5 way horn speaker system project - tapped, bass, mids and tweeter passive active

For now I am choosing a simpler path. A 2.5 way with a JA6681B in JMLC Iwata 300 and two pairs of Seas 10 inch Excel Nextel woofers. Those woofers are excellent at nothing special, and play almost as nice as an 8 inch woofer in the lower midrange, and goes as deep as you’d like - in the right cabinet. I have tried them in single driver 80 L ported cabinets, and now what I think is 65L sealed with four drivers. The sealed cabinet should have been a lot bigger, but this tight midbass I am getting fromt his smaller 65L cabinet is nice too. No need for deep bass all the time.
I have tweeter horns ready to be installed, that could play as low as 1400 Hz and up to 20kHz. I would need to attenuate them down. One thing I want to try is double JA6681B in each horn. Or even triple. It would strengthen the lower midrange.
For XO I went as simple as possible. A cap and a coil. I found that my DAC, the Gustard X16, has signal coming from both the RCA and the Balanced outputs, so I am just feeding one of them to the Papa V-Fet lottery amp at a fraction of 10W and then 500W x 2 class D amps for the woofers. I haven’t even had the time to measure them, but they sound just fine. Always ready to play hits for the kids, with moderate bass, just like how I like it.

That sounds good.
Bit simpler life🙂
I hear very good things about the X16.
The Sonore Ultrarendu + Gustard U18 really upped the front end streaming for me.
  • Like
Reactions: 1 user
Just an fyi that time aligning the first positive peak usually won't give you the correct alignment. Not sure if that's what you're doing since you say you 'adjust for the horn length'. If you look at the impulse responses of a filtered response curve, the peaks of the the impulses of the lower frequency sections arrive later in time. Here's an example plot for a 4 way system with LR4 acoustic slopes where the impulses are normalized in level for easier visualization:

View attachment 1220129
I've tried various timings and alignments over the years.
Even progressive changes from alignment to misaligned forwards and backwards to see what that does to what I hear.

You are right, currently I 1st peak positive align in Holmimpulse, but then take in to consideration the time the soundwave takes to travel the length of the each horn, respectively. As they are all differing lengths, the factors are tailored to each horn pair.
This has given a stunning realism / presentation.
Switching between standard diaphragm 1st peak positive alignment to the above Mouth Alignment (as I call it) demonstrates the change in presentation.
It's not subtle.

Do you have another way I can try?
I'll happily test it.
Last edited:
Speedysteve what a creation.real men listen to horns cause horn speakers are real speakers.
Hehe thanks!

I'm doing it while I have the space the inclination and finding it fantastic.

I was a bit in the wilderness with this room for several years.
It was good and I improved things steadily, but I always felt it could be better. I missed the previous room's sound somehow. Perhaps that was a myth of my mind making though!?

With the recent changes I am flying high, I find the realism and quality of every aspect of the sound fulfilling.

(There's one BUT coming - that I'll explain next) 😂
Having run with the new room config for several days (tapped horns across the end wall firing directly at me), listened, measured and tweaked it in all ways I know how..
It failed to live up to the previous setup.

Before I tried it it was a
"I gots to know" moment (Dirty Harry).
Had to be done.

So what was wrong?
The magical upper mid bass / mid / upper mid cohesion and stunning realism was lost.

No disrespect to hifi, but it sounded like a playback hifi, not an enthralling performance anymore.
Horns done right should wow, should sound right across the board, should do everything impressively.

The magic was lost!

The bass worked well enough. Not as much corner loading for the tapped horns of course.

I don't know why it didn't work?
Perhaps the change in proximity of the side walls to the upper mid horns, doesn't work like the mid horns do - frequency thing?

I didn't mind the slightly higher presentation of the mids. That was actually quite beguiling.
I missed the width a bit though.

But it was the combined realism of the horn grouping that was lost somehow!

So, 500kgs shifted around again, filled and painted the tapped horns a bit, vacuumed the floor, cleaned the skirting boards too, before putting it all back as it was.


Put the Thunderbird 2 reflectors on the tapped horns again🙂 Painted them beforehand too.
Loaded the old DSP preset for this room setup.

