That matches my understandings with regards to linear-phase pre-ringing potential.If you listen in an anechoic chamber, on the design axis, none of this matters but in a normal room where there are many reflections from all different angles where the HP + LP sum doesn't cancel preringing it obviously has to color the sound. Thus my personal conclusion after playing around with such systems, it's just not worth it. It creates more problems than it solves and just added complexity and additional processing to the system with no sense of improved sound quality. I just haven't been able to ever convince myself that if I start with a system that sounds good, global FIR phase correction makes it sound better.
I think anytime there is not acoustically complementary output from drivers spanning the same frequency range, pre-ring potential exists.
So off-axis lobing is a potential pre-ring source for crossover regions, even for those that are fully complementary on-axis.
That said however, I think it's pretty safe to exclude room reflections as a source of audible pre-ring simply due to their scattered attenuation vs direct sound.
I also strongly agree that a well designed, good sounding system, is very unlikely to benefit from global FIR.
And I think a poorly designed system is even LESS likely to benefit....due to all the potential pre-ring issues magnified in a poor acoustic design...
So in other words, to heck with global FIR 😛
Third agreement 🙂Anyway, it's fun to play around with this stuff but I don't think you it should be taken to seriously.
I have fun working with DSP, and making measurements,.... almost as much as physical speaker construction.
But I'm a strong advocate of linear-phase, and not really because i'm convinced linear-phase sounds audibly better.
I think linear-phase simply makes it much easier to get outstanding measurement results.
I am totally convinced that flat mag and phase quite simply achieve all the typical goals beyond flat frequency response,....like transient perfect response, minimizing group delay, etc. (obviously excluding goals regarding distortion or non-linear response)
Flat mag and phase simply equals technically correct, imo.....and sounds quite awesome to me. Why do anything else if the tools are available.
I really wish more folks would try linear-phase FIR on a driver by driver basis...it's sooo dang easy and works sooo dang well.
Especially when using steep xovers....pre-ring potential thru xover regions get minimized to the point of being immaterial...
Anyway, I've been up on this soap box too many times...enough said....
Mark that statement pushed me to say I agree, and look here to see my results -I think linear-phase simply makes it much easier to get outstanding measurement results.
I am totally convinced that flat mag and phase quite simply achieve all the typical goals beyond flat frequency response,....like transient perfect response, minimizing group delay, etc. (obviously excluding goals regarding distortion or non-linear response)
Flat mag and phase simply equals technically correct, imo.....and sounds quite awesome to me. Why do anything else if the tools are available.
I really wish more folks would try linear-phase FIR on a driver by driver basis...it's sooo dang easy and works sooo dang well.
Especially when using steep xovers....pre-ring potential thru xover regions get minimized to the point of being immaterial...
https://www.diyaudio.com/community/...m-correction-etc.349818/page-189#post-7213813
Attachments
@wgh52
As a minidsp user , i can tell you flex 8 won’t get fir down low enough to remove any phase wraps much below about 90hz
That said , I would use all the IIR in your dsp to make the crossover work the best they can minimum phase….
After you Haare a good crossover, use the fir to do your gloabal minimum phase eq and get into the fine tuning, as you’ll have unlimited eq down to 100hz in fir
Then remove any phase wraps between 80-1000hz , and that’s about it….
That should get you a pretty dang good sound, just because the fir is limited doesn’t mean it’s not powerful.
Best of luck
As a minidsp user , i can tell you flex 8 won’t get fir down low enough to remove any phase wraps much below about 90hz
That said , I would use all the IIR in your dsp to make the crossover work the best they can minimum phase….
After you Haare a good crossover, use the fir to do your gloabal minimum phase eq and get into the fine tuning, as you’ll have unlimited eq down to 100hz in fir
Then remove any phase wraps between 80-1000hz , and that’s about it….
That should get you a pretty dang good sound, just because the fir is limited doesn’t mean it’s not powerful.
Best of luck
My comments are specifically about altering phase in this region because if the room was minimum phase there would be no need to adjust for it 🙂Next we have the room to listener transfer function, RL(w). Assume RL(w) is minimum phase.
The things that I find particularly audible are not any of the things you have been discussing or demonstrating. In general I agree with you that most phase manipulation ends up making prettier graphs and no real improvement in sound quality to justify the complexity, but there are circumstances where it can. For me they are lower frequencies where the room is becoming a factor.I just haven't been able to ever convince myself that if I start with a system that sounds good, global FIR phase correction makes it sound better.
A phase wrap is a graphical representation of a large phase change in a short period to make it easier to see the general trend. Are you really trying to correct the wraps? I would suggest you window it more or ignore the wrap and adjust for the trend not beat the response into submission.As a minidsp user , i can tell you flex 8 won’t get fir down low enough to remove any phase wraps much below about 90hz
@fluid
Not necessarily wraps as it can be either delay in the measurement, or it can be a specified amount of phase shift, like counting wraps to determine how many degrees of phase and the amount of GD or the wavelength calculated from the wrap s
So , it depends, no not just arbitrarily making wraps flat, thanks for questioning that so he knows because that could have been misleading
I strictly meant the crossover wrap which has allpass behavior, and when viewed in unwrapped phase won’t go downhill like delay in the measurement would….(as you know)
So yes some care has to be taken, the term is loosely thrown around like a dirty horr … sorry my apologies….
Just the crossover twists, sorry I thought maybe I eluded to that when I said make the crossover behave itself 😁 furthermore, the crossover “wrap” (if cyclic) or half wrap?? (Lol such a word?? ((2nd order filter))) may not be where it’s expected to be, once things are summed , he may have a perfect 80hz alignment, but once summing with everything it could get pushed up or down… it should be close to tho….
I read an article on how summing different speakers can move the phase based on a lot of variables, some being unequal path lengths, distance, , something about harmonics and a few other things…. But I’ve experienced it, so I know it’s a thing.. on mine was an 80hz alignment, once summed in the system (like a flex would need) , it’s allpass was at 53. Strange right…..
One could investigate why, or linearize it and be done… I did and it worked great.
I’ll try n find that article, was very good and a fun read 🙂
Not necessarily wraps as it can be either delay in the measurement, or it can be a specified amount of phase shift, like counting wraps to determine how many degrees of phase and the amount of GD or the wavelength calculated from the wrap s
So , it depends, no not just arbitrarily making wraps flat, thanks for questioning that so he knows because that could have been misleading
I strictly meant the crossover wrap which has allpass behavior, and when viewed in unwrapped phase won’t go downhill like delay in the measurement would….(as you know)
So yes some care has to be taken, the term is loosely thrown around like a dirty horr … sorry my apologies….
Just the crossover twists, sorry I thought maybe I eluded to that when I said make the crossover behave itself 😁 furthermore, the crossover “wrap” (if cyclic) or half wrap?? (Lol such a word?? ((2nd order filter))) may not be where it’s expected to be, once things are summed , he may have a perfect 80hz alignment, but once summing with everything it could get pushed up or down… it should be close to tho….
I read an article on how summing different speakers can move the phase based on a lot of variables, some being unequal path lengths, distance, , something about harmonics and a few other things…. But I’ve experienced it, so I know it’s a thing.. on mine was an 80hz alignment, once summed in the system (like a flex would need) , it’s allpass was at 53. Strange right…..
One could investigate why, or linearize it and be done… I did and it worked great.
I’ll try n find that article, was very good and a fun read 🙂
Last edited:
Thanks for this encouragement! Actually, the DEQX does/did not provide FIR filtering much below 300Hz (resulting in acceptable processing latency), so I'm used to that anyway and can confirm your sound quality findings!As a minidsp user , i can tell you flex 8 won’t get fir down low enough to remove any phase wraps much below about 90hz
That said , I would use all the IIR in your dsp to make the crossover work the best they can minimum phase….
After you Haare a good crossover, use the fir to do your gloabal minimum phase eq and get into the fine tuning, as you’ll have unlimited eq down to 100hz in fir
Then remove any phase wraps between 80-1000hz , and that’s about it….
That should get you a pretty dang good sound, just because the fir is limited doesn’t mean it’s not powerful.
Best of luck
Thanks to all for accepting my side-tracking which is now finished.
Thanks and Regards,
Winfried
Especially when using steep xovers....pre-ring potential thru xover regions get minimized to the point of being immaterial...
@mark100, This is exactly correct. With steep crossovers there is always a caveat. With minimum phase there is high GD variation through the x-o region. With linear phase the magnitude of the ringing increase as the slope of the roll off increases.
My comments are specifically about altering phase in this region because if the room was minimum phase there would be no need to adjust for it
@fluid, Whether the room-listener TF is minimum phase or not depends on the room, source position and listener position. But, in general, the room response tends to minimum phase at higher frequency while potentially non-minimum phase in the modal region. So, assuming you are referring to the modal region, how would you approach this for multiple woofers?
Very nice results Wirrunna.....thx for showing your project as an example of the kinds of measurements folks can expect.Mark that statement pushed me to say I agree, and look here to see my results -
https://www.diyaudio.com/community/...m-correction-etc.349818/page-189#post-7213813
And very generous posts you made on ASR, on how to implement Camilladsp on a RPi.
I like your headphone amp idea, driving the DCX464.....may have to give that try...
Not side tracking, happy to try to help...🙂Thanks for this encouragement! Actually, the DEQX does/did not provide FIR filtering much below 300Hz (resulting in acceptable processing latency), so I'm used to that anyway and can confirm your sound quality findings!
Thanks to all for accepting my side-tracking which is now finished.
Thanks and Regards,
Winfried
Yep, if you're familiar with DEQX, you already realize the FIR time (latency) it takes to work on lower frequencies.
Unfortunately, I think miniDSP pretty much ditched the user-defined FIR file products/market, some time ago. Nothing they make really offers sufficient FIR time imo, for the DIYer.
I agree with Oabeieo, 2048 taps, especially at 96 kHz, isn't going to do much down low. (WTF miniDSP, why 96 kHz?....)
If you do want to try global FIR and stick with miniDSP, I'd recommend putting an OpenDRC-DI in front of the Flex 8. (and run it's 6k taps per channel @ 48 kHz, like dirac live does to gain its low freq resolution)
But repeating myself, I wouldn't bother with global FIR.
Imo/ime, the trick to making FIR and linear-phase worth the effort, is to jump up to multi-channel FIR, where xovers are all complementary linear phase.
I think anything short of that leaves the potential goodness (and ease/understanding ) on the table.
It's simply so dang easy to correctly time and phase align drivers with linear phase xovers, as compared to IIR xovers.
(again, especially when lin-phase xovers are steep)
Sorry... I guess I just can't stay off the soap box...lol
Hope this helps...
Thanks for this encouragement! Actually, the DEQX does/did not provide FIR filtering much below 300Hz (resulting in acceptable processing latency), so I'm used to that anyway and can confirm your sound quality findings!
Thanks to all for accepting my side-tracking which is now finished.
Thanks and Regards,
Winfried
Didnt you say flex 8
Where is deqx coming from, the deqx is Uber powerful…. Well the new one is…. It’s the best thing ever made
Sounded to me like he's familiar with DEQX....Where is deqx coming from, the deqx is Uber powerful…. Well the new one is…. It’s the best thing ever made
Have you seen a Manual or at least hard specs, on the upcoming new models? Only way to know how powerful it really is, in terms of FIR capability...
Current/existing stuff wasn't as strong as hoped, when I looked under the hood....here's an example from their manual...
I read the 200+ page manual for the new deqx that was released in December that uses the cortex ARM and ADSP215x with 1Ghz ..
I’m not sure if that manual shows the same thing if that’s the same manual or if you’re looking at the older one but the new one is oh my goodness powerful
And who in their right mind uses 96db slopes (I’m kidding) that’s too much attenuation in my personal opinion
So I’m not sure again if that’s the same manual or the same unit we’re talking about
I’m not sure if that manual shows the same thing if that’s the same manual or if you’re looking at the older one but the new one is oh my goodness powerful
And who in their right mind uses 96db slopes (I’m kidding) that’s too much attenuation in my personal opinion
So I’m not sure again if that’s the same manual or the same unit we’re talking about
What i posted is defintely the old manual...and why i asked if you've see the new one...do you have a link to it
Then I guess you've missed what I've been talking about....
And who in their right mind uses 96db slopes (I’m kidding) that’s too much attenuation in my personal opinion
Then I guess you've missed what I've been talking about....
