posted 26 March 2001 10:14            

auralexciter.kym

Attached sound is a stereo aural HF exciter. It works by adding harmonics of a sound that has passed through a high-pass filter to the original sound. You can select between even or odd harmonic generation, and you can bypass, and audition the effects alone.

Some people complain that odd harmonic generation doesn't sound musical. Well, to my ears it sounds better than even harmonic distortion, but what the heck... this one gives you a choice.

It works by high-pass filtering the incoming L/R channels and passing along to a multiplying waveshaper that uses a cosine table. The result is sent through a DC blocking high-pass filter and then onto a gain stage and mixer with the original material.

The principal is the same as FM with a DC carrier. The waveshaper produces sidebands at even multiples (since we used a cosine table) of the incoming frequency, in proportions determined by the BesselJ functions of even order with modulation index Pi*ampl where ampl is the drive level. In order to get the odd harmonics, we simply multiply again by the incoming signal. To get just the even harmonics we multiply by a constant instead of the incoming signal.

Cheers!

posted 26 March 2001 11:47            

There isn't normally much energy at high frequencies and the output of the filter may not excite as many harmonics as you want. Putting a gain stage after the filter allows more of the wavetable range to be utilized, and generally provides more energy in the higher harmonics. But the exact analysis of amplitudes is complicated by the Bessel functions. More gain in the wavetable doesn't necessarily mean more energy in any one particular harmonic.

... er... well I just plotted the amplitudes of the first 3 harmonics (J2, J4, and J6) with the modulation index from 0 to Pi. It turns out that these are monotonic in this range, so increasing the waveshaper gain will provide more energy in all the harmonics...

- DM

posted 26 March 2001 16:29            

I've been looking for one for some time.

posted 26 March 2001 20:10            

auralexciter.kym

It also provides some waveshaping injection gain that can be backed off by the post gain. The net effect is to excite more harmonics, but without significantly increasing the components at the fundamental frequency (unless you don't back off by a compensatory amount on the output).

There are 3 gain controls in the system:

Level -- controls the input level to the high-pass filter.

ExcGain -- controls the boost going into the waveshaper for development of higher harmonics.

Gain -- controls the output gain from the exciter. This amplified/attenuated signal is then subtracted from the original signal.

- DM

posted 27 March 2001 02:25            

auralexciter.kym

You want to tune this so that you don't clip the input signal using the excitation gain. (injection level = 0 dB) I try to stay lower than -3 dB going in. These are peak reading meters, and the persistence can be set with the release time fader.

The input level (in dB) is for attenuation of the high-pass filter input. The following excitation gain (in dB) serves to boost the injection signal to the waveshaper to encourage the creation of more harmonics.

The output gain is for setting the audible level of the HF excitation. Adjust to taste.

I found that getting good performance for both even and odd harmonics requires the use of both sine and cosine wavetables without multiplying by the signal. So this version includes a pair of waveshapers on both input channels, selectable with the "OddH" toggle. (I still think odd harmonics sound better... I tuned it up on the Kurdish woman singing. Odd harmonics make her voice sound absolutely tantalizing!)

The output DC blocking filters have a "DblF" switch to select between your input cutoff frequency, or twice this value. For even harmonics it makes little difference, but for the odd harmonics setting "DblF" cuts out the fundamental and leaves only harmonics 3, 5, 7, etc.

Wish list item for SSC: Some peak holding LED bar meters in the VCS panel. Perhaps with a slow decay in the peak level, plus instantaneous readout of the peak levels.

I promise to drop this topic now...

- DM

posted 28 March 2001 00:31            

auralexciter.kym

Final posting on this topic (from me!)...

Subtle improvements and complete tuning instructions for total novices like me...

Attached sound now includes a version with odd-only harmonics and 2-pole input filtering (I think it sounds better).

VCS Panel:

Cutoff frequency reads 5 - 10 (KHz) as the frequency of the 3rd harmonic (the main enhancement).

Level is the attenuation into the high-pass filter (2-pole) and runs from -20 dB to 0 dB.

ExcGain runs from 0 to 20 dB and is the boost going into the waveshaper.

Gain (-10 dB to +10 dB) is the enhancement level of the enrichment. This gain control has already been compensated for the input Level and ExcGain. (In other words, you can change the Level and ExcGain, and the volume output level stays fixed.)

PeakIn Meter = peak level between L/R channels going into waveshapers.

PeakOut Meter = Peak output between L/R channels after recompensation by Level and ExcGain but prior to Gain application. In other words, the PeakOut meter shows the actual level going out at these frequencies before excitation Gain is applied. Since the dominant energy here is the fundamental, this meter shows a pretty good indication of the peak levels present at the cutoff frequency in the original signal.

Release is meter relaxation time in seconds. Longer release times hold the peak reading longer, but in any case the attacks are 1 ms for immediate rise response.

EffOnly button - toggle to hear only the effect or its combination with the input sound.

Bypass button - toggle to turn off the excitation

Tuning Up...

