EnABL - Technical discussion

[soongsc]

For my part, I have a calibrated microphone, LAUD audio software, an appropriately quiet computer, with a DSP based sound card, modified to be flat and stable for low frequencies, a battery powered mic preamp and really, quite a bit of confusing data. Changes in distortion components that were never repeatable. Frequency response tests that showed what certainly looked and overlaid like, the same measurement, also different from day to day, no dedicated space for careful measurement of physical relationships between driver and mic and eventually an understanding that my skills either were not up to the task, or there really was nothing but self deception going on.

...

Bud

Sound's like my experience with cables. However, since there was a trend showing cables aging and change in characteristic above 10K Hz, also presenting reliable data on what characteristic sounded best, it was quite easy to pinpoint what to do. I just wish I had data on when cable characteristic becomes more stable.

 

[BudP]

mondogenerator

I have no doubt enable in some form will improve transient behavior of a loudspeaker cone, but i find it difficult to believe that with listening tests alone, one cannot optimize the pattern or distribution of pattern. I read the 'tap test' method, but i found my Al cone drivers rang most at about 2/3 cone dia and the dust cap point was non existent or maybe under the cap itself. Also i would probably prefer to put dots on the inside cone surface, which may defeat the object entirely. I also have to agree with some poster (sorry cant remember names) who riffed that to maximize the diffusion of boundary wave resonances the 'dot' profile would surely need to be greater that a 0.1mm paint dot.

I am going to assume that your aluminum cone has a curved profile. The tap test for all curved cones shows that the curved portion of the cone does not alter in character until some where around 1/2 to 2/3 of the radial distance measured from the voice coil, so your findings are suitable. You might go back and redo them and listen for a very slight "subduction" of the sound, off of the cone, as you tap very lightly. If you find one at about 1/2 way up the cone, a ring set will be needed there. Then tap further out, beyond the major ringing. There will likely be two or three more quite subtle changes in the character of the tapped noise. This can be as a sensed direction of decay, with a narrow null in this directivity, as the center for a pair of rings. Another potential subduction may be found, with a pair of rings needed to bracket the beginning, or more likely, another elevated taping response, usually out just before surround, with a pair of rings needed over this point also.

The end product you are after is an entire cone surface where the decay of the all tapped tones is sensed to head towards the surround and the initial or leading edge of the tap is the same across the radial distance. There may also be more than one "tone" change, in fairly broad steps, as you go up the surface. These steps do not appear to alter the way the cone emits into the air like the treatment zones do.

The tape you mentioned is likely to be 3mils thick and should be quite suitable. The typical paint block is 2 mils thick in it's center and 1 to 2 mils on the ends. To remove most tapes, with a releasable adhesive, isopropyl alcohol and a Q tip works best for me.

If you can provide me with a circumference for the center of any location you want to place rings I can provide you with a paper guide that you can lay on the cone just smaller than where you want to place the tape blocks.

Further discussions on these topics should be moved over to the techniques thread.

Bud

 

[Ed LaFontaine]

There will be a get together in Frankfort, KY on Nov. 20, 2010. The purpose is for area speaker builders to show their projects, swap stories, catch up with one another, etc.

CKYDIY

Among the speakers I plan to bring are identical pairs of sealed 2-ways. One is "naked", the other is EnAbled. There will be a receiver with A & B outputs, remote control and source set up for these. Playing from behind a veil will be available to enhance your listening pleasure.

 

[planet10]

Among the speakers I plan to bring are identical pairs of sealed 2-ways. One is "naked", the other is EnAbled. There will be a receiver with A & B outputs, remote control and source set up for these. Playing from behind a veil will be available to enhance your listening pleasure.

You should 1st check, and then adjust if necessary, relative levels. Driver to driver variations mean they are unlikely to all have the same efficiency to start + EnABL will reduce efficiency at least a small amount.

dave

 

[mige0]

Hi
( in continuation to http://www.diyaudio.com/forums/mult...ng-impressions-techniques-11.html#post2343690 (pretty complicated with this split threads ))


Did the first comparative measurements today. As we got fresh snow yesterday, its indoor measurements (roughly 2ms valid).