@mark100, This is exactly correct. With steep crossovers there is always a caveat. With minimum phase there is high GD variation through the x-o region. With linear phase the magnitude of the ringing increase as the slope of the roll off increases.
john k... didn't mis it.
No he didn't, not at all 🙂john k... didn't mis it.
And he even went on to amplify what the tradeoff is with using steep (always tradeoff in audio, huh?)
I was hoping to run a few measurements before replying, where i deliberately miss time-align linear phase LR xovers of different orders. (like off-axis lobing does)With linear phase the magnitude of the ringing increase as the slope of the roll off increases.
I figured it would be cool to show what you're talking about, and frankly ...i've never thought to look at it.
Tis on the to-do list...but right now I'm happily in the garage adding bass-reflex ports to a big-**** synergy.
Mark, thank you for the comments, the headphone amps were ClaudeJ1's idea back on the Klipsch forums, it was the only way I could afford a THX certified amp in my rig and @fluid has shown me how to get Excess Phase from REW as a measurement into rePhase to simplify the flattening process. Just going through the exercise now to update my project documentation.Very nice results Wirrunna.....thx for showing your project as an example of the kinds of measurements folks can expect.
And very generous posts you made on ASR, on how to implement Camilladsp on a RPi.
I like your headphone amp idea, driving the DCX464.....may have to give that try...
CamillaDSP with a Motu Ultralite Mk5 is a very versatile DSP.
Modal and transition but modal is the most obvious. For a stereo signal a measurement of left and right channels at the listening position, window (FDW) the measurements down to concentrate on the early sound and apply the correction to the left and right channels. If the multiple woofers were summed into a mono channel the correction might have to be separated. I have no practical experience of how well that might work or if adjustments would be needed, it would depend on how the bass management setup is configured.@fluid, Whether the room-listener TF is minimum phase or not depends on the room, source position and listener position. But, in general, the room response tends to minimum phase at higher frequency while potentially non-minimum phase in the modal region. So, assuming you are referring to the modal region, how would you approach this for multiple woofers?
https://www.deqx.com/wp-content/uploads/pdf/DEQX-User-Manual.pdfWhat i posted is defintely the old manual...and why i asked if you've see the new one...do you have a link to it
Then I guess you've missed what I've been talking about....
https://www.deqx.com/products/hdp-5/
This guy here ^^^
So I imported a minimum phase generated average into left and right into smaart as a capture to follow, and used rephase live and 3 iterations have a pretty good match…
It sounds more real and transparent especially the bass, mids are about the same…. Can’t tell on HF
Overall , it is better….. yeah I get it now I think
The response and phase are almost an inverse and they slowly diverge as my response tilts , the bass phase falls just like the minimum phase version …..
I can really mostly only tell in the bass…. Just being honest …. Bass is more real sounding…. But it’s very similar to just zeroed phase
It sounds more real and transparent especially the bass, mids are about the same…. Can’t tell on HF
Overall , it is better….. yeah I get it now I think
The response and phase are almost an inverse and they slowly diverge as my response tilts , the bass phase falls just like the minimum phase version …..
I can really mostly only tell in the bass…. Just being honest …. Bass is more real sounding…. But it’s very similar to just zeroed phase
Last edited:
Modal and transition but modal is the most obvious. For a stereo signal a measurement of left and right channels at the listening position, window (FDW) the measurements down to concentrate on the early sound and apply the correction to the left and right channels. If the multiple woofers were summed into a mono channel the correction might have to be separated. I have no practical experience of how well that might work or if adjustments would be needed, it would depend on how the bass management setup is configured.
??? Not sure I follow. There is no low frequency content in the early sound. The window would have to be at least 50m sec to get down to 20 Hz. Anyway, just for fun, some stuff I found that I did years ago. A little passive x-o speaker with global FIR to make it flat response and linear phase. White = input, red = measured output at the listening position. Upper right Insert = simulated response of the uncorrected speaker The deviation from the input, particularly at lower frequency, is due to reflections. x-o point was 1500 Hz I believe.
8192 taps.
Last edited:
- Home
- Loudspeakers
- Multi-Way
- Why not IIR filters + a global phase linearization by FIR