HomeA Closer Look at Fender Optical "Vibrato" (Really Tremolo)

Introduction. My little warning about photocell speed in Fender optical "vibrato" (frequently correctly called tremolo) has generated considerable interest. In response, I have made a further, more careful study of this circuit. This recent work has confirmed the earlier conclusion: replacement optical couplers that employ slow-recovery photocells are ill-suited for use in the circuit. However, as I have corrected in red in the earlier article, neon glow lamp activity in the Fender optical tremolo circuit is not "all or none" as that earlier article implied; as I will demonstrate three different ways below (see Section 3), the lamp's "on" cycles are rounded-off pulses. First, we'll look at the classic Fender optical "vibrato" circuit and its variants, then investigate the shape of the light pulses, present static and dynamic test results of three optical couplers, and finally discuss which type of neon lamp the vintage Fender amps may have used, along with their aging and failure modes. This rest of this article is divided into the following sections:
And the investigation continues in my follow-up article: "How Linear is Fender Optical 'Vibrato' (Really Tremolo)?"

1. Is There A "Canonical" Fender Optical Tremolo Circuit?
I'm inclined to say "yes." Except for relatively minor variations, all vintage Fender amp models with optical "vibrato" use practically the same circuit. With the variations noted, this "canonical" circuit is shown in Figure 1. In a comparative analysis, I examined all 33 Fender amp schematics reproduced in Aspen Pittman's 1993 book, "The Tube Amp Book" (4th Edition, published by Groove Tubes in Sylmar, CA) that included optical tremolo. The Fender models I studied are listed at this link. Early Fender models used a direct-coupled tremolo circuit and are not considered here; nor are more modern models and re-issues, except in Section 7 where the 1994 Vibro-King model is compared with the classic circuit, helping to suggest which lamp type the classics probably used.

Figure 1: "Canonical" Fender optical "vibrato" circuit
Figure 1. The "canonical" Fender optical "vibrato" circuit. The variants fall into four categories, as indicated in the table at the lower right of the diagram. The Fender models in each category are listed here. Component numbers are specific to this article and are not meant to correspond to ones that may be given in Fender amp schematics. Power to the 12AX7 heater is not shown; the heater is wired in parallel mode and supplied 6.3 VAC in all Fender amps. See follow-up article for details about the audio channel.

The schematic diagram in Figure 1 is layed out more functionally than in original Fender amp schematics, which were drawn for maximum compactness. Triode V1A and associated components form a low-frequency oscillator (LFO), which sets the speed of the tremolo effect. This type of oscillator is called a "phase-shift oscillator." The R-C network formed by C2-C4 and associated resistors (R3-R5 and R7) feed V1A's plate signal back into its grid. If there were no feedback, any signal developed at the plate would be inverted relative to the signal at the grid, since this triode is hooked up as a common-cathode voltage amplifier. Inversion is functionally equivalent to a 180-degree "phase shift" (but I caution readers that it's generally poor practice, yet all too common among audio engineers, to call signals with opposing polarity "180 degrees out of phase;" see this link for my commentary about confusing polarity and phase). But the three R-C stages in the feedback loop shift (lag) the voltage signal through an additional 180 degrees. Thus, the signal on the grid is reinforced (positive feedback), sustaining oscillation. Potentiometer R3 ("Speed") sets the oscillator's frequency by tuning the impedance ot the first R-C stage. Voltage lag at each individual R-C stage may not be exactly 180/3 = 60 degrees, but the overall phase-shift for the three stages together settle at precisely 180 degrees, sustaining oscillation. An excellent resource for phase-shift oscillator theory, and how to design them for tube amp tremolo circuits, is found within Randall Aiken's web site.

When open, foot switch SW1 disables the LFO by biasing V1A's grid to negative potential via R6 and R7, cutting off current flow through V1A and eliminating any amplification there. Note that V1B is also cut-off in that condition, so the neon glow lamp within the optical isolator remains dark. When SW1 is closed, both R5 and R7 are grounded, voltages at each triode's grid quickly rise toward the operating range, LFO oscillation starts, and neon driver triode V1B is enabled. Indeed, the relatively large voltage transient caused by closing SW1 helps "kick" the LFO into nearly immediate oscillation, with no perceivable lag, for tremolo-on-demand. A similar mode of "kick-starting" a phase-shift oscillator is one of several claims made in Clarence L. ("Leo") Fender's U.S. patents 2817708 and 2973681, which are cited on most of the classic Fender amp schematics. The circuits and numerous claims in these two patents pertain to amps with direct-coupled tremolo; I did not notice where they specifically mention optically isolated tremolo, but the LFO is similar. (The third of Leo Fender's U.S. patents commonly cited in the classic schematics is 192859, a "design" patent covering some cosmetic aspects of the vintage amps.)

When the LFO is active, its output voltage is applied to the grid of the neon lamp driver triode (V1B). Referring to Figure 1, driver feed connection "A" is much more common (29 of the 33 schematics examined; see models listed by category) than connection "B." While an oscillator output could be taken from anywhere in the LFO's feedback loop, the amplitude is greatest at the plate and is successively attenuated at each R-C stage. Thus, connection"A" feeds a higher amplitude signal to the grid of the neon lamp driver triode (V1B) than does connection "B." Apparently, either connection can operate V1B, especially since cathode bias resistor R9 is 56K (instead of 100K) when connection "B" is used (circuit Category 4). Presumably, the smaller bias resistor used with connection "B" makes V1B's grid less positive (compared to circuit Categories 1-3), so smaller negative oscillator swings are sufficient to cut off current through V1B and the neon lamp. Another somewhat subtle difference in circuit configuration is whether V1B's cathode bypass capacitor C5 is 25 F (most common; Categories 1 and 4) or 5 F (less common; Categories 2 and 3). The value of C5 affects the time constant of the driver stage's dynamic response. This helps determine the shape of the neon lamp's light pulses, as discussed in Section 4 and demonstrated in Figure 6.

With the optical isolator's neon lamp in series with V1B, current in this triode's plate circuit determines current in the lamp. Ballast resistor R10 helps prevent excessive current (and, by measuring R10's voltage drop, provides a means for monitoring lamp current; e.g., Section 3). In parallel with the lamp and R10 is 10-megohm resistor R8; it keeps V1B's plate from being effectively disconnected from the B+ supply when the lamp is off. In that case, when an increasing plate current causes the voltage drop in R8 to exceed the lamp's breakdown voltage, current begins to flow in the lamp and its ballast resistor. As indicated in Figure 1, the schematics in circuit Category 3 (four of the 33 examined; see models listed by category) include C6, a 0.022-F capacitor in parallel with the neon lamp and its ballast resistor. Most likely, this is to suppress (smooth out) the effects of a possible current surge when the lamp's breakdown voltage is exceeded in each cycle and the lamp begins to glow (I call this a "breakdown surge"). As neon lamps age, their breakdown voltage may increase (more about this in Section 7), making the breakdown surge greater, which can cause transient voltage drops across the resistance of the power supply and ground wiring (and the internal resistance of the power supply itself). These transients can find their way into the audio signal path via the power supply and ground connections to the preamp stages. This can cause the famous "tick-tick-tick..." noise some Fender amps make when their optical vibrato is switched in. As exemplified by Gerald Weber in his book "All About Vacuum Tube Guitar Amplifiers" (2009, Kendrick Books, Kempner, TX, ISBN: 978-0-9641060-3-1), many technicians solder a 0.022-F capacitor in parallel with the 10-megohm resistor (R8) to fix a ticking Fender amp.

In the vintage Fender amps, the optical isolator assembly is simply a neon lamp (the specific type is discussed in Section 7) shrink-wrapped against a light-dependent resistor (LDR or photocell). This assembly may also be called an optical coupler, optocoupler, opto-isolator, or--owing to its insectlike appearance--a "roach" or "bug." In each of the 33 vintage schematics I studied, the LDR side of the opto-isolator is used in the same network. Referring to Figure 1, C7 AC-couples the audio signal from the plate circuit of the tremolo channel's final preamp stage to the top (clockwise or CW) terminal of "Intensity" (also called "Depth") pot R11, and via R12 to the first stage of the power amplifier (commonly but unfortunately called a "phase inverter;" it may more properly be called a polarity splitter; yes, I have a pet peeve about conflating polarity and phase--see this link). The LDR is hooked between R11's wiper and ground. Its resistance is very high, typically in the tens of megohms, after several seconds in darkness. Quickly, however, LDR resistance can drop to less than a kilo-ohm given sufficient light intensity. The setting of "Intensity" pot R11 determines the extent that changes in LDR resistance can affect the overall resistance between R11's CW terminal and ground: There is no effect at full counter-CW, increasing effect as the pot is turned CW, and maximum effect at full CW. You may think of the overall resistance of the LDR-R11 combination as the "bottom" resistance of a voltage divider (or "L-pad"), the "top" resistance being the preamp channel's output impedance; the output of this voltage divider feeds the power amp. The follow-up article gives more detail about the optical tremolo modulation network and its non-linearities.

At a frequency determined by the LFO, periodic light pulses cause LDR resistance to vary accordingly. Acting in the voltage divider described above, this modulates the audio signal by an amount that depends on R11's setting, creating the tremolo effect (see also follow-up article). Besides LFO frequency and the "intensity" or "depth" setting, interacting factors affecting vibrato character include the shape and intensity of the light pulses, their duty cycle, and the dynamic characteristics (response speed and magnitude) of the LDR. This is a pretty complicated set of factors to ruminate on while pouring over schematics, let alone to formally model. I'll leave the latter to folks smarter than me, perhaps those trying to precicely digitally model Fender-style tremolos. "At the end of the day" (usually literally, lately) I'm a "hands-on" kind of technician, so I decided to breadboard the canonical Fender optical vibrato circuit and see what it does.

2. A Breadboard Version of the Tremolo Circuit for Experimental Studies. Figure 2 is a group of photos showing my breadboard and example experiment setups. The breadboard is versatile enough to try different circuit categories and opto-isolators. I focused mainly on Category 1 and 2 circuits, covering most (25 of 33) of the original schematics studied (see Figure 1 and Table of Models by Category). The breadboard itself represents only the part of the tremolo circuit up to terminals for a neon lamp (i.e., LFO and driver stages); additional circuitry is needed for tests such as evaluating the performance of different LDRs, and will be described below.

Figure 2: Fender optical tremolo breadboard and example experiment set-ups
Figure 2. Photos of optical tremolo breadboard and examples of experimental setups. A: The breadboard. B: Typical experimental setup, in this case recording light output of neon lamp (see also "E"). C: Neon lamp rubber-banded to LDR from VTL5C1 Vactrol. D: Neon lamp rubber-banded to Radio Shack LDR. E: Neon lamp in setup for recording light intensity waveform, with light-tight blanket removed (blanket seen in "B"). Some individual items are labeled as follows: 1: LFO enable switch equivalent to SW1 in Figure 1. 2: Series of five resistors substituting for LFO "speed" pot R3 in Figure 1; the clip lead (with green insulators) sets frequency; see Figure 3 for tap position versus frequency. 3: The Sovtek 12AX7WXT tube used in all experiments. 4: Three 9-V batteries in series serve as the negative bias supply (-27V) to disable circuit operation when SW1 is open; a single 9-V battery is also sufficient. 5: The new-old-stock NE-2H lamp used in many experiments. 6: Variac (variable auto-transformer) used to set B+ voltage. 7: B+ supply, consisting of a power transformer, full-wave rectifier, and ample filtering by those two large green cans, which came from a Marshall amp, each containing two 50-F 500-V capacitors; B+ was set at 425 VDC for most experiments. The white twisted pair running from the breadboard off to the left past the Variac leads to a 6.3-VAC filament transformer. 8: Heathkit IP-28 regulated DC power supply hooked to the photobias LED when recording neon lamp output waveforms. 9: Breadboard, whose close-up photo is in "A". 10: Light-tight blanket used to shield experiments from ambient light; made with layers of black cloth on either side of an aluminum foil layer, and the sandwich taped down to bench. 11: Part of channel used for recording light intensity waveforms (see my "Infrasonicon" web page); the interface unit (left) and data converter (right) are visable; the ribbon running from the latter past the AC outlets is the serial connection to an old PC, which is storing the data into a WAV file. 12: LDR cut from a VTL5C1 Vactrol; see also Figure 7A. 13: A Radio Shack LDR; see also Figure 7B. 14: TSL230 light-to-frequency converter chip in optical probe of "Infrasonicon" project, seen here bathed in red light from the photobias LED (#"15"). 15: "Ultrabright" red LED photobiasing the optical probe's TSL230 chip, typically used at a low constant current of less than 1 mA.

Instead of "Speed" pot R3, the breadboard uses an equivalent series of five fixed resistors that add up to 3.0 megohms. Figure 3 shows the values and arrangement of these resistors (which I call R3A through R3E), along with the LFO frequencies measured when the array was tapped at different positions to simulate different pot wiper settings. Note that the breadboard has R4 on the clockwise (CW) side of the pot-equivalent, while in the canonical circuit (Figure 1) it is on the pot's counter-CW side. This makes absolutely no difference to circuit function. R4 limits the maximum LFO frequency. As Figure 3 shows, I observed a frequency range of 3.5 to 11 Hz, a 1.65-octave span.

Figure 3: R3 equivalent on breadboard, w/ LFO frequencies
Figure 3. Schematic diagram of R3 equivalent for breadboard. Fixed resistors R3A through R3E, along with a clip-lead for setting the array for various total resistances, substitute for the 3-megohm reverse-audio-taper potentiometer (R3) in the canonical Fender optical tremolo circuit (Figure 1). C2, C3, and R4 correspond to like-numbered parts in Figure 1 (R4 is on opposite side of R3 but this makes no difference). Frequencies were determined by analyzing waveform data recorded in WAV files (e.g., see Figure 6.)

Initial experiments with this breadboard made it clear that the Fender tremolo LFO/neon lamp driver generates light pulses that are rounded, not squared-off as suggested in my earlier post. Since I want to set the record straight on this matter, evidence regarding light-pulse shape is presented next.

3. Smooth Shape of Neon Lamp Pulses in Fender Optical Vibrato Circuit. I can't say exactly what lead to my incorrect statements about neon lamp pulse shape in the previous article. I can say that I probably spent all of five minutes looking at the oscilloscope back then, and perhaps I was not using a slow enough sweep for the low frequencies involved. (My 'scope does have trouble sweep-triggering at certain very low frequencies, I've found.) But there is no excuse for stating, "the...neon lamp cycles on/off according to voltage thresholds (i.e., all versus none is the neon lamp's predominant mode...)." This is very misleading, to be kind. There is a voltage threshold involved when a neon lamp starts, it is the breakdown voltage. But once a lamp starts, and throughout its normal operating range, its light output is directly proportional to current. (For almost everything you'll want to learn about glow lamp theory and characteristics, I refer you to General Electric's Glow Lamp Manual, either the First Edition [1963] or the Second Edition [1966], and a 1979 Signalite Glow Lamp Manual.) Aside from expecting that current through the Fender circuit's lamp driver triode is cut off during part of the LFO cycle, one expects current to change fairly smoothly during most of the "lamp on" phase in this analog circuit, unlike a binary "all-or-none" function.

For comparison, I wanted a circuit that does switch a neon lamp on and off in a binary way. I used an SCR triggered by a square wave (or abrupt-edged pulses with other than a symmetrical 50-percent duty cycle) from a function generator. A schematic and photo of this setup is shown in Figure 4.

Figure 4: Circuit for operating neon lamp with square waveform
Figure 4. A: Schematic diagram of circuit used to cycle neon lamps on and off with square waveforms. Shows oscilloscope connection used to observe voltage drop in 100K resistor and verify that current is constant during lamp "on" phases. B: Photo of SCR driver setup, in this case being used for dynamic performance test of an optical isolator, which will be described in Section 6. Direct light pulse shape recording employed a different lamp (a new-old-stock NE-2H) and the optical probe pictured in Figure 2E, and is not shown here. The following items are labeled: 1: SCR, a Motorola MCR106-3. 2: A Fender-style replacement optical isolator purchased from Antique Electronic Supply. 3: DC analog of audio signal path used to evaluate LDR dynamic performance (see Section 6 and Figure 10).

Note that both the SCR-gated neon lamp driver and the Fender tremolo circuit have a 100-K ballast resistor in series with the neon lamp. In both cases, the voltage dropped in this resistor is proportional to current through the lamp. When the lamp is off, the ground-referred voltage measured at the lamp's anode (the terminal connected to the resistor) is at the high-voltage power supply (B+) level, but drops in accordance with current through the lamp, which in turn is proportional to the light being generated. For example, a 100-V drop represents 1.0 mA in the lamp. My measurements used a "10X" oscilloscope probe (and DC-coupled 'scope input), which introduces a 10-megohm load, and hence just a one-percent measurement error (100K divided by 10M). Since light output is proportional to current, the voltage waveform at the anode of the neon lamp is the inverse of the light waveform. (Within the lamp, light comes from the "corona," which is glowing gas near the cathode. The electrodes of neon lamps are interchangeable, so the way they are hooked into a DC circuit determines the ones to call anode and cathode.)

I also used an independent, direct approach to see the neon lamp's light output waveform. This was to record light intensity using an infrasonic (low-frequency) channel that I built and dubbed an "infrasonicon." Its main purpose is to record phenomena that have periodicities below the audio band, and then make them audible by high-speed playback, just to hear how their time-compressed signals sound. However, since the channel is linear within defined limits, it can also be used as an analytical tool--a low-frequency digital oscilloscope. The particular signal path I used for recording neon lamp activity was as follows: light sensor, a TAOS TSL230 light-to-frequency converter chip in an "optical probe" (see Figure 2E); frequency down-converter (divider) in "interface unit"; phase-locked loop (PLL) demodulator, fourth-order 50-Hz low-pass anti-aliasing filter, and 12-bit A-D converter sampling at 268 Hz in "data converter" (see Figure 2B); serial link to an old PC running QBasic code in DOS to accumulate data in a WAV file, whose playback sampling frequency was specified as 44.1 KHz, thus the "time-compression factor" is 165X. Time- and frequency-domain results displayed in audio editing software are scaled by this factor (manually, of course) to recover the original periods and frequencies. Lately I've been using Audacity, which is excellent open-source freeware.

Using photodiodes, the TSL230 responds fast (unlike a LDR) and is linear across an ultra-wide light intensity range (many decades). With such a sensor, the infrasonicon channel lacks sufficient dynamic range to record a neon lamp from total darkness to any relevant light output level. Therefore, I needed to bias the TSL230 with a constant light source; I used a red LED powered by a constant-current DC supply (see Figure 2B and 2E). This, however, raises a potential issue: neon lamps have a lower-voltage breakdown threshold in ambient light than in darkness. This is part of their "dark effect." Many neon lamps, including the one I used in this experiment, and probably the bulbs used in vintage Fender optical vibrato, contained small amounts of a "radioactive additive" to seed ionization and reduce the dark effect. I'll have more to say about this in Section 7; see also my sidebar demonstrating the dark effect. With the new-old-stock NE-2H lamp I used, as driven by the Fender circuit, the oscilloscope did not reveal any difference in the voltage drop waveform at the photobias level employed here versus in the dark (data not shown).

The top and middle rows of images in Figure 5 show, respectively, typical voltage drop (inverse of lamp current) and direct light output waveforms. The images came from different experiments and I made no attempt to line up time scales or make amplitudes internally consistent. The point of Figure 5 is just to show that the Fender optical vibrato circuit (Category 1 in this case) makes quite rounded light pulses compared to an "all-or-none" circuit running at similar frequency and duty cycle. In the voltage drop display, close inspection of the Fender pulse's leading side reveals an initial current surge (an abrupt discontinuity between the "dark" voltage level and onset of more gradually dropping voltage). This is a breakdown surge as mentioned in Section 1, and explained in more detail in Section 7. A corresponding feature cannot be resolved in the Fender circuit's light waveform, because of the surge's low magnitude and the recording channel's restricted (50-Hz) bandwidth. This bandwidth limit also accounts for sloping light-pulse edges and slight overshoot recorded for the SCR-driven lamp (in contrast to the voltage drop waveform, where the oscilloscope has a 20-MHz bandwidth). Unlike with the new-old-stock NE-2H used here, the Fender circuit-driven neon lamp in a new replacement opto-isolator obtained from Antique Electronic Supply did not have a detectable breakdown surge (see Figure 11). Thus, these two lamps have somewhat different characteristics, as discussed further in Section 5 and Section 7.

Figure 5: Fender light pulses vs. square light pulses
Figure 5. Light pulses from a Category 1 Fender opticlal vibrato circuit (left column) compared to light pulses from a SCR-switched neon lamp (right column). A new-old-stock NE-2H was used throughout this figure. Top Row: Photos of analog oscilloscope display of the lamp's anode voltage, which (due to a series 100-K ballast resistor) represents the inverse of current through the lamp. In the case of the Fender circuit-driven lamp, a small current surge is detectable at breakdown, and the inset enlarges one example. Middle Row: Light intensity of neon lamp recorded using a TSL230 light sensor and "Infrasonicon" channel; waveforms screen-captured from Audacity software. Bottom Row: Photographs of pulsing neon lamp placed against black cloth with cathode side toward camera; with its shutter held open, the camera was transited ("panned") perpendicular to long axis of bulb.

The third method I used to document neon lamp pulse shape was to move the camera while photographing the flashing lamp (Figure 5, bottom row). While certainly less precise than the other two methods, the results do seem to support my conclusion; you may look at the photos and judge for yourself. A final approach, which is completely subjective, is simply to look at the flashing lamp. I don't know if I represent the majority of humans, but I can't honestly say I can perceive the difference between abrupt and "smooth" lamp pulses, even at the tremolo breadboard's minimum LFO speed of 3.5 Hz.

4. Light Pulse Shape Differs Between Circuit Categories 1 and 2. In Figure 5, the Fender tremolo breadboard used a 22-F bypass capacitor in the output triode's cathode circuit (C5 in Figure 1), and the circuit was otherwise hooked up as in "Category 1" (specified as 25 F, C5 is only slightly larger in the actual vintage Fender amps of this category, so I consider this breadboard essentially like Category 1). Each light pulse is fairly symmetrical, starting with current (and light intensity) increasing at a decreasing rate until a maximum (peak) is reached, then falling off in a complementary manner. In fact the waveform resembles a half-wave-rectified sine wave, except the "off" (dark) phases linger beyond 50 percent of each cycle's period. Specifically, the dark phase of each cycle ranges from 73 to 65 percent from the lowest to the highest LFO frequency, respectively.

Making C5 4.7 F instead of 22 F essentially turns the tremolo breadboard into a "Category 2" circuit, and this affects light pulse shape as observed in Figure 6. Most noticably in the lower LFO frequency range, but present throughout, the pulses are asymmetrical--light intensity rises to a peak more quickly in Category 2 than in Category 1, and then falls off more slowly. The corresponding asymmetry is also seen in lamp-anode voltage drop waveforms on the oscilloscope (data not shown). Note that Category 3 amps also specify C5 as 5 F (Figure 1); I did not test it, but there is no reason to expect that lamp pulses for Category 3 are very different from those of Category 2.

I offer the following explanation for the light pulse shape differnce: The amount of current through V1B (and hence the neon lamp) depends on the voltage difference between its grid and cathode. If C5 is absent altogether, any increase in current due to the grid swinging more positive is immediately offset by the cathode going more positive too, because Ohm's Law forces the voltage across the cathode resistor (R9) to increase as the current increases (this is what happens in a cathode follower; "follower" means the cathode voltage follows, or tracks, that of the grid). But with bypass capacitor C5 present, the rate of cathode voltage swing is decreased (delayed; lags) because some current is diverted into the capacitor as it charges. At some point, the grid-cathode voltage difference, and hence the lamp current, must reach a maximum. This point occurs sooner in Category 2 than in Category 1 because a smaller capacitor (C5) charges faster (smaller time constant). Meanwhile, the input signal at the grid is the same in each category, and apparently keeps the grid sufficiently positive long enough to allow current in the lamp for about the same total period in each case. It's just that the larger capacitor in Category 1 delays the current peak. I would expect the Category 3 circuit to deliver asymmetrical light pulses also, since it uses the lower-value of C5, as in Category 2.

Figure 6: Light pulse shape in Category 1 versus Category 2
Figure 6. One-half-second clips of neon lamp light intensity waveforms of breadboarded Fender tremolo circuits representing Category 1 (top row) and Category 2 (bottom row), with LFO frequecy noted in the upper-left for each clip. While the light intensity scale is relative, it is linear and the same for each clip. Data recorded using a TSL230 light sensor and "Infrasonicon" channel set for 50-Hz bandwidth as described in Section 3; waveforms were screen-captured from Audacity software.

Why might Fender have manufactured some amps using the Category 2 (and 3) tremolo circuit? Could such asymmetrical light pulses offer some advantage? Actually, seemingly it would tend to make the tremolo "choppier" than in Category 1, especially since most photocells respond faster to an increase in light intensity than to a decrease (at least, that's true for all three LDRs I tested; see Section 6). Given that, one supposes it would be better to have the opposite kind of asymmetry--a slow increase in light intensity then rapid decay--if the goal is imparting maximum symmetry to the tremolo modulation. Perhaps the goal of the Category 2/3 design was to even out the peak intensity of light pulses across the LFO frequency range (tremolo speed). In Figure 6, notice how peak light intensity in Category 1 falls off with increasing frequency throughout the range. However, in Category 2 peak light intensity is maximum at 5.9 Hz, near the middle of the LFO's range (octave-wise). Maybe this gives an impression of more even tremolo depth while tuning the "speed" knob.

