Correction of tri-amplified horn loudspeakers

hulkss

Active Member
Thread Starter
Joined
Feb 12, 2020
Messages
248
I'm moving on to correcting my main loudspeakers after working on the subwoofers in this thread: Correction of a multi-sub array
I learned a lot from Bernt and others in that thread and I expect to here also.

My main speakers are PSE-144 two-way horns. These are combined with Klipschorn bottoms on the right and left and a La Scala II bass horn in the center to complete the three-way loudspeakers.

I've modified the PSE-144's by switching to Celestion CDV1-1757 compression drivers. It's a new special 1" driver designed for crossover as low as 800 Hz at 12 dB/oct. I lost a little top end but they really sound better to me.

Here's a pic of my center speaker as it sits on top of my Infinite Baffle (IB) subwoofer:

39337
 

hulkss

Active Member
Thread Starter
Joined
Feb 12, 2020
Messages
248
The first step is to measure the drivers with gains set for maximum safe output or to your amplifier's power limit on each driver at 0 dBFS. I measured with a -40 dBFS signal using Room EQ Wizard. Here's the Left channel in a center listening position:

39338


Here's the Center channel:

39339


  1. The midrange has a very evident bandpass behavior.
  2. Before I changed compression drivers there was almost no useable Mid-Hi driver overlap.
  3. The Constant Directivity (CD) horn has the expected high frequency roll-off.
In general, a great opportunity for Audiolense to show what it can do. The main purpose of this exercise was to pick the crossover frequencies. Since my subwoofers are up front and center I can go a little higher than typical on crossover and not worry about localization issues.

I selected 105, 350, and 1000 Hz for the crossovers. Here's the setup:

39340
 

hulkss

Active Member
Thread Starter
Joined
Feb 12, 2020
Messages
248
For a Target Room Response I have found that guidelines published by Floyd Toole and Harmon International are pretty good. I'll go with a linear slope down from 20 Hz - 20 kHz being -12 dB at 20 kHz. That may sound like a lot of drop in high frequencies but that's what I need in my room.

The direct sound has plenty of high frequency content for me. I guess this giant bass absorber I built is sucking up the highs that would normally be bouncing around some. I do listen at pretty high volume.

39341
 

hulkss

Active Member
Thread Starter
Joined
Feb 12, 2020
Messages
248
Let's give AL a spin. I'm going to use a minimum delay correction to keep latency low for watching internet shows & movies.

I also now use the Audiolense Convolver. It has some fancy partitioned process that eliminates convolution latency (keeps the CPU busy) and is largely unaffected by filter length. My min-phase filter set (with min phase XO too) has only 8 ms of latency. There is also some latency from the audio buffers. I've got 256 sample buffers working now at 48 kHz.

Here is the correction procedure. I am using partial correction and no bass offloading on the surrounds. They are passive XO speakers and the response varies a lot depending on listening position. No point in doing an extensive correction on them. I shortened the high end cycles to 3 to smooth the filters and avoid over correction.

39342
 
Last edited:

Jiri_ch

Member
Joined
Jun 7, 2020
Messages
10
Hi hulkss
I'll be watching this with interest as I mentioned before I use to own Pse 144 and on several occasions roughly 2 years apart and now thinking to buy 3rd pair if they come available as I do missed their sound
As you know even the original CD roll off a bit so I was trying to use 2.2uf cap and the FR plot of the horn was pretty good .
I did not use it with audiolense but friend of mine is using it now in 4 way and with audiolense so keep us posted
J
39343
 

hulkss

Active Member
Thread Starter
Joined
Feb 12, 2020
Messages
248
Oops, got to measure first. Here's the measurement screen. Check out the delay on those folded bass horns (channels 1 & 4). I would not want passive XO on a Khorn.

39344


Time alignment is a great feature of digital XO. Now this says the mids and tweets are almost dead nuts aligned in time. I think AL is flipping driver polarity to do this. If I measure with REW I see a mid - tweeter delay close to 180 degrees at 1000 Hz. Not sure what to think about that?

Bernt, when does AL do polarity correction? Can I turn it on and off after measurement while looking at it impulses in "remove measurement noise".

Edit: OK, I looked close at my REW measurements. The polarity is correct (same on both drivers), however, the REW delay measurements do not reference the same point on the Mid and Hi impulse waveforms.
 
Last edited:

jjazdk

Member
Joined
Aug 17, 2018
Messages
95
Here is the correction procedure. I am using partial correction and no bass offloading on the surrounds. They are passive XO speakers and the response varies a lot depending on listening position. No point in doing an extensive correction on them. I shortened the high end cycles to 3 to smooth the filters and avoid over correction.

