What would you do for a system HP if openDRC had 16k taps? (at 48kHz).
Linearize or not, and if linearize, linearize exactly what?
Or say even 65k taps, ala PC?
My interest here is purely in terms of avoiding pre-ringing, and unnatural corrections.
Your vision on this has to be much deeper than mine
I second that motion...it seems optimization is the last frontier after you've spent your money and assembled what is to be your dream system...
I'm still grasping the nomenclature...I think the amount of taps limit how steep of a slope you can create? I really don't know....and then you guys throw in sampling rate and I'm thinking....do minimum phase eq's have resolutions and if no why not? Some products claim to not create pre ringing and unnatural corrections....it appears that FIR is perfection of all things (letting go of the delay increase) and that for a person like myself who is subscribed to the one man show with the equilateral triangle....and for everyone else who is normal (lol)....outside of delay....with proper technique....things could get seriously accurate.
Please pour into our cup ye wise POS =) (and add auto flatten to the phase eq pla-eeaasse lmao!!!!)
Reading this thread for more than the last few pages or 20 seconds with a search engine would have given you this information but to save you that effort there is a reasonable explanation here in the digital FIR filter properties section
http://www.minidsp.com/images/documents/fir_filter_for_audio_practitioners.pdf
We’re not as good at keywords as you I suppose
I’ve looked and read everything that wasn’t complete Chinese to me.
I always love a good read.
Thanks for the link! I enjoyed it very much!
I missed the sarcasm. Lol I’m very serious person and technical and sometimes miss the innuendos of life as it passes me by.
I always love a good technical paper or white paper that isn’t completely 1960s I really wish there was a lot more papers on FIR that don’t go into the mathematics of coefficients and all the DSP aspect of things.
It would be really cool if Thomas would do a white paper and submit it to the audio engineering Society. And in it a full blown explanation of FIR filters in modern DSP multi-way. I honestly think that he would make a good contribution the house value, just wait in the next 15 years there’s a lot of things that are going to be linear phase coming out OEM.
But I think that all depends on guys like us and they consumers to get **** out there.....
All I’m seeing online is that it’s a useful tool but not necessary blah blah blah blah when in fact when it comes to low frequency audio reproduction it’s a goddamn necessity!
I always love a good technical paper or white paper that isn’t completely 1960s I really wish there was a lot more papers on FIR that don’t go into the mathematics of coefficients and all the DSP aspect of things.
It would be really cool if Thomas would do a white paper and submit it to the audio engineering Society. And in it a full blown explanation of FIR filters in modern DSP multi-way. I honestly think that he would make a good contribution the house value, just wait in the next 15 years there’s a lot of things that are going to be linear phase coming out OEM.
But I think that all depends on guys like us and they consumers to get **** out there.....
All I’m seeing online is that it’s a useful tool but not necessary blah blah blah blah when in fact when it comes to low frequency audio reproduction it’s a goddamn necessity!
My apologies for the delay. I am afraid I don't have that much wisdom to share on the matter
It is all about preference and practicality, and both seem to be personal things...
If you want to keep an objective approach then you have to consider if you want to be true to the acoustical source, in which case you might want to linearize the HP, or if you want to be true to the engineer/artists decisions, in which case you'd probably want to keep it minimum-phase to better reflect what they actually heard and made decisions upon.
Of course this is a simplification, and some might use closed box designs, or simply a much lower HP frequency...
It is all about preference and practicality, and both seem to be personal things...
If you want to keep an objective approach then you have to consider if you want to be true to the acoustical source, in which case you might want to linearize the HP, or if you want to be true to the engineer/artists decisions, in which case you'd probably want to keep it minimum-phase to better reflect what they actually heard and made decisions upon.
Of course this is a simplification, and some might use closed box designs, or simply a much lower HP frequency...
Thank you !
Your thoughts seem to match mine.
