linear phase tests have you conducted?.
Please describe the equipment, software used, test signals, documented test results and performance curves of your equipment, including EQs.
And most importantly, please provide acoustical test results of a square wave of several frequencies, played by your speakers.
Best Regards,
Bohdan
Agreed, but please, pretty please, if those tests are not directly related to rePhase, publish them in another thread.
No offense meant, but I would really like to keep this thread on topic, that is a reference thread to help users find information about rePhase (which has no documentation outside of forum threads...) rather than general assumptions and experiences about linear-phase speakers
Hi Pos,
No problem.
The title of this thread is “rePhase, loudspeaker phase linearization, EQ and FIR filtering tool”. Since we are talking about loudspeaker phase linearization, therefore this conversation sits very well within the title and scope of this thread.
This conversation is not off-topic.
It would only be fair and courteous to give Jean-Michel and opportunity to explain his “hands-on” involvement in linear-phase loudspeaker testing, rather than cutting him off, half-way through the sentence.
Best Regards,
Bohdan
I think I get his meaning. He left out a single letter in the topic that would make it more clear. The title he intended, I suspect, should have been
“rePhase, a loudspeaker phase linearization, EQ and FIR filtering tool”.
My impression is that he's attempting to focus on his tool, not the topics in general.
Dave
Hello Bohdan,
I never spoke about rephase itself but about the few experiments abaout phase linear filtering, phase linearization I am not originator of, but to which I attended. (During the European Triode Festival or during a meeting of the asssociation Melaudia I am member of which, during private listening sessions, etc.).
A friend of mine gave me the IR of his phase linearized system which show some long very low frequency pre-ringing. I'll see if I can share that IR.
But my own experience on such subject is not zero as few seems to suggest, as I was probably in France one of the first to experiments with the tools developped by Angelo Farina long time ago (convolution, Kirkeby, inverse filtering...) .
My opinion is that the benefits of phase linearization is more audible with bad loudspeakers and crossovers. With good one I don't hear so much improvment... I only post my opinion here as some food for further thoughts. Do what you want of it.
Finally I would say that we can do a parallel between phase linearization and directivity control. They both provide some benefits, but there is so many and more basics improvments of an audio system that can lead to greater benefits and that should be done before ...
Best regards from Paris.
Jean-Michel Le Cléac'h
Already posted in this thread
loudspeakers with passive xover linearized with rePhase.(at 4 feets)
measured at listening point (10 feets).
loudspeakers with passive xover linearized with rePhase.(at 4 feets)
measured at listening point (10 feets).
An externally hosted image should be here but it was not working when we last tested it.
Hi Dave,
I thought the "tool" word at the end (without an "s") made the sentence clear enough, but it appears it was not the case (language barrier, sorry...).
Thanks for that. I edited the title in the first post, but it looks like the title thread did not change (yet). I will let some time for the server to reindex the title thread and will eventually contact a moderator to get it modified if this does not work automatically...
I am hacking on my ADSP-21469. Currently I am trying to get the FIR accelerator to work, I will be using it for downsampling. In order to get some filter parameters i downloaded rePhase.
Can I use the output as FIR coefficients?
Is low-pass/reject high good enough for anti-aliasing?
It would be nice if you added "c array" as an output format option. Also a "reverse c array" would be nice since the FIR code wants the coefficients in reverse order.
Can I use the output as FIR coefficients?
Is low-pass/reject high good enough for anti-aliasing?
It would be nice if you added "c array" as an output format option. Also a "reverse c array" would be nice since the FIR code wants the coefficients in reverse order.
Hi
For anti-aliasing with a minimum number of taps (keeping power for actual processing) you should try the steepest possible filter (eg 1000dB/oct LR), a nuttall window (best ripple rejection most of the time), see what you get for a given (and small) number of taps, and lower the cutoff frequency until you get a decent rejection at targeted Nyquist, and a flat enough response within the intended range.
In fact in this case it is the number of taps that will dictate the slope.
Of course the larger the number of taps, the better (better rejection, and flatter inband amplitude)
Here are some simulations: (in french sorry for that, but the screenshots alone should be self-explanatory)
Le sujet Najda: une carte DSP pour applications HP - Page 4 du forum Correction active et logiciels de mesure sur Homecinema-fr.com - 30034920 - 1291
Regarding the output format, the closest to a C array that you can use will probably be the txt format. You should be able to transform that in a C array by automatically turning the carriage returns into commas with a good text editor.
I will add the C array (and possibly the reverse one, if it is a common practice... even if that feels a bit strange??) in the next version, but I don't know when I will release it yet.
For anti-aliasing with a minimum number of taps (keeping power for actual processing) you should try the steepest possible filter (eg 1000dB/oct LR), a nuttall window (best ripple rejection most of the time), see what you get for a given (and small) number of taps, and lower the cutoff frequency until you get a decent rejection at targeted Nyquist, and a flat enough response within the intended range.
In fact in this case it is the number of taps that will dictate the slope.
Of course the larger the number of taps, the better (better rejection, and flatter inband amplitude)
Here are some simulations: (in french sorry for that, but the screenshots alone should be self-explanatory)
Le sujet Najda: une carte DSP pour applications HP - Page 4 du forum Correction active et logiciels de mesure sur Homecinema-fr.com - 30034920 - 1291
Regarding the output format, the closest to a C array that you can use will probably be the txt format. You should be able to transform that in a C array by automatically turning the carriage returns into commas with a good text editor.
I will add the C array (and possibly the reverse one, if it is a common practice... even if that feels a bit strange??) in the next version, but I don't know when I will release it yet.
Last edited: