Updated with mods March 2010
"Univibe" is a registered trademark which was purchased by and is now owned by Dunlop. "MXR" and "Phase 90" are registered trademarks purchased and owned by Dunlop.
If you're an electric guitarist, you have to have been living under a rock to not know what a Univibe is. This is one of the fabled "Jimi" pedals, as Hendrix used it on a number of songs. Robin Trower's "Bridge of Sighs" is practically a Univibe anthem. These things have passed into legend and original ones now command prices of $300 to $800.
The Univibe, for all the hype, is a phaser or phase shifter. It's one of the earliest of the footpedal phasers, and was implemented in discrete transistors, instead of the opamps that most later phasers like the MXR Phase 90 would use. The phase shifting stages used in the 'vibe have a heritage reaching even further back into electronic history; only with the 'vibe did this approach get mixed in the right proportions, however.
It is probably the imperfections of the discrete, non-opamp implementation
and the residual distortion of the transistor stages that give it its unique sound.
The Univibe doesn't sound just like other phase shifters, and it's got a control
marked "chorus", which may account for it's being considered something unique.
Basic Blocks Inside the Univibe
The preamp section accepts two signals from the two input jacks. These are mixed in the two 22K resistors and 47K resistor to ground, and applied to the input of the preamp. While there is a relatively high input impedance at the preamp input, the 22K/47K resistor chain that the input jacks loads each input signal with a 69K nominal resistance. This is low enough to cause significant treble loss to single coil pickups, and probably should be modified for best results. More suggestions about this in the mods section.
The preamp is a fairly tangled set of three direct coupled transistors. The first one has a high 1.2M collector load; it's emitter is tied to the split emitter resistor of the third transistor, and the second transistor base is tied directly to the collector. The second transistor connection and the 1.2M resistor leading back to the base of the first transistor will be familiar to students of the Fuzz Face. This is a voltage feedback stage with an emitter follower buffer. The buffer, Q2, is also set up for gain with a collector resistor, as in the Fuzz Face. In addition, it has Q3 tied to it's collector,
Q3 has a 4.7K collector resistor and a split emitter resistor of 3.9K and 1.2K, for a total of 5.1K. With the emitter resistor unbypassed, Q3 has a gain of about one at both the collector and emitter - it's a phase inverter, kind of. Then there's that connection back to the emitter of the first transistor. Ack! What a mess.
So what's really happening here? The secret is in the split emitter resistor on Q3 and its connection to the emitter of the first transistor. We know that a bipolar transistor's emitter can be thought of as a second input, a non inverting one compared to a signal at the base, where the base itself is an inverting input. The emitter has a low input impedance, but if you can drive the load, the emitter offers us a summing junction in the ordinary bipolar transistor.
There is a lot of gain in the Q1/Q2 circuit, as we know from the high gain of the voltage feedback connection. That gain is carried through Q3 with a gain of 1, so there's still a lot of gain at both collector and emitter. The split emitter resistor couples back a divided down fraction of the output voltage to the emitter of the input transistor. The gain from the first transistor base to collector is inverting, from there to the second transistor collector inverts again for a net positive gain, and from there to the emitter of the third transistor still positive.
However, if we feed the emitter of the third transistor back into the emitter of the first transistor, that voltage opposes the signal at the base of the first transistor. The emitter of the first transistor looks like an inverting input, the base looks like a non-inverting input, and the emitter of Q3 looks like a low impedance output. Hey! this thing is a discrete opamp!
In fact, this is the case. The gain from input to output as set up is about four, which we can calculate from the classic opamp gain equation as 1 + Rf/Ri = 1 + 3.9K/1.2K = 1+3.xx = 4. That 330pF capacitor is the phase compensation capacitor, keeping the thing stable with feedback. [!!Note - the 330pF cap may not be as benign as I thought back in 1998. See mods, below.!!] Things start to make sense now.
The 100K/1uF/1.2M on the collector of Q1 is to make sure there is no
ripple or signal feedback from the B+ line into the input. The 1uF/6.8K on
Q2's emitter is for biasing and current level setting. The 4.7K in the
collector of the third transistor is for producing the out of phase signal to
be used for driving the variable phase stage after the preamp.