Bam - magic returned.

Layout itch scratched. Never again!

Musically threw everything at this set up over a good while. Love it.
Last edited:
  • Like
Reactions: 1 user
Joined 2008
Paid Member
Would you say it was the old presets for this room that captured it?

I was going to suggest that if you were to measure several positions on the room and average them, it's good if your listening position response is the position that more follows the trend of the room.
  • Like
Reactions: 1 user
then take in to consideration the time the soundwave takes to travel the length of the each horn, respectively. As they are all differing lengths, the factors are tailored to each horn pair.
Could you detail a bit more your correction (or did you somewhere else in this thread)? I can't think of any logical way to implement this. If you are measuring the arrival times and aligning off that, then you have already accounted for the lengths of the horns.

The method I use is to calculate the minimum phase response for each of my drivers with crossovers in Matlab. I will have actually already done this as part of my system design. I have a set of target curves and when I set up the system I EQ each driver to the corresponding target curve. Then I calculate the impulse response for each of those calculated targets (like in the previous plot I showed). That shows me what the alignment should be between impulse responses. Then I compare that to the impulse responses of the measured responses and adjust time delays as needed, finishing up checking the summed responses. In reality it usually gets a bit more complicated as the response of the drivers might be tough to completely accurately model, so some experimental adjusting of the delays for the best response might be needed. I also typically do this in an anechoic chamber, so doing it from in-room measurements will probably be more difficult.

I do this because the alignment between impulses varies with the type of crossovers you use, the bandwidth of each driver, whether you compensate a driver's response for the roll offs of the drivers above and below it, etc. For example, 1st order acoustic crossovers require zero delay between sections, but higher order filters will require more. LR4 acoustic slopes typically need around 0.3-0.4 of a wavelength at crossover delay between the peaks of the impulse responses, for example. If you're off by a cycle at crossover, you can get them to sum but the response won't be perfectly smooth. It will usually have a little ripple on both sides of the actual crossover frequency. This can easily be masked by a variety of issues - rough response from the drivers, non-ideal measuring conditions, not compensating for the roll offs of adjacent drivers, etc.
Last edited:
  • Like
Reactions: 1 user
My understanding of the alignment I use is,

When you align the diaphragms be it physically, or delay in DSP, you are essentially hearing the diaphragms 1st peak positive soundwave at the same or very similar time at the mic position.
In my system I can see the fly screens of the drivers (some way behind that is the diaphragm, depending on driver construction) / see the mid bass cone.

What I do is subtract from this delay value, the time in ms it takes for the soundwave to travel the length of the respective horns.

They may be 30cm, 60cm or 100cm long in my case.
The alignment does not begin at the diaphragm with clear line of sight to the mic, it begins earlier, the diaphragms make a soundwave, it travels the length of the respective horns and they emerge into the room aligned as one (well, truly as one - that's perhaps fantasy, but it a nice image in my head:) ).
The soundwaves spill out into the room in unity.

I first tried this alignment method with the horns where they were mounted - for convenience, neither diaphragms aligned physically, or mouths aligned physically.
It was different, had some appealing characteristics, but I won't say night and day better than standard diaphragm alignment.

I had tried physical mouth alignment before, then with standard diaphragm alignment in DSP. I didn't like the presentation much.

After trying the horn path length adjusted alignment, I tried physically aligning the mouths and then the horn path alignment adjustment.
It was a hit. They horns sing in unison.

I also tried adding the horn path length time to the delay. Same result, the horns are mouth exit aligned. To my mind that would take the alignment farther away from time 0 (tapped horns) though.
I tried playing around with the delay to include horn length of the tapped horn, but at those frequencies couldn't hear an difference. So currently leave as is, time 0.

Note: I didn't come up with this. Hornydude mentioned it on another thread and I tried it..

I don't think your method of alignment would be within my technical reach 🙂
Unless there is a tool or tried and tested user guide.
  • Like
Reactions: 1 user
Joined 2008
Paid Member
Consider that the delay may vary for the same horn over different frequencies. This itself should cause you to question how the start of the peak guarantees what you describe.

