• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

KEF LS50's converted to active with brickwall linear phase crossovers via rephase and camilladsp!

tomchr

Addicted to Fun and Learning
Audio Company
Forum Donor
Joined
Nov 5, 2018
Messages
984
Likes
2,607
Location
Calgary, Canada
How do you get the cabinet apart? Is the front just attached with some push-in gizmos?

Tom
 
OP
Razorhelm

Razorhelm

Active Member
Joined
Jan 31, 2020
Messages
160
Likes
316
How do you get the cabinet apart? Is the front just attached with some push-in gizmos?

Tom
They have 4 long bolts that go all the way from the back to the front. They are under some caps (which are sadly glued on and can get damaged when you take them off). Once unscrewed the front just pulls off, the drivers are mounted to the front.
 

gnarly

Addicted to Fun and Learning
Joined
Jun 15, 2021
Messages
988
Likes
1,390
They have 4 long bolts that go all the way from the back to the front. They are under some caps (which are sadly glued on and can get damaged when you take them off). Once unscrewed the front just pulls off, the drivers are mounted to the front.
Hi Razorhelm, again, nice project! (spilling over from current thread about active sounds better)

Here's some of my 2c thoughts/findings using linear phase filters, particularly in regards to pre-ring.

To start with disclosure.....my bias/limitations....i must admit I've never hear pre-ring, ......other than maybe a pre-echo like sound when using really long FIR filters (>65k taps @ 48kHz), and used wrongly. (will describe in a moment).

Ok, first looking at electrical xovers only,
it's easy to measure/see that complementary linear phase crossovers do not show any evidence of pre-ring when the two sides of the xover are summed together.
Impulse and step response look as if no xover is in place. And there will be flat magnitude and phase response across the spectrum, no group delay,.....yada.
Same for multiple xovers.
A 5-way, with 4 complementary linear phase xovers, will still have 'perfect response'. (this might not be true if more than two driver sections are summing, like with shallow xovers... ...i dunno because i always use steep.)

Now on to acoustic xovers. IF there is complementary acoustic summation, there will also be no pre-ring.
So this is generally true right on-axis, but can fail off-axis due to geometric lobing etc.
Like said, I've never heard evidence of this...but it's certainly there mathematically. Seems to me we're asking our ears to hear subtle pre-ring, and in lobing zones! ...i dunno Lol
Anyway, my solution for this potential issue is to simply use steep xovers and minimize the width of the frequency range for which this potential problem exists.

Some folks say steep xovers cause directivity mismatches, but I believe the opposite is true.
Directivity matching is a function of the acoustic design, the physical design. It's electricals' job to match that acoustic design up as smoothly as possible.
I've found using steep xovers helps being able to use the exact center frequency that fits the acoustic design. It's much easier to get two drivers to behave perfectly over a narrower range than over wider. (I typically use 96 dB/oct LR xovers.)

I've played with both shallow and steep linear phase xovers quite a bit, and can confidently say steeper has always given me better spinorama results outdoors.
And it's a shit ton easier to implement :D



Ok again, on to what i think does cause pre-ring. The used wrongly part.
Linear phase system high-pass (for a sub), or system low-pass for a tweeter.
There is no complementary summation in those cases to negate the one sided effect.

here's an example
It's the acoustic measurement of a 4-way MEH/synergy using 3 complementary liner phase xovers.

The blue trace has a 100Hz linear phase high pass, and the step response ramps down a ton before time = 0. A Premature -ringer Lol
The green trace has a minimum phase Butterworth 100Hz high-pass. No pre-ring.

4-way MEH 100Hz hpf  BW3 min phase vs LR16 lin phase.jpg


Ok, third time...
I've already talked too much, but the last thing i do to avoid (theoretically audible) pre-ring, is try not to adjust phase on individual drivers independent of their magnitude.
All driver EQs for flattening in-band and out-of-band response used prior to adding xovers, should be minimum phase. Where fix one, mag or phase, fixes the other.
Phase only EQs, like system high-pass and low pass, have no complementary offsets...is my thinking.

So anyway , a lot of fwiw, and ymmv !
 
OP
Razorhelm

Razorhelm

Active Member
Joined
Jan 31, 2020
Messages
160
Likes
316
Very interesting thanks! Try as I might I have never heard any pre ring, I now have a low distortion IEM system I use to compare and I have not noticed any difference, if there is any pre ringing is there it's subtle!

I have a system high pass and low pass so I might try changing those to the minimum phase Butterworth in your example.
 

gnarly

Addicted to Fun and Learning
Joined
Jun 15, 2021
Messages
988
Likes
1,390
Cool. Interesting that even with IEM's, you don't hear pre-ring with system hfp and lpf in place.
The few times I thought I've heard a pre-echo like sound...like a 'suction in the air sound' that a strong firecracker/explosion can make right before it POWS....has been with 65-131k taps, and a steep system high-pass at about 30Hz. Dunno what to think about that really...other than avoid it..

I think IEMs & headphones, or even coaxes and synergies...are the least likely to have off-axis pre-ring potential, because geometric lobing is minimized in all those devices.

Frankly, i think pre-ring mainly falls into the domain of the naysayers...
 
OP
Razorhelm

Razorhelm

Active Member
Joined
Jan 31, 2020
Messages
160
Likes
316
I am inclined to agree, I definitely need to do more reading about it but my gut feel and experience is pre-ringing is a pretty minor issue
 
OP
Razorhelm

Razorhelm

Active Member
Joined
Jan 31, 2020
Messages
160
Likes
316
I found an example I can actually hear!


Using headphones can hear the difference! To me it is subtle, if I stop focusing at all I lose it but can consistently hear it when attentive.

Hopefully this is a worst case example because in that case it is definitely not big issue in music for me.

However I might change the EQ filters to minimum phase and see if it has any effect.
 

dasdoing

Major Contributor
Joined
May 20, 2020
Messages
4,208
Likes
2,673
Location
Salvador-Bahia-Brasil
To me it is subtle

If you use linear phase filters for a while the linear phase example in the video will sound like a drunken man singin. I can identify it playing on my TV while having a shit in the bathroom lol

Hopefully this is a worst case example because in that case it is definitely not big issue in music for me.

it will be an issue very very fast.

BUT WAIT:

you guys are actualy talking about crossovers? in a crossover the HP and LP ringing are inverted and cancel each other out. you still don't want to use brickwall filters since this is only true in the perfect position.
 
OP
Razorhelm

Razorhelm

Active Member
Joined
Jan 31, 2020
Messages
160
Likes
316
Currently it is both EQ and and crossover,

But following the conversations here I am going to switch the EQ to minimum phase
 

Phorize

Major Contributor
Forum Donor
Joined
Apr 26, 2019
Messages
1,533
Likes
2,060
Location
U.K
Nice job, I’m not sure they fit in with the lifestyle aesthetic of the range anymore though;)
 

dasdoing

Major Contributor
Joined
May 20, 2020
Messages
4,208
Likes
2,673
Location
Salvador-Bahia-Brasil
But following the conversations here I am going to switch the EQ to minimum phase

what are you using to EQ the speakers? comercial "idiot proof" solutions like Dirac shouldn't produce any audible artifacts out of the box. it's only when you get to the limits of correction that those appear.
 
OP
Razorhelm

Razorhelm

Active Member
Joined
Jan 31, 2020
Messages
160
Likes
316
REW for measuring and generating the EQ, then rephase to generate the convolution file and camilladsp to apply it
 

dasdoing

Major Contributor
Joined
May 20, 2020
Messages
4,208
Likes
2,673
Location
Salvador-Bahia-Brasil
REW for measuring and generating the EQ, then rephase to generate the convolution file and camilladsp to apply it

just make sure that "excess group delay" of the resulting meassurement doesn't go below zero. actually you can take it a little below (by ear), but if there is nothing below there is no pre-delay at all
 

gnarly

Addicted to Fun and Learning
Joined
Jun 15, 2021
Messages
988
Likes
1,390
I found an example I can actually hear!


Using headphones can hear the difference! To me it is subtle, if I stop focusing at all I lose it but can consistently hear it when attentive.

Hopefully this is a worst case example because in that case it is definitely not big issue in music for me.

However I might change the EQ filters to minimum phase and see if it has any effect.
Hi, i think that FAB filter comparison is a ridiculously contrived marketing piece.
I mean look at the EQ in place +20dB gain, with a Q of maybe 20 ? That's plain nuts to use anytime, FIR or even IIR..
 

gnarly

Addicted to Fun and Learning
Joined
Jun 15, 2021
Messages
988
Likes
1,390
you guys are actualy talking about crossovers? in a crossover the HP and LP ringing are inverted and cancel each other out. you still don't want to use brickwall filters since this is only true in the perfect position.
With brickwall however, the xover range that has theoretically audible pre-ring potential is very, very narrow. Good luck trying to hear any pre-ring off-axis.
I use 96 dB/oct instead of brickwall, simply due to tap limitations, and with 96 dB/oct i have never heard any pre-ring off-axis (or on ).
Haven't heard it even outdoors listening to a single speaker in mono, for the most revealing hearing test i know.

I've tried typical lower order complementary linear phase xovers, like 12-24 dB/oct, but have never listened for pre-ring, because off-axis polars degrade so quickly compared to using steep. So I ditch the low order without bothering to listen much.

If folks are getting pre-ring from linear phase, which I definitely believe they can, i think it is due to improper implementation.
The most common mistakes I see are:
- linearizing the phase of system high-pass and low-pass, as i tried to show in #23.
- trying to counter the phase rotation of existing xovers
- and trying to linearize phase rolloffs of drivers with phase EQ.
All three of those will show various degrees a step response diving below zero (like in #23)
And are easy mistakes to make with any program that is trying to linearize an existing speaker.

I really think tuning individual drivers with minimum phase EQs, then add complementary linear phase xovers is the way to go.
Ime/imo, any attempt to linearize phase when that foundational work hasn't been done, is band-aids on top of band-aids.



Hey, it just hit me, if someone wants to measure whether complementary linear phase xovers cause potentially audible pre-ring off-axis,
they should simply take on and off-axis step responses. I'll try that next time I'm set up for outdoor polars.
Maybe step will show some appreciable pre-ring potential...
that said, even if it does i won't care, cause I sure can't hear any.
 

dasdoing

Major Contributor
Joined
May 20, 2020
Messages
4,208
Likes
2,673
Location
Salvador-Bahia-Brasil
With brickwall however, the xover range that has theoretically audible pre-ring potential is very, very narrow. Good luck trying to hear any pre-ring off-axis.
I use 96 dB/oct instead of brickwall, simply due to tap limitations, and with 96 dB/oct i have never heard any pre-ring off-axis (or on ).

I went way steeper even

sdf.jpg


I ended up with these by trail and error. they are actually brickwall which I rolled-off more and more by lowering taps. I tested with dirac impluses, kick drums and Billy Jean's intro which for some reason reveals pre-ringing very obviously. the steepness is close to what Brugeman from Acourate uses in his paper about crossovers actualy
 

gnarly

Addicted to Fun and Learning
Joined
Jun 15, 2021
Messages
988
Likes
1,390
I went way steeper even



I ended up with these by trail and error. they are actually brickwall which I rolled-off more and more by lowering taps. I tested with dirac impluses, kick drums and Billy Jean's intro which for some reason reveals pre-ringing very obviously. the steepness is close to what Brugeman from Acourate uses in his paper about crossovers actualy
Gotcha. I'll try Billy Jean intro next time I'm outdoors...just tried inside and heard nothing...

I have been using 96 dB/oct LR due to longstanding tap limitations. Had a bank of openDRCs that gave 6k taps per channel @ 48kHz, then went to a Q-Sys Core110 that can do 8k per channel, but not for 8 channels. More like 4k per channel for 8 channels.

Btw, your graph looks like you lowered the brickwall's taps down to something in that 4k-8k tap zone? Am i close?
Did the loss of frequency resolution from lowering taps cause issues?

I'm at 16k taps per channel (48kHz still) and sometimes think I hear better bass resolution.
This reply exchange has motivated me to try brickwalls in place of my LR96's. Got the taps now, just haven't bothered because I've been happy, with no audible problems.
Who knows, maybe measured polars from my DIY synergies will improve some more.

Q-sys has drag and drop FIR high-pass and low-pass, that allow 100dB stopband with transitions that can vary from 1 to 1/4 oct...so a nice way to implement brickwalls without a sharp corner cuttoff.
FirDesigner also has the abiltiy to build brickwalls with varying transition implementations.
I like those ways with softer knees/corners way, as it still allows the full use of available taps for frequency resolution
 

MCH

Major Contributor
Joined
Apr 10, 2021
Messages
2,580
Likes
2,197
I am thinking of doing something similar to the op with a svs 1000sb sub. Well, the opposite actually, from active to passive. Thoughts welcome!
 
Top Bottom