More blathering about octave dividers and the Blue Box

[Mark Hammer]

Finally, got my Blue Box working last night after probably a year sitting around nonfunctional.  I changed the transistors from some Radio Shack "3904 equivalents" with an hfe of <80 for some 2N4401's, following up on some earlier threads, but ultimately the reasons for its nonworking were embarassingly trivial: a fried 4558 and a misbehaving blend pot.  What clued me in was finding that there was about 50mv of AC signal on pin 3 of the op-amp when I strummed, but nothing showing up on pin 1 (the output).  Once the chip was replaced, I'd get decent octave volume for an instant and things would cut out.  In an effort to be "neat" I had used a small cable tie and gathered a bunch of wires neare the blend pot.  That ended up placing some strain on the solder lugs, creating very intermittent contact.  I clipped the cable tie, desoldered the pot, popped the back off the pot, pinched the rivets holding the solder lugs with some pliers, applied some Strabilant contact enhancer to the conductive element, reassembled and reinstalled the pot, and sweet octavity was mine.

The sole mod at the moment is a tone control (treble-cut) for the octave-down side, a 100k variable resistor and .022uf cap to ground from the R17/R19 junction.  It works fine and has the interesting property of appearing to move the lower note to the background.  Or rather, making it easier to treat the octave down as texture or ambience rather than another note.  I was actually pleasantly surprised at the quality of the fuzz tone itself, without the lower tone.  Actually pretty decent sounding (I used a BA15218 dual op-amp).  The only flaw, of course is that the envelope control intended to gate out sputtering from the octave section also fades out the distortion as well.  This suggests that perhaps a worthy mod for those who also like the fuzz tone on its own might be to insert a ground lift switch for R15.  This 56k resistor drains C8, resulting in a quick decay as the note starts to die out.  If R15  is unable to drain C8, then the decay is much much slower.  Too slow to be an effective noise gate for the octave-down tone, but long enough to allow a finger vibrato applied to a fuzz solo to actually lead somewhere useful.  Alternatively, one might consider replacing R15 with a 33k fixed resistor and 500k log pot to achieve variable decay rates.  At one end, a very short decay can "tame" misbehaving tracking.  At the other end, longer delay can be used for those occasions where the tracking is really good, or where you just want to use the fuzz on its own.

I had suggested to someone who contacted me offline that one possible fix for poor tracking was to adjust the gain of IC1a (we're using the Tonepad Caja Azul as our reference here).  Apparently that helped a little, but it got me to thinking afterwards that the gain of IC1a sets the behaviour of the gate, but the signal reaching the flip-flop IC2b is a function of the joint gain of IC1a (x471) and IC1b (x101).  Altering either of them changes the product.  So, if you want to boost the first stage to get a little more envelope signal for keeping the gate open longer, you probably need to trim back on the gain of the second stage a bit.  Though it is not guaranteed to work flawlessly, one simple way to accomplish this is to swap the location of C5 and R5 (it should work the same way despite the change).  Instead of R5 and R23 going directly to Vref (Vb), they now go to the outside lugs of a trimpot and the wiper of the trimpot goes to Vref.  So, as you reduced the leg of the trimpot tied to R5, you increase the leg of the pot tied to R23.  Presumably the original values of R5 and R23 have to be reduced, so that when they sum with the trimpot, the values are reasonable.  I haven't worked out the fine details, but in theory it ought to work.

This brings me to the meat of the posting, and that is the general theme of the synchrony/asynchrony of the octave and nonoctave tones.  What is the "ideal" time relationship between them?  Should the octave down start EXACTLY when the regular note does, or should it be staggered in time?  If it needs to be staggered, by how much?  I have a bunch of 3207/3102 combos sitting here doing nothing, and I'm wondering if one of them ought to be applied to "lining up" the octave-down and normal/fuzz signal in some manner.  Obviously one fixed delay willbe a rough compromise since the stagger would not be constant across all note., but it might improve the "punch" in the resulting tone.  I'm taking my inspiration here from the BBE Sonic Maximizer and its cousins, that use the time alignment of bass and treble to achieve a more pleasing and coherent sound.

Quite frankly, I'm not sure which of the sounds ought to be delayed, the divided or normal.  And with that query, I turn to yu, my trusted advisors, for your opinions.  Let the mélée begin.  I'm off to watch the Grey Cup!

[Processaurus]

Thats funny, I've read a bunch of tips and mods you've posted on the blue box, you just now got one of your own .   Good idea for the tone control for the octave.  Coincidentally this last week I've been working on a Blue Box modded beyond recognition, its stretching a DD size box right now.   I've thought about what you're talking about for it, delaying, or phase shifting the octave.   This idea came up from playing with the ms-20 synth, where you can never tune the two oscillators exactly right, so they kind of flange in a neat way when they're tuned close and mixed equally.  Delaying it with BBDs would probably sound great.  You wouldn't need to worry about any biasing or aliasing filters at all, you could probably even clock the BBD much slower than you would normally for slapback delays. Alternatively I wonder if you could use a simple shift register type thing to delay the octave, rather than a BBD, since its a logic signal at that point.  As for "punch", I don't know if the delay would add that as much as it would "richness" or "depth" (awful terms, I know). 

Some other fun things to try with the Blue Box is using the shmitt inverter section of Tim Escobedo's PWM to change the pulse width of either the distorted signal or the octaves.  Also feeding the distorted signal (from the collector of q1) and the 2nd octave down (pin 1 on the 4013) to a XOR gate (and the output of that straight to R11, rather than the 2nd octave down) gives another neat, synthier tone (an idea from the ring modulator section of the MS-20). 

[R.G.]

A-synchronous is more natural.

The problem with octave boxes (other than the distortion and unnatural tone) is that each wave is not only frequency locked but time locked to the wave that generated it. I'm thinking that if you delayed the clean tone a fixed amount and the octave an almost equal but randomly varying amount of time, that it would sound much more like your idiot brother-in-law who is sitting in because your regular bass player's kid gave him strep throat.

Well, OK, maybe it's just more like a live human that's different from the person who played the original guitar note. The delay and frequency shift on the bass would (maybe? probably?) sound more like a bassist instead of an effect. The notes would be neither perfectly in tune every time nor in perfect time unison.

Just guessing.

[Paul Perry (Frostwave)]

  What is the "ideal" time relationship between them?  Should the octave down start EXACTLY when the regular note does, or should it be staggered in time? 

Well, I think neurophysiology tells us it should not matter a row of beans. If we accept that the hairs in the cochlea resonate when energised by a particular frequency. I doubt there is any way for the ear to detect phase relationships between notes at different frequencies.

[R.G.]

I doubt there is any way for the ear to detect phase relationships between notes at different frequencies.

]
Audiologists have proven that there is not.

[alteredsounds]

Slightly O/T but did you read my post about my odd Scrambler build?  I have an old Blue Box and it sounds VERY similar but less chaotic.

[Mark Hammer]

Can phase relationships be detected?  Well, it would vary with frequency.  The last time I looked, there were different subsystems, or rather coding systems, for low frequencies and all other audible content.  Not to get too pedantic about it (like THAT ever stopped me  ), but neurons have physical limits to how fast they can fire, based on the speed of propogation along the nerve axon, but also based on the speed with which certain chemical processes occur.  Tis a bit like skiers going down a ski jump - you have to wait a certain period of time before the next one can go to make sure they don't crash into each other.  In the case of neurons, you can't expect to ever get more than maybe 100-200 pulses out of them per second on a good day with the wind at your back.  That *potential* firing frequency allows such cells in the ear to respond to every wave peak, provided the frequency is low enough, and the speed of firing of such cells provides a neural code for the frequency of the sound entering the ear.  I.E., if the specific cell is firing at 30 times a second then it is a 30hz tone.  When it comes to 2000hz tones, no individual cell will be able to fire for every peak.  Consequently, amplitude of higher frequencies is coded for by the average response of a group of cells.  I.E., if the activity of those cells that can respond to 2000hz has generally increased, then there must be a 2000hz tone entering the ear right now.  What neurophysiologists propose is a dual coding system where frequency coding (how *fast* certain cells are firing) is used for low frequency content, and place coding (*which cells are firing*) is used for content above that.

Since we are discussing original notes that are more than likely going to depend on place coding exclusively for being sensed, and fundamentals for the sub-octave that might turn to frequency coding for detection, quite frankly I am bamboozled as to how much phase relationships *could* be sensed.  I will merely say that it is neurophysiologically possible.  Whether it is necessarily *probable* in this instance, I am not nearly enough of a neurophysiologist or audiologist to say.  Just don't count it out when it comes to very low tones the way you should for the remainder of the audible spectrum.  Bass plays by different rules AFAIK

As for imposed delays, I'm thinking largely in terms of using 1024-stage devices, since these are most plentiful, especialy in my case.  The sort of delays I'm thinking of would be relatively short, on the order of 1-4msec or so.  Certainly short enough that I'm not even sure whether I'd want or need an LFO, and short enough that lowpass filtering needs would be minimal due to the ultrasonic clock rate.

