Free State FX - FSFX 106: Son Of Storm Tide Stereo Flanger
2.75" behind the expertly arranged and appointed 4MU panel
Alpha Pots and Switches
Now World-Famous FSFX build quality.
Get Yours before it's gone Somewhere in Time..."
Jurgen Haible says:
"My latest effect device (as I'm writing these lines) is an analogue flanger. There are more Flanger boxes on the market than I could possibly count, some Flanger stomp boxes pretty unexpensive, and I already own a few delay FX boxes (both analogue and digital) which can produce a Flanging effect - so why build another one ?
There's a couple of reasons, because there are some special features you won't find on many Flangers, which I wanted to try, and - if found pleasant - implement them in one device. And indeed I did find pleasant things, many of them coming from the "Instant Flanger", a classic from Eventide Clockworks...
I'm using the SAD1024 dual BBD chip, which is out of production and, while not ultra-rare, a bit hard to come by. You could use two single BBD lines like the TDA1022 or Panasonic "MN" BBDs with a slightly lower limit of clock frequency. (I'm clocking the SAD1024 up to 1.5MHz, just to test how far I can go; it's the first time I used the SAD1024.) TDA and MN devices are no drop in replacements, so you should know how to modify the circuit before you take that path. There's a large number of trimpots in my circuit. Many of them are not necessary, but I used them to test the limits of the circuit. The circuit is given as I built it, with all the trimmers rather than optimal resistance values.
Features I wanted to try, and results I got so far:
1. "Theta Processor"
This word was coined by an article in Electronotes for a combination of Flanging and Phasing. Flanging, on the one hand, involves constant (i.e. independent of frequency) delay, i.e. linear phase. Phasing involves frequency dependent phase shift, which also means frequency dependent delay. The main shortcoming of Flangers is that the peaks and notches (when the processed signal is mixed with the dry signal, or when feedback is applied) are spaced linearly over frequency, which means that you get more and more of them per octave for higher frequencies. They are are also located at integer multiples of Hz, so when one peak of the Flanger's response courve hits a harmonic of an input signal, all the other harmonics will be hit by other peaks of the flanger at the same time. This is the reason for the "metallic" sound of flangers.
Phasers, on the other hand, have a more "musical" distribution of peaks and notches. The spacing can be controlled by design (by detuning individual all pass filter stages), but even with several identical stages you get get very different spacing as with a Flanger. The big drawback is that for a "deep" sounding Phaser, you need a lot of voltage controlled all pass filter stages. And even a 14-stage Phaser as found in the ARP Quadra will only produce 7 notches.
The idea is to combine a delay line (Flanging) with all pass filters (Phasing), to get the huge number of peaks and notches, but a "musical" (notches per octave rather than notches per Hz) spacing as with Phasers. The Electronotes article proposed a combination of a few dozen *fixed* all pass filters (that is, just an opamp, 3 resistors and one capacitor each) and a modulated delay line BBD). I never tried this circuit, but the article showed that with 50 ... 100 all pass filter stages the goal would be achieved quite perfectly.
The Eventide Instant Flanger uses a similar approach, but with much less stages. There are two BBD delay lines (clocked at the same rate, so it's more like a single center-tapped line), and as few as 4 all pass stages. As it's practice in Phasers and Flangers, the input signal is split into two paths. One goes into the BBD lines, one goes into the all pass filter chain. The Eventide creates a stereo output signal like this: One channel is a mix of the first half of the BBD delay and the first couple of all pass filters. The other channel uses the full BBD delay and the 4 all pass filter stages. Now that looks much more sensible than 50 or 100 all pass filters, and I wondered how effective it could be. I definitely wanted to try it.
