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

Hi Pierre,

I am glad you found the eBook useful. Like you say, while I use Acourate, the principles can be applied using most DSP software.

Interested folks can browse the Table of Contents and read the first couple of chapters by clicking on the link in my sig below and then clicking on Look Inside.

Kind regards, Mitch

Recent developments of REW/rePhase in the field of room correction might give you the opportunity to extend the scope of the next version of your eBook ;)
 
Recent developments of REW/rePhase in the field of room correction might give you the opportunity to extend the scope of the next version of your eBook ;)

Yes, I have been watching and that is the plan for next edition :)

Keith_W has additional suggestions: DSP tutorial book written by mitchco

In addition to REW/rePhase development, if folks have additional interests using this software or other DSP, please add your suggestion using the link above.

Thanks pos for your excellent software contributions, looking forward to further developments.

Kind regards, Mitch
 
Mitch,

Yes, I have been watching and that is the plan for next edition :)
Cool :)

In addition to REW/rePhase development, if folks have additional interests using this software or other DSP, please add your suggestion using the link above.

- You have mentioned the subjects of pre-post ringing earlier. Bohdan who is active on this thread has published an interesting paper on this: http://www.bodziosoftware.com.au/pre_post_ringing_ir_and_pulses.pdf
I would be interested if you could elaborate on best practices to avoid pitfalls in this respect and how to chose between min-phase and linear phase filters in various occasions
- You quickly touched 'beamforming' (https://www.researchgate.net/public...esponse_measurement_using_beamforming_methods) as a way to average measures and get rid of room influence to reach close to anechoic measurement in your e-book.
A famous provider in the industry in promoting 9 measurements at the edges of a cube+center. This does, according to my experience, provide together with REW pulse algebra, quite interesting inputs for REW/rePhase usage. I would be interested if you could dig into this kind of direction in order to provide more robust measurement strategies.

Thanks in advance,
Kind regards,

Pierre
 
Last edited:
Administrator
Joined 2004
Paid Member
My question for you in your particular situation with the driver you are dealing with is even with a fast slope at 28Khz what is the attenuation at 20Khz or even at 14Khz with the given slope you have chosen? .....An input filter on many amps and other devices would seem to limit the possibility of exciting that 28Khz mode, besides the fact that most music has little to no information in that range.

I'm with you there. I wonder just how much actual musical content is up there to bother the tweeter. And if the filter is before the input, that doesn't stop anything downstream from creating noise or harmonics up at 28K. Certainly a 192dB/octave filter is super steep, but if there is really content in the files causing tweeter resonance, wouldn't a notch filter be a better idea?
 
I'm assuming that we are talking about some type of metal dome tweeter that would even have a possibility of producing 28Khz. As you say Pano what music information would you ever see that high. Not going to happen with most source material. Now if you have a harmonic at 14Khz or some higher order harmonics you could possibly excite the dome but I don't expect much energy to get the dome going that high. A notch filter is one option, depends on what other hash is happening up that high. I almost want to say we are talking about Ti.

Sorry about going off topic here.
 
Last edited:
Hello!

Thank you for your answers so far. I can understand that some of you would like to discuss other solution for this problem.
Since this is the rePhase thread I would just like to discuss about such FIR filter only. When I know the advantages and disadvantages of such a filter, I want to compare it to other solutions like a bell eq filter, or doing nothing at all because there is no content at that frequencies.
I think it is mandatory to gather all aspects of different solutions together before making a decision.
In general it would be nice to know what special FIR problems could occur, when designing filters with rePhase and how to find them.
 
For my current project I would like to use a lowpass filter at 22 kHz to prevent that the 28 kHz resonance of the tweeter modulates into the lower frequencies and increases multisine distortion.
For this I plan to create an linear phase FIR filter with rePhase but I am not sure, how steep this can be without causing any problems. I created this filter with 400 taps at 96 kHz and 192 dB / octave.
Can such filters cause any audible problems like pre-ringing or something else? I changed the scaling to 50 kHz and -150dB and the rolloff was totally smooth.

Thank you for your answers so far. I can understand that some of you would like to discuss other solution for this problem.
Since this is the rePhase thread I would just like to discuss about such FIR filter only. When I know the advantages and disadvantages of such a filter, I want to compare it to other solutions like a bell eq filter, or doing nothing at all because there is no content at that frequencies.
I think it is mandatory to gather all aspects of different solutions together before making a decision.
In general it would be nice to know what special FIR problems could occur, when designing filters with rePhase and how to find them.

Hello

In your situation adding a brickwall-like linear-phase filter around 22kHz is not much different that having a sampling rate converter to 48kHz or the like.
These things are everywere in digital devices, and imply antialiasing filters.
Some are linear-phase, some are minimum-phase, with different possible slopes and imply different trade offs between stop band rejection and impulse ringing.
Many DACs will let you choose between several of them (eg ESS), and some (audiophile) people have expressed preferences for one or another type (minimum-phase and/or gentle slopes, most of the time).
That said industrial guys often add these option to give the user (or integrator to be more precise) the option of reducing the delay (minimum-phase implies a shorter delay), not really for sonic qualities...

Both pre and post ringing are probably not audible at these frequencies (what is? :D), and phase shifts in these HF frequencies is also not something to worry about, so... But when you need to integrate a DAC the delay it implies or the phase shifts it will add are something you might want to be able to modify.

With rephase you can generate any filter you want: linear or minimum phase, or a mix of the two, and with any slope. This is what soekris did for his DAC.
I would recommend using a brickwall filter and then play with the different Albrecht windowing algorithms in rephase: the more terms you use the deeper the noise floor (up to 290dB for the 11-term variant when applied on a brickwall filter (which is a worst case scenario)), and the gentler the slope (which you can increase by adding more taps). I really like these Albrecht windowing functions!
You can also combine a linear-phase filter and a minimum-phase one, possibly at different frequencies, to get a tradeoff between preringing, phase behavior and rejection. In this case you do not want the windowing function to dictate the slope.

Regarding your specific case, with the 28kHz breakups and peak, you should also consider the fact that even if you remove the offending frequencies from the input signal by filtering or EQing it, distortion products from the lower range might still trigger the peak (eg 3rd harmonic distortion generated at 9kHz). Active filtering cannot do anything to address this, and all you can do is either use acoustic filtering (eg reticulated foam) to reduce the output of the driver (in opposition to reducing the input), or EQ/filter you input and the address the distortion problem by using current drive (a simple resistor inline with the tweeter will do the job) to reduce the 3rd distortion generated at 9kHz.
 
Mitch, SwissBear,

I am quite happy that thise REW/rePhase bridge is effective because automated EQ is somthing that many people requested, but certainly not somthing I wanted to add into rePhase.
rePhase is not meant to manipulate measurements (averaging, gating, smoothing, or combinations of the above like temporal windowing), and this is off course something mandatory when dealing with automated EQ. I feel like this is a more natural workflow to let the measurement software handle all this.

At some point I wanted to add an option to average several imported measurement into one, but that also would be out of scope. What I would like to add is the option to import several measurements (with meaningful description) and be able to go from one to the other.
You can do this today by drag and dropping measurement files onto the rephase window, but that implies that all those measurement have the same time and magnitude offset, and you also cannot save all of them in the ".rephase" file...
 
Hello!

First of all I would like to say a big thank you for this great tool.
For my current project I would like to use a lowpass filter at 22 kHz to prevent that the 28 kHz resonance of the tweeter modulates into the lower frequencies and increases multisine distortion.
For this I plan to create an linear phase FIR filter with rePhase but I am not sure, how steep this can be without causing any problems. I created this filter with 400 taps at 96 kHz and 192 dB / octave.
Can such filters cause any audible problems like pre-ringing or something else? I changed the scaling to 50 kHz and -150dB and the rolloff was totally smooth.

Thank you!

Don't know if you aware but audio domain visuals is possibility multitask into some of the great free programs, below are as example a perfect impulse response 400 taps at 96kHz created in Rephase and saved as 32bit IR-wav file and then imported to REW, below that 96kHz perfect impulse is same data for your steep 22kHz filter. As pos posted it exactly looks like ringing that many modern DACs output, can say this because have two USB DACS with Cirrus silicon in 100-130€ area that looks exactly as below or worse except when they run full throttle at 192kHz, where another Essense II card and good old AP192's reconstruction looks close to perfect clean IRR domain impulse.

As in spirit of above Rephase is also perfect creating target curves for measurement software so one can see how a given perfect pass-band will look verse the often less quality acoustic real world data. As example create perfect pulse is most simple one with no correction at all but adjust sample rate/output file type/file name, and create IRR or FIR pass-band pick among build in HP and LP filters and generate output file. Tip to create target curve for missing Bessel 2.order use either one LR2 or cascade 2x BW1 and add a IRR paremetric EQ at same frq as slope and set Q0,58 amplitude +1,3dB.

When in REW "All SPL" tab one can manipulate in "Control" various IR-wav files imported from Rephase with math as A times B or A divided B and etc, also if IR is offset in "Impulse" tab one can offset them relative in time just as with real drivers. XSim can also work with these Rephase created pass-bands but they have to go thru REW first to get exported as frd-files. Second attachment (post 1032) and third (post 910) is examples of simulations that was started as target curves into Rephase and then manipulated by REW math or into XSim that can output square wave performance.
 

Attachments

  • 1.png
    1.png
    712.6 KB · Views: 379
  • 1.jpg
    1.jpg
    535.5 KB · Views: 364
  • 2.png
    2.png
    82.1 KB · Views: 363
Last edited:
Mitch, SwissBear,

I am quite happy that thise REW/rePhase bridge is effective because automated EQ is somthing that many people requested, but certainly not somthing I wanted to add into rePhase.
rePhase is not meant to manipulate measurements (averaging, gating, smoothing, or combinations of the above like temporal windowing), and this is off course something mandatory when dealing with automated EQ. I feel like this is a more natural workflow to let the measurement software handle all this.

At some point I wanted to add an option to average several imported measurement into one, but that also would be out of scope. What I would like to add is the option to import several measurements (with meaningful description) and be able to go from one to the other.
You can do this today by drag and dropping measurement files onto the rephase window, but that implies that all those measurement have the same time and magnitude offset, and you also cannot save all of them in the ".rephase" file...

@pos, that's great! Congrats again on your very useful software!

@SwissBear, will send a PM to continue discussion.

Kind regards, Mitch
 
Hi ustas,

Not sure what you expect. The process is straightforward. Here is an oversight: Using REW and rephase to generate amplitude and time domain corrections
From REW:
- take both channels measurements, with Psycho smoothing and no FDW, go to 'EQ' tag, select rePhase as your EQ, and generate your filters according to your prefs. Save filters to .xml files
- export both measurements with full time window (no FDW) and higher resolution (1/12th octave) to text files
- export both measurements with 6 cycles FDW to text files
In rePhase:
- draw one measurement in full time window onto the interface of rePhase. Use Gain EQ and import the filters from REW. Make further adjustments
- draw the measurement with 6 cycles onto rePhase interface and adjust phase. If you do not know the crossover frequencies of your LS, make a 10cm measurement of each driver and superimpose in REW. This will give you a good approximation.
- linearize the crossovers with appropriate function (try various slopes)
- gently adjust phase near crossover frequencies with phase EQ.
You can add a subsonic filter (high-pass filter). You can choose from min-phase (no pre-ringing but more post-ringing) or lin-phase (a little bit of pre-ringing, not audible).
And you are all set.
If you want to post charts of your measurements, you can do it. It is easier to help you do things than to do for you :)
 
Last edited:
  • Like
Reactions: 1 user