Of course, you realize this now calls for an experiment in which an octave divider has separate fundamental and octave outputs provided, and each is run through an external fixed short delay (e.g., through a BF-2 set to minimum depth with fixed delay time set by the manual control) before being mixed back together.

And, just so we're absolutely clear on this, I am NOT talking about any sort of frequency-divider arrangement where low bass and everything else are separated by means of an active filter and THEN time-staggered or delayed in some manner.  I'm also not convinced just yet that it WILL improve anything.  Rather, the idea crops up simply because people do rave about how BBE makes everything so much more intelligible and sonically coherent, so I started wondering if it was worth applying in an analogous manner to octave division.  Keep in mind that octave division predates BBD-based effects (manufactured ones, that is; I am confident Ton can find a patent demonstrating that delays predate octaves!  ), so it is all too easy for octave-division to have adopted an entrenched approach to doing things without considering other options.  Stuff like that happens.

[A.S.P.]

I doubt there is any way for the ear to detect phase relationships between notes at different frequencies.

Audiologists have proven that there is not.

but a "Phaser" shows that the ear can detect (e.g.: shifted) phase relationships between
the fundamental of a note and it`s harmonics.
Me thinks that`s how those M*xim*zers work.
The E-H Worm is also capable of such "adjustable static phaseshift".

[Mark Hammer]

Actually, unless you meant something different, that's not a good example.  In phasers, the relationship between 2 copies of the same signal results in electronic cancellation before the listener's ears are even involved.  The phase relationship being discussed here has more in common with how big a time delay difference is needed for the "copy" heard by each ear to result in the perception that a sound is off-axis.

[MartyMart]

Thanks Mark for those great tip's ( along with many others )   
I've "just" got my vero-version working after leaving it for a few months !
I had awful "bad tracking" and a kind of "ramping up" of signal, which
was making it all but useless ....... until ......
I just swopped out the Ne5532 for an RC4559 IC and it all works great now !!!

What was that all about  ?? ( answers on a postcard to .... )

Any way, time to try some "mods" on it now, in particular the " 0ne octave down"
mod

Cheers,
Marty.

[R.G.]

but a "Phaser" shows that the ear can detect (e.g.: shifted) phase relationships between the fundamental of a note and it`s harmonics.

Actually, no it doesn't. Phasers work not by having the ear detect phase differences, but amplitude differences. Check in and read "The Technology of Phasers and Flangers" at GEO for an explanation. We do not hear the phase difference, but the null in amplitude when the phase shift makes waves cancel. The human ear is very sensitive to amplitude variations. Adjusting phasers for best notch depth makes the audible effect strongest.

There was detailed research work, good research techniques and peer reviewed, etc. on the audibility of absolute phase and on relative phase of harmonics, and it turns out that the ear is distinctly insensitive to them.

[A.S.P.]

how about wavefront and directivity,
when I listen to the dry signal through the left,
and the wet signal through the right channel,
in a stereo-setup of a (unmodulated) splitted phaser?
(not wanting to interfere with Marks "hard delay all frequencies by the same amount)"

[Mark Hammer]

There was detailed research work, good research techniques and peer reviewed, etc. on the audibility of absolute phase and on relative phase of harmonics, and it turns out that the ear is distinctly insensitive to them.

Again, I'll simply note that what counts as a "phase difference" at 2khz turns into a time difference of a couple of milliseconds at much lower frequencies.  If I invert a 200hz waveform, the positive peak of the inverted one arrives 2.3msec after the positive peak of the noninvertedr.  More than enough time for their arrival difference to be detectable.  Of course that is in air, with one ear hearing one sound and the other ear hearing the other sound.  Should these two signals be mixed to mono and then fed to the same ear, the capacity to recognize that there is a delay between them would be expected to be diminished.  This is further complicated when dealing with natural sounds where the precise waveform is not kept steady over time, and synthetic sounds like an oscillator. With natural sounds where the signal is constantly changing, a brief delay may be far more detectable.  The basic answer to the question "what can we hear?", is "What is the ear capable of overall, and what does it have to work with at the moment?".  The capoabilities change, depending on frequency band and signal properties.

Here's something completely different I would like to understand more about.  When the Bluebox mistracks, it tends to jump up an octave, to provide one octave down rather than two octaves down.  What I would like to get a firmer grasp of is why that predictable and orderly "disorder"?  Is the device responding to the first harmonic of the original and generating 2 octaves down from that, or what?  I can't seem to find any point in the design that might roll off lows in a manner that would give such harmonics an advantage in triggering a note, so I'm baffled.