5. The Optical Couplers Tested, and Their Static Characteristics. I tested three opto-couplers in this study: Two used a new-old-stock NE-2H held close to either a LDR cut from a Perkin-Elmer VTL5C1 Vactrol or a Radio Shack-supplied LDR, and the assembly shielded with a light-tight blanket (Figures 2C and 2D show these improvised couplers held together with elastic bands). I have designated these respectively as "NE-2H-->LDR of VTL5C1" and "NE-2H-->R'Shack LDR." The third was a pre-assembled Fender-style replacement optical coupler purchased in December, 2010 from Antique Electronic Supply (AES; their item number R-VOP1020; called somewhat confusingly "Switch, Vibrato LDR for Fender Amps" in their online catalog last checked in June, 2013); it appears in Figure 4B; I'll designate it "AES Replacement."

Figure 7A shows one intact VTL5C1 and one sawed in half so that the peice containing the LDR (at left in photo) can be coupled with a neon lamp instead of the red LED with which it was originally paired (this was how I obtained a fast photocell in my earlier article). Figure 7B shows an assortment of five photocells from Radio Shack's item number 276-1657, purchased at a local store in December, 2012; there are three diameters: 6.5 mm (2 each, at left in photo), 5 mm (2 each, at right), and 3.5 mm (1 each, lower center-right). The arrow points to a 6.5-mm unit, which was used in NE-2H-->R'Shack LDR. As I will demonstrate in Section 6 and Figure 11, this Radio Shack LDR recovers high resistance in darkness too slowly for use in optical tremolo. I did not test the other LDRs in the assortment, but since this first one was too slow, I can't recommend using any photocells from Radio Shack's 276-1657 package in a "roll-your-own" tremolo coupler--many users will not be equipped to screen the assortment for response time.

Figure 7: LDRs and neon lamp used for two of the three optocouplers tested
Figure 7. Components used in improvised optical couplers "NE-2H-->LDR of VTL5C1" and "NE-2H-->R'Shack LDR." A: One intact Perkin-Elmer VTL5C1 and one sawed in half; the LDR half at left was used in the experiments. B: Five LDRs supplied as Radio Shack item number 276-1657 and their packaging; arrow points to one of the 6.5-mm diameter units used in the experiments. C: A "new-old-stock" NE-2H lamp next to its original Radio Shack packaging (item number 272-1102), probably purchased in the 1980s (see text); the leads were tarnished copper before I cleaned and tinned them; its package partner was the one actually used in the experiments and can be seen in Figure 2.

The NE-2H lamp used in the experiments is from a package of two I found in my late father's electronic parts stash, which of course I inherited. One of the lamps and the original package is shown in Figure 7C. Notice that the package calls the lamps "N-2H." This is a typographical error, because the item is listed in all of the old Radio Shack catalogs as "NE-2H." (By the way, this web site with virtually all of the Radio Shack catalogs is an extensive and fast-browsing resource for researching items supplied throughout Radio Shack's history.) Pairs of NE-2H lamps packaged as item number 272-1102 first appeared in 1970 (at $0.39), but the $0.69 price tag on this one dates my Dad's purchase to between 1982 and 2000. (That's a long run without a price increase; I'll bet The Shack was not renewing their inventory with fresh lamps during most of that period.) Knowing my Dad, he probably picked up the lamps in the earliest part of this time frame (early- to mid-1980s). The main reason I'm concerned about neon lamp age is that many types, including the NE-2H, contained a radioactive additive to reduce the dark effect; I'll discuss this below, in Section 7.

Referring to the schematic inset in Figure 8, "static" (steady-state) optocoupler tests were done as follows: At different B+ voltages, the voltage dropped across the neon lamp's 100K series resistor was directly measured on a digital multimeter (DMM), whose input impedance was 10 megohms. This reading was used to calculate lamp current (which is directly proportional to light output for a given lamp, as noted previously). Several seconds after each current level was set, the photocell resistance was read directly on a DMM, and the result plotted against the corresponding lamp current. The resulting characteristic curve for each of the three optocouplers is given in Figure 8. (In the follow-up article, another optocoupler characteristic is investigated: harmonic distortion.)

Figure 8: Static characteristics of three optocouplers
Figure 8. Static characteristic of three optical couplers. Inset shows the test setup, in which LDR resistance was measured at various lamp currents. Curves approximate best fit to the data points (small circles). The vertical arrows indicate lamp current when the supply voltage (B+) is just sufficient for breakdown (i.e., the lamp's striking threshold). The dashed portion of each curve applies only when the lamp has struck and current is later decreased. Dark resistance (at zero lamp current) for each coupler's LDR was more than 40 megohms (beyond the DMM's range).

As shown in Figure 8, the Radio Shack LDR developed the lowest resistance at any given neon lamp current. The LDR in the AES Replacement optocoupler generally gave the highest resistance readings, except at current inputs below 1.8 mA, where the LDR from the VTL5C1 had greater resistance. The low-current portion of each characteristic curve (dashed) was accessible only by decreasing the current once the lamp had previously lit (or "struck"). The vertical arrows indicate observed lamp current after the supply voltage was slowly increased to the breakdown threshold, when the lamp lit (detected as an sudden increase in voltage across the 100K resistor), and then was held constant to read the current. Current at breakdown was lower for the AES Replacement coupler's lamp (about 0.18 mA) than in the NE-2H (about 0.26 mA). Since each used the same value ballast resistor, this difference in current represents a difference in the lamps' maintaining voltage (voltage dropped by lamp, once struck and operating in its normal glow range; see Section 7) as well as a difference in breakdown voltage. As shown in a separate experiment, the AES Replacement had a lower breakdown voltage, 70 VDC versus 95 VDC for the NE-2H in darkness (see Section 7 and Table 1). This significant difference suggest that the lamps are not only different types, but probably represent the two major classes of neon lamps--the NE-2H is in the "high brightness" class, and the bulb in the AES Replacement coupler is likely one of the "standard brightness" types.

I'll have more to say about the differences between standard and high brightness neon lamps in Section 7 of this article. For now, it's important to note that lamps in the high brightness class are more efficient than standard brightness ones--they emit more light per milliamp. Therefore, at least part of the higher resistance of the AES Replacement coupler (compared to NE-2H-->LDR of VTL5C1) at input currents greater than 1.8 mA could be due to it using a standard brightness lamp. A difference in photocell characteristic is also likely, because this is the best explanation for the curves crossing (at 1.8 mA); the lamps' absolute light intensities must differ (by a constant representing their efficiency difference) at every given input current, since each has its own linear relationship between current and light intensity. For comparison, in Figure 9 I've copied a graph of two neon lamp-->LDR characteristic curves published in the 1963 G.E. Glow Lamp Manual. The lamps differ in efficiency but the LDR is the same in each case, and the curves do not intersect. In fact, from this data one can conclude that the NE-2H (high brightness) is about four to five times more efficient than the NE-23 (standard brightness) across nearly a three-decade current range (the NE-23 requires four to five times the current as the NE-2H to yield any given LDR resistance).