I sometimes get torn between wanting to be true to the acoustical source, and wanting to replicate what engineers/artists heard.
But then i think about how impossible it is to replicate all the various studio bottom end group delay curves.... from all the different high-pass (and low pass) filters in play.
So i linearize the HPF as best I can for the sub in use, and leave the sub's natural rolloff alone.
My personal compromise... (plus it ain't easy finding enough taps to do much otherwise Lol)
Your thoughts seem to match mine.
I sometimes get torn between wanting to be true to the acoustical source, and wanting to replicate what engineers/artists heard.
But then i think about how impossible it is to replicate all the various studio bottom end group delay curves.... from all the different high-pass (and low pass) filters in play.
So i linearize the HPF as best I can for the sub in use, and leave the sub's natural rolloff alone.
My personal compromise... (plus it ain't easy finding enough taps to do much otherwise Lol)
My apologies for the delay. I am afraid I don't have that much wisdom to share on the matter
It is all about preference and practicality, and both seem to be personal things...
If you want to keep an objective approach then you have to consider if you want to be true to the acoustical source, in which case you might want to linearize the HP, or if you want to be true to the engineer/artists decisions, in which case you'd probably want to keep it minimum-phase to better reflect what they actually heard and made decisions upon.
Of course this is a simplification, and some might use closed box designs, or simply a much lower HP frequency...
This isn’t not a generalization... at all!!!
The recording is key and what went into it...
Although...... in the manner of making “cherry pie out of dog doo” and making a system sound good where normally it wouldn’t with the use of DSP is a whole different aspect ..... just sayin (like in a car)
After skimming and searching through this thread I had a few very basic questions that I was hoping someone could answer.
1) If a speaker has a passive crossover already, how much of a benefit will there be using rePhase only to correct the phase issues? I know that the results may very, but I am wondering if there is a general consensus for the typical user / speaker.
2) My understanding is that rePhase can also equalize the speakers and many people talk about exporting it to JRiver. Can this be exported to Equalizer APO? Furthermore, if one was just wanting equalization, would there be a benefit in using rePhase over REWs generic equalization optimizer and then exporting that into Equalizer APO or a miniDSP?
Once again I just heard about rePhase and I certainly can see that it would be an invaluable tool if designing an active crossover or for general speaker design / DIY. I am just wondering how applicable it is to someone who has pre-assembled speakers who is looking into general EQ options.
Finally, if anyone could pass on any good resources / links for a first time user / beginner that would be greatly appreciated!
Thanks for any info!
1) If a speaker has a passive crossover already, how much of a benefit will there be using rePhase only to correct the phase issues? I know that the results may very, but I am wondering if there is a general consensus for the typical user / speaker.
2) My understanding is that rePhase can also equalize the speakers and many people talk about exporting it to JRiver. Can this be exported to Equalizer APO? Furthermore, if one was just wanting equalization, would there be a benefit in using rePhase over REWs generic equalization optimizer and then exporting that into Equalizer APO or a miniDSP?
Once again I just heard about rePhase and I certainly can see that it would be an invaluable tool if designing an active crossover or for general speaker design / DIY. I am just wondering how applicable it is to someone who has pre-assembled speakers who is looking into general EQ options.
Finally, if anyone could pass on any good resources / links for a first time user / beginner that would be greatly appreciated!
Thanks for any info!
johnp98,
1) If your speakers are well behaved (complementary and coherent crossovers over a wide angle) phase linearization should be fairly easy (mainly using the "filter linearization" tab, and typically one or two gentle phase EQs, if at all). If the crossover is not coherent then it is best left alone.
As for the benefits, it is up to you to try and decide
2) You can load the FIR directly in Equalizer APO, but then of course you wont be able to modify it directly from its interface like generic IIR filters.
For the second part of the question, if the question is IIR vs FIR for minimum-phase corrections, then I would say FIR if the convolution engine has enough taps for the task. Convolution is guarantied to work as expected 100% of the time, whereas IIR is implementation-dependent, both for the transfer function and for the potential quantization noises.
