rePhase, a loudspeaker phase linearization, EQ and FIR filtering tool

By the way John, I have a question I wanted to ask you for some time, and this is the perfect opportunity: is a BR (Helmotz resonator) a [minimum] phase thing? Or is there something special about the ringing?

BR is just a minimum phase thing. Ringing is typical of a 4th order HP alignment. Nothing special.

The Four Audio stuff seems way over priced. Sort of like the DQEX. It seems to be aimed at the typical audiophile who has more money than brains. I think a lot of audiophiles don't realize that dsp is more computer science and mathematics than audio once you are between the A/D and D/A converters. I certainly no expert here but good algorithms for FFt convolution aren't exactly a mystery.
 
By the way John, I have a question I wanted to ask you for some time, and this is the perfect opportunity: is a BR (Helmotz resonator) a phase linear thing? Or is there something special about the ringing?
pos, Yes. A Bass Reflex (including the double ported & KEF versions) are all Minimum Phase.

It's very difficult NOT to have a minimum phase speaker at LF though some back & front loaded horns try very hard.

Practically all the phase-distortion in a speaker is from evil xovers & drive unit (mis)alignment. :eek:

John K.., I'm impressed by your Dipole speakers having dabbled with them in the previous millenium. :)
 
Thierry, do you know of any plugin that does direct convolution?
I have played with SIR but found some annoying artifacts... (on specific test signals)
If the PC is powerful enough it looks like direct convolution is the way to go. I have no idea how much power it would require though...

i've read this thread,direct FIR core i7

a nuclear factory !:p
a powerfull graphic card should be implemented to run 8 convolutions of 8000 taps...

for HTPC or PC HIFI,fft convolution allows passive heatsink and modest processor.no fan noise
i've checked THD an IMD with convolutions in the loop,it's ok too,nothing to notice.

why direct convolution could be a better way ?
 
BR is just a minimum phase thing. Ringing is typical of a 4th order HP alignment. Nothing special.
pos, Yes. A Bass Reflex (including the double ported & KEF versions) are all Minimum Phase.

Thanks!
So if phase can be corrected there is no downside for using a BR!
I am thinking about a BR for my lowbass driver (TAD TM1201) which is seriously excursion-limited (distortion skyrocket right under 400Hz). A passive radiator would even suppress the problem of the hi mids leaking through the vent...
 
John K.., I'm impressed by your Dipole speakers having dabbled with them in the previous millenium. :)

You should build a pair of my newNote II RS speakers. Relatively inexpensive and very, very good. :)


Also regarding the comment about direct convolution. There is no difference between direct and FFt convolution, if the algorithms are correctly implemented, other than the reduction in latency which is a big deal if you are trying to sync with video.
 
The Four Audio stuff seems way over priced. Sort of like the DQEX. It seems to be aimed at the typical audiophile who has more money than brains. I think a lot of audiophiles don't realize that dsp is more computer science and mathematics than audio once you are between the A/D and D/A converters. I certainly no expert here but good algorithms for FFt convolution aren't exactly a mystery.

It was made in 2007. I think it is not aimed at the end user, but for professional integrator and speaker builder (thus the price of the software).
I think it was used in some H&K/Neumann products.
 
What are you thinking of?

If the phase is corrected (which is one of the purposes of the tool presented in this thread) and the amplitude slope can be taken into account (especially for a lowbass driver that will cross to a sub...) then the only possible downsides I see are:
- sound leaking through the port, that can be suppressed using a PR
- cannot fill the whole encolure with damping material, so the of shape of the box might be more critical to avoid standing waves
- need to add another HP filter (linear phase) to prevent excursion problems under resonance, so final target slope will likely be around 48dB/oct
 
What are you thinking of?

If the phase is corrected (which is one of the purposes of the tool presented in this thread) and the amplitude slope can be taken into account (especially for a lowbass driver that will cross to a sub...) then the only possible downsides I see are:
- sound leaking through the port, that can be suppressed using a PR
- cannot fill the whole encolure with damping material, so the of shape of the box might be more critical to avoid standing waves
- need to add another HP filter (linear phase) to prevent excursion problems under resonance, so final target slope will likely be around 48dB/oct

Provided the cut off is below any frequency to be reproduced. If there is content below the cut off point it will introduce preringing. A ported system with 25Hz cutr off should be pretty good.
 
If the input signal has any frequency content below 200 HZ then that may introduce a pre-response. Look at a linear phase high pass filter response to a square wave. As long as the fundamental is above the cut off frequency (well in the flat band region actually) the square wave is perfectly reproduced. If the square wave fundamental is below the cut off the wave form is distorted. Here is a picture.

An externally hosted image should be here but it was not working when we last tested it.


Notice that for the 20 Hz signal the minimum phase response is causal but the linear phase is not. It has a pre-response. At 125 Hz, above the HP cut off, the minimum phase is still distorted but the linear phase is not. Neither is very good below the cut off but above the cut off the linear phase system gets it right. ( The slight droop in the linear phase 125 Hz result is because the filter isn't flat at 125 Hz so the fundamental of the square wave is slightly attenuated. )
 
Ok, but this is true for any linear phase filter, not only for a phase-corrected BR.
If I EQ and filter a BR to obtain a acoustical LR 48khz (for example), then (as you confirmed it) it will behave exactly as a normal LR48 would have. Whether or not I subsequently correct its phase or not will have the same impact as any other LR48, right? The steeper the slope, the more preringing will be seen on the impulse, especially if there is not symmetrical LP filtered driver (which is not the case here as I want to use a BR for the midbass as part of the midbass/woofer crossover).
 
Ok, but this is true for any linear phase filter, not only for a phase-corrected BR.
If I EQ and filter a BR to obtain a acoustical LR 48khz (for example), then (as you confirmed it) it will behave exactly as a normal LR48 would have. Whether or not I subsequently correct its phase or not will have the same impact as any other LR48, right? The steeper the slope, the more preringing will be seen on the impulse, especially if there is not symmetrical LP filtered driver (which is not the case here as I want to use a BR for the midbass as part of the midbass/woofer crossover).

Yes, absolutly true. It doesn't matter if it electrical, acoustical or a combination of both. If the final output is linear phase it behaves as a linear phase filter of the same amplitude respponse because that is what it is.
 
I am not convinced steep slopes are the way to go: they are more prone to preringing as soon as the high-passed et low-passed drivers are not perfectly summed in phase (which can happen quite quickly off axis, even if the system is perfectly in phase on axis).
What slopes have you tired in your system Thierry?

The main advantage I see with a pure FIR crossover, in addition to a linear phase, is the opportunity of using any complementary shape you want for your crossover, such as Horbach-Keele slopes (special MTM) for example.
 
we can strongly lowering the tweeter cutoff frequency.
and join the mid section in a constant directivity region,avoiding too much electrical power in the tweeter.

it's interesting with 2 ways 15"+2" compression and coaxial driver.

i've tried 48 and 96 dB/oct at several frequency.
last choice is 96 dB at 140 Hz for low/mid-low cutoff.
always with passive at 2400 Hz.

i'm looking toward for a 8"+1" coaxial driver for mid/high,with a 1500 Hz 96dB/oct crossover point.

also FIR EQ seems to be more efficient.
:eek: so much possibilities...it needs time,measurement and standing back to aim better choices.
 
I am not convinced steep slopes are the way to go: they are more prone to preringing as soon as the high-passed et low-passed drivers are not perfectly summed in phase (which can happen quite quickly off axis, even if the system is perfectly in phase on axis).
What slopes have you tired in your system Thierry?

The main advantage I see with a pure FIR crossover, in addition to a linear phase, is the opportunity of using any complementary shape you want for your crossover, such as Horbach-Keele slopes (special MTM) for example.

Well all those theoretical crossovers are really nothing more that acoustic targets as it is the acoustic response of the driver that must match them. That is what is so nice about the Bodzio UE. It allows you to specify the acoustic target and it then generates the filter transfer function required to achieve it based on the measure driver SPL data.
 
Of course, the natural frequency response (amplitude and phase) needs to be taken into account, or corrected.
I don't like the "automated" approaches: I prefer manual corrections based on different measurements, to avoid correcting things that should really be left alone (measurement dependant).

For now with rephase the user needs to import the generated impulse in HOLM (or REW) and do a convolution with the measurement(s) (C=A*B in the "manipulation" menu in HOLM).

So by going back and forth between rephase and HOLM (adjust correction, generate, impot in HOLM, do the convolution, adjust correction, ...), the user is able to first "linearize" both the amplitude and the phase of a driver within and around the passband (1 or 2 octave, depending of the slope that will subsequently be applied), and then apply any filter slope he wants.
So this is a two-steps thing, and always manual.

I will include the possibility to import measurements in rephase to directly see the effect of a correction in realtime, but it will remain manual (no automated inverse correction). I will try to make it easy to load multiple measurements and either average them or make it easy to switch between them during corrections.
 
Last edited:
Your manual approach is pretty much the way the UE works deep down inside. You start with a measurement. That measurement can be a lot of things: the on axis response, the smoothed axial response, a spatially averaged response,..... Then, the user specifies a frequency range over which minimum phase equalization is applied to flatten the response. Then a text book type filter is applied where in the net acoustic output will exactly match the text book response over the eq'ed range. Over the frequency ranges where the eq is not applied the response will be the text book filter timed the raw SPL data. Then the final response is made linear phase (on option).

So if you want to make a tweeter have a linear (or minimum) phase LR4 type response at 3k Hz but only want to equalize the response in the actually crossover region and leave the response above say 3.5k un equalized, you can do that.

[edit] Then the response can be measured and if you are not happy you can change it. Since the UE has has the convolution engine built in you never have to dick with stuff. You know exactly what you will have because you measure the result and can compare the measured result with the predicted result.
 
Last edited: