THAM15 - a compact 15" tapped horn

djim:
A little extra info on mouthpieces, and a pic of one of my favorites... a berg.

Changing the baffle angle even a degree is audible on a mouthpiece. This is true on bari and bass sax, as well as the rest of the family. The guys in the 50's used to take files to their mouthpieces, creating vortex generators, different baffle angles, etc. Many times a guy's "tone" was largely due to weeks or months spent slowly modifying mouthpieces. A berg vs a link mouthpiece is night and day difference...

Now is this related to putting in a baffle/reflector on the 'throat side' of a TH segment? who knows.....
 

Attachments

  • berg.JPG
    berg.JPG
    30.7 KB · Views: 1,066
Brian we all know it’s about phase differences when dips become peaks and visa versa, that’s logic.

Your questions:
“1.) how can changes to the bends affect the overall response, 2.) when these changes are many times smaller than the "acoustic dimensions" required to affect the wavelengths being considered.”

Answers:

1.) The Flute theory fits and explains that it’s possible, no question about it. This is also confirmed trough HornResp. I used L23 as an open tube, calculated its harmonics and looked at HornResp. Al the harmonics show up at exactly the points of important dips and peaks. Coincidence? From a scientifically point of view not very likely since all harmonics, not just a few, showed up at these places.

2.) The flute theory fits again and explains that it is possible. Just read all about it. We also know that funny ¼ wave boxes can produce moving air into full wavelength tones… For people who don’t know these funny boxes, just call them ...…:D

FACTS:

Findings of JBell, Anders and even you - Dips that are deeper then predicted.

Cause: Probably building limitations causing a different path from the ideal path predictions in HornResp.

Question: Do these dips have anything in common with the suggestion that just a section can work as an individual open tube?

Answer: Could be, if you look backwards you will see that the lengths of certain sections (especially those who end with 90 degrees) produces the same harmonic (following the flute theory of open tubes) frequency as the dip that shows up in real live measurements. Coincidence? Don’t know yet because I only used the Tham so far…

Question: is there a theory? Probably, although I haven’t found the right one yet but all brass instrument makers describe the same thing…

(for instance read page 75 of Bart Hopkin’s book : Practical information for instrument making).

“Short, fat air columns are generally poorer in overtones than long skinny ones.”
“A straight tube has some acoustic advantages. The problems associated with bends should be negligibly small, though, if all bends are reasonably gentle and gradual”

Does it prove something even without knowing the right theory behind it? Of course it does or you deny to accept Jbells and Anders (with many more) findings and the ancient knowledge about ‘wind’ instrument making?

Do we need to know the right theory?

Not for making TH's although I would like to hear...

My advise: "just try to think out of the box" ;)
 
Last edited:
djim:
A little extra info on mouthpieces, and a pic of one of my favorites... a berg.

Changing the baffle angle even a degree is audible on a mouthpiece. This is true on bari and bass sax, as well as the rest of the family. The guys in the 50's used to take files to their mouthpieces, creating vortex generators, different baffle angles, etc. Many times a guy's "tone" was largely due to weeks or months spent slowly modifying mouthpieces. A berg vs a link mouthpiece is night and day difference...

Now is this related to putting in a baffle/reflector on the 'throat side' of a TH segment? who knows.....

Thanks Jbell, I'm learning a lot. Actually got inspired by you...and learning more and more... For instance I need to quit smoking....
009.jpg


Just picked up this old Besson...
 
As I stated earlier:

“If you use the rules of harmonics in tubes, both ends open and see L23 that way and suddenly all dips in HornResp show up at predictable points, time after time (lucky guesses right) The only thing I haven't figured out why some are peaks instead of dips...
Theoretisch%20model%20TH.jpg

Actually the answer is with in the diagram, I think. The two lower resonances that end up being peaks IMO is the cause of the ‘virtual’ second driver. While it backfires a positive resonance into L23 it cancels the dip. While it backfires it also fires forwards a positive resonance at the same freq and ends up as a peak.
Does this make sense?
 