[R.G.]

Directional sensing is based on time differences between the sounds hitting the ear. Since phase shift equals time shifts for a cycle or two, I can postulate that there is a time delay of high frequency transients in a phaser and that might make a difference in perceived direction in a setup like you say.

But from everthing I've ever read (I didn't do the research myself) the ear cannot hear phase difference. Apparently one of the tests was to run a waveform consisting of a fundamental and harmonics into an allpass filter - this is the same as a non-varying phase shifter line. This shifts the harmonic phases relative to the fundamental, and the waveform coming out of the shifter looks dramatically different than the waveform going in. But listeners could not detect differences in sound between the original and the phase shifted version of the sound with any better than random guessing accuracy.

[R.G.]

When the Bluebox mistracks, it tends to jump up an octave, to provide one octave down rather than two octaves down.  What I would like to get a firmer grasp of is why that predictable and orderly "disorder"?  Is the device responding to the first harmonic of the original and generating 2 octaves down from that, or what?  I can't seem to find any point in the design that might roll off lows in a manner that would give such harmonics an advantage in triggering a note, so I'm baffled.

The input signal is putting out more second harmonic than fundamental at some point in the decay and the input waveshaping feeds the dividers a first harmonic (2xF0) to divide.

There is nothing in the dividers to do that. The guitar signal is fooling the circuit that cleans it up into a square wave to trigger the dividers.

Over the years I've kept meaning to build a real time filter bank with less than 1/2 octave outputs for situations just like this. Imagine if you had a bank of filters each 1/3 octave wide, and each of those had a signal to indicate "signal here, Boss!" and a final output signal gate that could let signal through or not. A dozen filters would cover the guitar's four octave range of fundamentals.

You could then use priority encoder logic to look at all the "signal here" bits and turn on only the lowest one. To the limits of the accuracy of the filters, this would then rid you of everything above the fundamental. That would make for good bass synthesis, because you could then phase lock to it to generate harmonics up from it and make an accurate triangle or sawtooth out of it.

[Mark Hammer]

Your comments about detection of phase differences makes perfect sense.

Here's the twist that undermines its relevance to this discussion, as near as I can tell. The flip-flop output, being a square wave, has its own harmonics, separate from the harmonic content of the original unsynthesized tone.  The question, among many, that I'm asking is whether there is any preferred alignment of these two sets of harmonics.  Again, if the BBE re-alignment of harmonic content and fundamentals was pure audio mojo, I wouldn't raise the question.  But the re-alignment of spectral content has demonstrable results.  Of course the other alternative is that I have completely misconstrued what it is BBE does.

If it should be the case that there is no basis for suggesting any analogy whatsoever between alignment of original ansd synthesized tone and the alignment of broad-spectrum frequency content, then I'll shrug and stick it in the "kooky ideas" bin.

As for priority detection by means of 1/3 octave filters, that's not exactly a starter project, is it?   Maybe the solution is to have a very shallow-slope lowpass filter, starting way down low, that persists in giving the advantage to the fundamental over the 1st harmonic.

[R.G.]

The question, among many, that I'm asking is whether there is any preferred alignment of these two sets of harmonics.

Ah. OK, got it. That would depend on whether you got reinforcement/cancellation. THAT would be audible!

Hmmm... if the original string fundamental was F0, then the two-octaves down tone is F0/4 and the one-octave down is F0/2. From the first square wave you get 3F0/4, 5F0/4, 7F0/4... and from the one-octave down you get 3F0/2, 5F0/2, 7F0/2... It looks like none of those are harmonics of the original F0 since you never get to an integral number of F0's.

Of course, guitar strings don't put out perfect harmonics either, as a result of stiffness and friction.

I don't know what time alignment would do or not do.

As for priority detection by means of 1/3 octave filters, that's not exactly a starter project, is it?

Well, maybe not, but it's not as onerous as it sounds. A 1/3octave bandpass is a one-opamp deal. An envelope detector for such a small frequency variation could easily be a single opamp deal and have good attack/decay characteristics.

So a beginner would have to make the same thing a dozen times to get the filter bank running. Boring, but not unlike what my teachers did to me back then. 

[puretube]

some inspiration...

[R.G.]

How about that. And it expired two years ago, free for the use.

I wonder why they did the half wave rectifier in front of the filters? I'll have to read the thing, I guess.

[puretube]

to emphasize the fundamental, rather than the 2nd harmonic...

(I wouldn`t have posted the link, if it where a currently protected file, btw...)