First Three Phase Stages
The Univibe is a phase shifter, and a lot of the circuitry inside it is taken up in phase shifting stages. Three of the four are identical to the illustrated one here.
The gain stage is an emitter follower driving a directly coupled phase inverter. If you leave alone the collector resistor on the second transistor, you have a darlington emitter follower, which has a very high current gain, a voltage gain of about 1, and a very high input impedance, about the product of the current gains of the transistors times the emitter resistor of the second transistor. In addition, there is a capacitor from the emitter that connects to the biasing string of 100K/47K.
This capacitor from the emitter to the biasing voltage point is a "bootstrap" capacitor that, in conjunction with the 100K resistor to the base of the actual transistor, raises the input impedance to very high levels. It does this by applying the signal from the emitter, which is almost equal to the input signal, to the other end of the 100K biasing resistor. The bias resistor still lets the same DC current flow to bias the transistor, but with both sides of the resistor at very nearly the same AC signal, the resistor "looks" much bigger to the input signal. With bootstrapping, the input impedance of this stage to the signal is probably well in excess of 1M, in fact probably in excess of 5M. So - we have a very high input impedance phase splitter.
For a nominal 15V supply, the voltage at the junction of the 100K/47K is about 4.8V, and this means that the output emitter is 1.4V lower at about 3.4V. A similar voltage drop appears across the collector resistor since it is equal valued, so the collector sits at about 11.6V. This means that the stage can handle a signal of about 3.4V peak before running into power supply limits and clipping on both emitter and collector.
The actual phase shifting is done in the phase capacitor connected from the collector of the preceeding stage and the resistance between the emitter of the preceeding stage (DC blocked by the 1uF capacitor) and the base of the phase stage buffer. Since the phase stage buffer is a very high impedance, we'll assume that it is infinitely high compared to the phasing components, and will ignore the loading.
The phase stage is best understood by considering what it does at frequency extremes. If the signal frequency is so high that the capacitive impedance is much lower than the resistance of the resistive leg, the signal from the collector of the previous stage dominates, and the signal is effectively the collector signal from the previous stage. At frequencies low enough that the capacitor impedance is high compared to the resistive leg, the signal from the emitter of the pervious stage dominates, and this stage is 180 degrees out of phase with the signal from the previous stage collector. At a frequency where the capacitor impedance equals the resistive leg, the signal is 90 degrees phase shifted compared to BOTH inputs, and has an equal amplitude because of the way not-perfectly-in-phase sine waves add.
The frequency where there is 90 degrees of shift is F0=1/2*pi*R*C.
What this means is that for a given capacitor and resistor, as you sweep a signal in frequency, the signal at the input to the phase stage buffer is constant in amplitude and changes from the phase of the emitter of the previous stage to the phase of the collector of the previous stage as the frequency goes up.
Now add to this the ability to change the resistance of the resistive leg by changing the light on the Light Dependent Resistor (LDR) and you have a way to move the set of frequencies where the phase shift is taking place up and down the audio spectrum as the resistor varies.
Although it is clear how the phase shift stages work, it is not at all lear why the phase capacitor values that exist in the univibe were chosen. The correct values are 0.015uF, 0.22uF, 470pF, and 0.0047uF. These values are NOT any progression of values that I can come up with that would be meaningful in an acoustic sense. I'm fairly sure that they were not chosen randomly, but I can't figure out why they are what they are. I've corresponded with a number of people with ideas about why they think the cap values are what they are, but haven't yet found an explanation that satisfies me.
This will probably turn out to be one of those things that will be painfully obvious when I do find out.
The output side of the phase stage is fairly uneventful, as the signal at the input is buffered with great fidelity at the emitter of the phase stage buffer. Because the buffer output transistor has a collector resistor equal to the emitter resistor, and because almost exactly the same current goes through the collector of the output transistor as through the emitter, a signal 180 degrees out of phase with whatever the signal is at the input of the buffer also appears on the collector of the buffer stage. The emitter and collector signals are once again ready to drive another set of phasing capacitor and resistors for the next phase stage.
There are three identical phase stages, all in a series chain. The
first takes its input from the phase splitter on the output of the preamp,
and the third provides inputs to the final phase stage.
Final Phase Stage
The final phase stage is different from the others. It has only one transitor, and has only one output, at its emitter. The phasing network at its input, the biasing network and the bootstrapping arrangement from its emitter work exactly as described for the other three phase stages. The bias point is at a slightly higher voltage as it has only one output and does not need to have a large output voltage swing at both collector and emitter as the earlier phase stages do.
The phase shifted signal at its input is merely buffered at the final
phase stage's output emitter, and from there sent to the output mixer.
From the earlier discussion of the phase stages, the thing that makes the phase shifts move is the change in value of the LDR's. This is done by the Low Frequency Oscillator (LFO) and lamp driver section.
Overall, the LFO generates a sub-audio sine wave that is coupled through the Amplitude control to let a variable amount of LFO drive the lamp driver. The lamp driver varies the current through the lamp which then changes brightness, changing the resistance of the LDR's - what we wanted to do in the first place.
The LFO is an odd implementation of a phase-shift oscillator. It uses two transistors connected as a darlington for very high current gain. The base of the darlington is biased by the 3.3K/4.7K resistors to about 8.6V on the base and about 7.2V on the emitter, about the middle of the supply. The emitter drives the output of the LFO into the amplitude control through a 10uF capacitor, and also drives a mess of resistors, capacitors, diodes, and the speed control pot. This is where the LFO gets the feedback necessary to oscillate.
A normal phase shift oscillator uses three RC networks from the collector of a transistor back to it's base, the gain of the transistor making up for the losses of the phase shift network. In this implementation, the stack of three capacitors and the two resistive legs of the 220K/4.7K/100K pots provide a net voltage GAIN (this is mildly astonishing to an EE, as you in general can't get gain from passive components) back to the base of the LFO darlington. This satisfies the criteria for oscillation, gain of greater than one and in phase with he input, although in a non standard way.
The values of the three capacitors and the equivalent resistance of the 220K/4.7K/100K pot are the time constants that control the speed of oscillation, so varying the control pot changes the speed of oscillation. The 4.7K limits the upper speed, the parallel combination of the 220K and 100K pot controls the lowest speed.
In the original Univibe, the speed control pot is in the footpedal; the original Univibe will not oscillate, or provide any audible effect if the footpedal is not plugged in. I understand that some of the clones on the market today do provide a panel mounted speed control that works with or without a footpedal.
The taper of the control pot is another issue. For a panel mounted pot, the speed control must be a reverse-log taper dual pot so that speed increases as the pot is turned clockwise. The original Univibe used a dual 100K log taper pot in the footpedal that was mechanically turned the correct direction.
The two diodes across the center capacitor limit the size of the LFO output waveform. An oddity of this particular way of building a Univibe is that the LFO amplitude goes up with increasing speed, even with the diodes there.
The amplitude control selects a fraction from 0 to 1 of the LFO output voltage, and provides this to the input to the lamp driver.
The lamp driver does three functions: (1) it sets the static level of current to the lamp; that is, the current which the bulb sits at when the amplitude is turned to zero; (2) it buffers and drives the LFO waveform into the bulb so the brightness varies; and (3) it provides a Cancel function, which is used in lieu of any bypass. The Univibe is another of those pedals that has no bypass at all - the Cancel turns off the light to all the LDR's, which means that the signal goes almost entirely through the phasing capacitors and no phasing is audible. The original box would be much better off with a real bypass. This will cause problems with the static level; we'll talk about that later.
The static current of the bulb is important. It is set by the fixed and adjustable resistors in the emitter of the lamp driver transistor. The trimmer is adjusted until the lamp glow a medium yellow-orange, about "halfway" to normal brightness. Some original models of Univibe have only the trimpot, some have only a fixed 150ohm resistor,
With the proper bias current through the bulb and max amplitude, the bulb goes from almost dark to very bright in almost-flashes.
The transistors of the LFO and lamp driver are under considerable stress, and are the most common cause of failure.
The bulb and LDR's are under a shiny metallic shield in the middle of the board. The shield not only blocks the ambient light so it does not affect the LDR's, it also acts as a light mixing cavity to even out the light intensity on the LDR's.