Speaker A un-treated :

http://www.kinotechnik.edis.at/pages/diyaudio/enabl/8W4P-A_un-treated.wav


Speaker B un-treated :

http://www.kinotechnik.edis.at/pages/diyaudio/enabl/8W4P-B_un-treated.wav


Speaker C treated :

http://www.kinotechnik.edis.at/pages/diyaudio/enabl/8W4P-C_treated.wav


Speaker D treated :

http://www.kinotechnik.edis.at/pages/diyaudio/enabl/8W4P-D_treated.wav


Mic distance was 1m, SPL was set at roughly 80dBA, MLS, apodizing 0.02
As expected, there is not much conclusions to draw from mere IR and CSD analysis - especially as we do have not a before / after comparison here (except possibly that the second peak is lower and the dip following got wider).

Will see how I can dig into EnABL any deeper

Michael

 

[planet10]

Michael,

Can you generate the waterfalls with a periods time axis?

dave

 

[mige0]

Michael,

Can you generate the waterfalls with a periods time axis?

dave

Hi Dave,

in ARTA one can do „Burst Decay“ analysis. This is probably what you are looking after.
Its possible to format Burst Decay as „waterfall“ as well, but I find it easier to look at those plots in „sonogram“ formatting (no hidden valleys)


Speaker A un-treated :

Speaker B un-treated :

Speaker C treated :

Speaker D treated :

And here is a plot how to translate above Burst Decay sonogram into ms :
Hope Ivo will provide this as an overlay grid some time

Michael

 

[planet10]

I agree that the oblique view of the waterfall does obscure some stuff... does it do a version where the curves are plotted and looking from the front. Instead of having to interpret the colour, that gives a direct view of the ring-down.

Here are the 1st 2 images (1st stock vrs 1st treated) in a blink test for easier comparison

Enable Tests

dave

 

[mige0]

Looking straight from the front would not really be helpful, as the most (for me) interesting part is to detect discontinuity of the spectral distribution along the time line, and how that changes by treatment.
Obviously those up and downs of the "mountain ranges" would be hidden in pure front view.
So all thats left is either common waterfall or top view.

Nice blink pix you made - but please put a remark that its actually two different chassis.

As the Burst Decay plots form ARTA sadly can not be normalized, we see some changes in the room reflections that should not be there (lower left from 2ms).
Those stem from different leveling (peak is always set at 0dB).


Michael

 

[planet10]

Looking straight from the front would not really be helpful, as the most (for me) interesting part is to detect discontinuity of the spectral distribution along the time line, and how that changes by treatment.
Obviously those up and downs of the "mountain ranges" would be hidden in pure front view.

The ups & downs are not hidden if the curves are not filled. The way that the subsequent curves decrease tells a whole lot about the ring down. A "perfect" driver would have all the curves "parallel" to the one above & below and a monotonic difference in differences between curves.

Ideally one could grab that waterfall and rotate it to look show any view, switch fill on & off, and adjust the transparemcy of slices.

Similar for impedance magnitude & phase would also be useful.

dave

 

[mige0]

The ups & downs are not hidden if the curves are not filled. The way that the subsequent curves decrease tells a whole lot about the ring down. A "perfect" driver would have all the curves "parallel" to the one above & below and a monotonic difference in differences between curves.

dave

I'd say an *ideal* driver has a drop down in time frequency plots like a cliff. period.

Any montonic ring down is a result of a resonance - meaning a "TRUE resonance" = a resonance form a "compact mass / massless spring system". The angle of ring down is set by the Q of that resonance.

Any non- monotonic decay basically is a CMP behaviour not based on "TRUE resonance" but on constructive and destructive interference with delay involved (or possibly coupled systems with delay involved).

In contrary - a pipe is also said to be "resonating".
In fact, this particular mechanism is more kind a standing wave which has nothing to do with the behaviour of a "compact mass / massless spring system" if looked at more closely.

In this context its interesting to see that the term "stored energy" has gained some popularity - which actually does not tell us anything more - whereas above distinction has not been emphasized on until now (at least as far as I have seen).


Michael

 

[planet10]

I'd say an *ideal* driver has a drop down in time frequency plots like a cliff. period.

