Jean Michel on LeCleac'h horns

[Elias]

Hello Jean-Michel,

How is the measurement done, what is the quality of the anechoic room? For me it looks like the data is contamined with room reflections, and thus the ripple in phase and therein in group delay. Then, considering this what we actually see from this picture?


Another thing I'd like to know is what is the order of the high pass transfer function of these horns? Because that will naturally define the group delay of the system at low freqs.


- Elias

Originally posted by Jmmlc
You'll find in attached file:

1) the response curve of a Le Cléac'h Basshorn unequalized (the response is a bit limited in LF due to ARTA's FFT window the max length of which depends on the length of the signal window... )

2) the phase curve (the "true" phase curve not a useless "minphase curve")

3) the group delay curve (the "true" group delay not the one calculated from the minphase)

[Jmmlc]

Hello Elias,

I guess you are joking ;-)

How would you that we move those bass horns inside an anechoic room

http://nsm01.casimages.com/img/2009/...5253677910.jpg

as they are conceived for the ground and lateral walls of the auditorium to be their natural extension.

A 130Hz high Q annulation in the response could eventually correspond to the interference between 2 signals having a diffference of path of 3,9meters which value don't correspond to anything precise in the auditorium. Eventually it can be due to the large rear chamber.

You'll find the wave (in TXT format) attached to that message.

I guess to derive the transfer function of such a bass horn will be a difficult task (the rear close chamber cancels partly the throat reactance of the horn), better to use Hornresp or Akabak to study their behaviour. (While Hornresp mouth area is limited...).

Remember also that those bass horns have very low cut-off frequency (lesser than 30Hz) and that the acoustic impedance becomes rapidly purely resistive above that cut-off .

Best regards from Paris, France

Jean-Michel Le Cléac'h

[soongsc]

Originally posted by Lynn Olson
It's entirely possible the subjective impression is a combination of group-delay variation (if excessive and not masked by room standing-waves) and IM distortion.

The major categories of bass reproducers are going to have very different combinations of group-delay vs IM distortion profiles, as follows:

Open-Baffle with (heavy) Equalization: This offers the potential of smooth group delay vs frequency, but the IM distortion will be the highest of all, thanks to equalization. Remember, a closed-box speaker is a constant-acceleration device, which means excursion increases at a rate of 12 dB/octave as frequency is lowered. An open-baffle operating below the baffle peak has another 6 dB/octave term, so the excursion of an equalized OB increases at a rate of 18 dB/octave.

Closed-Box: with the Q set to a moderate value around 0.57, group-delay variation will be low. Unfortunately, closed-boxes trade Q against distortion - the lower-Q alignments have more excursion, and higher-Q alignments have less. These are 2nd-order highpass filters with user-selected Q settings (box volume and suspension compliance control the Q of the highpass filter). As mentioned above, excursion increases at a 12 dB/octave rate in the passband, and below F3, becomes constant. Very deep LF content can intermodulate higher frequencies, even the direct LF is not itself audible, since it is below the passband.

Vented-Box: Excursion sharply drops at the box frequency, which lowers excursion in the octave centered around Fb. The tradeoff is significantly worse group delay, since all vented-box alignments are 4th-order highpass filters, and some of them have additional unwanted group delay near the corner frequency.

There are a couple of unwanted characteristics of vented-boxes: parameter shift under drive, which detunes the rather sensitive 4th-order highpass under dynamic drive conditions (this can be minimized by selecting a Bessel or Gaussian highpass), and an unwanted zero in the response in the vicinity of 1 Hz (this is caused by box leaks, or Ql). This leads to the in-and-out "breathing" under dynamic conditions sometimes seen in vented-box systems.

Below the passband, the vented-box driver is essentially unloaded, and excursion is only limited by the driver compliance. The previous comment about very deep LF content intermodulating higher frequencies applies with greater force to a vented-box system. As a result, additional electronic highpass filtering (raising the system highpass function to 5th or 6th-order) is sometimes needed to protect the driver from damage from extreme LF thumps and pulses.

Bass horns: These have the lowest distortion of all, thanks to resistive horn-loading in the passband. If I recall right, they are constant-velocity instead of constant-acceleration, so excursion only increases at a rate of 6 dB/octave instead of the 12 or 18 dB/octave of the previous devices. With great care in horn design, group-delay can be controlled, as JMMLC's measurements show.