There are a lot of things to think about first.

Which/who's HR models are you using?

Is said model exact enough? If it isn't 4 sections it isn't. ;)

How far off are the FR graphs?

How are you accounting for the extra volume in the corners?

How are you accounting for the 180deg bend with no real taper?

If using the MKII did you adjust the path lengths?
 
Last edited:
I've got a possible "fix" for the THAM15 notch. Well, my rough tests to attempt to fix a similar "problem" with my POC#1 shows some promise.

Here's a before/after FR graph from 30 Hz to 300 Hz - before in blue, after in green. Looks like a rough stab at the fix fills in a good portion of the notch, at the expense of a dB or so @ the low end of the passband (which you can likely get back by decreasing S1).

With some fine-tuning, it should be possible to smooth out the FR at the upper end of the passband, particularly considering the crude method I used to alter the FR that way.

The fix (1) doesn't require any major changes to the box dimensions, (2) doesn't involve any sort of "horn voodoo" that's not backed up by known theory, and (3) can quite likely be modelled by Akabak, but probably not HornResp (not enough "taps" along the line's path to model the effect).

For the moment, I'll call it the "Brian Steele's Dog Food" method. Just wish Akabak had a nifty realtime modelling wizard like HornResp to make the iteration process a bit easier.

Interested? :)


Here's my first run at using the technique to reduce the notch in my POC#2, which shares the same folding topology as the THAM15. I'm still fiddling with it to see if I can improve it any further. Green is the original response, blue is the "dogfooded" response. I think that dip at 70Hz might be a measurement artifact, as it didn't show up on previous tests.
 

Attachments

  • response3.jpg
    response3.jpg
    146.3 KB · Views: 1,010
Here's my first run at using the technique to reduce the notch in my POC#2, which shares the same folding topology as the THAM15. I'm still fiddling with it to see if I can improve it any further. Green is the original response, blue is the "dogfooded" response. I think that dip at 70Hz might be a measurement artifact, as it didn't show up on previous tests.

Second run at it. This is about as close as I can get with the rough method I'm using. The first image is the raw response, the second is 1/6th octave smoothed. Perhaps a smooth 50 Hz (or even lower) to 250 Hz should be possible :). Again, I think the response at the low end can be boosted by narrowing S1 a bit, but that remains to be tested.
 

Attachments

  • response4.jpg
    response4.jpg
    118 KB · Views: 971
  • response5.jpg
    response5.jpg
    110.7 KB · Views: 230
Are you going to explain your "Dog Food" method, or just wait for us to guess? ;)

LOL. I was spending a bit more time trying to research the effect, empirically :).

It's simple, really. Remember when I did my experiment to demonstrate that modifying the bends in my tapped pipe had little or no effect on the passband if the cross-section around the bend was not significantly altered?

Well, after doing that, I just messed around with the cards a little at other locations in the pipe to see if they would produce any significant response changes at other locations. I noticed that, at one particular point in the pipe, if I adjusted them to have a significant effect on the cross-section in that area, the "notch" started to fill in, without significantly impacting the rest of the passband. The effect would significantly reduce, if not disappear, if I moved the obstruction caused by the cards only a inch or two forwards or backwards in the pipe. I also placed the mic *in* the pipe in that location, and set TrueRTA to Rel mode, the graph started to show a peaked response around the same frequencies that the notch appeared in the tapped-pipe's response curve. Coincidence? Perhaps.

I thought that perhaps I could remove the notch even more if I used something more solid than the matt-board cards to act as an obstruction. But what to use...?

Then I saw the cans of dogfood that my eldest daughter left on the kitchen counter.

The pictures below are self explanatory :).

One shows the change to the tapped-pipe that seemed to produce the best results, and the other shows the change to the tapped-horn that seemed to produce the best results also (luckily I didn't permanently attach that particular panel!). In both cases, I'm pretty sure the effect can be modelled in Akabak as constrictions in the path at those points (no "horn voodoo" involved).

