Hi Oabeieo,
If you ask me one should never linearize phase for systems stopbands, they are natural because amplitude falls off then phase also starts it turning (90º per order) and so behave natural sounds in real world inviroment, you free to do it but it should not sound natural and more or less ruin original timing in music (harmonics line up).
So if you ask me only linearize that LR4@80Hz filter for your subwoofer but as we talked about in past remember also add stopband from subwoofer to HP section of your midbass bandpass and linearize for LR4@80Hz filter but not linerize that @45Hz natural system stopband of Q0,83 2nd order butterworth, difference for slope is not small seen in below where red is a clean linear phase filter HP LR4@80Hz and blue is the cascaded linear phase LR4@80Hz filter plus that @45Hz stopband minimum phase filter of Q0,83 2nd order butterworth : )
Hard thing if you ask me is get a system with a natural HP stopband of @45Hz Q0,83 2nd order butterworth to sound natural when inside a real room, it should maybe sound exelent outdoor or out in space but inside a real room response most often gets a huge bit hotter and ragged as in below example for raw room gain in red curve and blue is that red curve times a system having a HP stopband of @45Hz Q0,83 2nd order butterworth. In my experience blue curve below is really much too hot ragged and unnatural, can one get it smoother and high passed somewhere around 20Hz point then sound quality will improve to natural.
If you ask me one should never linearize phase for systems stopbands, they are natural because amplitude falls off then phase also starts it turning (90º per order) and so behave natural sounds in real world inviroment, you free to do it but it should not sound natural and more or less ruin original timing in music (harmonics line up).
So if you ask me only linearize that LR4@80Hz filter for your subwoofer but as we talked about in past remember also add stopband from subwoofer to HP section of your midbass bandpass and linearize for LR4@80Hz filter but not linerize that @45Hz natural system stopband of Q0,83 2nd order butterworth, difference for slope is not small seen in below where red is a clean linear phase filter HP LR4@80Hz and blue is the cascaded linear phase LR4@80Hz filter plus that @45Hz stopband minimum phase filter of Q0,83 2nd order butterworth : )
Hard thing if you ask me is get a system with a natural HP stopband of @45Hz Q0,83 2nd order butterworth to sound natural when inside a real room, it should maybe sound exelent outdoor or out in space but inside a real room response most often gets a huge bit hotter and ragged as in below example for raw room gain in red curve and blue is that red curve times a system having a HP stopband of @45Hz Q0,83 2nd order butterworth. In my experience blue curve below is really much too hot ragged and unnatural, can one get it smoother and high passed somewhere around 20Hz point then sound quality will improve to natural.Attachments
Last edited:
Hi Oabeieo,
If you ask me one should never linearize phase for systems stopbands, they are natural because amplitude falls off then phase also starts it turning (90º per order) and so behave natural sounds in real world inviroment, you free to do it but it should not sound natural and more or less ruin original timing in music (harmonics line up).
So if you ask me only linearize that LR4@80Hz filter for your subwoofer but as we talked about in past remember also add stopband from subwoofer to HP section of your midbass bandpass and linearize for LR4@80Hz filter but not linerize that @45Hz natural system stopband of Q0,83 2nd order butterworth, difference for slope is not small seen in below where red is a clean linear phase filter HP LR4@80Hz and blue is the cascaded linear phase LR4@80Hz filter plus that @45Hz stopband minimum phase filter of Q0,83 2nd order butterworth : )
Thank you BYRTT! That makes sense. So the closed box linearization, from what your saying maybe should be used with prudence, depending on the environment. Your room gain charts really help at a lot, and in a car the cabin gain is upwards of +25db at 38hz. So even more intense of a room.
The in room graph says a lot. And I do agree that using one linearization (weather it be the box linearization or the filter linearization) sound the best but wasn’t sure, the amount of room gain I think is screwing me up.
Thanks a ton
I would favor tools from the "filter linearization" tab over phase EQ, when possible.