Any montonic ring down is a result of a resonance - meaning a "TRUE resonance" = a resonance form a "compact mass / massless spring system". The angle of ring down is set by the Q of that resonance.

My words aren't getting the idea across.

When i talk about ideal driver i talk about one that we can actually make... they will always have a finite fall time.

If the curves looking into the time axis are "parallel" and the driver has flattish" FR, then a monotonic change of the distance between curves means the resonant behavior is well controlled (you can't actually pick out a specific resonance. Monotonically increasing distance is the preferred direction.

Any bunching of the distance between curves is a sign of a specific resonance

dave

 

[Ed LaFontaine]

Originally posted by planet 10:
You should 1st check, and then adjust if necessary, relative levels. Driver to driver variations mean they are unlikely to all have the same efficiency to start + EnABL will reduce efficiency at least a small amount.

Concern for the above led me to "Comparing Speakers" by Festler:

http://www.audioxpress.com/reviews/media/ferstler110.pdf

Dave, can you suggest whether series resistors are suitable for this purpose, and the range of value?

 

[planet10]

Series resistors will change the Q of the system. You really need to have an adjustment at the amplifier.

For my comparison units, i fortunately have 10s (at least) units to choose from, and with standard production spread i've been able to choose before & after drivers that are within a tenth of a dB efficiency.

In the (slow) development stage is a variable transimpedance amplifier with variable gain on 2 inputs to allow adjustment at the pre-amp stage.

dave

 

[mige0]

Weather conditions today have been good enough to do promised outdoor measurements (short, calm time between blasts and rain)

Resulting wavelet analysis are not exactly as I'd have expected, but have a look yourself:



Selenium 8W4P “driver A” - un-treated:

http://www.kinotechnik.edis.at/pages/diyaudio/enabl/8W4P-A_un-treated-od.wav



Selenium 8W4P “driver B” - un-treated:

http://www.kinotechnik.edis.at/pages/diyaudio/enabl/8W4P-B_un-treated-od.wav



Selenium 8W4P “driver C” - treated:

http://www.kinotechnik.edis.at/pages/diyaudio/enabl/8W4P-C_treated-od.wav



Selenium 8W4P “driver D” - treated:

http://www.kinotechnik.edis.at/pages/diyaudio/enabl/8W4P-D_treated-od.wav

Measurements were taken at roughly 1m mic distance with the nude speaker “swingin' mount”.
Again, keep aware this is not a before / after comparison !


From the listening impression with the wooden stick which, to me, sounded kinda more “snappy” for the EnABL'ed speakers, I would have expected to see overall decay changes.
This – in my eyes – does not happen.

My guesstimate regarding less CMP behaviour – based on John's measurements – does not happen the way I'd have expected, as well.

What happens – as far as I can see and especially when looking at the treated driver “C” - is that compact reflections (at roughly time slot 1.2ms) get split into many.

I do not have a really good explanation for that. We might wait until more EnABL measurements are available to compare with.
Most interesting I find that back wave diffraction around the rim seems not to be dominating results.
For comparison here is a measurement of my 180deg honker (NEO3 in round OB) where the diffraction around the rim is most obvious.

My summary on the basis of what I've measured so far:

1.a pronounced “around the rim” diffraction overlay at a discrete time interval simply may not happen with nude speakers
2.the reflections we see may be mainly cone brake up – which basically is a CMP reflection process too (modes due to standing waves)
3.EnABL seems to break down (spread out) strong break up modes into several ones
4.The EnABL pattern applied at the SELENUIM 8W4P does not seem to have a *cone dampening* effect at first hand (as IMO is seen in John's “EnABL” measurements)


But – of course - I could be totally wrong too


It might be interesting to do such analysis along each step of EnABL procedure to better pin point what comes from what step of treatment
You know - wavelet analysis software is available for free too


Again Bud - thanks for all the nice work you've put into this !

Michael

 

[planet10]

Thanx Michael,

Blink test drivers A & C are here: Enable Tests

dave

 

[john k...]

Thanx Michael,

Blink test drivers A & C are here: Enable Tests

dave

Compare A to D and B to C. Pretty obvious that the treated case really smears the response. This is particularly evident if you overlay B with C. I don't know how anyone would call that an improvement. It may sound different. Some may say it sounds better. Some may say it sounds worse. But while neither is close to perfect, in the A to D and B to C comparisons it is apparent that untreated appears more actuate with less time smear. I choose those comparisons because at low frequency A and D, and B and C appear more similar and I would expect such treatment to have little effect at low frequency.

