mxr blue box: theory or operation?

[duck_arse]

I know I'm supposed to be doing something else, don't know what it is, because this weeks bright and shiny distraction is the mxr blue box. I have it on breadboard, working, modified. it's a weird thing, and I can nearly even explain how it works.

with reference to the following:
http://www.tonepad.com/getFile.asp?id=73

the first opamp stage converts everything to a square wave because of its gain, and shoves this into the second stage (wired as a comparator?), producing a 6V5 pp squarewave to feed Q1 for level shifting to 9V for the digital and Q2. with the input grounded, I measure about 1V2 pp of 50Hz at pin 1, which gets compared to ground in the second stage, producing a 6V5 pp 50Hz square wave, which it then happily feeds to Q1.

so whether signal or no signal, the Q2//IC2//Q3 section always has a 9V squarewave clocking away, and no signal leaves behind an ugly 50Hz when the rectifier stops feeding the Q2 and 3 collectors, audible as a clicking buzz mix of f and f/4 at the output. and with mods, f/2.

is this part of the blue box "charm", the normal mode of operation, the trade-off made for the circuit simplicity? I've been shoving caps between R22 and Vb, which works to kill the 50Hz, but does nothing for the output decay. (I would say it goes obnoxious splatty gate-y, but that's all the time normal.) is there a better way to kill the unwanted 50Hz, short of the differential dual rectifiers of other designs, or just live with it, "crackle OK"?

[R.G.]

the first opamp stage converts everything to a square wave because of its gain, and shoves this into the second stage (wired as a comparator?), producing a 6V5 pp squarewave to feed Q1 for level shifting to 9V for the digital and Q2.

Pretty close. The first opamp both amplifies and filters the signal to produce a much larger signal with some of the high frequency grak removes. The second opamp is a Schmitt trigger comparator - it has some hysteresis to produce a snap action comparison. This is where the square waves are produced.

The amplified, filtered input is passed through C7,D3,D2,R15,C8 which makes a DC (ish) level out of it. The DC level is the "envelope" of the input signal, and this is chopped by the action of Q2 and Q3 into the output signal. The square waves at the output of the Schmitt drives Q1, which provides both the input to the non-divided signal at Q2 and the input to the divider network in IC2. IC2 divides it by 4, then feeds Q3 for the divided-down bass signal. R20 feeds a mix of 1x frequency from Q2 and 1/4 F frequency from Q3 to the output.

with the input grounded, I measure about 1V2 pp of 50Hz at pin 1, which gets compared to ground in the second stage, producing a 6V5 pp 50Hz square wave, which it then happily feeds to Q1.

The circuit is operating incorrectly. It should not oscillate - or pick up 50Hz hum, which is a possible alternative explanation if you live where AC mains are 50Hz.

is this part of the blue box "charm", the normal mode of operation, the trade-off made for the circuit simplicity? I've been shoving caps between R22 and Vb, which works to kill the 50Hz, but does nothing for the output decay. (I would say it goes obnoxious splatty gate-y, but that's all the time normal.) is there a better way to kill the unwanted 50Hz, short of the differential dual rectifiers of other designs, or just live with it, "crackle OK"?

It's busted.

There's a bad connection, wrong component, miswiring, something. The Blue Box is sputtery and ugly, with poor decay, etc., but it doesn't gargle at 50Hz.

By the way, IMHO, the Blue Box is junk for what it does. The PAIA Rocktave divider is a vastly improved and more flexible way to do the same thing.

[duck_arse]

thanks rg.

and bugger. so another rebuild, w/ diff components. and yes, 50Hz is the local choice of mains.

[Mark Hammer]

Crap!  Now you've got me thinking....never a good sign.

IC1b serves two purposes.  The principal purpose is to output a square wave that reliably exceeds the minimum needed to trigger the flip-flop into division, for much of the note's lifespan.  But...while we're in the neigbourhood, that same high-amplitude square wave is used to provide a source of a "fuzz" signal to be blended with the sub-octave.

What would it take to convert/repurpose that square wave into a frequency-doubled signal that could be blended in with sub-octave and fuzz?  Or would that best be done by tapping the output of IC1a instead?

[R.G.]

What would it take to convert/repurpose that square wave into a frequency-doubled signal that could be blended in with sub-octave and fuzz?  Or would that best be done by tapping the output of IC1a instead?

We're still stuck in the same old problem. It's easy enough to make a square wave from a signal, but doubling it is much harder than dividing it. The options are still to convert it to some other waveform that can make one full cycle from each edge, both positive and negative-going, or use a PLL and up-vert it by multiplication.