In the case of the tapped-pipe, the pipe is being constricted about 2/3rds of the way along its length. In the case of the tapped-horn, I'm not sure what would be the best way to model the effect - the 2/3rds point appears to occur right on the bend - maybe the arrangement of the cans converts that area into a resonant chamber that notches out the frequencies that cause the cancellation measured at the mouth. In any case, it seems to work.

In both cases, shifting the cans just an inch or two back and forth results in the effect being greatly diminished. It's VERY position dependent.

This weekend, if I get the opportunity, I'm going to try to replace the cans with something else a bit more fitting to see if I better the response graphs I previously posted.

Tuning horns with dogfood cans - who knew? :)
 

Attachments

  • 20110126-dogfood.jpg
    20110126-dogfood.jpg
    137.3 KB · Views: 1,224
  • 20110127-dogfood.jpg
    20110127-dogfood.jpg
    142.3 KB · Views: 1,219
I figured it was this. I was wondering if you were using cans, or shaping it with your hands. ;)

My question is what happens as you turn it up? I would expect the lowend to continue to possibly suffer more and more attenuation. Can you check without upsetting the neighbors?
 
I figured it was this. I was wondering if you were using cans, or shaping it with your hands. ;)

My question is what happens as you turn it up? I would expect the lowend to continue to possibly suffer more and more attenuation. Can you check without upsetting the neighbors?

Had a simple test sim going earlier based on the one I posted in your spreadsheet thread.

Code:
System 'DEMO1'

Def_Driver 'Driver'
dD=26cm dD1=11cm tD1=6.5cm |Cone
t1=1.1cm
fs=39Hz Vas=92L Qms=7.5
Qes=0.31 Re=5.2ohm Le=0.37mH ExpoLe=0.618


Def_Const
{
S1 = 4.8e-2;
S2 = 6.3e-2;
S3 = 8.2e-2;
S4 = 9.6e-2;
S5 = 10.9e-2;
S6 = 10.1e-2;
S7 = 10.1e-2;
S8 = 11.8e-2;
S9 = 13.1e-2;
S10 = 12.1e-2;
S11 = 12.1e-2;
S12 = 14.5e-2;
S13 = 17.1e-2;
S14 = 16.1e-2;
S15 = 17.3e-2;
S16 = 15.8e-2;
S17 = 13.6e-2;
S18 = 13.6e-2;
S19 = 14.8e-2;
S20 = 13.4e-2;
S21 = 11.8e-2;
S22 = 13.6e-2;
S23 = 16.3e-2;
S24 = 15.3e-2;
S25 = 16.3e-2;
S26 = 17.7e-2;


L1 = 20e-2;
L2 = 26.7e-2;
L3 = 2.5e-2;
L4 = 3.3e-2;
L5 = 2.0e-2;
L6 = 33e-2;
L7 = 3.0e-2;
L8 = 3.9e-2;
L9 = 2.5e-2;
L10 = 30.5e-2;
L11 = 4.0e-2;
L12 = 5.2e-2;
L13 = 3.7e-2;
L14 = 4.0e-2;
L15 = 5.4e-2;
L16 = 4.0e-2;
L17 = 17.0e-2;
L18 = 3.0e-2;
L19 = 4.7e-2;
L20 = 4.0e-2;
L21 = 4.2e-2;
L22 = 4.8e-2;
L23 = 3.4e-2;
L24 = 13.9e-2;
L25 = 20.9e-2;
}


Driver Def='Driver''Driver'
Node=1=0=100=3

Duct    'Dogfood'  
Node=101=102
dD=10cm  Len=1.8cm




Waveguide 'Horn segment 1'
Node=3=2
WTh=41.4cm HTh={S1}
WMo=41.4cm HMo={S2}
Len={L1}