 

[mige0]

...Pretty obvious that the treated case really smears the response. This is particularly evident if you overlay B with C. I don't know how anyone would call that an improvement. It may sound different. Some may say it sounds better. Some may say it sounds worse.

Lets possibly stay with : no one has stated "from this measurements" that this particular EnABL treatment actually *has to* sound better or worse.

To state any conclusion *if* time smear may be for the better or for the worse "in such cases" (cracking up pronounced cone modes) may take some more investigation as has been done, I'd say.
But this is not the point *I* ever will make (though, having my own thoughts on this of course).

But while neither is close to perfect, in the A to D and B to C comparisons it is apparent that untreated appears more actuate with less time smear.

First hand it seems obviouse, but for above outlined reasons I'd not "bet the ranch" that you are actually right in this.

What I have contributed (if at all) is the mere fact that there seems to be a measurable reason that EnABL changes audible presentation.

Besides your measurements (and Soongs that I have not seen or forgot about) this - for me - is more pin point than I think has been revealed yet.

Besides - am I right in my assumption regarding your measurements, that you have added mass with some "dampening effect" (soft material with some thickness) - in contrary to EnABL technique that does not exactly "dampen" membrane modes with its minimum thickness Acrylic dots?

Michael

 

[BudP]

Compare A to D and B to C. Pretty obvious that the treated case really smears the response.

Don't see how you can claim that John. The untreated drivers have very obvious secondary and tertiary ringing. With significant amounts of time between each peak. Plus, these exhibit quite a wide smear for each of the ringing peaks. A very obvious display of Dave D's "hall of mirrors" effect.

Within the same time window and with a measurably narrower energy peak up in the upper registers, where it is most important, we see a very even and natural decay of the energy in the cone, with barely any evidence of the wide peaks of ringing. Instead we find very narrow time bands, right at the threshold of comprehension, with a very even spacing. This sort of parabolic loss of energy is going to allow the brain to fix on that initial pulse, with it's power peak that has a narrower time window, and interpret the rest of the energy as a resonant decay of that initial strike. The untreated driver is going to sound like a hall of mirrors by comparison, with no clearly comprehensible envelope to the sounds being produced.

Remember John, the ultimate arbiter for the human in the loop is not how fast something is but how intelligible it is and our test gear is far faster than our rate of information uptake. Plus, we are hardwired to accept the initial strike as the descriptor of a sound and when we are presented with multiple sonic descriptor images, at near the same level and broad enough in time to be comprehensible as another discreet strike, this activity will not promote comprehension of the strike as a single event.

Wonderful work Michael. I had hopes for the wavelets and they are bearing fruit. Just as with John's blink comparison, we are seeing exactly what EnABL patterns are doing.

Thanks for putting the wavelet's into a blink comparator Dave. This really does show what the differences are.

The damping material on the dome center and on the back of the cone are there to remove a very prominent "honk" in a very narrow bandwidth, one that colors the sound from an EnABL'd driver far more obviously than it does with an untreated driver. Tap the area on the back side of the untreated driver cone, down near the voice coil joint and then in a radial line out towards the surround. You will find a pronounced ringing on the area covered by the dust cap. Now perform the same activity on the treated cone.

I have found that nude drivers have a very closely held null zone, with that of an untreated driver at the front edge of the basket and at the back edge for a treated driver. Once you are a foot away it just sounds as though the front half of the emissions are "painted" onto those from the back half. I have a pair of Lowther PM6A's, 15 ohm silver voice coil, with a 3mm XMAX that nude, on a stand, are the most natural sounding speakers, with the closest to true real life dynamics, of anything I have come across, including Mille' Nestorovic's model 16 super systems.

Bud

 

[planet10]

Here i did them all

Impulse: AC | AD | BC | BD
Waterfalls: AC | AD | BC | BD || AB | CD
Spectrograms: AC | AD | BC | BD
Wavelets: AC | AD | BC | BD

Hopefully it all works

dave