Also make sure polarity is right beforehand (or does REW automatically handle this when deducing excess phase? I must confess I have never used that feature.)
Taking proper phase measurement is not too different from taking proper magnitude measurements: most of the time you will need several measurements with different mic distances and impulse gating for different frequency ranges.
Smart windowing techniques might help you make the most of a single measurement, but IMHO taking multiple specific measurements always leads to better results.
Hi BYRTT, Hi POS,
I don't get why you guys say this.
It seems to me, the recording already possesses any natural phase rolloffs (or worse depending on mastering hpf, xovers, etc)....
.....and by not linearizing the stopbands we are increasing natural order rolloffs ..???
Would truly love to hear your thoughts here... thx, mark
This is exactly my thinking as to why I asked , I was looking at both arguments and couldn’t exactly tell what is proper.
Not doing the linearization for the stop band does sound more natural as far as how the speaker sounds like in the case of my sealed box. But adding it also has a good sound as well, adding it almost to me gives me the impression that the bass is “sped up” or moved ahead in time more. Knowing the casualty of the filters used I assumed it was correct. But I didn’t know what is actually right.
And like byrtt was suggested, like my rolloff starts naturally at 45hz but in car responce is mostly flat and the actual f3 is around 16hz .......
So that lead me to ask, where really is the “natural spot”? 45 or 16? And should I be even doing it at all and does my filter linearization take care of box linearization if they are close simultaneously
Last edited:
When I wrote "Phase EQ" I was trying to say a filter that changes phase without affecting amplitude
The polarity should be correct as it is whatever phase deviates from the minimum phase, making that approach 0 should get you back to minimum phase when applied to the original
The idea of the tutorial seems to be aimed at an overall room based correction/target on an already finished speaker
The spatial averaging from using multiple measurements will reduce the room phase hash quite significantly and go a long way to showing speaker response with persistent room influences added
I think the tutorial was suggesting a 15 cycle Frequency dependant window and 1/6 octave smoothing
The more you smooth and window the the measurement the less chance there will be to make the correction too position dependant
It depends on what you are trying to achieve as to which is the best way to view the response
When I wrote "Phase EQ" I was trying to say a filter that changes phase without affecting amplitude
The polarity should be correct as it is whatever phase deviates from the minimum phase, making that approach 0 should get you back to minimum phase when applied to the original
The idea of the tutorial seems to be aimed at an overall room based correction/target on an already finished speaker
The spatial averaging from using multiple measurements will reduce the room phase hash quite significantly and go a long way to showing speaker response with persistent room influences added
I think the tutorial was suggesting a 15 cycle Frequency dependant window and 1/6 octave smoothing
The more you smooth and window the the measurement the less chance there will be to make the correction too position dependant
It depends on what you are trying to achieve as to which is the best way to view the response
Thank you, and thank Pos for your replies.
SwissBear guide is aimed to room correction, correct? That means that If I do this guide, I should not correct each speaker additionally? Or should I start with a "corrected" speaker?
All my questions are considering a finished speaker with a passive crossover, I am still too far away from active XO.
...relax
i did write one is free about it ...A simple answer is because then i can get sound from recordings to sound as was it the analog happening itself, whenever i allow any excess phase to whatever audio band performance it can sound good but always then it sounds as a recording flavoured with a more or less modern kind of synthetic signature add on, guess having varius numbers of excess phase didn't happen in the old days before any dac or modern compensated amps was out, at that time excess phase was only to find as a cosequence in multiway speakers from their summing XO points and those we can more or less linearize nowadays with up to very good or fully transperent results.
Now i hate above is nothing but my own subjective conclusion on stuff but think a logic point is if we manipulate natural physics we get a effect proccessor and effects belong to recording or mastering process not to a reproduce chain that should be as natural as possible a reference system.