Waveguide 'Horn segment 2'
Node=3=4
WTh=41.4cm HTh={S2}
WMo=41.4cm HMo={S3}
Len={L2}

Waveguide 'Horn segment 3'
Node=4=5
WTh=41.4cm HTh={S3}
WMo=41.4cm HMo={S4}
Len={L3}

Waveguide 'Horn segment 4'
Node=5=6
WTh=41.4cm HTh={S4}
WMo=41.4cm HMo={S5}
Len={L4}

Waveguide 'Horn segment 5'
Node=7=6
WTh=41.4cm HTh={S6}
WMo=41.4cm HMo={S5}
Len={L5}

Duct 'Duct segment 6'
Node=7=8
WD=41.4cm HD={S6} Len={L6}

Waveguide 'Horn segment 7'
Node=8=9
WTh=41.4cm HTh={S7}
WMo=41.4cm HMo={S8}
Len={L7}

Waveguide 'Horn segment 8'
Node=9=10
WTh=41.4cm HTh={S8}
WMo=41.4cm HMo={S9}
Len={L8}

Waveguide 'Horn segment 9'
Node=11=10
WTh=41.4cm HTh={S10}
WMo=41.4cm HMo={S9}
Len={L9}

Duct 'Duct segment 10'
Node=11=101
WD=41.4cm HD={S10} Len={L10}

Waveguide 'Horn segment 11'
Node=102=13
WTh=41.4cm HTh={S11}
WMo=41.4cm HMo={S12}
Len={L11}

Waveguide 'Horn segment 12'
Node=13=14
WTh=41.4cm HTh={S12}
WMo=41.4cm HMo={S13}
Len={L12}

Waveguide 'Horn segment 13'
Node=15=14
WTh=41.4cm HTh={S14}
WMo=41.4cm HMo={S13}
Len={L13}

Waveguide 'Horn segment 14'
Node=15=16
WTh=41.4cm HTh={S14}
WMo=41.4cm HMo={S15}
Len={L14}

Waveguide 'Horn segment 15'
Node=17=16
WTh=41.4cm HTh={S16}
WMo=41.4cm HMo={S15}
Len={L15}

Waveguide 'Horn segment 16'
Node=18=17
WTh=41.4cm HTh={S17}
WMo=41.4cm HMo={S16}
Len={L16}

Duct 'Duct segment 17'
Node=18=19
WD=41.4cm HD={S17} Len={L17}

Waveguide 'Horn segment 18'
Node=19=20
WTh=41.4cm HTh={S18}
WMo=41.4cm HMo={S19}
Len={L18}

Waveguide 'Horn segment 19'
Node=21=20
WTh=41.4cm HTh={S20}
WMo=41.4cm HMo={S19}
Len={L19}

Waveguide 'Horn segment 20'
Node=21=22
WTh=41.4cm HTh={S21}
WMo=41.4cm HMo={S20}
Len={L20}

Waveguide 'Horn segment 21'
Node=22=23
WTh=41.4cm HTh={S21}
WMo=41.4cm HMo={S22}
Len={L21}

Waveguide 'Horn segment 22'
Node=23=24
WTh=41.4cm HTh={S22}
WMo=41.4cm HMo={S23}
Len={L22}

Waveguide 'Horn segment 23'
Node=25=24
WTh=41.4cm HTh={S24}
WMo=41.4cm HMo={S23}
Len={L23}

Waveguide 'Horn segment 24'
Node=25=26
WTh=41.4cm HTh={S24}
WMo=41.4cm HMo={S25}
Len={L24}

Waveguide 'Horn segment 25'
Node=26=27
WTh=41.4cm HTh={S25}
WMo=41.4cm HMo={S26}
Len={L25}

Duct 'Du1' Node=100=26
WD=41.4cm HD=30cm Len=1.7cm

Radiator 'Horn mouth' Def='Horn segment 25'
Node=27

That looks pretty good, however the "Dogfood Duct" is perhaps a little bit longer, there are actually two with different effective cross-sections in POC#2, and their lengths should be subtracted from the waveguide sections in those areas.