Figure 9: Figure 4-11 from the G.E. Glow Lamp Manual (1st Ed., 1963)
Figure 9. Image copied from page 74 of General Electric's Glow Lamp Manual (1st Ed., 1963) which shows static characteristic of two different neon lamp types illuminating a certain LDR (sometimes called a "photoconductor" in 1963-speak). Compare with my results in Figure 8 (note that the current scale in Figure 8 is arithmetic while in Figure 9 it is logarithmic).

6. Dynamic Performance of Optical Couplers. I wanted to evaluate LDR response speed under conditions similar to those of a Fender amp's "vibrato" channel. I used the network shown in Figure 10, which was built on a small perf-board as seen in Figure 4B. For this DC analog of a typical Fender signal path, a constant DC input voltage (I used 5.0 V) stands in for the audio signal. Viewed on an oscilloscope, the circuit's low-frequency output waveform shows how an audio signal would be modulated when the tremolo intensity knob is full clockwise (maximum). This test circuit assumes that the final tremolo-channel preamp stage's output impedance is 38K, and the input impedance of the first power amp stage (the "phase splitter") is one megohm. Actual impedances can diverge from these values somewhat without affecting the results significantly, as long as they are "in the ballpark." Regarding the expected error compared to actual AC (audio), calculations show that inserting 0.1-F coupling capacitor C7 does not cause a significant loss of low-frequency respone in the guitar-relevant band at any LDR resistance (see follow-up article, which focuses on AC performance). Thus the DC analog should predict audio performance quite well.

Figure 10: DC analog of Fender tremolo channel
Figure 10. DC analog of Fender optical tremolo signal path used for dynamic test of optocouplers (results are shown in Figure 11). The resistor numbers (R1-R4) are not intended to match component numbers in Figure 1. The network was assembled on a small perfboard, as pictured in Figure 4B (where it is labeled item "3").

Dynamic test results for the three optocouplers that were described in Section 5 are shown in Figure 11. For this test, the couplers' lamps were driven either by the Fender tremolo breadboard (configured as Category 1) or a symmetrical square-wave-gated SCR (as depicted in Figure 4). For the oscillographs shown, the Fender LFO was operated at 7.6 Hz or the SCR clocked at about 8 Hz. (My analog oscilloscope-digital camera combination offered the most presentable results in this frequency range, requiring both consistent triggering of chop-mode 20-msec/division sweep ['scope] and shutter opening for nearly exactly one sweep period [camera]). In each photo of the oscilloscope display, the upper trace is the neon lamp's anode voltage referred to ground (the inverse of lamp current and hence light intensity; see Section 3), and the lower trace is the output of the signal path DC analog drawn in Figure 10. While the raw vertical scale for the lower trace is 0.5 volts/division, on either side of each photo I have transformed the scale to show attenuation (in dB relative to infinite LDR resistance) on the left, and LDR resistance on the right.

Figure 11 shows that signal attenuation maxima (and LDR resistance minima) caused by neon lamp pulses is generally consistent with the static test results given earlier (Section 5 and Figure 8). That is, in order of increasing attenuation depth, the tested optocouplers were: AES Replacement, NE-2H-->LDR of VTL5C1, and NE-2H-->R'Shack LDR. On a timescale relevant to tremolo, the LDRs in all three couplers responded rapidly to increasing light intensity (leading side of a light pulse); the LDR from the VTL5C1 was slightly slower than the others, which is most easily discerned in the SCR-switched lamp results. However, at the other side of the light pulse, recovery of high LDR resistance (and hence low signal attenuation) after the lamp goes dark was very slow for the Radio Shack LDR. Both the VTL5C1 LDR and the LDR in the AES Replacement allowed recovery to nearly 0 dB attenuation during the dark phase at the test frequency, but the Radio Shack LDR only mustered recovery to about -5 dB before the next light pulse (in the Fender tremolo circuit). This is a direct demonstration of how a slow-recovery photocell causes a perceivable loss of volume when the optical tremolo effect is engaged, which was the subject of my previous article. My advice: don't "roll rour own" optocoupler with a Radio Shack LDR.

Figure 11: Dynamic test results for three optical isolators
Figure 11. Dynamic performance of three neon lamp-LDR optical couplers. Neon lamps driven by a Category 1 tremolo circuit at 7.6 Hz (left column of photos) or a square-wave-gated SCR at about 8 Hz (right column). In each photo of oscilloscope screen, the upper trace shows voltage drop across the lamp's 100K ballast resistor, and represents the inverse of current through the lamp (and its light output). The lower trace is the voltage output of the Fender tremolo DC analog network diagrammed in Figure 10. The vertical offset between the two traces is not necessarily the same among these photos. The lower traces' attenuation and LDR resistance scales were calculated by applying elementary DC network math to the known resistor values and 5.0-V input as given in Figure 10; the reference level for 0 dB attenuation is the output voltage when LDR resistance is infinite (2.244 V; actual LDR resistance after several seconds in darkness is sufficiently high that it can be modeled as infinity without introducing significant error). As described in the sidebar at this link, the six oscilloscope display photos in this Figure and the two in Figure 5 were graphically transformed with a 4-degree vertical skew to help compensate the effects of a strong magnetic field that affected the analog 'scope during photography.

Relative to the speed an LDR acquires low resistance after illumination, recovery of high resistance in darkness is slower; this appears to be a general characteristic of photocells. This could be one reason Fender optical tremolo circuits were apparently designed to keep the neon lamp dark 65 to 73 percent of the time: LDR recovery lag insures that the audio is attenuated somewhat longer than the length of the light pulses. But what kind of modulation envelope is optimal? A precise theory would require some knowledge about the psychoacoustics of tremolo. Specifically, how does the modulation waveform affect perception of the tremolo effect? What is the maximum percentage of the modulation cycle's period that the signal can be attenuated without a perceived loss of volume when tremolo is engaged? I will not presume to guess the answers here, other than to state what seems obvious: the signal must return to near zero attenuation for some part of each modulation cycle to avoid loss of perceived volume when the effect is turned on. That's why the Radio Shack LDR is not appropriate for tremolo. More insight into perception of tremolo may be found in the follow-up article, which shows that Fender optical tremolo modulates fundamental tones deeper than their triode-generated harmonic distortion products.