Deviation from my view is fine and admit real world it could be hard agree my point probably because multi way speaker systems acoustic sum and possibility for more or less distorting errors can be a complicated animal in itself and also the varius temporary indoor enviroment we benchmark in to establish our subjuctive conclusion is also not easy to make precise controlled from each trial setup to the next trial setup, what helped me most get a feel and probably better conclusion on phase is execute various phase distortons or tweaks into a head phone domain and Rephase have been the fantastic tool to do these exercices thanks to pos.
Sorry if I didn't sound relaxed...cause i really am...was just my way of saying I don't understand
And thanks for the reply.
I can see why we can get used to the sound of stopband rolloff and prefer to keep it.
I guess i do see that as a preference...a preference for a certain level of group delay.
I'd think logically, there is already group delay in the recording, and then we add some more with playback driver's natural rolloff (and crossover in systems not using linear phase xovers).
So two stages of group delay..recording and playback. And if we get used to that, we may well want to keep it all.
My question is: technically, why would we want to double up on final net group delay??? (Assuming I'm correct in thinking there are two stages introduced)
And thanks for the reply.
I can see why we can get used to the sound of stopband rolloff and prefer to keep it.
I guess i do see that as a preference...a preference for a certain level of group delay.
I'd think logically, there is already group delay in the recording, and then we add some more with playback driver's natural rolloff (and crossover in systems not using linear phase xovers).
So two stages of group delay..recording and playback. And if we get used to that, we may well want to keep it all.
My question is: technically, why would we want to double up on final net group delay??? (Assuming I'm correct in thinking there are two stages introduced)
POS, the CROSSOVER section of the "Filters Linearization" tab appears to work well for my LR96 XO's, but I have a couple of additional questions about the "Filters Linearization" tab.
1) Can it address an internal mid/tweeter asymmetric 6/18dB XO (if so, how) ?
2) Are there any specific settings for OB/Dipole speakers. I just see CROSSOVER, BOX(closed, vented) and SUBSONIC sections.
TIA
Last edited:
Sorry if I didn't sound relaxed...cause i really am...was just my way of saying I don't understand
And thanks for the reply.
I can see why we can get used to the sound of stopband rolloff and prefer to keep it.
I guess i do see that as a preference...a preference for a certain level of group delay.
I'd think logically, there is already group delay in the recording, and then we add some more with playback driver's natural rolloff (and crossover in systems not using linear phase xovers).
So two stages of group delay..recording and playback. And if we get used to that, we may well want to keep it all.
My question is: technically, why would we want to double up on final net group delay??? (Assuming I'm correct in thinking there are two stages introduced)
Maybe because we dont want to improve the group delay or timing isolated on its own is because the physical amplitude signal really isnt there and we then end break physics for timing in natural sounds, know its too simplistic but take below theoretical example from a recording chain, red curve is raw bandwidth of a acoustic bass guitar with 1st order stop bands at 41Hz and 7kHz, and if we feed or say cascade that bandwidth thru a very good microphone bandwidth with 1st order stop bands at 6Hz and 20kHz we end up a bit limited domain in the blue curve. Amplitude and timing (phase) for that new blue curve is what the other musicians plus mix and master process will base their cooperation or work on and therefor will guess if we change blue curves original blue phase to be the red phase we add a timing distortion because the wider amplitude performance of red curve we never get back because of the cascaded chain.
Attachments
Here are my "initial" (after a few tries) results using REW's AutoEQ (cuts only, not boosts) and RePhase's "Filters Linearization" tabs. I had to estimate some XO's and then slightly change them to get the predicted phase plots to bend accordingly. I did not use the phase sliders, only the "Filters Linearization" tab.
These plots are of a stereo 3-way active XO to integrate stereo OB/dipole subs, bi-amp planar/ribbon speakers and to correct their passive internal mid/tweeter XO.
Constructive critique, criticism and suggestions is very much welcomed.
My main question is the left side (blue line) of the after Step Response. Is this downward dip before 0 expected/correct ???
Here is the before(green) and after(blue) frequency response with "cuts only".
Here is the before(diagonal/blue) and after(horizontal) phase correction for the left, right and combined channels.
Here is the before(green) and after(blue) Impulse Response.
Here is the before(green) and after(blue) Step Response.
Here is the before(green) and after(blue) Group Delay.
Here is the before(green) and after(blue) Impulse Response reflection peaks.
These plots are of a stereo 3-way active XO to integrate stereo OB/dipole subs, bi-amp planar/ribbon speakers and to correct their passive internal mid/tweeter XO.
Constructive critique, criticism and suggestions is very much welcomed.
My main question is the left side (blue line) of the after Step Response. Is this downward dip before 0 expected/correct ???
Here is the before(green) and after(blue) frequency response with "cuts only".
Here is the before(diagonal/blue) and after(horizontal) phase correction for the left, right and combined channels.
Here is the before(green) and after(blue) Impulse Response.
Here is the before(green) and after(blue) Step Response.
Here is the before(green) and after(blue) Group Delay.
Here is the before(green) and after(blue) Impulse Response reflection peaks.
Last edited:
Here are my "initial" (after a few tries) results using REW's AutoEQ (cuts only, not boosts) and RePhase's "Filters Linearization" tabs. I had to estimate some XO's and then slightly change them to get the predicted phase plots to bend accordingly. I did not use the phase sliders, only the "Filters Linearization" tab.
These plots are of a stereo 3-way active XO to integrate stereo OB/dipole subs, bi-amp planar/ribbon speakers and to correct their passive internal mid/tweeter XO.
Constructive critique, criticism and suggestions is very much welcomed.
My main question is the left side (blue line) of the after Step Response. Is this downward dip before 0 expected/correct ???
Here is the before(green) and after(blue) frequency response with "cuts only".
Here is the before(diagonal/blue) and after(horizontal) phase correction for the left, right and combined channels.
Here is the before(green) and after(blue) Impulse Response.
Here is the before(green) and after(blue) Step Response.
Here is the before(green) and after(blue) Group Delay.
Here is the before(green) and after(blue) Impulse Response reflection peaks.
Looks like we have another expert gang,
Okay I’m jel
Looks like we have another expert gang,
Okay I’m jel
If that is sarcasm, please let me know what I did wrong.
I am trying to learn.
No! No!
Not at all. That looks great. Believe me buddy I’m more a novice than you, although I can promise you my room is definitely more complex to work in (a car) which is where I get stuck on a lot of things, but that looks great! It’s better than what I can do on my own.
Last edited:
For main quistion will say that is normal and expected because when we manipulate some typical excess phase lag in time domain it has to be a pre operation view into impulse or step response graphs that would cost some overall processing systen lag to repair.
In general looks you have a good feel on stuff
As some general tip for where you are now say everything is really perfect based, then remember than any wish for other house curve adjustment or left verse right channel calibrations has to be global adjustments. What i mean is stay away any house curve adjustments per selective band pass unlesh you really happen find some errors there, because even small EQ changes per pass band will often or probably need a new time allignment setting and that operation can often be a big workload.
Those graphs look good
Remember to listen to the correction and see if you like it, I have made many graphs that look great and sound terrible. So you have an active crossover between dipole subs and main speakers with an unknown passive crossover?
As Byrrt said the change in the step response before the peak and the increased preringing in the impulse is the result of the phase manipulation. You have gone from 1700 degrees of phase turn to 360 degrees 20-20k so something has to give.
Are these in room averaged listening positions measurements like in SwissBears tutorial?
If so a flat room response will tend to be too bright. You could try adding in a room curve in REW to see how you like it. By using minimum phase EQ at that stage you should not upset the rest of the phase correction too much as you only used filters linearization,
Here is an example to start with
REW settings to get that curve, change the target level to suit your measurement
