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

mark100,

Slopes into REW overlay plot, textbook target curves is imported IR-wav created in Rephase.
 

Attachments

  • 93.png
    93.png
    57.6 KB · Views: 357
This is only terminology indeed, but the distinction is important :)
The final acoustical filer is the one to consider, and is the combination of the "natural" (whatever that means, box+driver+passive components) and "electrical" (passive or active, analog or digital, IIR or FIR) filters.

Yes, totally onboard.
And sorry for the term 'natural'...what i was trying to say was the response of the driver + enclosure, without any form of crossover at all...
 
To put it simply, convolution on the input cannot do anything about relative magnitude and/or phase differences between drivers.
Those will remain no matter what.

Hi pos,
Sure, and....
Since a speaker is a collection of drivers, ...and since magnitude and phase differences between drivers cannot be solved by input convolution....
....doesn't that mean the most optimal way to tune a speaker is with individual driver by driver alignment ?
 
Hi guys. I posted a thread with this question but thought perhaps it would be more suitable here as it involves using rePhase.

I have recently finished making a 3-way crossover using the IIR and PEQ capabilities of my MiniDSP Open DRC-DA8. The result is a flat amplitude response (using appropriately gated measurement to eliminate reflections) at the measuring point, 1m from the baffle at tweeter level. The gated overall phase response is not linear, but integrates well between drivers and has an overall gradual slope from +180deg to -360deg from left to right over the frequency range.

I want to use the FIR capability in tandem with my current IIR crossovers and PEQ filters in order to carry out phase only correction using said FIR filters for each driver on my DRC-DA8 to linearise the overall phase of the loudspeaker.

To do this, do I:

Measure the individual gated response of each driver with the current IIR filters, PEQ filters and time delay applied (as shown in screenshots below)? Each driver is measured individually from the same point (1m from baffle and in line with the tweeter) as is the overall system response. Once this is done, import the results individually into rePhase, then use the phase EQ tabs to change the phase of each individual driver to be a flat line over the usable frequency band (Also im not sure if the phase should be flat/linear at the same point for each of the three drivers, ie. they all must be 0 degrees or all 90 degrees etc as long as they are the same). Then, use the rePhase generated output for each and put this into the FIR filter for each individual driver on the DRC-DA8 plugin.

Presumably once this is done, I would measure the response of each driver again, measure the overall combined response and hopefully find that the response is still linear in amplitute throughout the loudspeakers frequency range 30hz-20khz but now also linear in phase.

Is this the correct method where my goal is to linearise the phase of my loudspeaker after already applying a conventional IIR 3-way crossover and PEQ filters?

Note all measurements attached are gated to 3.2ms to remove reflections so low frequencies <300hz are inaccurate.

Many thanks in advance for your advice.

OVERALL SYSTEM


WOOFERS (2x scanspeak revelators 6.5")


MID (Morel EM428)


TWEETER (Scanspeak D2905/9500)


Solidworks Render of my speakers
 
AndrewUK1990,

Nice build there, traced your plots in below picture-1 overlay compared to textbook system stopbands at BW2 20-25kHz and LR4 XO points at 600/3000Hz.

Short version think sounds as you got it right, drivers are minimum phase devices (IRR) and you at present correct system with reverse IRR EQ, then if measurements are right plus correction for acoustic center offset between W/M/T you just need the phase correction that Rephase create on "Filters Linearization" tab setting one filter to correct LR4 600Hz and another filter to correct LR4 3kHz. This filter will neutralize IRR 2x 360º phase turn caused by summing your system with 2x IRR XO points so that after correction only phase turn left back is the two IRR BW2 stop bands that in these simulations is set to 20Hz at LF roll off and 25kHz for HF roll off.

Above correction will fail if measurements or acoustic offset isn't right because phaseturn isn't where we expect it to be but think no harm to try and see result, if your DSP need one correction that include both IRR and FIR correction in same config file or IR-wav file there is more work to be done or maybe better to use compensate mode as pos has guided.
 

Attachments

  • 90.png
    90.png
    63.6 KB · Views: 150
  • 105.jpg
    105.jpg
    535.5 KB · Views: 198
Hi guys. I posted a thread with this question but thought perhaps it would be more suitable here as it involves using rePhase.

I have recently finished making a 3-way crossover using the IIR and PEQ capabilities of my MiniDSP Open DRC-DA8. The result is a flat amplitude response (using appropriately gated measurement to eliminate reflections) at the measuring point, 1m from the baffle at tweeter level. The gated overall phase response is not linear, but integrates well between drivers and has an overall gradual slope from +180deg to -360deg from left to right over the frequency range.

I want to use the FIR capability in tandem with my current IIR crossovers and PEQ filters in order to carry out phase only correction using said FIR filters for each driver on my DRC-DA8 to linearise the overall phase of the loudspeaker.

To do this, do I:

Measure the individual gated response of each driver with the current IIR filters, PEQ filters and time delay applied (as shown in screenshots below)? Each driver is measured individually from the same point (1m from baffle and in line with the tweeter) as is the overall system response. Once this is done, import the results individually into rePhase, then use the phase EQ tabs to change the phase of each individual driver to be a flat line over the usable frequency band (Also im not sure if the phase should be flat/linear at the same point for each of the three drivers, ie. they all must be 0 degrees or all 90 degrees etc as long as they are the same). Then, use the rePhase generated output for each and put this into the FIR filter for each individual driver on the DRC-DA8 plugin.

Presumably once this is done, I would measure the response of each driver again, measure the overall combined response and hopefully find that the response is still linear in amplitute throughout the loudspeakers frequency range 30hz-20khz but now also linear in phase.

Is this the correct method where my goal is to linearise the phase of my loudspeaker after already applying a conventional IIR 3-way crossover and PEQ filters?

Note all measurements attached are gated to 3.2ms to remove reflections so low frequencies <300hz are inaccurate.

Many thanks in advance for your advice.

OVERALL SYSTEM


WOOFERS (2x scanspeak revelators 6.5")


MID (Morel EM428)


TWEETER (Scanspeak D2905/9500)


Solidworks Render of my speakers

Hi Andrew,


Can you please explain:

1. Dynamic range of your measurements is only 20-23dB.
2. Phase response of is rising towards higher frequencies. How is this possible in minimum-phase systems?.

Best Regards,
Bohdan
 
Hi guys, thanks for the info. Brytt - the above measurements include time alignment of each driver to allow for acoustic offset to the measuring point (1m from baffle in line with tweeter). The impulse peak is also almost exactly the same for each driver at this point following the time alignment.

Bohdan - to clarify:

1. Dynamic range is relatively low on these measurements because I live in a shared house and could not measure at high volumes. I will carry out another set of measurements with greater dynamic range when i get the chance.

2. Phase response is possibly rising on the tweeter after 10khz because I have a mild high shelf filter to compensate the rolloff of the soft dome (or accidental lack of accurate measurement at high frequencies?). If this is a problem i can easily remove it from my PEQ filters.

Thanks for the advice,

Andy.
 
Last edited:
2. Phase response is possibly rising on the tweeter after 10khz because I have a mild high shelf filter to compensate the rolloff of the soft dome (or accidental lack of accurate measurement at high frequencies?). If this is a problem i can easily remove it from my PEQ filters.

Thanks for the advice,

Andy.


Hi Andy,

No, phase response in minimum-phase systems does not work like this.

If your measured SPL rolls-off at higher frequencies (and it must roll-off at high-end, because all loudspeakers are band-pass devices), the phase response must follow the roll-off accordingly to Bode's principle.

-6dB/oct roll-of -> phase must go to -90deg at high end.
-12dB/oct roll-off -> phase must go to -180deg at high end.
-18dB/oct ->.....and so on....

I can not see this behaviour in your measured responses.

Best Regards,
Bohdan
 
Thanks Bohdan. The tweeter does roll off in amplitude beyond 20kHz but my mic isn't accurate beyond that anyway so I cut the measurement at 20. I would presume the phase will do the same. The phase shown in the diagrams above is the direct input from REW following measurement, with the only post processing being that it is gated. The "generate minimum phase" hasn't been shown here. Not sure what the difference is between these two in REW?
 
Thanks Bohdan. The tweeter does roll off in amplitude beyond 20kHz but my mic isn't accurate beyond that anyway so I cut the measurement at 20. I would presume the phase will do the same. The phase shown in the diagrams above is the direct input from REW following measurement, with the only post processing being that it is gated. The "generate minimum phase" hasn't been shown here. Not sure what the difference is between these two in REW?


Hi Andrew,

No problem.

If you focus on your midrange driver, it will be easier for you.

REW seems to have a problem with "generate minimum phase" function. I explained this in post 925 on page 93 of this thread.

Measuring minimum phase response is actually quite difficult task.
This paper http://www.bodziosoftware.com.au/Min_Phase_Appr_Deriveration.pdf explains some of the issues.


Best Regards,
Bohdan
 
Hi guys, on quick question :
I looked around but could not find the optimized parameters for exporting the filters (windowing, centering, optimization). Any link where I could find this ?

Thanks

Hi

It depends on what you are looking for.
Some parameters will obviously be dictated by your convolution engine (number of taps, sampling type and freq) others might be dictated by the delay you can accept (number of taps, centering, ...), and others by the type of correction you are trying to achieve (centering, windowing, etc.). For example phase only correction, minimum phase EQ or linear phase filtering will all call for different settings.

If you tell us more about the context of your correction we might be able to provide generic guidelines. But in the end you will have to try different settings and pick the ones for which the result curves track the target one close enough for you needs.
 
Last edited: