Welcome, Guest
Username: Password: Remember me

TOPIC: Rephase (for dummies)

Rephase (for dummies) 4 years 4 months ago #24656

  • denver8me4dinner
  • denver8me4dinner's Avatar
  • Offline
  • Gold Boarder
  • 3038564518
  • Posts: 261
  • Thank you received: 52
  • Karma: 6
Hello everyone,

As topic states, I'm sorta a dummie to all this, but understand that idea of it and can put use to most of it .

I am wondering if anyone can help me and explain it in a way a 12yr old would understand. lol. :)

I have 2-2x4hd's with up to 2048 taps divided two ways where I can allocate taps to each channel. Each unit has 4096 taps total into 4channnels. I'm running a 4way system (in a car) consisting of 2-compression drivers, 2-pr170mo 2- Beyma 10G40 and a 15" sub.

So far I've successfully linerized some crossovers. And it makes a huge difference.

Here's my questions.

I have no clue what any of the windowing types mean. I have just been trying all of them to get the best predicted result. Can anyone shed some light on windowing a bit.

Also I can tell there's a way for it to center the impulse. I don't know what any of the options mean or what to do there. Any links or advice there would be great. I love reading about this stuff.

I have been using 44k to make best use of taps (I can't hear diffrance in higher samples so I don't care-I'm streaming 320kbs so whatever) using that nice low (normal) setting I can't get a good prediction with filter linerizing on some crossovers. Also on crossovers I can get good prediction, when I want to add a touch of phase eq to it as well it won't work in the prediction (-160° Q-20 @759hz) in fact when I use the phase eq at all it has Horrable prediction....what am I doing wrong) being I have speakers on my dash there's a lot of combfiltering and reflection so I know I can't do anything with that. I'm just trying to ajust the only spot that the audax goes 180° Relitive. Everything under 130° I ignore.

Given my equipment and available taps and stuff what/how would you do it ?

I have horns crossed at 1.6k
Audax 300-LR12 1.6kLR24
Beyma 80LR24-300LR24
Sub 80-LR48

I'm also using delay on left speakers ...to get a center image. It works .
But with 4FIR banks all doing something a bit diffrent how do I calculate (for dummies) how much delay that will be subsequent from each convolution? Is the "latency" time the number I offset in the signal delay part of DSP to realign after convolve?

Cheers,
Andrew
Last Edit: 4 years 4 months ago by denver8me4dinner.
The administrator has disabled public write access.

Rephase (for dummies) 4 years 4 months ago #24662

  • pos
  • pos's Avatar
  • Offline
  • Platinum Boarder
  • Posts: 350
  • Thank you received: 196
  • Karma: 39
Hi Andrew,
denver8me4dinner wrote:
I have no clue what any of the windowing types mean. I have just been trying all of them to get the best predicted result. Can anyone shed some light on windowing a bit.
When doing phase linearization only recangular windowing often give the best results.
But you can try other algorithms and check the resulting curves: what you see is what you get.
Also I can tell there's a way for it to center the impulse. I don't know what any of the options mean or what to do there. Any links or advice there would be great. I love reading about this stuff.
Not sure what you mean here.
The "middle" centering setting will make sure the peak of the correction impulse is centered.
I have been using 44k to make best use of taps (I can't hear diffrance in higher samples so I don't care-I'm streaming 320kbs so whatever)
You have to use the sampling rate of your convolution device, which I believe is 96kHz for the c2x4HD with the current plugin.
The sampling rate of your source does not matter as everything will be resampled to 96kHz in the device prior to applying the FIR.
using that nice low (normal) setting I can't get a good prediction with filter linerizing on some crossovers. Also on crossovers I can get good prediction, when I want to add a touch of phase eq to it as well it won't work in the prediction (-160° Q-20 @759hz) in fact when I use the phase eq at all it has Horrable prediction....what am I doing wrong) being I have speakers on my dash there's a lot of combfiltering and reflection so I know I can't do anything with that. I'm just trying to ajust the only spot that the audax goes 180° Relitive. Everything under 130° I ignore.
Not sure I am following you here.
Beware with phase EQ, and do not try to linearize comb filtering :p
Given my equipment and available taps and stuff what/how would you do it ?

I have horns crossed at 1.6k
Audax 300-LR12 1.6kLR24
Beyma 80LR24-300LR24
Sub 80-LR48

I'm also using delay on left speakers ...to get a center image. It works .
But with 4FIR banks all doing something a bit diffrent how do I calculate (for dummies) how much delay that will be subsequent from each convolution? Is the "latency" time the number I offset in the signal delay part of DSP to realign after convolve?
The delay that rephase gives you is to be substracted to the one you would normally set to align your drivers.
If you use the same number of taps and the "middle" centering then all these delays will be the same and you can ignore them.
But if you want to make the most of the taps at hand then you will want to distribute them unequally across channels, and use either the "energy" centering option of place the center yourself. In those cases you absolutely have to take note of the final delay implied by the FIR (ie the position of the peak within the FIR) and take it into account when aligning your drivers.
https//wavetracing.com | rephase.org
The administrator has disabled public write access.
The following user(s) said Thank You: devteam

Rephase (for dummies) 4 years 4 months ago #24694

  • denver8me4dinner
  • denver8me4dinner's Avatar
  • Offline
  • Gold Boarder
  • 3038564518
  • Posts: 261
  • Thank you received: 52
  • Karma: 6
Pos,
Thank you for the help. I spend most the day figure it out. I'm really getting the hang of it now.
It really works very good!

So just to be clear, after I generate a file on the bottom it says the filter delay in ms. For instance I have a filter with a 11.61ms delay, so the output on that channel is now delayed 11.61ms?
I think I'm starting to understand why minidsp has 80ms of delay on board!

So the sub filter I generate has delay of 22.28ms . The mid & midbass is 11.61ms and horn has no fir so it has no delay.

So I should delay the horn 22.8ms in the minidsp and the mid/midbass 10.67ms and all will have 0delay at outputs correct?

Also, linerize crossovers works very good, but when I make a linear crossover in rephase the graph and predicted grap lines up and shows a filter , (the red is on top of the blue) it looks good, but when I import it into minidsp the minidsp plot graph shows a different outcome than what rephase shows. It's sorta the same shape but has totally diffrent slopes. The rta seems to agree with the minidsp graph. Is this a known bug or am I not doing something right.

The 2x4hd has been working at 44.1, it says it will take all types of sample rates. So I'm going with 44.1 unless you can think of a reason not to.
I only have 80ms of alignment delay so longer filters might not work.

Last question (thank you) if the prediction shows a good result with a lower number of taps will it perform as good as a filter with a higher number of taps. Again given the prediction looks the same.

Cheers,
Andrew
The administrator has disabled public write access.

Rephase (for dummies) 4 years 3 months ago #24895

  • pos
  • pos's Avatar
  • Offline
  • Platinum Boarder
  • Posts: 350
  • Thank you received: 196
  • Karma: 39
Hi Andrew,

Sorry for the delay.
I was on vaccatrion and still have a lot of catching up to do...
denver8me4dinner wrote:
So just to be clear, after I generate a file on the bottom it says the filter delay in ms. For instance I have a filter with a 11.61ms delay, so the output on that channel is now delayed 11.61ms?
That is correct. But remember in the end the total delay does not matter (unless you need to sync with something that cannot be delayed, or you are doing a live recording/monitoring). What matters is the relative delay between channels, so you only have to compensate for the differences.
So the sub filter I generate has delay of 22.28ms . The mid & midbass is 11.61ms and horn has no fir so it has no delay.

So I should delay the horn 22.8ms in the minidsp and the mid/midbass 10.67ms and all will have 0delay at outputs correct?
Not sure what you mean here :)
Adding delay to delay can only increase it (well, I think :D)
As said above what matter in the relative delay, and of course you have to take into account the geometrical distance between emitting surfaces.. And that is only if you end up with perfectly complementary crossovers, which is not that easy in real world scenarios.
Once you have compensated for delay differences (both from the FIR and from geometry) the best solution is to use the "reverse polarity" trick on one channel and search for the deepest null on your listening axis. That will give you the flattest summation and off axis behavior when you turn everything back to normal polarity.
Of course you have to do it a pair of drivers at a time.
Also, linerize crossovers works very good, but when I make a linear crossover in rephase the graph and predicted grap lines up and shows a filter , (the red is on top of the blue) it looks good, but when I import it into minidsp the minidsp plot graph shows a different outcome than what rephase shows. It's sorta the same shape but has totally diffrent slopes. The rta seems to agree with the minidsp graph. Is this a known bug or am I not doing something right.
Looks strange. Did you bypass the measurement before looking at the correction curves in rephase?
The 2x4hd has been working at 44.1, it says it will take all types of sample rates. So I'm going with 44.1 unless you can think of a reason not to.
It can accept 44.1kHz signal, but it will convert any sampling rate to its internal sampling rate, which is 96kHz in the current firmware/plugin.
So that is the frequency you have to aim for when generating the FIR in rephase.
This might be the reason for the difference you see in the rephase curves and in the miniDSP interface...
I only have 80ms of alignment delay so longer filters might not work.
Luckily 80ms is far beyond the current FIR capabilities of the 2x4HD ;)
2048 samples at 96kHz is only 21ms
Last question (thank you) if the prediction shows a good result with a lower number of taps will it perform as good as a filter with a higher number of taps. Again given the prediction looks the same.
Yes
https//wavetracing.com | rephase.org
The administrator has disabled public write access.
The following user(s) said Thank You: devteam, denver8me4dinner
Moderators: devteam