I thaught this was a very interesting and informative essay on what happens during PWM. The main part of the article then appears to reveal an amazing and arcane trick to simulate this sound: Detuning two sawtooths! (?) Is it me, or did I miss something...

Also...

None of the string ensembles I've used employed PWM. The waveform is normally a very 'short' pulse softened with filters. This provides a light overall sound which doesn't step all over everything else in a mix, like full blown saws might. 99% of the sound though is due to the noisy but silky analogue 'ensemble' effects they used. That is the whole reason you'd get one in todays environment. Just try any of the String Ensemble emulations available to see why... not even close. I nearly burst a blood vessel trying to disect and get a vaguely passable sound in some solina patches.

Well to be honest you probably missed domething. This is one of the oldest modular tricks in the book. The saws should be a saw and an inverted saw. You will notice that on some synths that lack PWM, there's a choice of saw and inv saw, for example Waldorf Microwave and Modulus Monowave. I'm not sure how common this is, I just mention two synths that I've used a lot.

As you say it IS the oldest trick in the book, and I was just amused that it was presented as an almost revolutionary concept. It must be one of the first 10 things a synth novice tries (especially with the press it's had since the VA's came in) But I wondered if I had missed a subtle point - evidently not.

The article however does claim that a ramp is not nesescary.

I thaught this was a very interesting and informative essay on what happens during PWM. <snip> Is it me, or did I miss something...

No, it is true, provided one of the saws is phase reversed. So, one saw is rising and the other is falling.

Just make a sound recording of the attached example in e.g. Soundforge or a similar program and have a look at the waveform.

In fact the effect of two sawwaves or deep pwm (0% <-> 100%) is quite similar, only deep pwm is a stronger effect, if the pulsewidth goes through 0% there is no sound. However, nothing like that is used in e.g. the Solina or the Arp Omni or similar string ensembles. They all use a chip that generates the twelve frequencies of the top octave as squarewaves. Each frequency is divided down by multiples of two to get all the other frequencies. These frequencies have square waveforms but by using some logic chips these are combined to generate pulses of 25% and 12.5% for the avaiable registers. The combined waveforms are filtered a simple fixed filterbank for each register and only after that stage the ensemble effect is created with an ensemble unit consisting of three delaylines with variable delaytimes between 10 and 40 msec. So much longer than the NM can do, but also much more noisy So, the timbre of the Solina is mainly produced in the fixed filter banks for each register together with the ensemble unit. The trick I used that got me closest is to put a 50% pulse wave through a tracking highpass filter and ringmodulate the output of the HP filter with the original pulse. If I listen carefully to the sound of your patch there is a definite problem in the 10kHz up region of the soundspectrum, there is just too much buzzy sound energy over there which does not attribute at all to the ensemble sound. Remember that the spatial effect of a sound is produced in the midrange of the spectrum, if the harmonics in the 10kHz up region are chorused a high buzzy sound is produced that has no spatial effect whatsoever, instead it kills the spatial effect of the midrange. The reason is that the human mind uses the phaseshifts between the left and right ear of roughly the 500Hz-3500Hz band to locate a sound in the stereo field. For other frequencies the distance between the ears is either too large or too small to locate a sound (or so it seems). To create a spatial ensemble effect it is best to apply the effect only to this 500Hz-3500Hz region. So some careful filtering is needed, both on the pre-chorus timbre and in the ensemble unit. Also listen to how the effect of phasing caused by the two delaylines and the effect of the chorus module work together to get that effect of all those dancing midrange harmonics in the ensemble sound.

Simple guideline is that when making an ensemble unit or using the chorus module it is best to prevent higher frequencies to get a chorus effect, but instead feed these clean to the output. That will increase the 'vintage' spatial effect and prevent the high frequency buzz. A 6dB LP filter set to 2.74 to 3.90 kHz before the unit will in general do the job but steeper filters might be used to get more special effects, albeit subtle.

That's a great trick Rob, I just had a go, (at different cutoffs it really makes some fat saw shapes too)

If I listen carefully to the sound of your patch there is a definite problem in the 10kHz up region of the soundspectrum, there is just too much buzzy sound energy over there which does not attribute at all to the ensemble sound.

And as you say it does need careful filtering, my difficulty is, I filter to lose the 'swirling' effect, yet when I compare the real solina, it has lots more high end *and* richness. OK, I *know* it's analogue, but I'm sure it's possible to get the modulation and filtering damn close.

Rob, I've taken the liberty of sending you a sample of the Solina chorus (100k) as you may not have a 'raw' one handy.

Solina C1 - C4

Here's some stuff I dug up on the Solina 'Chorus', some from the patent: (which I haven't stuck rigidly to in many patch permutations)

For the fx section, the solina uses 3 delay lines, but not 3 lfo´s to drive each delay, it uses 2 3-phase LFO´s to modulate the delay and the output is not mixed with the dry signal.

A circuit for producing a chorus effect in an electronic musical instrument.

The circuit includes N separate channels, where N is an integer greater than one, with each channel having an analog delay line to which a tone signal is applied. Each delay line frequency modulates the applied tone signal at a subaudio rate in response to changes in the frequency of clock pulses applied to the delay lines.

The delay variations in one delay line are out of phase with the delay variation in every other delay line by a selected amount which is normally 360°/N.

Clock pulses are generated by means including a nonlinear circuit to compensate for the nonlinearity in the frequency interval between tones in the musical scale.

The outputs from the delay lines after filtering of the clock frequency components are utilized to produce the desired chorus effect output from the instrument. Each of these outputs is applied through a separate voltage controlled amplifier (VCA) with a common control voltage being utilized for all of said VCA's. The magnitude of this control voltage is determined by the amount of light from a light source which impinges on a photo-resistive device controlled by a mask positioned between these elements.

What is claimed is:

In an electronic musical instrument, a circuit for producing a chorus effect comprising:

N separate channels where N is an integer greater than one, each of said channels having an analog delay line; means for applying a tone signal as an input to each of said delay lines; and means for generating clock pulses for said delay lines, the clock pulse variations for the delay lines of each channel being out of phase with those for each other channel by a selected amount,said means including first and second N stage counters, means for stepping said first counter at a first predetermined relatively low frequency and for stepping the second counter at a second frequency, the second frequency being several times greater than the first frequency, means for combining the output from each stage of the first counter with the output from a selected stage of the second counter and applying each of said combined outputs to a single line, a filter means for each channel, a voltage controlled oscillator for each channel, means for applying the combined output on each of said lines through a different one of said filter means to control the output frequency of the corresponding oscillator, and means for utilizing the output from each oscillator to produce the clock pulses for the corresponding delay line; the outputs from said delay lines being utilized to produce a chorus effect output from said instrument.

Here's two more string ensemble patches that layer very well together (on an expanded NM). When changing the voices take care that no voice is assigned to the DSP where the CVA resides, these are SlotD-bug prone patches!