If the question is about automated (REW) vs manual (rePhase), then this is again a personal choice with pro and cons. Manual corrections can be seen as more time consuming, but most of the time and effort has to be spend on the measurement(s) and their post processing and analysis. That does not change with automated corrections like the one REW provide. Manual corrections give you more control on what you want to correct (eg not addressing specific artifacts, etc.), which can in fact simplify the measurement phase. You can also use both, as you can for example generate an automated correction in REW, export it as rePhase filter settings, and tweaks them in rePhase before generating the final correction in FIR...
As for tutorials, you will find some on the rePhase - Official Site - Free FIR filtering tool page.
1) If your speakers are well behaved (complementary and coherent crossovers over a wide angle) phase linearization should be fairly easy (mainly using the "filter linearization" tab, and typically one or two gentle phase EQs, if at all). If the crossover is not coherent then it is best left alone.
As for the benefits, it is up to you to try and decide
2) You can load the FIR directly in Equalizer APO, but then of course you wont be able to modify it directly from its interface like generic IIR filters.
For the second part of the question, if the question is IIR vs FIR for minimum-phase corrections, then I would say FIR if the convolution engine has enough taps for the task. Convolution is guarantied to work as expected 100% of the time, whereas IIR is implementation-dependent, both for the transfer function and for the potential quantization noises.
If the question is about automated (REW) vs manual (rePhase), then this is again a personal choice with pro and cons. Manual corrections can be seen as more time consuming, but most of the time and effort has to be spend on the measurement(s) and their post processing and analysis. That does not change with automated corrections like the one REW provide. Manual corrections give you more control on what you want to correct (eg not addressing specific artifacts, etc.), which can in fact simplify the measurement phase. You can also use both, as you can for example generate an automated correction in REW, export it as rePhase filter settings, and tweaks them in rePhase before generating the final correction in FIR...
As for tutorials, you will find some on the rePhase - Official Site - Free FIR filtering tool page.
Last edited:
When designing filters for EQ APO with rePhase and REW I find something strange.
Measured summary of bass and mid band SPL line depends from which one filter is inverted in rePhase General/Output Settings/polarity.
On first image below red is when both filters are not inverted, obviously responses are out of phase on 215 Hz..
Green line is when mid filter is inverted, blueish is when bass filter is inverted in rePhase.
On impulse graph is visible that impulse is not only inverted but has also little different shape in range -2ms to 0.
What can cause this?
Can some filter Impulse settings influence this?
Bass filter had 100 000 taps, mid 10 000, delay is aligned in EQ APO.
Measured summary of bass and mid band SPL line depends from which one filter is inverted in rePhase General/Output Settings/polarity.
On first image below red is when both filters are not inverted, obviously responses are out of phase on 215 Hz..
Green line is when mid filter is inverted, blueish is when bass filter is inverted in rePhase.
On impulse graph is visible that impulse is not only inverted but has also little different shape in range -2ms to 0.
What can cause this?
Can some filter Impulse settings influence this?
Bass filter had 100 000 taps, mid 10 000, delay is aligned in EQ APO.
Attachments
Would rephase ever have anything that can sorta do the correction with some assistance.
What would be cool is if there was a way to do the entire correction with a single filter that has the entire all pass and magnitude correction without the use of “eqs” or peak type filters handling the correction
Like a inverted correction that knows where not to invert and can separate all of it into fir and iir coefficients to be exported into a Minidsp or the like....
Will there ever be a invert switch that allows us to invert everything that should be inverted.
What would be cool is if there was a way to do the entire correction with a single filter that has the entire all pass and magnitude correction without the use of “eqs” or peak type filters handling the correction
Like a inverted correction that knows where not to invert and can separate all of it into fir and iir coefficients to be exported into a Minidsp or the like....
