TOPIC:

So guys I thought I would share a few pointers to help…

I have 3X 2x4hd in a multi-way… I’ve been RePhasing since 2015 with them. A few pointers to make taps work with Rephase.

1. use rectangular windows 
2. For LF crossover linearizarion you can try 1024 taps , rectangular window, and 9ms centering (just type -9ms) all lower case in centering box
3. 1024 taps will “play” down to 94hz so all phase eq or eq  needs to be above 94hz with 1024 taps 
4. Do sub eq in the peq and save fir eq for mids and highs as you don’t need to make phase changes below 100hz 
5. If you have extra response under midbass crossover you don’t have linearize a sub to midbass to sound good….  If you turn on let’s say an 80hz crossover on a midbass and make the responce flat below 80hz it will be linear phase, so don’t make it flat to 20 , just make it flat to right above fs of the midbass then let it naturally drop off, then add more taps to the sub and linearize it, center it at 15ms or so with about 2000taps and it will sound phenomenal.
6. Make close mic measurements to speaker gated at 4cy. Then make measured phase flat in rephase and send that to speaker.


the 2x4hd has just barely enough power to get it done, but it won’t pre ring and will sound excellent! You don’t need a million taps to get a good fir. 256 taps gets you fir power down to just under 400hz. There shouldn’t be any reason you can’t do it in a 2x4hd. The fir is powerful, and once you get the hang of it , it will blow Dirac out of the water. RePhase is a extremely powerful tool and makes very good sounding fir for the 2x4hd! 

make sure to donate to POS for his hard work at rephase.org 

andrew 

I think I’d much rather have a smaller sample rate like when you use Dirac on the 2x4hd it swaps to 48khz and therefore can ‘correct’ down to 49hz with 1024 taps

i did consider using less taps on two channels for my tweeters in my active setup, and more on the bass end to allow corrections at lower freqs, maybe get down to 70hz and only to 2khz or something for tweeters with less taps

1024 at 96khz is 98hz
512 is 196hz
256 is 392hz
128 is 800hz (rounded up)
64 taps is 1600hz
32 taps is 3200hz

so maybe 48 taps on the tweeters for roughly 2.5khz and then use the ‘spare’ taps for getting the midbass lower in correction with 2000taps on midbass so maybe 55hz… but if it was a 48khz sample rate it would be good to 24khz freq response… and then correction down to 28-30hz would be theoretically possible within its processing power

i could also use all pass to help correct the phase in the top end of the tweeters with maybe a low q all pass in the right place using smaart to place it and maybe two cascaded

Yeah that sounds about right 

I have had better results using rectangular windows as well…. 

and yes I also use some all pass in the low almost subsonic range sometimes to pull the midbass around to match up to sub that’s on another dsp 


you can also canter the impulse forward to get better use of your taps, in the low frequencies 

if your trying to get correction to 70-80hz (ie linearizarion) you can try centering the impulse forward , like for example with 1024 taps and a 80hz linearizarion, you can center the impulse around 9ms, 9.33ms perhaps , and adjust taps and centering until happy with 

 

dumdum wrote:

I think I’d much rather have a smaller sample rate like when you use Dirac on the 2x4hd it swaps to 48khz and therefore can ‘correct’ down to 49hz with 1024 taps

i did consider using less taps on two channels for my tweeters in my active setup, and more on the bass end to allow corrections at lower freqs, maybe get down to 70hz and only to 2khz or something for tweeters with less taps

1024 at 96khz is 98hz
512 is 196hz
256 is 392hz
128 is 800hz (rounded up)
64 taps is 1600hz
32 taps is 3200hz

so maybe 48 taps on the tweeters for roughly 2.5khz and then use the ‘spare’ taps for getting the midbass lower in correction with 2000taps on midbass so maybe 55hz… but if it was a 48khz sample rate it would be good to 24khz freq response… and then correction down to 28-30hz would be theoretically possible within its processing power

i could also use all pass to help correct the phase in the top end of the tweeters with maybe a low q all pass in the right place using smaart to place it and maybe two cascaded

I’ve tried short firs for tweets, they do work , however , I realized it worked better to be able to smooth it all the way into the stop band , example I’m using a 256tap fir right now for a 4th order filter at 3000hz , it lets me correct into the stop band down to about 315hz , 

I’ve noticed a better result if the speaker is corrected pretty far into the stop band 


another thing would be natural rolloff, you can add a compensate filter for example a midrange may have a natural rolloff at 5khz that’s a natural 24db rolloff 

if you add a 24db minimum phase filter right over top of it and put in compensate mode , the projected responce is flat and the phase of the filter is flat , then use linearizarions to farther unwrap any phase and phase eq 


understand that that compensate filter could add as much as a 50db  boost as it gets near ultrasonics, so it’s important to add the filter you desire after the drivers natural responce is corrected.

in that example you could add an arbitrary 3000hz low pass , and that will pull down that 50db  boost , but also make it play flat phase all that way into the niquist. 


the compensate feature works tremendously good, and having the stop band phase made flat as well makes a big difference and then alignment of the drivers becomes simple 

geometric alignment with a basic tape measurement distance pretty much nails it every time… no weird off axis stuff that a minimum phase filter would have otherwise had  

 

How do I limit the frequency spectrum in rephase?  I’d like to do 250Hz and up?  Do I just import a measurement starting at that time?   I’d like to have the lightest weight filter I can make

I’m not sure I understand 

ate you asking how to make a filter that only can correct 250hz and up


take fs / 96000 / frequency 

so in your case you need 384 taps if you only want to correct down to 250 (and that means correct into the noise floor 

if you are wanting to cross over at 250 , depending on how steep you want to cross , you’ll want more taps for the stop band ,

so if your 250hz crossover is a 4th order, at 250hz , and noise floor is at 40db and your measurements (listening is at 100db) then your noise floor is -60db  so you would need more taps to get down into noise floor theoretically…
You can simulate it in rephase very simple 

best of luck 
            

Yeah, 250 and up.  Perhaps I don't understand how these work.  But I have a miniDSP HD and use it with video so I want to eliminate as much latency as possible.  They speakers I have crossover at 360Hz (LR 24dB/oct) and 2700Hz (LR 24dB/oct).  I only need to "fix/flatten" the mid and the tweeter for these particular speakers.  I figured limiting the amount of frequency range would allow that.  Perhaps I'm assuming this works differently? 

The 2x4hd with 1024 taps is only 5.33ms 

even if you did 2048 taps that’s only 10.66ms 

that won’t cause lip sync issues with video 


it’s a non issue with this dsp , I would go for 1024 taps , in rephase set your tap length to 1024 
then use IEEE-754 mono.bin as your file type 

use a rectangular window and center at closes perfect impulse 


then make the filter you want , hit generate , as long as the red is on top of the blue it will work just fine even if it’s a tiny bit off it will still work just fine

once you generate the filter in the bottom  under the generate button it will tell you the latency of the filter

as long as your tap lengths  are the same on across the channels The latency should be the same but sometimes it’s not and if it’s not just use the delay on your DSP to offset the values

If you take the time to be patient and learn how to use a rephase you’re going to be astounded how good your filters are going to sound

I would go for a linear phase filter for the highs and low pass 

you certainly don’t have to cross it at 250 with 1024 taps , you can pretty much sit your crossover anywhere above 100 and as steep as 4th order 


 

nvm.  After running it, I do see that it goes backwards and as it "runs out" of taps it is unable to adjust to lowest.  I was actually able to get it down to 256 at 48khz which works for the miniDSP HD and like you said, 384 at 96khz for the Flex. 

This is exciting.  Looks like adding adjustments to the Paragraphic EQ adjustments adjusts phase.  So if my AVR is making paragraphic adjustments, I should take my post EQ measurements, correct? 

You have to be at 96k 

Even if Audyssey and Dirac downsample to 48kHz?

Your not using Dirac , in flex 2x4hd mode native sample rate is 96,000

in flex ddrc24 mode native sample rate is 48,000


your fir in 2x4hd mode has to be at 96k 

otherwise your 200hz filter will be a 100hz filter and any peq will be weird , because it will sample a 48k filter , it just won’t behave properly 

Awesome!  This is helpful!

Do the measurements I import from REW need to be 96 as well or is it just the output from rephase?

Only rephase