Some chorus/vibrato questions (3007-based)

[midwayfair]

A present has forced me to ease into learning a bit about bucket brigade devices, which I've been avoiding. I'm just making a simple chorus with sine-wave modulation and 15V operation, essentially a simplified CE-2.

I have three four questions:

-Does anyone have a good guess as to the typical delay time at idle in a CE-2 or similar design, or if that simply doesn't matter? The chip does 5.2-51ms, but obviously it has to wiggle, and I expect that the last quarter or so of that is noisy. I get that the delay time is set by resistance between pins 5-7 of a clock chip in delays, but in the Madbean Pork Barrel (http://www.madbeanpedals.com/projects/PorkBarrel/docs/PorkBarrel.pdf, page 4), there is no pin 6 connection, and pin 7 is basically controlled by the modulation.

-Does anyone have any experience using the vibrato mode of a design like that where they could tell me how noticeable the latency is?

-Are 3007s anywhere near as sensitive to bandwidth and headroom as 3005s, or is this not as much of a concern since the chip isn't being squeezed for every available mS of delay?

-Does anyone have a handy link of required reading? A forum search turned up surprisingly little except a couple fun PT2399 designs that I'm not sure will work.

[Mark Hammer]

Welcome to the dark, clocky side. 

I wouldn't take the datasheet specs too literally.

As for "idle time", it is whatever you want it to be, provided you have either a) the current drive on the clock to make it go that fast, or b) the lowpass filtering to cope with making it go that slow.

The teeny storage cap cells inside the BBD will leak.  Not lightning fast, but fast enough that they won't hold a signal forever.  So there will be limits to how slow any BBD can be clocked before the signal you're trying to delay drips out and becomes unrecognizable.  Conversely, there is capacitance on the clock input pins on the BBD, that magically turns nice square clock pulses into things that are more traingular than square if the frequency gets too high.  This buggers up the bucket-brigade aspect of smooth seamless cell-to-cell handoffs of signal by inserting blank periods as the clock pulse rises to the level that will trigger the switches inside the BBD.  Happily, buffers that pump a little more current with those clock pulses can overcome that particular limitation.

The specs you see on the datasheets are predicated on using THEIR clock chip, with nothing else between clock chip and BBD.  While the clock chip is an elegant and space-conscious solution, it does not possess the needed current drive to clock very quickly.  So those specs are for a somewhat limited circuit, rather than an "assisted" one.

The "vibrato mode" HAS no latency.  Yes, the Boss VB-2 vibrato (and its Behringer clone) has a variable rise time built in, but vibrato, as such, is simply use of a time-modulated signal with no real-time signal combined.  Wherever the LFO is at the moment is where it is, and whatever rate it is at is the rate.

Headroom for any of the BBD chips can be increased by means of companding.  You will see that invoked when there is any sort of appreciable delay involved.  When the delay is short enough, though, as in chorus and flanger pedals, more often than not you'll see the "poor man's Dolby", which is treble pre-emphasis (usually in the 1st or 2nd op-ampstage) with complementary treble de-emphasis on the op-amp output stage.  That's more for noise control than for headroom, though.

BBDs also provide one of those instances where higher supply voltage improves sound quality.  Not always, but often enough.  Most folks will build a CE-2 clone to run off a battery, but many will likely stick to an external supply.  I imagine you could easily run the thing off 12-15v, provided you keep an eye out for cap voltage ratings, and set your bias properly.

There is a ton of stuff on BBDs here.

[midwayfair]

Mark, thanks for the great explanations.

The "vibrato mode" HAS no latency.  Yes, the Boss VB-2 vibrato (and its Behringer clone) has a variable rise time built in, but vibrato, as such, is simply use of a time-modulated signal with no real-time signal combined.  Wherever the LFO is at the moment is where it is, and whatever rate it is at is the rate.

This confuses me. The pitch vibrato is created by varying the time of a repeat. You feed in a guitar signal, it runs through the delay, which outputs a signal with a delay of X+-Y mS, then it goes to the output. The dry signal never makes it there. So surely there must be X+-Y mS of latency in the signal compared to the (inaudible) dry signal? I get that it might not be noticeable when we're talking about a dozen or so milliseconds. Are we just using latency differently?

[Mark Hammer]

Well, we ARE using latency differently.  My bad.

But keep in mind that the audio signal you feed in is continuous, and the output of the BBD is continuous.  So, even though the output may occur a few milliseconds after you play (and most chorus pedals generally don't stray much out of the 4-12msec range), you won't notice it "beginning late", the same way you might notice latency on a gutar synth, or using a delay line with the mix set to 100% wet.

[Seljer]

Also, keep in mind that the speed of sound is 1.1feet per millisecond. So you get the same 'delay' as you do from a vibrato just from standing 10 feet in front of your amp.

[armdnrdy]

In a delay circuit you will have a "base" frequency that you set the clock at.

In a flanger or chorus the clock signal is always moving....so if there are adjustments, you set the upper and lower frequencies. (sweep)

You'll notice in the CE-2 circuit there are no adjustments for the clock.

The circuit in between the LFO and the clock sets the range.
Since the LFO would provide a pretty linear voltage to the clock, the circuit (Q4, Q5, D1, D2) provides a log or antilog (not sure which one at a glance) feel to the sweep.

[midwayfair]

Folks, thanks for the help.

Between this and some discussions elsewhere with Scruffie, I now have a bunch of stuff to test. Especially interested in the multivibrator method from the micro chorus, since that frees up a lot of space for some other things and makes for a less expensive and simpler build (and it's a lot easier for me personally to digest than the clock driver in the CE-2).

Hopefully I will end up with something good, small, and robust, and it will be sharable with the community.

[Scruffie]

Midway has already seen this but thought i'd post it here too as its a very useful analysis of the CE-2 and the information can also be extrapolated and applied to other chorus effects too - http://www.electrosmash.com/boss-ce-2-analysis#link8

[Mark Hammer]

Here's something I've never been clear on: what is the optimal, and most "musical-sounding" LFO waveform for a chorus?  We use triangle waves because...well, they're easy to make and seem good enough.  But is there anything that sounds nicer without requiring too much effort or cost/parts?  The poster child for this sort of thing is the CE-1, which uses different LFO waveforms for vibrato and chorus, but I rarely, if ever, see attempts to produce anything different.

Although, I have to admit that the various transistor constellations one sees surrounding MN3101/3102 chips are beyond the perimeter of my understanding.  bear in mind that ultimately, it is the behaviour of the clocking that is of interest.  Whether that clocking is a result of the direct control of the LFO waveform, or some lin/log/antilog conversion of a plain vanilla LFO triangle, is immaterial.

What I want to know is if there is some other manner of clocking BBDs in chorus effects that just sounds a little sweeter than the norm.  I'll end by noting that, in the case of flangers, the sort of waveform that sounds fine for fast gurgly sweeps is considered to be less optimal for long slow sweeps.  The so-called hyper-triangular waveform accentuates the lowest parts of the sweep...when and if it is slow enough.  But it does nothing particularly useful for fast sweeps and may even be undesirable for them.  The speed ranges that chorusses cover are not nearly as broad as those of flangers, so conceivably there is one sweep waveform that works well for all chorus speeds.  But I'm willing to keep an ear open to possible waveform differences that might make, say, slow Leslie simulation a little more palatable.

[midwayfair]

Here's something I've never been clear on: what is the optimal, and most "musical-sounding" LFO waveform for a chorus?  We use triangle waves because...well, they're easy to make and seem good enough.  But is there anything that sounds nicer without requiring too much effort or cost/parts?  The poster child for this sort of thing is the CE-1, which uses different LFO waveforms for vibrato and chorus, but I rarely, if ever, see attempts to produce anything different.

Although, I have to admit that the various transistor constellations one sees surrounding MN3101/3102 chips are beyond the perimeter of my understanding.  bear in mind that ultimately, it is the behaviour of the clocking that is of interest.  Whether that clocking is a result of the direct control of the LFO waveform, or some lin/log/antilog conversion of a plain vanilla LFO triangle, is immaterial.

What I want to know is if there is some other manner of clocking BBDs in chorus effects that just sounds a little sweeter than the norm.

If you check the LFO section in the article scruffy linked to, he described the difference between sine and triangle wave:
1. Sine: Pitch is always moving.
2. Triangle: sounds more like a flipflop with two pitches.

Sine would sound more natural for vibrato, but with chorus ... I actually kind of would think that the triangle might be better for straight-up chorus, because it'll give more of the impression of just having multiple voices.

Guh. Thanks for bringing this up. Now I have to reconsider the idea of sticking exclusively to a sine wave. i'm going to need an extra breadboard just for five different LFOs ...

[Jdansti]

Maybe you've already seen this, Jon, but here's some more food for thought:

http://www.synthdiy.com/files/2003/MN3004+MN3011+MN3101.pdf

http://www.electrosmash.com/mn3007-bucket-brigade-devices

[Mark Hammer]

Sine would sound more natural for vibrato, but with chorus ... I actually kind of would think that the triangle might be better for straight-up chorus, because it'll give more of the impression of just having multiple voices.

Guh. Thanks for bringing this up. Now I have to reconsider the idea of sticking exclusively to a sine wave. i'm going to need an extra breadboard just for five different LFOs ...

Although vibrato, chorus, and flanging can all have the same fast speeds, vibrato effects are nearly inaudible at the sort of slow sweep rates that makes chorus sound nice, and flanging sound sort of swirly.

So, to stretch out my earlier comment about hypertriangular LFOs, some LFO waveshapes may have benefit within certain sweep speed-ranges, but not in others.  And the narrower the sweep-rate range (e.g., 1-6hz vs .06-6hz), the greater the likelihood that more than some alternate or tweaked LFO waveform might be needed for different ends of that speed range.  In the case of vibrato, the speed range tends to be so narrow that it is hard to imagine the need for more than one type of LFO waveshape.  We have already established that flangers can benefit by having one waveform at the slowest end of the sweep range, and another at the fastest end.  What remains unclear - to me at least - is whether choruses are improved by having something different than they already have, whether that involves alternate waveforms, or simply one waveform better suited to chorus.

The hypertriangular circuit I scanned from POLYPHONY ( http://hammer.ampage.org/files/hypertriangleclock.gif ) suggests that switching between hypertriangular and sine is as simple as connecting/lifting one end of a diode.  Of course, the clipping that produces that sinusoidal end of the sweep also reduces the LFO amplitude, so one needs to be mindful of that.  I have one of these things ( http://hammer.ampage.org/files/SEWAR.PDF ) etched and stuffed, but not yet wired up.  I really have to try it out one day and see if waveform makes a difference.