1. Start with Level at max (0 dB), ExcGain at min (0 dB) and Gain at midlevel (0 dB).

2. Inject sound and watch the PeakIn meter. It should be less than 0 dB. If not, then reduce Level until it is.

3. Boost ExcGain as desired, but keep PeakIn reading below 0 dB. It will never show more than 0 dB even when the system is overloaded because of digital clipping.

4. PeakOut shows the amount of Gain headroom you have. You can adjust Gain to no more than the negative of this before the output stage clips digitally. Generally at these high frequencies you can go all the way if you really want to because the energy here is so small.
-----------
These are the nominal extreme tunings. To get more or less enhancement at the higher harmonics use more or less ExcGain.

The output Gain control is akin to a Dry/Wet Mix ratio control. Lower gain settings leave the signal more dry. Higher gain settings wet it more.

Finally, (and I have to learn this one...) use your ears and tune to suit.

- DM

posted 28 March 2001 11:21            

Others might find the comments interesting too!

I just did all this because I have always been curious about what these devices do and how they work. Most of my Kyma toiling is for discovering how things work...

posted 29 March 2001 12:28            

This whole topic got my engine started... I went to the Aphex site -- home of the better known Aural Exciters, to read about the real units. They list their patent number in their brochures.

A check of the US Patent Office
www.uspto.gov/patft/index.htm

Patent number: 4,150,253 granted in 1979, reveals the ancient patent cited by Aphex. It is an amusing read, and shows that to get their basic performance, one needs to send the processed signal through a simple 2-pole Butterworth HPF, and then into a one-sided clipper to generate both even and odd harmonics. Back then it was all tube circuits.

I played around a little with Kyma in this regard, and generated a simple half-wave rectifier transfer function to be used by a Waveshaper module. It does indeed generate both even and odd harmonics, but I am not sure it sounds better than using the Sine and Cosine modulation transfer functions.

Aphex exclaims the virtues of a phase shift incidental to their clipping circuit and HPF. Aside from the phase shift induced by our HPF, I'm not sure how to get a similar effect on the Kyma, nor how important it really is.

The ProAudio guys claim that it is very important to get both even and odd harmonics.

FWIW... I think, if it sounds good on the Kyma, we can thank the engineering at SSC for doing such a remarkably good job in their analog and conversion stages. Without a doubt, the Capy is the best piece of gear in my own studio in this regard.

BTW -- a quick check of the Kurdish singer on the spectrum analyzer shows, without a doubt, that this sound was regenerated on a bank of oscillators. Very impressive indeed!

- DM

posted 09 April 2001 11:35            

The suggestions I have seen recently indicate that for digital recordings, a bit of touch up high EQ is more useful.

Do all you audio pro's agree with this sentiment? How do you tend to use things like aural exciters?

posted 11 April 2001 01:57            

auralexciter5.kym

Not really using an "anti-compressor" on output, but rather a pair of VCA's driven by the outputs of the waveshapers, and amplitude followers on the input HPF's, to restore the output loudness levels to the [relative] input levels present ahead of the distortion waveshapers.

There is a fudge in the output amplitudes driven by the combination of the amplitude follower gains (100x) and the overall output effects gain. I chose these so that the range of effects gains from -10 dB to +10 dB would cover nearly every need. Feel free to change these to suit.

The amplitude followers have e-folding time constants of 10 ms, suitable for these higher frequencies.

The wavetable polynomial coefficients were derived from a 5th order Tchebyshev series approximation for a unipolar clipper with a 10% threshold.

The input compression ratio of 200:1, with a threshold of -30 dB, virtually guarantees constant distortion levels regardless of input amplitude. The ExcGain (Exciter Gain) following the compressors is used to nudge the input levels to the waveshapers as close to zero dB as you wish to go.

The use of a unipolar clipper [like the Aphex units] generates both even and odd harmonics. The input filters are tuned according to the frequency of the third harmonics desired, from 5 KHz to 15 KHz. (this is a leftover from previous experiments, so feel free to change the frequency expression in the HPF's).

In an effort to emulate the Aphex exciters, I have placed 1 ms delay lines in both output channels. There is a phase reverse button (leftover from earlier experiments) that has virtually no effect unless these delay lines are substantially shortened.

There are gain indicators in the VCS panel showing the waveshaper injection levels in dB, and the available output margin in dB.

You'll see an input mix between channels 1&2 and boosted 3&4. I frequently use Kyma effects like these with either S/PDIF channels 1&2 or Analog Effects sends into channels 3&4. Similarly, the outputs are sent to both S/PDIF channels 1&2 (on my system) and analog outputs 3&4.

For some reason my unbalanced analog effects sends into channels 3&4 need 12 dB boost instead of 6 dB as I would have thought. That's why you see the input gain after channels 3&4 inputs. Any ideas as to why that is?

[I'm confused about this... half the input voltage swing should be 6 dB regardless of input gains. Shouldn't it?]

- DM