Sample Results

Measurement Setup

Here we show a more typical test setup and results for harmonic distortion and intermodulation distortion (IMD) product generation with the System 5000. This page is intended for the more technically inclined viewers...

These results are indicative of what you could expect with pretty severe corrections. As your required corrections become more benign, the corresponding distortion products rapidly diminish, always staying well below your threshold of audibility.

The image below shows our calibration GUI as we had it set up for this test. The test consists of sending a -14 dBFS peak 1 KHz sine tone into the System 5000 with rather extreme settings for hearing corrections as you can see from the lower two panels of the GUI.

(click on image for full size preview)

HD and IMD Measurements

And the next image below shows the results of this test obtained with a spectrum analyzer. The spectrum analyzer display has a bias of -6 dB (it apparently only counts the power in positive frequencies...) and that is why the sinewave test tone shows a peak amplitude of -20 dBFS.

Also shown on the analyzer display are a (generously low) estimated masking threshold in Yellow, and the user's own threshold elevations in Orange. (The Orange curve is not quite correct at the lowest and highest frequencies, because our analyst failed to account for the Fletcher-Munson corrections... hmmm...) But in the vicinity of the test tone frequency, the threshold curve is close to being correct.

The Magenta line is the absolute threshold of perfect hearing, something we never encounter in the real world. Instead, the normal audiological threshold of perfect hearing is the horizontal blue'ish line elevated about 20 dB above that absolute threshold. If your audiology reports perfect hearing, that blue'ish line represents your hearing (... sans Fletcher-Munson corrections).

As you can see, the Harmonic Distortion products (those lowest spurs shown) are running below -130 dBFS. The two larger spurs, located at 2.5 KHz, and 3 KHz are IMD products resulting from the amplitude followers in the System 5000 algorithm.

Of the two IMD products, only the spur at 2.5 KHz would be barely audible to perfect hearing, as it is both masked by the test tone, and barely rises above the absolute threshold at 2.5 KHz. None of these distortion products are anywhere near to being audible to our test subject.

Swept-Sine Measurements

The next image is another kind of test result, produced by pushing a log-swept sinewave, embedded in pink noise, through the System 5000. The system corrections setup was the same as shown above for the previous test. The sweep runs from around 200 Hz to 16 KHz in 10 seconds.

This is the ultimate torture test for any sound restoration system. If noticeable artifacts are going to show up, they will certainly show up here. This kind of test simulates musical performances, but elides all the confusion sources present in real music, so that real details of the processing can be discerned.

This test result illustrates several things...

1. There is a trough along the carrier, scooped out of the noise background, that represents the combined effects of loudness masking compensation, and amplitude follower release.
   
   As a loud carrier is sensed, the gain in that corresponding Bark band is reduced to accommodate it. The effect is to keep the sound at the correct level for the listener at all times, and also to suppress apparent noise signal levels in the vicinity of that signal frequency.
   
   
2. There is a remarkable cleanliness about this sweep as it progresses from one analysis band to the next. Simpler systems would be showing numerous vertical streaks in this sonogram, at band spacings, indicating abrupt gain level changes, as the tone sweeps from one band to the next.

But here, our System 5000 passes this test with flying colors. There are zero artifacts when the test tone moves through each band and into the next. What this means for the listener is that portamento sweeps, and legato passages, of highly tonal instruments will be heard as smooth and artifact free. This is a notoriously difficult accomplishment.

Our System 2000 also exhibits this same pristine cleanliness. No point showing the results here because they appear almost identical. If you want to see them for yourself, send us a note.

(click on image for full size preview)

The next image is the corresponding amplitude sweep for this test. As promised - no abrupt gain transitions can be seen. The "hair" on the signal is simply the pink-noise background.

The input signal had constant amplitude, approximately equal to that shown at the extreme left end, where the bass corrections were minimal.

What you are seeing here is the combined effects of gain corrections changing with frequency and the EQ de-emphasis, in this case corresponding to a Behringer post-EQ. As you can see, the de-emphasis allows us to remain in-bounds with the available digital headroom.

The resulting processed spectrum is slightly "whiter" than it was on input, while we rely on the outboard re-emphasis EQ to provide the lion's share of the overall correction gains. The DSP processing handles the fine details.

Here is a snapshot spectrum of the log-sweep sinewave signal being applied in this test. As you can see the signal itself is about 40 dB above the pink-noise background. You can see the masking effects here as well.

The right end toward 20 KHz illustrates the noise reduction capabilities of the System 5000 algorithm. As sounds fall below perception levels the gain in those Bark bands is reduced in order to avoid excessive noise amplification. Elsewhere, the pink noise background is well above the no-sensation level, and so it is fed to the listener.

And just to show that nothing is up my sleeve... Here is the same log-sweep test performed with a -6 dBFS (= +8 dB K-14 RMS-Wide, i.e., very loud!) sine and no noise background. Just to be sure that nothing is hiding behind that noise.

This signal is actually 8 dB more intense than in the previous test, thereby being even more likely to trigger gain change artifacts.

As you can see there are no vertical streaks showing up in the sonogram that would indicate abrupt gain changes as the signal moves from analysis band to band.

The scale on this image places the black level at -120 dBFS. There are some faint artifacts showing up down in the bass region that correspond to weak distortion products. But otherwise this image couldn't be any cleaner.