What did you choose in the partial correction, and why? (it is not on the screenshot :cool:)

I like the idea of smoothing out the filters by limiting the cycles, as I don't believe in correction too many small wrinkles.
 

juicehifi

Audiolense
Staff member
Joined
Feb 5, 2018
Messages
697
Oops, got to measure first. Here's the measurement screen. Check out the delay on those folded bass horns (channels 1 & 4). I would not want passive XO on a Khorn.

View attachment 39344

Time alignment is a great feature of digital XO. Now this says the mids and tweets are almost dead nuts aligned in time. I think AL is flipping driver polarity to do this. If I measure with REW I see a mid - tweeter delay close to 180 degrees at 1000 Hz. Not sure what to think about that?

Bernt, when does AL do polarity correction? Can I turn it on and off after measurement while looking at it impulses in "remove measurement noise".

Edit: OK, I looked close at my REW measurements. The polarity is correct (same on both drivers), however, the REW delay measurements do not reference the same point on the Mid and Hi impulse waveforms.
Yes, you can turn polarity correction off ... the measurement menu in the main form.

Glad to see that you are ising partialcorrection here, btw.
 

hulkss

Active Member
Thread Starter
Joined
Feb 12, 2020
Messages
248
What did you choose in the partial correction, and why? (it is not on the screenshot :cool:)

Here's the screen. I chose to do only "Use target as EQ in no correction zone". This works great for me with just a sloped line as a surround speaker target. The surrounds vary too much from seat to seat to do much else.


39345
 

hulkss

Active Member
Thread Starter
Joined
Feb 12, 2020
Messages
248
Here is the frequency response from the above correction settings and sloped target, frequency correction only, minimum phase XO. Sure looks better than that mess in the first post of this thread. Kudos to Audiolense :hail:

39346


Here's the step response:

39347


Bernt, what is that "noise" ahead of the main step above? I have determined that the crossover filters are the cause of it. It is primarily the bass offloading XO. Narrower width filters are definitely worse. Would a different filter type be better for minimum delay applications? Here is the exact same frequency correction with linear phase XO:

39348
 

Omid

Member
Joined
May 28, 2017
Messages
131
I will I was just wondering why you didn’t chose TTD per driver (do you get better results without it?).
 

hulkss

Active Member
Thread Starter
Joined
Feb 12, 2020
Messages
248
I will I was just wondering why you didn’t chose TTD per driver (do you get better results without it?).
TTD is the best correction (sometimes with the per driver option). Initially I'm NOT using linear XO, linear targets, or TTD. I'm working on the best low latency correction (for use with internet streaming movies & shows where lip-sync is an issue). It will then be easy to add TTD for use when latency is not an issue (playback from audio or video digital recordings). Latency with linear filters and TTD can be well over a full second.
 

juicehifi

Audiolense
Staff member
Joined
Feb 5, 2018
Messages
697
Hulkss,

It's hard to tell why you get that rugged step response above. Maybe it has something to do with how you have bundled your subwoofers. The two combos you showed me had pretty nasty impulse responses. Perhaps you are dealing with an impulse here that is literally impossible to correct in the time domain.

You are probably better off with partial correction in the top for front, left and center, btw.
 

jjazdk

Member
Joined
Aug 17, 2018
Messages
95
I am surprised that the step response with/without TTD are so close to one another. There is obviously something I am missing, as I thought the whole point of TTD was to get the phase behavior correct to get an optimal step response, hmm..
Bernt what am I missing/overlooking here?

P.S. On another note. I redid parts of my system over the weekend, and toyed with AL6.16. It finally seems like I can get a really good correction without loosing too much gain (and with sort of similar load on my subwoofers). Guess I should bite the bullet with a purchase very soon :cool:
 

jjazdk

Member
Joined
Aug 17, 2018
Messages
95
Hulkss,

It's hard to tell why you get that rugged step response above. Maybe it has something to do with how you have bundled your subwoofers. The two combos you showed me had pretty nasty impulse responses. Perhaps you are dealing with an impulse here that is literally impossible to correct in the time domain.

I was thinking the same.
If it where me, I would try bulking the 4 Epiks into one sub, and have the IB sub as a second sub. The 4 Epiks should behave nicely when measured in MLP as one, at least that is what I would expect. But the IB sub is sure to have a very different phase behavior than the Epiks, as it is a totally different construction/tuning. Also the IB has at least as much output as the sum of the 4 Epiks, that also speaks in favor of my proposed bundling.
 

hulkss

Active Member
Thread Starter
Joined
Feb 12, 2020
Messages
248
I am surprised that the step response with/without TTD are so close to one another. There is obviously something I am missing, as I thought the whole point of TTD was to get the phase behavior correct to get an optimal step response, hmm..