Note all four designs have significantly different combinations of group delay and IM distortion profiles - and these are inherent to the design, not the choice of driver. (If anyone can find a distortionless bass driver, I'd like to know about it.) The reason I mention IM distortion instead of harmonic distortion is the potential for intermodulation with higher frequencies, which will be much more audible than fairly benign 2nd-harmonic. Since excursion at lower frequencies is always greater than at higher frequencies (regardless of alignment), the intermodulation is always in an upward direction - LF dirties the upper-bass frequencies.

In addition, the spectral tilt of the IM distortion is greatest for OB, then CB, then VB, with bass horns having the least tilt of all. This implies a significantly different spectral distribution of distortion that affects the upper octaves of bass - in other words, the loading on the diaphragm controls the resulting distortion spectrum it will create. If the driver was distortionless, we wouldn't care, but drivers are not distortionless, and contaminate the spectrum with IM products that are dispersed upwards in frequency.

Lynn, thanks for organizing this information. I do wonder, considering the room modes and reflections, how can we tell what difference we hear is caused by what? I am constantly asked questions of this sort in a local forum here.

[Lynn Olson]

This is a perceptual difference. Above a certain critical frequency (I'm guessing 300~500 Hz) the ear starts to distinguish between room-sound and the direct arrival. Above this frequency, time-decay performance (stored energy) of the loudspeaker affects the timbre and coloration of the music, and sensitivity to IM distortion increases, reaching a peak in the 1~3 kHz region.

In the 100~300 Hz region, loudspeaker and room sound start to merge together, partly because the wavelengths are getting long enough that room standing-wave modes are starting to dominate both perception and measurement (which is why it is industry practice to "splice" curves at 300 Hz, using nearfield measurements below that).

Loudspeaker coloration is a single, unitary perception, and what "sounds" fast might have nothing at all to do with transient response. It could any of several things - lack of stored energy in the loudspeaker cabinet, avoiding room modes, or low IM distortion. They could all sound alike, or subtly different.

This is why measurements are necessary, because correlation between what you "think" you heard and what's actually there may have a rather complex and non-obvious relationship. I've worked enough with amplifiers and speakers to confuse the sound of each - low distortion can sound "fast", and good impulse response can sound like good spatial resolution and a realistic ambient impression, but not "fast" in a subjective sense.

I think people are somewhat misled by harmonic distortion. 2nd harmonic, as generated by a test circuit, is really pretty hard to hear, even in the midband. 3rd sounds like a more "edgy", sharp sound, but again, that's in the midband where it has peak audibility. Audibility of THD in the bass region pretty much follows the Fletcher-Munson curve - very low below 100 Hz.

But IM distortion is another story. Even though the mechanism creating THD and IM are the same, the distortion itself has a quite different audibility, because it "thickens" the spectrum with harmonically unrelated tones. The denser the spectrum, the more sum-and-difference tones are generated. The BBC did a study a few years back and calculated with as few as three tones, the amount of IM distortion exceeds THD terms, and with more tones, it goes up geometrically. With a complex spectrum from music, IM distortion dominates by far, and the more complex the music, the more IM distortion is present. In the simplest model, it acts as a dynamic noise floor.

IM distortion is particularly relevant for loudspeakers, because below-passband stimuli, although not directly audible by themselves, crossmodulate up into the audible band, and create unrelated sum-and-difference spectral artifacts. One of the most important functions of the tweeter crossover is to protect against below-band energy creating IM distortion in the passband - and the passband of the tweeter typically falls in the region of peak distortion audibility.

Think of 10 Hz and 1 kHz going through a single vented-box midbass driver; the 10 Hz isn't audible at all, since it is well outside the passband of the VB system, as well as falling off the far edge of the Fletcher-Munson curve (10 Hz has to be very loud to be heard at all). But the 990 and 1010 Hz distortion sidebands are really there, and do no favors to the midrange. Now add 10 Hz, 20 Hz, 1 kHz, 2 kHz, and 3 kHz - we're starting to get a real thicket of sum-and-difference terms that are not present in the original musical material.

[Lynn Olson]

There is also a third distinction between Open Baffle, Closed-Box, Vented-Box, and Basshorn: CB and VB are very close to omnidirectional below 300~500 Hz (depending on size), while OB is a dipole down to the lowest working frequency, and horns are directional down to the cutoff frequency (the pattern spreads out below cutoff).