It's Ok though because I believe the discrepancy can be reconciled in phase. So could delay though. Therefore your overall (group) delay picture between methods has to be revealing.
I know very little sbout the subject, but I have noticed that around 40-700Hz I get peaks and valleys depending on where I am in the room. I think you should aim to correct one tiny area and try to place your head there every time. And then use PC hardware with enough power to calculate the best types of IIR or FIR filters that you can find from those who knows how to set that up.

What I don’t get is how the filter in the computer can know for example that the midrange should not compensate for dips treble or bass. It would seem sensible to combine the crossover and room correctionnon the same filter.
Last edited:
  • Like
Reactions: 1 user
All rather unobtainium for me I fear.
These things sound good, but do I have the where with all to implement them.


I realised I can edit the earliest posts in this thread, so I can put the photos back!
These were lost when PhotoBucket debacle occurred.

Takes me back to 2011, the early days of the conical system construction.

I'll be adding more and more over the next few days..
  • Like
Reactions: 1 users
Joined 2008
Paid Member
SpeedySteve, I think what I'm saying is that a plot of group delay may sometimes show areas for improvement, if you keep in mind the issues that you mustn't compromise but which don't show up in group delay, such as cone breakup for example.

This region is often affected by baffle step, or more accurately is below it enough that the room interaction increases.

is how the filter in the computer can know
Manual control is usually the most successful. Acoustic issues must be fixed before you can expect the most from EQ.
I've been measuring and focusing on the @jtalden procedure.

Aligning in Holmimpulse of my upper mid horn
  • Vitavox S2s on 550Hz Le Cléac'h horn crossing at around 1000Hz 1st order and rolling of no X/O, and
  • mid horn JBL 2482 on T200 horn 2nd order L-R at 300Hz and 750Hz.

I get these first positive response and phase plots;

Green Upper mid horn, Red mid horn


Help interpreting much appreciated.

Esp. this that jtalden wrote, many moons ago

"Adjust the Offset (delay) of the MR Horn as needed to achieve close phase tracking of the direct sound throughout the XO range. [I expect a very clean phase trace in this case but sometimes, particularly when XO <500Hz, there can be room modes or reflections that make phase chart more difficult to properly interpret.]"
Last edited:
I haven't read this thread nor used Holm for many years, so my comments don't consider your project background or specifically how to make the adjustment using Holm.
In General:
I always use the TW as the overall system reference time and adjust the next lower driver (upper mid) to align it. The next lower drive (mid) is then aligned to the upper mid and so forth to the lowest driver.

Regarding your chart:
Considering the upper mid as the reference, we can use Holm options to change the timing of the mid to overlap the phase traces of the 2 drivers as closely as possible. [Note: It is critical that Holm captured the relative timing between these 2 measurements (stream was kept active).] If so, then we can adjust the lower mid in Holm options to create the best overlap and identify the amount of delay change needed.

Per your chart as a ballpark estimate. It appears there is about an 80° offset required to overlap the phase traces in the XO range. Manually estimating the needed adjustment, it is roughly:

0.22 ms delay reduction of the mid. This corresponds to about a 75 mm increase in mid driver distance from the mic located at the MLP.

A remeasure after the change is needed to confirm the result.
  • Like
Reactions: 1 user

After years of the earlier pics on this thread gone missing - due to various image hosting shutdowns,
I've been able to rebuild all the pictures (well 99% of), on this build and evolve a 5 way horn system thread.

Many thanks to moderator Jason and others for making it possible.

I'm now working through the more recent images that are fine and visible, but are not uploaded / attached. So they will be future proof too.

If you fancy looking back at the wood work madness of the early days / and passive to digital evolution etc etc, go back to post one and forwards🙂
  • Like
Reactions: 1 user
That's exactly what I've done👍
Attached all the images.
Hard to get rid of all the invisible when you edit, stubs of missing pics though.
All pics are there though.

Ah .jpg invisible stubs were only on the very 1st post - left over from a previous update by Pano I think?, many moons ago!
Managed to delete them all..pesky things being invisible, until the post was saved!

All my recent updates seem clear.
Last edited:
  • Like
Reactions: 1 user