That's another reason I mentioned the Rocktave. I've listened to both a Blue Box and a Rocktave, and the Rocktave front end does a much better job of producing a waveform that can be tracked and used by digital stuff later. It's output also has much fewer sputters, crackles and mis-starts. I was mildly surprised at how the first prototyper got it working with the PLL up-multiplication. Thinking about that, I should not have been, because the Rocktave has a good conversion process for making relevant edges for the digital stuff to follow.

If I had to do an up-verter without a PLL, I'd do something like an Octavia to get full-wave-rectified octave up, then filter harder to get it back to one transition per doubled-frequency, and then run that into digital-ish stuff.

[Mark Hammer]

Yeah, I guess that first op-amp stage applies way too much gain to have something one could rectify into a doubled wave.  In which case, one would likely need a few more op-amp stages, one of them being a unity-gain buffer to feed the stock first gain stage, plus an auxiliary stage to prepare the signal for FWR.

But let's say we had that buffer and auxiliary op-amp stage available to us via another dual op-amp.  Couldn't one take that additional gain stage's output, hang a single transistor phase splitter with appropriate diodes on it, à la Foxx Tone Machine, combine those rectified outputs and feed that to yet another transistor, whose collector is controlled by the same side chain that controls the transistors which gate the fuzz and sub-octave?

[wilrecar77]

If you converted the square wave of the original signal into a triangle wave using the escobedo "square wave shaper", couldn't you run that through a window comparator to get square pulses at twice the original frequency? Specifically, I'm thinking of something like this:


And here's an output you could get out of something like that. It would need gating, but I think it could work. I'm about to breadboard it.

[R.G.]

I can hear "As Time Goes By" playing.    I spent a fair amount of the middle 70s pondering this set of issues while being bored with my day job.

Yes, all of the above. It's so tempting to mess with this because the rest of the pedal says "octave... octave... you can have an octave...".
But the problem is the same as getting an octave up from any signal - the conversion to square waves and pulses for the dividers are completely uninvolved with the question of generating an octave up.

The problems with doing octaves up lie mostly in the time domain with waveform-shaping schemes like this. You can easily enough make a large signal from a small one, and amplify/filter it up so you can FWR and get an Octavia-like or Super Fuzz-like, or Fox Tone Machine-like, or Fender Blender-like octave up. Once you've done that, you have to cope with the issues from those similar rectification approaches. Those are that guitar signals are not sine waves, and often have more than one zero-crossing per cycle. More than one zero crossing (which is the same as one bias voltage crossing) per cycle confuses comparators a lot, into putting several up-and-down edges where it more musically would have only one, and these edges are distributed by the time-interaction of the waves, not the frequency interactions of the waves. That's what a lot of the prefiltering of signals before FWR octave-up is about - cut that wobbling down so the FWR doesn't have funny wobbles in it too. In an odd aside, the diode drop that's the bane of FWR based doublers helps this, as the diode drop is subtracted off the bottom of the waveform, which occupies where the zero-crossing grak used to be. It actually helps clean up the result.

However, the FWR process itself generates a slew of harmonics and intermodulation. That's the fuzziness that gets more or less filtered out of FWR doublers, and that would be there even if you fed it a sine wave.

Generating an FWR octave up would work OK-ish if you could use an adaptive-tracking threshold for the comparator that would track the points on the FWR result so the comparator gave a better middle-of-the wave slicing of high and low for its output. But adaptive analog tracking is tough. I tried it once, and the complexity of the tracker went beyond the FWR circuit almost instantly. This stuff has to work with varying signal levels too.

Another approach is to pre-compress the input to a constant level to make it hold still in terms of amplitude so you can work on it with waveform shaping stuff. You need a good compressor, and a good noise gate to do anything approaching good on that, and the complexity explodes again.

As to the window voltage thing - sure, you can generate a pulse on each zero crossing, which is what the circuit shown and others do if you tweak them in. The problem becomes - what's the duty cycle of that pulse, and what does it do as the amplitude of the incoming waveform diverges from the neat, tidy, one-zero-crossing-per-cycle that these circuits are always shown with?

In the waveforms shown, if the triangle gets smaller, it gets smaller either centered on zero, or asymmetrically in one direction. If it gets smaller around zero, the pulses stay nicely evenly separated in time, but the duty cycle goes from zero when the triangle peaks are less than the window size to a beautiful 50% duty when the triangle is so much bigger than the window that the signal spends almost no time in the window as it shoots through.  At the points where triangle peaks just peek above and below the window, you get little signal blips up and down.