This means that there are three main groups of directivity patterns (omni, dipole, and quasi-conical), and they excite room modes in completely different ways, making comparisons between types in different rooms very difficult, since the room/loudspeaker interaction is going to be unique to each combination.

So there are three distinctions between the four main types of bass reproducer: group delay variation, IM distortion magnitude and spectral distribution, and LF dispersion pattern. No wonder they sound different - they are different in significant ways, and will interact with different rooms in different ways.

The loudspeaker group delay variation merges with the room modes at the lowest frequencies; as a result, the LF dispersion pattern has a strong influence on the overall loudspeaker/room group delay performance. Sure, you can take an OB, CB, VB, or basshorn outdoors and measure it, but do you plan to do all of your listening outdoors as well? Once you put it back in a room, the LF dispersion pattern matters a great deal to how it sounds.

Note that I haven't mentioned drivers - the differences between drivers (distortion spectra, et al) are much smaller than the essential differences between the four types of bass reproducer.

[Jmmlc]

Hello

Michael (mige0) asked me some time ago to help him to design a horn in order to load a ribbon tweeter AMT both at its front and at its rear ("dual horn").

He wanted that the horn could be inserted in an existing open baffle. Only a rectangular place of 30centimeters (width) x 15centimeter (height) was usable on the OB.

I calculated fro him a "Le Cléac'h quasi cylindrical wavefronts horn".

You can see a beautiful 3D view of that dual horn at:

http://www.diyaudio.com/forums/showt...60#post1839560

And read the message:

http://www.diyaudio.com/forums/showt...92#post1837992

The response curves at 0°(equalized ) , 10°, 20°, 30° and 40° are visible at:

http://members.aon.at/kinotechnik/di...-30-40-deg.gif

Additionally the response of the original tweeter taken as a 0dB reference Michael calcualted the gain of that dual horn (gray cruve) compared to open baffle (yellow curve).

As you can see that the dual horn provide an intersting gain in the frequency range 700 - 3000Hz. This is due to a better acoustical loading of the ribbon. It lead to less ribbon displacement for the same SPL and thus to less distortion.

http://members.aon.at/kinotechnik/di...-horn-gain.gif

Best regards rfom Paris, France

Jean-Michel Le Cléac'h

[Lynn Olson]

Hmm ... could the "Le Cléac'h quasi-cylindrical wavefront horn" be scaled up to match an Altec 414 12" driver? I'm thinking of something like the short horn built into the Altec A5 or A7 cabinets that would cover the range from 150 to 800 Hz. I know it would end up being about the width of the A5/A7 in order to have enough mouth area, but I'm sure a Le Cléac'h profile would have much less diffraction than the Altec original, while retaining aspects of the Altec sound.

[cuibono]

Originally posted by Jmmlc

Michael (mige0) asked me some time ago to help him to design a horn in order to load a ribbon tweeter AMT both at its front and at its rear ("dual horn").

The results are impressive! Thank you for doing this.

Sometime soon, I'd like to do the same for a pair of Neo3PDR. Is the contour Michael used scalable? Or how does one calculate it?

[mige0]

Jean-Michel

Yes - what was intended as an experimet only with rather modest expectations from my side turned out very very well.

Big thanks ! and my compliment to your contour - as I think it makes all the difference between "only" good loading and directivety control and a wonderful sound.

I have played around with your spreadsheet but - being only a monkey pressing buttons when it comes to horn contour design - havent found the right ones.

Maybe you could do an english text for the parameters in question as well and underlay the fields which have to be individually adjusted in one color and the results of calculation (coordinates) in an other color?

Is there a possibility given in your spreadsheet to calculate the coordinates of the wave front as this is somthing to play with for cylindrical horns especially



the sexy kissing lips I derivied from:
http://www.diyaudio.com/forums/showt...68#post1790568


Michael

[Pano]

Originally posted by Lynn Olson
... could the "Le Cléac'h quasi-cylindrical wavefront horn" be scaled up to match an Altec 414 12" driver

Do you mean something with a flat top and bottom like Michaels's ribbon horn, or the "under the floor" horns in France? If so, that would look a bit different from the A7 type horn, yeah?

The mouth termination of the Altec isn't great, but it's helped some with the side wings. Would be interesting to do a better horn.