Putting aside questions about the shape and symmetry of tremolo modulation cycles, let's return to the depth of attenuation per cycle. The data in Figure 11 show that the two acceptable couplers I tested differ in that regard: With the Fender Category 1 circuit at 7.6 Hz, maximum attenuation per cycle for the AES Replacement coupler was about -16 dB, but the NE-2H-->LDR of VTL5C1 reached about -24 dB. This means that, for a comparable effect, the intensity knob (pot R11 in Figure 1) would require a more clockwise setting (by about 8 dB) with the AES Replacement than with NE-2H-->LDR of VTL5C1. Also, the latter coupler would be capable of deeper tremolo at extreme clockwise settings; whether such effects are practical is unknown (although I would tend to err on the side of not limiting the options available to musicians). Unfortunately, I did not evaluate an original vintage Fender optocoupler in pristine condition for comparison (arguably such a device might be difficult to find in 2013). As mentioned in Section 5, some of the difference in illuminated LDR resistance between the AES Replacement coupler and NE-2H-->LDR of VTL5C1 may be due to a difference in neon lamp efficiency. This raises the question: what type of neon lamp did the vintage Fender amps actually use in their optical "vibrato" couplers (or, will any old neon lamp work as a replacement)?

7. Which Neon Lamps Do Vintage Fender Amps Use, And Are They Radioactive? There are basically two classes of neon glow discharge indicator lamps: standard-brightness (SB) and high-brightness (HB). According to General Electric's Glow Lamp Manual (First Edition [1963] and Second Edition [1966]), in terms of light intensity per milliamp, the HB types average three-fold greater efficiency than the SB types. But HB lamps are also rated for greater current and an equal lifespan, so General Electric (G.E.) states that HB lamps have about 10 times the light output of SB on an "equal life basis" (light output integrated over the useful life of these lamps; of course this implies that HB lamps can handle on average 3.33 times more current than their SB cousins). The candidate lamps we're concerned with here are all rated for an "average useful life" of 25,000 hours at their nominal RMS current for AC operation; for DC, which optical tremolo uses, the SB and HB types provide about 60 and 50 percent of this expected life, respectively. Lifespan is not consumed unless the lamps are drawing current, which is relevant for pulsed DC operation as in a tremolo circuit. When lit, the manuals say the lifespan of most SB types is inversely proportional to the cube (third power) of the lamp current up to 1 mA, and for HB types, an eyebrow-raising sixth power of the current up to 6 mA. This means that, especially for HB lamps, one is ill-advised to run a lamp very long at currents greater than its rated value. Nominal current of the NE-2H and NE-2U (HB types) is 1.9 mA; for NE-2E and NE-2V (SB types), it is 0.6 mA.

Why is the lifespan of DC-operated neon lamps 50-60 percent that of lamps operated on AC (at equal RMS current)? It's because degradation of the cathode is a major factor in lamp aging. Each electrode of a lamp running on AC serves as cathode half the time that the (designated) cathode of a DC-operated lamp does. Cathode "sputtering" consumes the lifespan of all lamps during use. Sputtering is caused by bombardment of the cathode by high-velocity positive neon ions; the electode's specially treated emissive surface is gradually lost, while some of it gets deposited on the inner side of the glass envelope, darkening it. The G.E. manual says that SB and HB types differ in the way they age and ultimately fail. SB lamps gradually get dimmer due to darkening of the envelope, while their breakdown voltage may change gradually (usually increasing) throughout their life; end-of-life is defined (somewhat arbitrarily) as 50 percent loss of the SB lamp's initial brightness. In contrast, the light output of  HB types is "quite constant throughout most of lamp life." Near the end of their lifespan, HB lamps undergo an abrupt increase in breakdown voltage, according to the manuals, which continues to increase until the lamp no longer strikes.

The G.E. manual states that "either standard brightness or high brightness glow lamps may be used with photoconductive devices." But it also says that, since HB lamps can emit a total of 10 times more light during their life than SB types, and since the HB's intensity per milliamp changes little during this period, HB lamps are "thus suited for use with photoconductors and other light-activated semiconductors." Indeed, at least in the mid- to late-1960s, Clairex (a manufacturer of photocells) made two neon lamp-->LDR optocouplers (CLM3120A and CLM4120A) which each used a HB lamp but differed in their LDR characteristics, according to their datasheet. Apparently, the tremolo circuit of at least one model of 1969 Guild guitar amp used a CLM4120A according to a thread in Music Electronics Forum. I have not located a schematic to see how it was used. The Clairex units were built using rigid tubes, not shrink-wrapped like the Fender couplers (which I suspect were assembled at the Fender factory). The price of the Clairex units was a hefty $2.00 each in quantities of 300-999 in 1968 dollars (about double the cost of a 12AX7 tube) according to my Dad's old Srepco/Pioneer-Standard Electronics catalog (8th edition). An amplifier manufacturer might have found it more cost-effective to assemble their own couplers (and/or they called for specific lamp and/or LDR specs not available in a pre-assembled unit).

If you're thinking I'm leading up to saying that Fender amps used HB lamps, don't jump to any conclusion yet; read on. Let's have a look at some candidate lamps available in the 1960's and 1970's. I will consider only "indicator" glow lamps; "circuit component" types are optimized for their electrical characteristics (e.g. for use as a voltage reference, as one might use a zener diode), and thus are an unlikely choice for optocouplers. I'm mostly using the old-style (but easier to remember) "NE-" type numbers in this article, giving the corresponding American Standards Association (ASA) lamp numbers in parentheses in this paragraph. Form factor offers a clue: vintage Fender tremolo couplers used a "formed tip" or "pip"-sealed bulb with wire terminals emerging from a "pinch" at the bulb's base. The outside length of the glass envelope is medium-sized, if I recall correctly; I don't remember it as being quite an inch long, so this eliminates the NE-2 (A1A), the archetypical SB type, and its radioactive (see below) "brother" NE-23 (5AB), and a handful of other types. Both the First Edition (1963) and the Second Edition (1966) of the G.E. Glow Lamp Manual specifies these as 1-1/16 inches long (max), and the 1979 Signalite manual lists them as one inch; besides, the NE-2 has a nominal current rating on the low side (only 0.5 mA) and the NE-23 even lower (0.3 mA). (If a vintage coupler did use a 1-inch bulb, it would probably be the NE-86, a higher current [nominal 1.5 mA] version of the NE-23.)  More likely, the candidate lamps are one of the 3/4-inch types (and this is the physical size of the lamp in the AES Replacement that I tested). This essentially narrows the field to NE-2E (A9A) or NE-2V (A2B) for the SB types, and NE-2H (C2A) or NE-2U (A3C) for the HB types. All except NE-2E contain a radioactive additive to reduce the dark effect (see below); NE-2U is like NE-2H except it contains more radioactive additive. The First Edition (1963) of the G.E. manual specifically lists NE-2U as a "suggested lamp type" for use as a "photoconductor activator."

