jh_scanner_vibrato

JH. Scanner Vibrato

shameless advertising (;->) :

my 3rd CD

Holiday in Purgatory

Ever since I had designed that Interpolating Scanner for the JH-3 Synthesizer, I wanted to use it for this very application which had inspired it in the first place: An emulation of the the Hammond Scanner Vibrato, without moving parts.
Now I've finally found the time to make some experiments, and I have built two veroboards, one with a "Line Box", and one with the Scanner.

1. The Delay Line

In a Hammond organ, the Line Box is an approximation of a long transmission line (to provide a delay of approx. 1ms), built from lumped elements, i.e. from inductors and capacitors. This can only be an approximation of a real transmission line, fairly good for low frequencies, but with a distinct upper frequency limit. Therefore a Line Box is a low pass filter; a line Box with 25 Inductors actally forms a 50-pole (!) low pass filter. Later Hammond used a shorter Line Box with 18 inductors, which were magnetically coupled, to get a similar delay time with less inductors. Or so I've been told; I never analysed this later version. Instead, I went straight for the 25-inductor version, which needs more components, but is much easier to handle mathematically.
As using 25 inductors in the 500mH range was out of the question for me, I had two options left:
a) electronic emulation: FDNR-filter, or
b) using smaller (readily available) inductors.

I did some simulations for option (a). And I found that an electronic version of this "simple" LC circuit of basically 25 identical L's and C's is not so simple at all! First of all, a 50-pole LPF is not trivial. The better of my two filter design programs allows for a maximum filter order of 25. There are not many applications that need a 300dB/Octave LPF, after all. (;->) Also, it's not enough to get the desired response at the end of such a filter. Every tap along the line must carry the right signal, i. e. delayed by a certain fraction of 1ms, and with a fairly smooth amplitude response. Both of these constraints cry for a FDNR implementation, where the topology of the original LC circuit is preserved. Just simulation the inductors with gyrators is difficult, as all these inductors are floating (not grounded). So we have to transform inductors to resistors, resistors to capacitors, and the (grounded) resistors to grounded frequency dependend negative resisors (FDNRs) a.k.a. "super capacitors". This would certainly have been possible, but it's an increadible demanding application for the opamps that are used. TL074's, which are certainly fast enough for ordinary audio applications, are way too slow for a FDNR LPF with a few kHz cutoff frequency. This will result in an overshot that is increased from stage to stage - no way to build a Line Box with 25 stages with these. I could have used much better opamps, but these would have been expensive, and fast as they are, a lot of decoupling capacitors would have been needed.
So this solution was discarded, and I chose option (b) instead.

So following option (b), I went shopping for affordable inductors.
I found some unexpensive, small, pcb-mount 33mH inductors.
Some quick calculations were encouraging:
Making an impedance transformation on the original Hammond circuit, a capacitor value of 47nF would fit for the 33mH inductors,
providing the same 1ms delay and the same cutoff frequency as the original, and resulting in a decreased line impedance of 840 Ohm.
Now 840 Ohm isn't that low really - it can still be driven from a simple emitter follower without much effort. I was uncertain about the maximum current the inductors could carry without distortion, and thus about the maximum signal voltage that can feed the delay line. I found that a signal of 4Vpp is no problem. And I didn't intend to go much higher anyway: As I'm using 4000 series CMOS switches in the scanner, a supply voltage of +/-7.5V was chosen anyway. So as a nice coincidents, all these factors fit together.
The remaining question was about the losses in the inductors. These losses are higher than in the original Hammond circuit. (No surprise when you look at these tiny inductors!) But Hammond had to take care of these losses, too. As the signal voltage decreases along the Delay Line, taps at the beginning of the Line were attenuated by resistors, and taps at the end of the line were fed into the scanner unattenuated. As the losses in my curcuit are slightly larger, the compensation for the individual taps must be different. I have simply fed a signal into the line, measured the level after each inductor, and calculated a correction factor.
At the moment (verobaord test circuit) each tap (after each inductor) is buffered with an opamp that compensates the loss. I did this because I also plan to experiment with multiple scanners scanning alongthe same delay line (trying to get a solina triple chorus effect). For a final Hammond-Vibrato version, these opamps can be omitted, and weighted resistor values to feed the scanner can be used instead. (For the VCAs in the scanner, the signal is divided down to 20mVpp anyway.)
Schmatics for this Delay Line will come soon. I have to make a clean drawing first.

2. The Scanner

In theory, I would just have to expand my old 8-stage Interpolating Scanner to 9 channels for a Hammond Vibrato emulation, and control it with a triangle LFO of carefully adjusted amplitude. Then the 9 Line Box taps would be scanned back and forth like the original, where a 16-stage rotary scanner was wired to form a linear 9-stage scanner.
But I think I have found a better solution. Instead of using 9 VCAs, I'm now only using two, connected as an ordinary crossfader that blends between two signals. Each of the two VCA inputs has a 4051 analogue multiplexer which is switched to a new input signal exactly at the moment when the connected VCA is completely off (and the other VCA is completely on). Further, as we don't have a scan with an arbitrary CV, but a continuous scan in one direction (at fixed rate, or even at variable rate), the control circuit can be much simplified by interweaving the LFO circuit and the MUX control:
A triangle LFO runs at 55 Hz with its triangle wave directly controlling the 2-VCA-Crossfader. As the triangle reaches its highest point, the schmitt trigger of the LFO is toggled, and also the counter for one 4051 MUX is increased. As the triangle reaches its lowest point, the schmitt trigger is toggeled back, and the counter for the other 4051 MUX is increased. For the counters, a 4520 is used, which allows opposite polarity for its two clock inputs.
Thus a 16-stage rotational scanner is achieved without a lot of components.The two npn pairs that form the VCAs must be selected for very low offset voltage (<<1mV), or trimmed (not shown). The pnp current mirror, shared by the two VCAs, can work with unmatched transistors, as the total current from both npn pairs is constant.
And there is another advantage from using just 2 VCAs: The shape of the crossfade function can be changed easily (because I only have to do it once for all channels). I don't know the exact geometry of the stator and rotor capacitor plates of a Hammond Scanner, and even if I did, it would be difficult to determine effects from th edges of the plates on the capoacitance. (The E field will not be homogenous there.)
So while ideally the crossfade function would be like a triangle, in practice it might be more than a trapezoid. The 20k trimpot across the emitters of current steering transistors will change that shape. I have set it such that I have a triangle function, and adding any resistance in parallel (eternal pot, switch, etc.) will change it more to trapezoid. (With the two points shorted, I get an almost-rectangular chopper effect.)

16-Stage Scanner Schematics

3. Audio Samples

Dry Organ sound (84kB *.mp3)
(A Korg CX-3 with 16' 8' 4' 2' 1' drawbars at 8, all others at 0, playing a Gsus4 chord: G3 + C4 + D4)

Organ with Scanner Vibrato in Chorus 1 (C1) mode (64kB)

Organ with Scanner Vibrato in Chorus 2 (C2) mode (82kB)

Organ with Scanner Vibrato in Chorus 3 (C3) mode (108kB)

Organ with Scanner Vibrato in Vibrato 1 (V1) mode (59kB)

Organ with Scanner Vibrato in Vibrato 2 (V2) mode (65kB)

Organ with Scanner Vibrato in Vibrato 3 (V3) mode (85kB)

Pad sounds from an OB-8 synthesizer thru Scanner Vibrato in C3 mode (1.1MB)

A held chord on the OB-8, Scanner Vibrato with variable speed switched in after ca. 10 seconds (960kB)

4. Picture of Prototype

5. No Inductors ?

Idea for Leapfrog Filter implementation (untested!)

Back to JH. homepage