Pity AkAbak doesn't have an optimization routine to determine the best size and location for the ducts :).

Good point on possible power compression. My neighbours seem to be out, so I may get an opportunity to do some SPL tests this weekend. To reduce the effects of compression, perhaps rounded ducts (think half-pipe sections placed tangential to the path) could be used. Think I'll pay me a visit to the hardware store tomorrow...
 
Last edited:
That looks pretty good, however the "Dogfood Duct" is perhaps a little bit longer, there are actually two with different effective cross-sections in POC#2, and their lengths should be subtracted from the waveguide sections in those areas.
That was just me playing around earlier today trying to see what you could be doing with dog food.

I figured you were either creating a restriction, or using the can as a stencil to add some tuning holes. The holes reduce a resonances output, and the restriction can help a null.

Looking at your pictures I would do something more like this to mimic it, quickly:
Code:
System 'DEMO1'

Def_Driver 'Driver'
dD=26cm dD1=11cm tD1=6.5cm |Cone
t1=1.1cm
fs=39Hz Vas=92L Qms=7.5
Qes=0.31 Re=5.2ohm Le=0.37mH ExpoLe=0.618


Def_Const
{
S1 = 4.8e-2;
S2 = 6.3e-2;
S3 = 8.2e-2;
S4 = 9.6e-2;
S5 = 10.9e-2;
S6 = 10.1e-2;
S7 = 10.1e-2;
S8 = 11.8e-2;
S9 = 13.1e-2;
S10 = 12.1e-2;
S11 = 12.1e-2;
S12 = 14.5e-2;
S13 = 17.1e-2;
S14 = 16.1e-2;
S15 = 17.3e-2;
S16 = 15.8e-2;
S17 = 13.6e-2;
S18 = 13.6e-2;
S19 = 14.8e-2;
S20 = 13.4e-2;
S21 = 11.8e-2;
S22 = 13.6e-2;
S23 = 16.3e-2;
S24 = 15.3e-2;
S25 = 16.3e-2;
S26 = 17.7e-2;


L1 = 20e-2;
L2 = 26.7e-2;
L3 = 2.5e-2;
L4 = 3.3e-2;
L5 = 2.0e-2;
L6 = 33e-2;
L7 = 3.0e-2;
L8 = 3.9e-2;
L9 = 2.5e-2;
L10 = 25.5e-2;
L11 = 4.0e-2;
L12 = 5.2e-2;
L13 = 3.7e-2;
L14 = 4.0e-2;
L15 = 5.4e-2;
L16 = 4.0e-2;
L17 = 17.0e-2;
L18 = 3.0e-2;
L19 = 4.7e-2;
L20 = 4.0e-2;
L21 = 4.2e-2;
L22 = 4.8e-2;
L23 = 3.4e-2;
L24 = 13.9e-2;
L25 = 20.9e-2;
}


Driver Def='Driver''Driver'
Node=1=0=100=3

Duct    'Dogfood'  
Node=101=102
WD=24.84cm HD={S10} Len=5cm

Waveguide 'Horn segment 1'
Node=3=2
WTh=41.4cm HTh={S1}
WMo=41.4cm HMo={S2}
Len={L1}

Waveguide 'Horn segment 2'
Node=3=4
WTh=41.4cm HTh={S2}
WMo=41.4cm HMo={S3}
Len={L2}

Waveguide 'Horn segment 3'
Node=4=5
WTh=41.4cm HTh={S3}
WMo=41.4cm HMo={S4}
Len={L3}

Waveguide 'Horn segment 4'
Node=5=6
WTh=41.4cm HTh={S4}
WMo=41.4cm HMo={S5}
Len={L4}

Waveguide 'Horn segment 5'
Node=7=6
WTh=41.4cm HTh={S6}
WMo=41.4cm HMo={S5}
Len={L5}