It's worse if the triange isn't symmetrical on the window. Then you get two pulses, same duty cycle issues as above, but the timing changes so you get pairs of pulses that are no longer evenly spaced, and the harmonic series of that is heavily non-related to the original waveform or any of its harmonics.

Again, you can do anything not forbidden by the laws of physics if you have enough money and time. You can compress, filter, etc, to construct the right waveform for the window comparator to give you the right waveforms. But complexity has exploded again.

I should be clear here - I do not know all of the possibilities, and there may be simple ways to make this work. But I never found one, and I looked pretty hard for quite a while. I'll be mortified if it turns up NOW.   

[duck_arse]

It's busted.

There's a bad connection, wrong component, miswiring, something. The Blue Box is sputtery and ugly, with poor decay, etc., but it doesn't gargle at 50Hz.

mine does.

well, I rebuilt w/ diff parts, on a different part of that board, then I got other parts, and rebuilt just the first stage with an lf351 or a 741, with the same result, amplified 50Hz with the input grounded, enough to trigger the comparator. with low-level signal, I got a 50Hz wave with low-level signal riding it, until signal got big enough to clip. I changed psu to batteries, same result. THE ONLY common in the end was the cro, as I swapped probes as well.

so it looks like I have too much gain in the front end (with the specified components) to stop me chucking this in the bin, and clearing some breadboards for the rocktave. surely, other builders must be getting amplified hum fed through their blue boxes???

"that which is forbidden is not allowed"

[RLawlor]

It's busted.

There's a bad connection, wrong component, miswiring, something. The Blue Box is sputtery and ugly, with poor decay, etc., but it doesn't gargle at 50Hz.

mine does.

well, I rebuilt w/ diff parts, on a different part of that board, then I got other parts, and rebuilt just the first stage with an lf351 or a 741, with the same result, amplified 50Hz with the input grounded, enough to trigger the comparator. with low-level signal, I got a 50Hz wave with low-level signal riding it, until signal got big enough to clip. I changed psu to batteries, same result. THE ONLY common in the end was the cro, as I swapped probes as well.

so it looks like I have too much gain in the front end (with the specified components) to stop me chucking this in the bin, and clearing some breadboards for the rocktave. surely, other builders must be getting amplified hum fed through their blue boxes???

"that which is forbidden is not allowed"

duck_arse, (lovely username btw) I know you've seen my recent thread(s) about the fuzz that was oscillating. It was heavily based off of the MXR blue box and I've realised the noise I've picked up is exactly the same 50Hz sound as you experienced. Did you have any luck in fixing it? If so how?

[duck_arse]

hello RL, and sorry, no, I consigned the parts back to the parts drawers, went no further at that time. possibly, there was no circuit problem as there was a cro probe had a bad ground wire about that time [and dodgy breadboard, too]. but I would have to try and re-bread it to confirm. or deny.

[ElectricDruid]

In an odd aside, the diode drop that's the bane of FWR based doublers helps this, as the diode drop is subtracted off the bottom of the waveform, which occupies where the zero-crossing grak used to be. It actually helps clean up the result.

However, the FWR process itself generates a slew of harmonics and intermodulation. That's the fuzziness that gets more or less filtered out of FWR doublers, and that would be there even if you fed it a sine wave.

I was thinking about this problem in terms of synthesiser waveshaping functions the other day. I completely understand why it might keep you busy for the best part of the middle 70s. If you were doing this in a synth, for a start you'd have a nice clean sine input, which makes everything easier, but you wouldn't use a simple full-wave rectifier to double it, since the output from that is a waveform with a sharp spike on the bottom (or top, if you do it that way up) and that means lots of harmonics and intermodulation. Instead, you'd use a U-shaped function rather than a V-shaped function (which is what the FWR does). That flatter bottom section smooths the "spike" out and makes it look more sinewavey. Which is were the diode-drop comes in - it gives that kind of effect. What it boils down to is that a bit of crossover distortion ahead of a FWR would actually *improve* the result and give a *smoother* effect, which is not what you'd generally expect.

Other problems remain though. If you use a U-shaped function, you've got an area around ground where nothing happens much, so your decay will just die when you get down there. Don't expect nice fade outs. The Roctave cunningly gets round this by artificially fading out the octave-down signal just before it gets to the splattery bit. Others have already mentioned the benefits of using a compressor as a front end. It's clever stuff.