The two step responses above are with min phase XO and then with Linear Phase XO, neither are TTD, both are to a min phase target. They are identical correction procedures to the same measurement except for the crossover filters.
 

hulkss

Active Member
Thread Starter
Joined
Feb 12, 2020
Messages
248
I was thinking the same.
If it where me, I would try bulking the 4 Epiks into one sub, and have the IB sub as a second sub. The 4 Epiks should behave nicely when measured in MLP as one, at least that is what I would expect. But the IB sub is sure to have a very different phase behavior than the Epiks, as it is a totally different construction/tuning. Also the IB has at least as much output as the sum of the 4 Epiks, that also speaks in favor of my proposed bundling.

Actually The IB + Epiks combine OK. Here is a plot of the combined Epiks (dotted magenta), the IB (dotted green), and the Combination (red) in the main listening position with no correction or gain adjustment. I'll try a two driver sub, 4 Epiks as one driver, and the IB as the other. They will sum the best that way as you suggest since they will get separate corrections.

39380
 

hulkss

Active Member
Thread Starter
Joined
Feb 12, 2020
Messages
248
Hulkss,

It's hard to tell why you get that rugged step response above. Maybe it has something to do with how you have bundled your subwoofers. The two combos you showed me had pretty nasty impulse responses. Perhaps you are dealing with an impulse here that is literally impossible to correct in the time domain.

You are probably better off with partial correction in the top for front, left and center, btw.
Here is the impluse and step from the subs. I hope I we find the cause of the step response problem.

Could you please elaborate a little on the suggestion for top end partial correction? I am stopping the correction with the target after the response drops to cut off that peak at the very top end.

39381


39382
 

juicehifi

Audiolense
Staff member
Joined
Feb 5, 2018
Messages
697
Here's the correction filter of left tweeter after just running a full range correction. Lots of high frequency activity that is not to desire.

39384


Here's everything the same, but with no correction above 15 kHz. Much quieter tweeter.

39383


By tweaking the partial correction stop frequencies, and perhaps by using the target as eq above the no correcton frequency you should be able to achieve a reasonably quiet impulse response, yet one that gives you the frequency balance you want.

The step response with the minimum delay correction: You basically get what you get. High frequencies are front-loaded to avoid high latency and get a faster rise time, low frequencies are delayed to cross over in phase with the high f river.

You may be able to obtain a better result with even wider crossovers than you are using here. I tend to like 4 octaves at low frequencies for minimum delay crossovers.

You may get better results by using wider crossovers. The wider they get, the better they integrate as minimum phase crossovers.
 

hulkss

Active Member
Thread Starter
Joined
Feb 12, 2020
Messages
248
Thanks Bernt, I'll see if I can get a better low latency result and try the two driver sub. I do my critical listening from recorded media, so I'll move on to TTD corrections after I finish with minimum delay.
 

hulkss

Active Member
Thread Starter
Joined
Feb 12, 2020
Messages
248
I'm still wondering why the linear phase XO works so much better with the same min phase correction?
 

jjazdk

Member
Joined
Aug 17, 2018
Messages
95
Is it at all possible to have linear phase XO, without delay..?
With my limited digital filter knowledge it is counter intuitive, since a linear phase FIR filter must have a delay at least as long as the filter phaseshift of the corresponding minimum phase filter. Or at least, that is my understanding, which admittedly may be flawed.

Bernt, could you help out here. Minimum phase XO vs Linear phase XO, but for a real time streaming setup. Is there any difference between the two?
 

hulkss

Active Member
Thread Starter
Joined
Feb 12, 2020
Messages
248
Is it at all possible to have linear phase XO, without delay..?
With my limited digital filter knowledge it is counter intuitive, since a linear phase FIR filter must have a delay at least as long as the filter phaseshift of the corresponding minimum phase filter. Or at least, that is my understanding, which admittedly may be flawed.

Bernt, could you help out here. Minimum phase XO vs Linear phase XO, but for a real time streaming setup. Is there any difference between the two?
I have tried both. Linear XO added a lot of delay. I believe it is XO frequency dependent.
 

hulkss

Active Member
Thread Starter
Joined
Feb 12, 2020
Messages
248
OK, new results with 4 octave wide XO filters. The Left step response is much improved.

39385


The right and center not so much.

39386


If I use the left bass rerouting filter on right and center here's all three looking pretty good.

39387


However, the left rerouting filter does not work on right and center in the frequency domain :gah:

39388

Where is the answer hiding? :ponder:
 
Top Bottom