The bulb and it's time response are important to getting the right sound. The original bulb in all except one original unit I've repaired is a nominal 28V, 40ma bulb. The cold resistance is just over 100 ohms. I have sucessfully used 12V/40ma and 12V/80ma bulbs here. The bulb has a thermal time constant that means that as you try to turn it on faster and faster by twisting the speed control, it responds more and more to the average of the current through it, not the instantaneous value. This should not be a surprise, as at 60 Hz, the bulb does not respond to the instantaneous value at all, and its light output is proportional only to the average of the current through it.
There is some reason to believe that the odd configuration of the LFO was chosen to add a rising amplitude of LFO with frequency as a first order compensation for the falling response of the bulb.
The LDR's have been the subject of a lot of consternation and supposition. The original part numbers if read from an original Univibe do not show up in any current manufacturer's books - meaning only that they are/were custom made for Unicord. Like anything that is unavailable in the music world, this has caused a lot of mojo-rumors to spring up; for instance, "only the real original parts sound that way, no modern parts can do that" or "the original parts were hyper-matched by hand in Hungary and it's impossible to do that today". In fact, it's not clear that the LDR's need matched, or selected. This is an open question as far as I know.
I do know a procedure for matching LDR's, but I don't know if it's needed or not.
The output mixer in the Univibe is a fairly simple, but critical part of the Univibe. The buffered "dry signal" at the emitter of Q3 in the preamp section, in addition to driving the first phase shift section, also attaches to a 100K resistor in the mixer. The "wet" signal that has been through the phase shifting sections attaches to another 100K resistor, and the two 100K resistors are tied together, and that junction is connected to one terminal of the "Chorus/Vibrato" switch. The other side of the switch is attached to a resistive divider (47K/220K) which is driven only by the "wet" signal.
The throw of the switch is connected to the 100K volume control and then to the output jack. This setup selects either the purely "wet" signal from the phase line output or the mixed signal with about equal amounts of wet and dry. In the frequency domain, phase shifting and then mixing the signal with the original causes the signal to reinforce at frequencies where the signal as a phase shift of 1, 2, 3, ... times 360 degrees, making it twice (3db) as big. At frequencies where the phase shift is 1, 3, 5, ... times 180 degrees, the wet and dry signals cancel, producing a notch of a depth determined only by the degree to which the amplitudes match.
It is these notches we hear in phasers. In general, you can count on each phase stage to contribute 90 degrees of phase shift at some frequency, so you get one notch per pair of phase shift stages. The Univibe is a two notch phase shifter.
As I noted back on the mixer, the mixing of the wet and dry signals is important to getting a good phase sound. You need to diddle the 100K's to get as big a cancellation at the mixer as possible, as the signal drops a bit in each phase stage. This is usually in the range of selecting 100K resistors to pick the one that sounds best.
The power supply for the 'vibe is oddly sparse. The vibe is AC line powered, and uses an internal transformer of about 10VA (just guessing from the size). The secondary voltage seems to be 14 to 18 VAC on the units I've seen, with 16V a good middle of the road value.
The secondary voltage is half wave rectified, which makes for a lot more work getting the ripple out of the supply voltage, and which I find an oddity, as diodes were not expensive even when the 'vibe was designed. I take this as an indication that lowest possible cost was a key design objective, and that there are other places that the maker may have cut corners.
The single diode feeds a first capacitor of 1000uF, to make about 22VDC which goes to supply the LFO section and the incandescent lamp. The LFO and lamp do not carry any signal, so any ripple in their supply is not a sound quality issue. After the first filter, there are two RC filters, with 100 ohm resistors feeding another 1000uF and finally a 100uF capacitor. This setup produces a well filtered supply of about 16VDC to the signal path. The actual voltage varies a bit, depending on the transformer, lamp brightness, and component tolerances.
In the clone of the 'vibe I built, I skipped the extra RC stages and simply took a 15VDC three terminal regulator from the first filter cap and supplied regulated 15V to the signal path. This saved quite a bit of board space and was actually cheaper than the cost of the caps and resistors. I believe that three terminal regulators were more expensive when the 'vibe was designed. The regulator is a trouble-free improvement.