My version has the all pass filter stages switchable (with separate switches for the first and second pair). I also included a (switchable) send / return effect loop to try longer all pass filters, or fixed delays, or frequency shifters, or whatever. (See Thru-Zero-Flanging below). The result was really stunning. The 4 all pass filters have no big effect as long as the Flanger runs in mono, and they are certainly not enough to create the ideal of "musical" spacing of peaks described above. But as soon as you run it in stereo - BIG effect. Running the Flanger in stereo *without* the all pass filters is quite nice, but nothing really special; you have probably heard it many times before. But switching in the all pass filters sort of "unlocks" the stereo image. You're switching from an up-front metallic resonating stereo effect to a much smoother "space" sound ("space" as in "room", not as in "outer space"). Well, outer space is within reach as well. I have only made tests on the veroboard as I'm typing this, but I think I have a clue about Tomita's drastic Flanging effects now (;->).
2. Thru Zero Flanging
It is well known that the "original" Flanging (with two tape machines) can produce an effect that can't be emulated with a single electronic delay. Electronic flangers can vary the delay time between long and short, but the delay cannot become negative, i.e. it cannot go "thru zero". With two tape recorders, any of the two machines can be slower than the other, so you can manipulate the speed such that one half of the signal "overtakes" the other. For one moment of overtaking, both signals are in phase for *all* frequencies, which causes a special effect. To emulate this with electronic devices, you need two delay lines with their outputs being mixed (to to be mixed up with the two BBDs in series as in the Eventide), with one delay time fixed, and the other modulated shorter and longer than the first. Or with two delay lines modulated in opposite direction.
I have made such experiments with the BBDs of a modified Polysix FX board, and the results were not quite as desired. A friendly member of the synth-diy community has told me he had similarly unsatisfactory results. The problem is aliasing noise here. Delay lines (BBD or digital) are time discrete devices, and even with the two clock frequencies way above the audio range, you get interference from the difference frequency of both, which *will* be in the audio range at some point, when you're modulating one clock frequency against the other. Worst case is using two identical BBD chips, as the *fundamentals* of both clocks will interact during the most interesting thru-zero time.
My choice was not to integrate a second delay line in this box, but provide an insert path instead. So I can try various external delays, from Effectron to Dynacord, and see which will give the least interference. The insert path is inside the compander system of the Storm Tide Flanger, so the external delay line can always be operated at optimal input level.
3. Noise Reduction
There's a friendly person on the synth-diy list who will remind us that BBDs sound bad, from time to time.
(Hi Harry !) Compander-less BBD delays have no great SNR indeed, and require careful adjustment of input levels at least. Many BBD delays use a NE570 or NE571 for noise reduction. This is a great improvement, especially when it's done as well as in some of these big Dynacord devices. But these chips are rather slow (if you design them for low distortion, that is), so these boxes are still prone to overload from fast attack transients. I have chosen the simple but effective compander from the Phaser of the ARP Quadra. This is optimized for use with medium to high level synthesizer signals. (If you run a readily mixed ballad with voice and e-piano at low volume thru it, it's not that perfect, but it's great for all kinds of synthesizer sounds, and for a Wurlitzer at decent level as well.) No more hairy noise, Harry (;->).
Resonance can be positive or negative (center zero potentiometer), and the resonance loop can be chosen short (one BBD) or long (both BBDs). This provides good variations, but is pretty standard. The interesting part is the Resonance Limiter, which is also "inside" the Compander system (after Compressor, before Expanders). Not my invention but ARPs, I cannot stress enough how sweet this sounds. It's also taken from the Quadra Phaser, with levels adjusted to fit the BBDs. The idea, as I understand it, is that you have full control over the "balance" between input signal and resonating peaks. In many Phasers, Flangers and Synthesizer VCFs the resonance will either be too "weak", or it will "scream" into your face. Sometimes this is desired, but sometimes you have to carefully set the resonance amount to find a "sweet spot" - for one input signal, that is. If your input signal changes its harmonic contents (not so much a problem with synth VCOs, but with more complex sonic material), it might suddenly scream once again, when a new harmonic hits a resonant peak of your device. The solution is to limit the amplitude, which many circuits do, some better than the others. Often it's the *combination* of input signal and resonance loop that is limited, which leads to problems.
Imagine your device clips at 0dB (for ease of description). You may even have a soft clipper at 0dB to prevent too "harsh" distortion. Now feed in a signal with variable level, and increase the resonance. Let's start with a input level of -10dB. Without resonance, that's -10dB at the output as well. Now increase the resonance (with input level constant at -10dB) and adjust the delay time until the resonance peak "hits" a harmonic of the input signal. With increasing resonance, you'll get an increasing peak (or several peaks) in the frequency response, until your output amplitude reaches the soft clipping point at 0dB. Increasing the resonance further will increase distortion (sometimes pleasantly so), but it won't increase the peak level anymore. So you get a nice resonance of 10dB relative to the input signal.
Now repeat the very same experiment with an input level of -40dB. Increase resonance until soft clipping occurs at 0dB, only now your resonance has become a "scream" 40dB louder than your input signal.
When I was new to electronic music, and my ears were younger, I was fond of this, but meanwhile I see the benefits of smooth sounding circuits.
With the ARP type resonance scheme, you have full control of the "sweetness" (or harshness) of resonance. After the peak-limiting type compressor, you have a pretty constant signal input amplitude, which is then divided down to fit the BBD's voltage swing. For instance, with the BBD being capable of 1Vp, you can "partly fill it up" with the input signal to, say, 0.25V, (or -12dB) *regardless* of the actual signal amplitude before the compressor. Then you can adjust the soft limiter of the feedback path to 0.75V, so the sum of both will never exceed 1V. You're effectively limiting the resonance peaks to 12dB above the input level.
This can be adjusted for other values of course, by choosing a different divider factor between compressor and BBD, and by adjusting the resonance limit accordingly. It's just important that both add to the maximum allowed voltage swing of the BBD. This is the "big secret" of the ARP Quadra Phaser, IMO. The limiting factor there is just the level of the all pass filter transistor ladder rather than a BBD.
Now this is another Eventide feature. It's not the original circuit, but I have just used different components to avoid electrolytic capacitors, so all credit for this goes to Eventide here as well. The summed control voltage (Manual, Pedal, LFO, Envelope follower) is fed into a sub audio BPF, then into a nonlinear amplifier with positive feedback (not a Schmitt trigger, but similar), and then filtered by a second BPF. A variable portion of this is mixed to the straight CV to control the BBD's clock frequency. The effect is most prominent on single-shot, non-periodic CV changes, such as a fast single Manual of Pedal sweep. The single sweep is followed by smooth "echoes" that go in either direction (faster clock and slower clock). This is to emulate the "bouncing" tape speed of reel-to-reel tape machines when the friction from a thump against the reel is suddenly released.
This is a full wave rectifier in the CV path that prevents the clock frequency from going too low. Rather than just limiting the CV, it folds negative CVs back up, which gives a nice effect with the Bounce circuit and also with LFO waves that would go below zero (in combination with the manual CV, that is). I reckon that Eventide has used this to limit the clock rate (sampling theorem ...) and to enhance the Bounce effect. I have changed this such that the foldback threshold is variable, so you can limit the range to higher clock rates if you like. Just bounce between 500us and 300us delay if you want ... (My lowest limit, with the foldback pot fully ccw, is adjusted for 40kHz clock rate.)
7. LFO Waveshapes
With an external CV input available, the internal LFO can be left quite basic. No internal Sample and Hold or SAW waveforms, for instance. I have implemented two options, a linear Triangle, and a EMS-like unsymmetrical Sine wave that produces the famous "roller coaster" or "Dive" effect. The idea is old, but my single-OTA circuit may be not. Input level shifting, nonlinear bending and re-adjusting output DC level in a single 3080 stage.(The Switch performs a brute-force override from the triangle LFO when its closed.)
8. Anti-Aliasing Filter
Built around a single LM301, this has a 24dB/Oct rolloff. Taken from the Eventide Flanger, does not work with other opamps."//