Duct 'Duct segment 6'
Node=7=8
WD=41.4cm HD={S6} Len={L6}

Waveguide 'Horn segment 7'
Node=8=9
WTh=41.4cm HTh={S7}
WMo=41.4cm HMo={S8}
Len={L7}

Waveguide 'Horn segment 8'
Node=9=10
WTh=41.4cm HTh={S8}
WMo=41.4cm HMo={S9}
Len={L8}

Waveguide 'Horn segment 9'
Node=11=10
WTh=41.4cm HTh={S10}
WMo=41.4cm HMo={S9}
Len={L9}

Duct 'Duct segment 10'
Node=11=101
WD=41.4cm HD={S10} Len={L10}

Waveguide 'Horn segment 11'
Node=102=13
WTh=31.05cm HTh={S11}
WMo=31.05cm HMo={S12}
Len={L11}

Waveguide 'Horn segment 12'
Node=13=14
WTh=31.05cm HTh={S12}
WMo=31.05cm HMo={S13}
Len={L12}

Waveguide 'Horn segment 13'
Node=15=14
WTh=31.05cm HTh={S14}
WMo=31.05cm HMo={S13}
Len={L13}

Waveguide 'Horn segment 14'
Node=15=16
WTh=31.05cm HTh={S14}
WMo=31.05cm HMo={S15}
Len={L14}

Waveguide 'Horn segment 15'
Node=17=16
WTh=31.05cm HTh={S16}
WMo=31.05cm HMo={S15}
Len={L15}

Waveguide 'Horn segment 16'
Node=18=17
WTh=31.05cm HTh={S17}
WMo=31.05cm HMo={S16}
Len={L16}

Duct 'Duct segment 17'
Node=18=19
WD=41.4cm HD={S17} Len={L17}

Waveguide 'Horn segment 18'
Node=19=20
WTh=41.4cm HTh={S18}
WMo=41.4cm HMo={S19}
Len={L18}

Waveguide 'Horn segment 19'
Node=21=20
WTh=41.4cm HTh={S20}
WMo=41.4cm HMo={S19}
Len={L19}

Waveguide 'Horn segment 20'
Node=21=22
WTh=41.4cm HTh={S21}
WMo=41.4cm HMo={S20}
Len={L20}

Waveguide 'Horn segment 21'
Node=22=23
WTh=41.4cm HTh={S21}
WMo=41.4cm HMo={S22}
Len={L21}

Waveguide 'Horn segment 22'
Node=23=24
WTh=41.4cm HTh={S22}
WMo=41.4cm HMo={S23}
Len={L22}

Waveguide 'Horn segment 23'
Node=25=24
WTh=41.4cm HTh={S24}
WMo=41.4cm HMo={S23}
Len={L23}

Waveguide 'Horn segment 24'
Node=25=26
WTh=41.4cm HTh={S24}
WMo=41.4cm HMo={S25}
Len={L24}

Waveguide 'Horn segment 25'
Node=26=27
WTh=41.4cm HTh={S25}
WMo=41.4cm HMo={S26}
Len={L25}

Duct 'Du1' Node=100=26
WD=41.4cm HD=30cm Len=1.7cm

Radiator 'Horn mouth' Def='Horn segment 25'
Node=27

This is still pretty ruff, but it is closer to what you have done there. Did the restriction end up at about halfway in both horn paths?
 
Playing with it some more it seems like you are countering the dip between the 5th and 7th harmonics, or your third impedance peak depending on how you want to look at it.

This is where the horn is changing over from 1/4WL to 1/2WL operation.

Using some horns I am more familiar with it seems 1/2 the horns acoustic length is the place to start. So physical length + mouth correction, and divide by two. Looks like placing the initial resistance as close to this as possible, and them moving towards the mouth to dial it in for the flare rate used seems to work pretty well.