Will there ever be a invert switch that allows us to invert everything that should be inverted.
I think EQ APO buffers are not problem here, otherwise there must be all phases wrong, and impulses delayed With needed delays defined in EQ APO I get different bands with different number of FIR taps aligned as they must be. This problem I have only in left channel, on right channel all is OK.
But I can test same taps FIRs tomorrow, currently I already using middle centering.
That is probably not going to happen, sorry.
Automated corrections need to rely 100% on the measurements, which often implies a very specific measurement procedure that needs to be part of the whole process (think Diarc, Trinnov, etc.).
If you simply apply brute force phase inversion (or any other type of correction really) on an imported measurement you take the risk of correcting things that are just an artifact of the measurement itself (position-dependent issue, measurement post-processing, etc.).
Having some sort of "inverse measurement" (flat magnitude, flat phase, mini phase) on a given frequency range could probably be useful for some situations and to some users, but this would open the door to oh so many abuses and problems...
I already regret having added so much phase EQ points, with such a big amplitude, because I often see people using dozens of those to correct phase shifts, including phase shifts on top due to bad delay compensation, or sharp phase shifts due to comb filtering...
I also often see measurements with wrong polarity imported into rephase (with the well recognizable phase shape on top). That cannot be considered a valid base for any phase correction, let alone an automated one!
A properly measured loudspeaker (for a given frequency range) should produce a measurement that is easy to cleanup (gating and/or smoothing, preserving what is important, with correct polarity and delay compensation), the phase of which would in turn be easy to linearize using one or two "filter linearization" entries, and maybe a tad of phase EQ...
No need to automate that: this is as simple as it gets, and you can focus on the frequency range you know that particular measurement is valid in.
On the other hand what could (should?) be automatized are the level/delay/polarity settings when importing a measurement, with an optional "auto adjust" setting. That would need some work and could bring wrong results on weird measurements (but those are probably hopeless anyway), but that would potentially help to avoid some of the problems described above.
..
I absolutely love this response...
I’ve been thinking about it for hours now waiting to reply
So for a long time are used rephase crossovers whether linearization or linear phase crossovers... and then I used a Direc upstream to do everything else
I’ve been switching back-and-forth my sharcs going from open to direc making my own correction and using theirs back-and-forth and back-and-forth and there’s things I love about mine and there’s things I love about their‘s
I’ve read every single white paper Johannesen has put out. All about the mixed phase stuff end room correction etc etc ....
The further I get into using rephase for my room correction the more I’m finding a massive inclination and seeing what mixed phase corrections usefulness is all about.... I get it , and there right... there doing it the best way...
Their approach really is sound and solid..... and the unit circle and poles matter a great deal. Just to touch base on what you were saying about polaritys and etc.
I’ve been using your software for a few years now and I absolutely love it truly it really is very very good but I’m finding myself only using it to make crossovers and to get my rise times (with proper delay) all exactly the same and like you said with just a little bit of Phasee Q and then let him direc you the rest.
Because no matter how hard I try it seems to have some sort of better SQ in the end.... although like I said there’s some thing I love so much more by doing it all myself..... so I’m torn ...... still!
And I’ve been switching back-and-forth at least every month trying to figure out how do I get this to work just as good as that I’ve been looking at filter hose and looking at their approach but I already see flaws in their videos alone without purchasing a license
Soooo what you’ve done is absolutely amazing, but I think there’s a next step and I don’t know how to express it properly but I think you understand what I’m saying there’s got to be a way to allow us to manually do things the way that you want manually but also add that level of computation that gets us to that next level because I’m sorry less you’re an absolute genius like you and Johannessen we’re just not gonna get it and if there was some thing just even something that helps us get on the right track it would be so so amazing.
What is measurements were done from within rephase? Hint hint nudge nudge
Edit : and for god sake‘s dude you need to do a white paper you are absolutely brilliant you deserve proper recognition
Last edited: