S. Harsch XO

I don't know, I am not familiar with JMLC xo. Do you have a link?


JMLC quasi-optimal crossover

For a crossover frequency normalized for 1 kHz :

hi-pass : 18but at 1.1456 kHz (-3 dB), inverted polarity.
--
lo-pass : 18but at 0.8729 kHz (-3 dB), 220 µS time advance
(= 220µS time delay in the high pass section).


You'll find details in french here :

filtrage

You can play with this calculation sheet

http://nicolas.davidenko.perso.sfr.fr/outils/filtre_simul.zip

Many different quasi-optimal crossover configurations, the JMLC one being the first to be popularized, have been found with it.

There is a huge amount of discussions about them on the forums in France.

The whole idea with quasi-optimal crossovers is to get better group delay curves of the sum of the outputs than those obtained with standard configurations, and this without significantly impairing the axial responses (and preferably, with keeping the phase difference between the two outputs as small as possible).

Quasi-optimal crossovers are achieved by playing with different cut-off frequencies of the ways and a time delay for one of them.
Samuel Harsch's crossover is a typical example, it is known in France since 2009.
 
Last edited:
Founder of XSA-Labs
Joined 2012
Paid Member
In a previous life at a commercial speaker factory asymmetrical (passive) crossovers were the norm. Having said that, the Harsch, and XRK's derivatives, look like a neat and maybe repeatable solution for some applications. Can't wait to try this on my PAudio Coaxs !

Interesting that you bring up coax's: before I ever learned of Harsch's XO, I accidentally ran into a transient perfect step response from my coaxial Beta 8cx with CD in a dipole Z baffle using standard LR2 in an electrically symmetric setup! Acoustically it was asymmetric as can be see below:

491684d1435901325-rockin-kazba-dipole-k-aperture-z-baffle-dipole-kazba-beta8cx-xo.png


And here was IR and SR:
491687d1435901325-rockin-kazba-dipole-k-aperture-z-baffle-dipole-kazba-beta8cx-ir.png


More info on this alignment here:
http://www.diyaudio.com/forums/full...zba-dipole-k-aperture-z-baffle-dipole-19.html

It had me scratching my head for some time - but now I see that there are a family of asymmetric XO's that when you adjust delay, you can get a right triangle shaped SR.
 
Coax drivers are a nice platform for several different possible transient-perfect topologies. Many TP crossovers have bad lobing behaviour which is less of a problem with a coaxial arrangement.
The fact that the tweeter is usually a little delayed relative to the woofer can also be of help in this respect.

Here you can see the step response of a Beyma 8cx800Nd with a subtractive crossover with 2nd order HP and 1st order LP. There is an additional delay for the woofer using a Bessel lowpass at some kHz. Because the group delay of the woofer isn't Bessel-like it might be better to use another filter type for the lowpass that gives a better group-delay consistency but that would have to be found by experimentation.
The measurement is inverted however but that doesn't matter.


Here
Regards

Charles
 

Attachments

  • coax_step_xov.jpg
    coax_step_xov.jpg
    110.5 KB · Views: 1,679
Founder of XSA-Labs
Joined 2012
Paid Member
I am going to explore coaxials more now and see how a Harsch XO works on them. I would imagine it will look pretty good. The XO dependent lobing issue will be minimized and that is a bonus. Where is a good frequency for XO on an 8in coax and a 12in coax to have the off axis uniformity in dispersion? Typical recommended XO freq from manufacturer is in 2.5k to 3.5kHz range. Should it be lower?
 
X, great work as always, but I'm not convinced that there are benefits to having flatter phase. Have you done any testing that suggests you can hear the difference between perfectly flat phase and phase changes introduced by, say, LR4 crossovers? Look at rePhase. It will allow you to play with such things.

Most of the differences that people hear in different crossover topologies come from differences in the on- and off-axis frequency response.

Good points in words by your last line and to investigate would need some serious time and setups to put right and admit i haven't studied, but my subjective opinion for exactly the LR4 exercise you mention i can toggle on/off on the fly inside JRiver in the two way seen below and here must say i support X opinion it makes a difference to the better side. Some few tracks can be immune but the rest and most think really shine especially if i listen lets say an hour with and with out the IIR LR4 XO phase neutralized by the RePhase FIR filter. Subjective think it adds warm to the attacks, a little bit same as if my old pure DC coupled Sansui power amp input is feeded from a low offset DC pre sounds a little warmer than if feeded from a AC coupled pre.

(LINK: http://www.diyaudio.com/forums/full...3in-5in-drivers-round-2-a-46.html#post4397112)
attachment.php



Next stepup for impressive looking SR is those that linearize phase down low even FR fall off as the SR seen at the very bottom this post, this i also tried out and made it with RePhase, it can sound impressive but a thought here is that for old recordings they were probably not mixed with such filter so a little afraid it becomes a effect in that producer probably had more normal group delay when doing his mix. Below captured picture show bwaslo answer to xrk971 regarding the nearly DC coupled looking SR seen at the very bottom this post.

bwaslo (LINK: http://www.diyaudio.com/forums/full-range/273971-group-delay-questions-analysis-21.html#post4333659)
attachment.php





DIYA REAL WORLD STEP RESPONSES WHEN IT GETS CLOSE TO ART :D:


Barleywater (LINK: http://www.diyaudio.com/forums/full...-performace-testing-platform.html#post4371147)
490543d1435333812-fast-tc9fd-dsp-performace-testing-platform-ir-step-response.png


wesayso (LINK: http://www.diyaudio.com/forums/full...er-full-range-line-array-127.html#post4391729)
impulsestep.jpg


bwaslo (LINK: http://www.diyaudio.com/forums/full-range/273971-group-delay-questions-analysis-15.html#post4332804)
IIRd%20at%20LP%20sr.png
 

Attachments

  • FAST.PNG
    FAST.PNG
    88.4 KB · Views: 2,771
  • bwaslo.PNG
    bwaslo.PNG
    62.2 KB · Views: 2,769
Last edited:
Founder of XSA-Labs
Joined 2012
Paid Member
Bwaslo's SR seems to go to infinity! That is art! Can you imagine the punch of a tom tom you would feel with this speaker?

Wesayso's, Barleywater's, and Byrtt's SR's - all aided with FIR techniques are all art in my book. I have yet to do the FIR route but may be going there via a hardware route with miniSHARC approach one of these days. For how low cost and easy the IIR route with IIR filters is though, I am having plenty of fun and good listening at the moment with my speakers.

The great thing about all of this is, I am sure MOST (98%) of speakers sold commercially don't even think about trying to achieve this. When you hear it, you know it. If you have not heard it, you won't know you are missing it, but if you do happen to hear it... well all I can say is that your existing speakers that don't have it, won't seem so great anymore :eek:

It is not on all tracks, but it is an effect that gives the sense of listening to the actual instrument playing and not that of a good recording and good speaker playing that instrument.
 
The more artificial a recording is the less important is the transient accuracy. Minimalistic recordings of natural instruments profit the most. I.e. for more than 95 % of available source material it doesn't matter at all.
But since we know how to achieve it then why not try it ?

From the information theory point-of-view every deviation form the original input signal is a loss of information. In other words: An information channel with less phase-distortion has a higher channel capacity. But it is only of value if the whole chain from source to sink can actually make use of it.

Regards

Charles
 
Founder of XSA-Labs
Joined 2012
Paid Member
The more artificial a recording is the less important is the transient accuracy. Minimalistic recordings of natural instruments profit the most. I.e. for more than 95 % of available source material it doesn't matter at all.
But since we know how to achieve it then why not try it ?

From the information theory point-of-view every deviation form the original input signal is a loss of information. In other words: An information channel with less phase-distortion has a higher channel capacity. But it is only of value if the whole chain from source to sink can actually make use of it.

Regards

Charles

This could be why I really notice the impact on Eric Clapton's "Unplugged" album which appears to be a minimalist produced ambient live recording in a small venue.

Another recording I have "Live at Smalls" which are live recordings from a small jazz club also seem very good.
 
Founder of XSA-Labs
Joined 2012
Paid Member
Harsch XO compared with 2nd Order Linkwitz Riley XO

The Harsch XO was implemented at 400Hz electrical frequency but acoustic XO is closer to 500Hz. A 1.62ms delay between mid/tweeter and woofer was used:

496335d1438327415-s-harsch-xo-harsch-xo.png


The LR2 XO was implemented at 310Hz for the low pass on the woofer and 400Hz for the high pass on the mid/tweeter to get an approximate 500Hz acoustic XO frequency. A 0.22ms delay was used to time align the drivers:

496336d1438327415-s-harsch-xo-lr2-xo.png


Here is the comparison of the step response, the LR2 XO tweeter goes the wrong way (and this can be seen in the time series trace below):

496338d1438327415-s-harsch-xo-sr-compare.png


Here is a screenshot of the time series for the Harsch XO (top) and LR2 XO (bottom) - as you can see, the traces are sometimes 180 deg out of phase.

496337d1438327415-s-harsch-xo-harsch-vs-lr2-time-series.png


Here is the comparison of the phase for both - Harsch phase varies a lot less and doesn't wrap:

496339d1438327415-s-harsch-xo-phase-compare.png


Here is the comparison of the group delay - the Harsch has about 1ms less GD:

496340d1438327415-s-harsch-xo-gd-compare.png


The sound clips are below, change the .asc extension to .mp3 in order to listen. They are anonymously labeled A and B so you won't be biased when you listen.
 

Attachments

  • gd-compare.png
    gd-compare.png
    69.5 KB · Views: 2,388
  • phase-compare.png
    phase-compare.png
    55.6 KB · Views: 2,823
  • sr-compare.png
    sr-compare.png
    47.2 KB · Views: 2,840
  • Harsch-vs-LR2-Time-Series.png
    Harsch-vs-LR2-Time-Series.png
    112.8 KB · Views: 2,486
  • lr2-xo.png
    lr2-xo.png
    98.4 KB · Views: 2,866
  • harsch-xo.png
    harsch-xo.png
    94.7 KB · Views: 2,875
  • Clip-2-A-XO.asc
    1.7 MB · Views: 204
  • Clip-2-B-XO.asc
    1.7 MB · Views: 160
  • Clip-3-A-XO.asc
    1.7 MB · Views: 144
  • Clip-3-BXO.asc
    1.7 MB · Views: 138
Last edited:
I think the B XO sounds better. The plucking of the strings at the beginning of clip 2 and the punch of the toms in clip 3 sound more natural.

But then again I don't go to many live acoustic performances and don't play myself so it could be I'm just more used to the LR2 XO.

I listened using monitor audio s2 with NAD c356Bee and ppsl sub http://www.diyaudio.com/forums/subwoofers/268803-15-plenum-ppsl-3.html
 
Last edited:
Founder of XSA-Labs
Joined 2012
Paid Member
Unless you have transient perfect speakers - you should listen using good single driver full range headphones. Typical multiway speakers will not reproduce transient perfect like a full range headphone. I say full range headphone because they make multiway (2 & 3 way) driver balanced armature in ear monitors that have passive XO's.
 
Sound clips Harsch XO compared with 2nd Order Linkwitz Riley XO

Listened the sound clips via below cheap full range monitors with two mods one is input cap is shorted with a wire and second mod is the value for electrolytic output capacitor this power amp unfortunate needs : ( is upped several times with some think decent quality Panasonic FC series wired back to back, Player is JRiver MC 20, soundcard cheap UCA222, think difference is clear for me with this setup therefor too lazy rig some headphones : )

Result is in the attached txt file.

Subjective think better step response adds warm to the attacks as a low offset DC coupled amp chain can do compared same when AC coupled.
 

Attachments

  • EDIROL MA-5A.PNG
    EDIROL MA-5A.PNG
    384.5 KB · Views: 256
  • Harsch XO compared with 2nd Order Linkwitz Riley XO.txt
    151 bytes · Views: 238
Last edited:
Founder of XSA-Labs
Joined 2012
Paid Member
Listened the sound clips via below cheap full range monitors with two mods one is input cap is shorted with a wire and second mod is the value for electrolytic output capacitor this power amp unfortunate needs : ( is upped several times with some think decent quality Panasonic FC series wired back to back, Player is JRiver MC 20, soundcard cheap UCA222, think difference is clear for me with this setup therefor too lazy rig some headphones : )

Result is in the attached txt file.

Subjective think better step response adds warm to the attacks as a low offset DC coupled amp chain can do compared same when AC coupled.


Byrtt,
Apparently, a transient-perfect full range monitor helps because you picked it correctly. :)