As power is increased I would expect a non-linear increase in overall SPL, and more a pronounced reduction on the lower end SPL with a drop in Fc. Maybe you will get to test it out this weekend. If it goes well with your setup I'll get it added to some testing to be done in the spring with some monster horns.
 
(Big sigh) Many many thanks guys!!! Finally I’m getting some answers! Btw Soho, do you have some pretty diagrams of those horns in your collection which you are willing to share? Don’t see them around here.

But then again it still doesn’t answer or rule out the possibility that a section within the horn can produce harmonics as side effect of ‘rough’ bendings, is it?

(Brian, next time if you want to throw off a bite call it the old "brick method" for instance cause fish only take a bite in rocks they recognise… Lol, I really didn’t catch that although I admit it drove me bunkers; first mouth measurements, fluffy cards and then Dogfood?!?!:rofl:)
 
Last edited:
Playing with it some more it seems like you are countering the dip between the 5th and 7th harmonics, or your third impedance peak depending on how you want to look at it.

This is where the horn is changing over from 1/4WL to 1/2WL operation.

Using some horns I am more familiar with it seems 1/2 the horns acoustic length is the place to start. So physical length + mouth correction, and divide by two. Looks like placing the initial resistance as close to this as possible, and them moving towards the mouth to dial it in for the flare rate used seems to work pretty well.

Eyeballing it, it seems to be closer to 2/3rds down the path, but it's quite possible that my approximation may be off a bit. In the case of the tapped-pipe, it looks definitely like 2/3rds, as the constriction's best location is in the second half of the box.


As power is increased I would expect a non-linear increase in overall SPL, and more a pronounced reduction on the lower end SPL with a drop in Fc. Maybe you will get to test it out this weekend. If it goes well with your setup I'll get it added to some testing to be done in the spring with some monster horns.

It may be possible to achieve something similar by using instead an external chamber that resonates at the center frequency of the notch (in the case of the THAM15, this could be a chamber at the bottom, or even one on the 1st 180 degree bend, as I've found that even filling up almost half the bend with cans really didn't do much damage to the passband response. This should avoid any potential compression problems caused by putting a constriction in the path. Of course this is more difficult to test quickly :).
 
I prefer an approach (voodoo or not) that works at 2.83v AND at 28v and louder.... and doesn't require a can opener and a pet to make modifications. ;)

My ss15 for example has a dip exactly where hornresp predicts... but it's only 2HZ wide... not nearly what hornresp predicts. In fact it's such a narrow dip that smaart completely misses it at 1/24oct resolution. If I didn't make it a practice to listen to sine waves at every 1hz increments -- I'd never known it was there.

Considering 1/2 steps are 10hz apart at that frequency... a 2hz wide dip, that is halfway between notes will never be noticed.

The whole point (to me anyway) in this discussion, is to be aware of what path lengths 'can' create issues and at what frequency, and how to avoid making mistakes in the folding process. However, if you make a cabinet and have something you can't live with. (like my 2x2x4 cabinet) Knowing what to look for in trying to salvage a cabinet, gives you options to make it 'better.' In the case of that cabinet... 2 small reflectors made things better.
 
I prefer an approach (voodoo or not) that works at 2.83v AND at 28v and louder.... and doesn't require a can opener and a pet to make modifications. ;)

...and I prefer an approach that depends on known theory (no "horn voodoo"), and can be simulated by software that allows me to optimize the design before cutting a single piece of wood.

BTW, if it wasn't obvious, the dogfood cans (which were full btw, no can-opener required!) were only deployed temporarily, to determine the effect of changing the horn's topology at a particular point to see if the THAM "notch" can be effectively dealt with without significantly altering the design. As I fully expected, the effect can also be modelled (and optimized) via Akabak - no voodoo involved :).

Next step is to replace the cans with something a bit more permanent, and I can use Akabak to model the effect before ever cutting a piece of wood.

Whether or not the modification has any significant impact on the horn's response @28V remains to be tested.