I've seen a number of 'vibes now, and there are a few things that seem to break more often than others. The bugs I've found are (in order of frequency)
Mods !? On a UNIVIBE???? WHY!??
Isn't this like gilding a lilly - or even worse platitudes?
Not really. It turns out that like every piece of commercial gear, the Univibe had compromises in the design that affect the sound. As a tinkerer, you have the ability to spend a bit more for parts and some of your time. These compromises can give you an even better-sounding 'vibe.
Like all modifications on vintage gear, you have to make peace with yourself BEFORE making any changes, as this may affect the vintage value of a piece. The person who buys it from you (if you ever sell it) will give you much less money and curse your name for modifying it, and "experts" on the internet will roundly criticize you for your shortsightedness. After you apply the soldering iron, it's too late.
But if you are interested in the sound, not the resale, you might consider these:
The depth of the notches in a phaser is what makes things sound interesting,
and that depends on the dry and delayed signals cancelling in the output mixer.
In the 'vibe, this is done by the two 100K resistors feeding the
"chorus" side of the chorus switch. If the gains of the dry and delay
paths are not almost identical, the mix will not give you good depth. The trick
is to make those resistors balance the two to give a good null. You do this by
changing one or the other resistor to more or less than 100K to get them to
balance. The simplest way to do this is to change both resistors to 68K, 75K or
82K, and then use a 50K pot with its outer lugs connected to the resistors and
its wiper connected to the chorus switch. This lets the resistor increase one of
the "100K" mixers and decrease the other to balance the mix. +/- 25K
on each resistor is usually plenty to get good, deep phasing when adjusted by
listening while tweaking the pot. You can leave the pot in, or you can set the
pot just right, then measure the resistance of the pot plus fixed resistors and
solder in fixed resistors to match the measured values.
rebias the last phase stage to be the same as the other stages and
add a 4.7K collector resistor, changing the emitter resistor to 4.7K as well.
Take the collector signal off through a 1uF capacitor to another entire output
mixer and mix the
collector signal in where the emitter signal goes in the original output mixer.
The switch for vibrato/chorus needs to be changed to DPDT for this one, and
a new output jack added.
In "chorus" the second output has notches where the original output has
peaks and vice versa.
Increase to Unity Gain
The stock 'vibe has a slight signal drop from input to output when the volume control is at full volume. This is the bane of simple "true bypass" modifications to the 'vibe. The "cancel"function does not have this problem since the signal still goes through the same gain/losses that the effected signal does, and gets unity gain by default.
The key to making the 'vibe be unity gain is there in the preamp. The gain overall is controlled by the split emitter resistor on the third transistor, and is the reciprocal of the resistor divider ratio. To get a bit more signal through the signal chain to match the dry signal level in bypass, change the value of the lower resistor down slightly and the upper resistor up slightly. The only kicker is that once you have the gain level matched, you must make the collector resistor in the third transistor in the preamp be the same as the sum of the two emitter resistors to keep the phase inverted signal on the collector at the same magnitude as the signal on the emitter so the first phase stage operates correctly.
Update March 2010: The preamp of the vibe already has about 12db gain.
What's killing gain is the input mixer. By simply changing the input 47K
resistor to ground up to 2.2M or bigger, you add in a lot more signal level by
NOT dividing down the input. Give it a try. Simpler than dinking with the
Brightening things up - improving the input
Most 'vibe afficianados don't think of it this way, but the stock 'vibe suffers from some of the same problems that the unmodified crybaby does. The input resistance to either input is only 69K (22K plus 47K). This is low enough to significantly load the output of a guitar and dull the sound. In the stock vibe, you pay this price to a degree even when "bypassed", as the "cancel" switch is not a bypass, it merely turns off the lamp, and the signal still goes through the phase stages with the LDR's at very high resistance.
Update March 2010: I've found that the preamp of the univibe itself has a fairly high input impedance all on its own. What's poisoning things is the 22K/47K input divider. Those are there to mix two inputs (I think), but hardly anyone ever uses two inputs there today. To clean things up a lot, change the 47K to well over 1M, preferably something like 4.7M. This kicks the input impedance up to about 430k ohms, and there's then almost no treble loss. This pretty well makes input buffers unnecessary. So try the one-resistor change first.
One way to brighten things up some is to buffer the input. I've shown two ways to do that; one with a high gain transistor buffer, similar to the input buffer on the Ibanez series of pedals, and another with a JFET source follower. The JFET is a higher input impedance and a simpler circuit, but a lower allowable input voltage. The bipolar buffer is a lower input impedance (but probably good enough) and a slightly more complicated circuit.
I've recently messed with some simulation of the stock 'vibe circuit and come up with some interesting results.
1.The 330pF compensation cap, C4 on the Neovibe layout is from Q2
collector to ground. This is odd compared to more modern feedback amp practice.
I think it may have been selected that way to ensure that it swamps out the
parasitic capacitances on Q2 collector and Q3 base. However, it does give a
resonant peak out at about 900kHz to 5Mhz depending on what transistors, gain,
etc are in there. On a whim I removed it and replaced it with a 30pF from Q2
collector to Q2 base. The result was more phase margin, and no peak. Now the
gain rolls off nicely at about 2MHz, with some variation with the gains of the
transistors, primarily Q2. It's a major step to better stability on that input
feedback amp setup. Note that it is possible that some combination of
transistors will not work well with this, so this must be classed as likely to
work, but experimental. In fact, it looks like from my simulation that any
capacitance at C4's position makes stability *worse*, which I find distinctly
odd. And any capacitive loading on the emitter of Q3 can definitely make this
thing oscillate. Very strange.
2. Buffer, who needs a buffer? I figured out the input impedance at the input to Q1 independent of the 22K/47K input resistors. Turns out it's respectably high. Want to end tone sucking? Replace the 47K with 1M. This seems to work fine with my two "victim" boards as well as simulation. No loading, and it also makes the "unity gain mod" of questionable value since it gets to unity or above all by itself now that the input network is not loading down the signal source and dividing the input down. Again, experimental, but try it out.
3. I'm pretty sure I had another notch here somewhere. The 'vibe has the makings of a two-notch phaser, but the second notch never raises its head (lowers its feet?? ) in simulation. I had always put this down to the funny spread-the-phase distribution of the phase capacitors, which may have been to get a better vibrato sound. When I posed myself the question of what could be making this not show up, I looked at the low end response of the circuit, and noticed that the low end has several rolloff time constants overlaid. Hmm. What's cutting bass? Well, for one thing, the 1uF caps in series with the LDRs. These are nice, low impedance connections when the LDRs are up at a megohm, but what about when the LDR goes down to 500 ohms? I popped in 10uF for caps C6, C8, C11, and C14. The second, lower notch peeked out!
But it still wasn't very prominent, only about 12db of notch depth, which correlates to how audible phaser notches are. You need to get -20db to get real jet-plane-y.
So I thought "those bootstrap caps seem to be big enough, but why not just try it?". I subbed in 10uF for C7, C10, C13 and C16. Now the lower notch dropped much lower. Here's an oddity, though. C16 being left at 1uF with the others being 10uF was even better. I gotta do more thinking about that one. Messing with the other 1uF caps didn't seem to do much.
4. Dinking with the phase caps is a popular pastime amongst 'vibe afficionados I know; I got interesting results from making the phase caps be 22nF, 47nF, 1.5nF and 470pF(unchanged values, but rearranged in order, that is). Worth playing with in concert with the other cap changes.
One way to soup up the intensity of the 'vibe is to use feedback, like most
phasers after the EH Small Stone do. This takes the form of a resistive divider
fed from the emitter of the last phase stage through a 1uF capacitor that feeds an adjustable
fraction of the phase stage output back into the input of the preamp. This has the effect of super
sharpening the peaks and super-nulling the notches. It can be so intense that it even
causes oscillation if you use too much signal fed back.
The Ultra-vibe, a double length phaser in the Univibe style
One way of making the vibe sound more intense is to make it longer - that is, more phase stages. As an experiment, I laid out what I (modestly...) call an ultra-vibe, a circuit like the vibe, but with eight stages rather than four.