Full DSP Pluto Clone

[AndrewW]

Tweeter and woofer impulse responses are captured via swept sine technique:

raw spectra.png (28.43 KiB) Viewed 1289 times

Observation at 40Hz and 85Hz reveals resonance of 6”x39” woofer pipe. Gigantic peak at 20kHz reveals that tweeter is a Peerless 2” full range.


Inverse transfer function of impulses is generated with Cool Edit using Aurora pluggin Invert Kirkeby. A FIR filter crossover for 1kHz is generated and applied to the inverse transfer function. Another FIR filter is used to HP the woofer at about 15Hz. A third FIR filter is used to LP the tweeter at about 17kHz:

Woofer and Crossover Filter.png (26.67 KiB) Viewed 1289 times

Tweeter and Crossover.png (25.95 KiB) Viewed 1289 times

The filters are implemented using Convolver pluggin for Windows Media Player: http://convolver.sourceforge.net. Microphone is placed in front of speaker and a few seconds of drum percussion is recorded. The result is brought into Cool Edit, aligned with the source track, and the two tracks normalized for amplitude. In the following screen shots top track is source and lower track response recording. The first shot is about 100ms and the second shot is about 10ms:

Percussion 94ms.png (41.32 KiB) Viewed 1289 times

Percussion zoom 10ms.png (39.73 KiB) Viewed 1289 times

[dewardh]

peak at 20kHz reveals that tweeter is a Peerless 2” full range.

830983? (I've been playing with that one . . . really nice)

Totally down with using 6" sonotube for the woofer, too . . .

Looking forward to seeing more of your project, and results.

[ginopalletta]

so those resonances at 40 and 85 are what is left after you stuff the pipe with "sound absorber"?

[AndrewW]

dewardh,

These are using Peerless 830970. Specs are quite similar. Of course, Aura 2" works quite well too! I milled off flange and glued tweeter into pipe:

IMG_0342 Pluto Clone 320x274.jpg (55.31 KiB) Viewed 1191 times

ginopalletta,

Yes, the pipe is very well stuffed. I started with a lot of stuffing with similar outcome, so reduced. In prior post spectrum is with FFT of 65k Blackman-Harris windowing. At sampling rate of 44.1kHz it reveals detail of low frequencies fairly well. Often results are published with too small an FFT, sometimes intentionally because this smooths the result. Here is same result with FFT of 8k:

Woofer fft8k.png (18.46 KiB) Viewed 1191 times

[AndrewW]

Spectrum of system, tweeter, woofer, reverse null, CSD: Microphone is Earthworks OM-1 placed at 9 inches. Results are un-smoothed.

Impulse.png (24.26 KiB) Viewed 1153 times

Solo impulse.png (28.17 KiB) Viewed 1153 times

Reverse null.png (29.92 KiB) Viewed 1153 times

Impulse response ported into ARTA:

CSD impulse2.png (18.92 KiB) Viewed 1153 times

Microphone moved to 42 inches. A clear view of room mode behavior:

Impulse 42 inch.png (27.17 KiB) Viewed 1153 times

And the above gated to about 3ms. This captures direct sound at expense of low frequency resolution:

Impulse 42 inch gated 3ms.png (25.3 KiB) Viewed 1153 times

And square wave response around 1kHz, 66Hz, and 2kHz:

sq1k.png (16.22 KiB) Viewed 1153 times

sq 66Hz.png (18.64 KiB) Viewed 1153 times

sq 2k.png (20.35 KiB) Viewed 1153 times

[SL]

OK, so how does it measure off-axis at 9 inch?
Do the brickwall xo lowpass and highpass outputs still add ok in the frequency domain or is there pre-ringing in the time domain?
Up to what off-axis angle, horizontally and vertically, does the summation hold?

SL

[AndrewW]

Hello Mr. Linkwitz:

The system is very stable:

45 degrees off axis horizontal: Gated and un-gated results are shown. Rest of results are all gated.

45 degrees.png (43.12 KiB) Viewed 1083 times

90 degrees off axis horizontal:

90 degrees.png (39.95 KiB) Viewed 1083 times

30 degrees horizontal and 30 degrees vertical above:

30H 30V degrees.png (42.24 KiB) Viewed 1083 times

15 degrees below:

15 degrees below.png (41.79 KiB) Viewed 1083 times

[AndrewW]

PS to above: Changes in level are due to normalizing amplitude of all results to same peak value.

AW

[AndrewW]

....and all together.....

All together.png (27.67 KiB) Viewed 1077 times

AW

[SL]

Very nice.
And how does it sound?
Compared to the analog version?

SL

[AndrewW]

The sound is wonderful. Holographic.

I have gotten very good results with DCX2496, but have not built or heard with ASP.

Currently I am working on integration with sub. Here is link to ground work: http://waveformfidelity.com/arbitrary-s ... nment.html

Andrew

[SL]

In your subwoofer tests you are using both a dipole and a monopole cabinet. Are both driven simultaneously, in effect generating a cardioid like source?
Do I understand correctly that you show when the source is equalized to a flat response at the microphone location, then you have almost complete recovery of the modulation envelope and that even serious driver distortion has little effect upon the envelope? That is what I would expect. The loss of modulation depth appears to be a phase response issue, not so much an amplitude response issue of the unequalized response, unless the relative amplitude of the two sidebands changed sufficiently. In any case it is a linear, amplitude independent phenomenon. Is its effect/occurence greater for a monopole than a dipole?

Can you calculate/measure the before/after MTF?

SL

[AndrewW]

Monopole and Dipole woofer alignments:

An alignment run with separate treatments for monopole and dipole elements:

Log sweep 10 octaves generated from 2Hz to 2205Hz of 1048576 sample length at fs 4410Hz, 1/10th of CD sampling rate. Suitable inverse filter derived using Aurora/Kirkeby inverse plugin. Sweep is just under 240 seconds in length and thus is sweeping 24 seconds/octave. So 24Hz to 48Hz in 24 seconds averages to 1Hz/sec. It’s a very exciting sweep! Re-sampled to fs 44.1kHz for sound card and CD compatibility. Microphone placed about 44” from stack and floor. Sweeps of independent drivers and simultaneous drivers recorded with 26 volts rms drive signal. Very hot and exciting sweeps! After a few goes LP 1kHz applied. The aluminum Dayton 12” HF 4ohm with 26V across it at 1kHz-2kHz is really too loud, even with earplugs in.

Results re-sampled back to fs 4410Hz and each convolved with inverse sweep to recover impulse responses:

Raw Spectrums:

Monopole:

Monopole spectrum.png (16.56 KiB) Viewed 967 times

Dipole:

Dipole spectrum.png (16.79 KiB) Viewed 967 times

Cardioid system as Monopole and Dipole driving together:

Cardioid.png (17.1 KiB) Viewed 967 times

Cardioid as sum of Monopole spectrum and Dipole spectrum, sum is +6db shown result normalized for comparison to Cardioid spectrum above:

Cardioid v Cardiod as sum of Monopole and Dipole.png (18.4 KiB) Viewed 967 times

Kirkeby engine used to generate inverse function of each IR. Results band-passed 9-175Hz.

Candidate inverses convolved with copies of test sweep, up-sampled to fs 44.1kHz, run, and impulse responses derived as with test sweep:

Spectra with correction filters:

Monopole:

Monopole response spectrum with correction filter.png (14.53 KiB) Viewed 967 times

Dipole:

Dipole spectrum with correction filter.png (14.45 KiB) Viewed 967 times

Cardioid system as Monopole and Dipole driving together:

Cardioid system with correction filters.png (14.33 KiB) Viewed 967 times

Raw Impulse responses Monopole top, Dipole bottom:

Monopole top Dipole bottom raw impulses.png (22.16 KiB) Viewed 967 times

Impulse Responses with correction filters: Monopole top, Dipole bottom

Monopole top Dipole bottom impulses.png (30.77 KiB) Viewed 967 times

Monopole top and Dipole-Monopole bottom:

Monopole top v Dipole-Monopole bottom impulse dif.png (28 KiB) Viewed 967 times

The waveform fidelity performance within passband is similar to the Pluto Clone. The woofer stack and Pluto Clone may be integrated, but a better alignment perhaps would use Monopole <50Hz, Dipole passband 50Hz-100Hz. This would greatly reduce driver demands on the Seas woofer, reducing intermodulation distortion.

The last screen shot was made by subtracting the monopole wave from the dipole wave, and makes clear that the MTF of the monopole and dipole as conditioned must be very similar. I have looked into MTF, and am currently working on a graphical presentation of the complex MTF using Cool Edit.

Andrew