The HB types have a higher breakdown voltage (either in darkness or light) than the SB types. The same is true for the maintaining and extinuishing voltages (which do not depend on ambient light), respectively defined as the voltage drop across the lamp once struck, and the minimum voltage at which the lamp can remain lit. For any given lamp, the order of voltages from largest to smallest is: breakdown, maintaining, and extinguishing--all of which are fairly independent of current, in the normal operating range; current is determined by impedances in the external circuit. I focus mainly on breakdown because it is most relevant for optical vibrato circuit performance/reliability, especially as the lamps age. The manuals generally list specs for operation in ambient light, and do not give an average breakdown voltage for many indicator types, just a maximum: it is 90 VDC for our candidate SB types and 135 VDC for HB. For more perspective on this, I measured the DC breakdown voltages of different lamps, including the new-old-stock NE-2H and the lamp in the AES Replacement coupler, plus individual samples randomly picked from 10-packs of new NE-2H, NE-2E and NE-2 lamps bought in February, 2013 from Jameco Electronics. The results are given in Table 1. The relatively low breakdown voltage for the AES Replacement coupler's lamp suggests that it is a SB type, as noted in Section 5.

Table 1. Static DC Breakdown Voltage Measured for Different Neon Lamps in Darkness Versus Bright Light
Lamp TypeLamp ClassBreakdown Voltage, Dark (1)Breakdown Voltage, Light (2)
Unknown, in AES ReplacementProbably SB (3)70Not Determined (4)
NE-2H (new-old-stock) (5)HB9590
NE-2H (new, from Jameco)HB9287
NE-2E (new, from Jameco)SB7167
NE-2 (new, from Jameco) (6)SB7570
Notes: (1) With lamp covered by light-tight blanket and hooked in series with 100-K resistor, DC supply voltage was gradually increased (at 1V per second or less, near the threshold) until ground-referenced DMM with probe hung on lamp-side of resistor showed a sudden voltage drop, indicating lamp striking, then probe moved to power supply side of resistor and the voltage read. Each voltage shown in this table is average of 5 replicate tests (data range was generally less than +/- 1V) rounded to the nearest volt. (2) Same method as test in darkness, except light-tight blanket removed and a high-brightness near-ultraviolet LED (405 nm; Radio Shack part number 276-0014), operating at 20 mA DC and directed at lamp's cathode, was placed against the lamp. (3) Based on the relatively low dark DC breakdown voltage. (4) This test would have required cutting open the opto-coupler, which was not done. (5) This was the same lamp as used in the improvised optical couplers and for direct recording of light pulse shape. (6) These nominal "NE-2" lamps are only 7/16 inch long, not the expected 1 or 1-1/16 inch; this makes one wonder about their authenticity. Having taken over Chicago Miniature Lamp, Visual Communication Company, LLC still supplies many of the legacy neon lamp types, in their classic form factors, including the NE-2. Perhaps the ones supplied by Jameco are from a different manufacturer.

Note that Table 1 presents static breakdown voltages, where voltage ramps much more slowly than in most practical applications, such as optical tremolo. Dynamic breakdown voltage is always greater than static (both in ambient light and especially in darkness). This is because all lamps have a finite ionization time--the lag before full operating current can develop after applying a potential equal to or greater than the static breakdown voltage. Ionization time increases as ambient light intensity decreases, and/or as the applied voltage in excess of the static breakdown threshold decreases. For applications where the lamp must strike in ambient darkness (such as optical tremolo), both static breakdown voltage and ionization time are larger than if there were ambient light. This is called the "dark effect," and it increases the dynamic breakdown voltage. (I show a demonstration of the dark effect at this link.) The surface of a neon lamp's electrodes are specially treated to emit photoelectrons when exposed to ambient light, "seeding" ionization of the gas, and accelerating striking once the breakdown voltage is exceeded, compared to operation in darkness. "Dark-effect-reduced" lamps contain radioactive material to seed ionization as I will detail below.

The magnitude of the current surge when the lamp finally strikes depends on the applied voltage at that instant, the lamp's maintaining voltage, and all impedances in series with the lamp (not necessary just a ballast resistor). As explained earlier, lamp aging is another factor than can increase breakdown voltage and hence current surge. In Fender optical tremolo, surges may leak audible "ticking" into the preamp stages via the common power supply connection, as described in Section 1. (Another possible pathway is direct induction into the signal path apparently via the LDR and associated conductors, which is implied in a service manual for the 1993 Fender Vibro-King amp [whose tremolo circuit is discussed further below], which states: "If the Vibrato exhibits a clicking sound, the problem is caused by the position of the neon bulb in relation to the LDR's. Adjust the bulb by gently pulling on the leads. Test and repeat as necessary.")

As highlighted in the top left photo of Figure 5 and mentioned in Section 3, in my experiments the new-old-stock NE-2H lamp driven by the Fender circuit begins each light pulse with a detectable current surge. This dynamic "striking surge" is on the order of 0.05 mA, smaller than the current-at-breakdown of about 0.26 mA that I observed in static tests using a 100-K ballast resistor (the position of black and red arrows in Figure 8). (Although the vertical scale is not calibrated in Figure 5, you can also see the NE-2H's striking surge in Figure 11 [center and bottom photos in the left column] where the voltage scale is marked and current can be reckoned). Why is this lamp's dynamic "striking surge" smaller in the Fender circuit than in its static tests? It's due to additional impedances besides the 100-K ballast resistor which are in series with the lamp in the Fender circuit: the 100-K cathode resistor (R9 in Figure 1 for a Category 1 circuit) and the driver triode's own dynamic plate resistance. This "starves" the lamp's current-at-breakdown compared to conditions used in the static tests. Note in the left column of Figure 11 that the (likely SB) lamp in the AES Replacement does not have a detectable "striking surge" under identical conditions in which one is detected in the (HB) new-old-stock NE-2H. This is evidence that the former has a significantly lower dynamic breakdown voltage and maintaining voltage than the latter, in the Fender circuit. I would expect the AES Replacement is less likely to cause "ticking" noise in a Fender amp than a coupler made from the NE-2H.

At this point I will hazard a guess that the designer(s) of the Fender optical tremolo circuit (Leo Fender himself?) was aware of neon lamp striking surge, and was concerned with making the leading side of the light pulses as smooth as possible, to minimize noise leaking into the audio path. The use of C6 in circuit Category 3 (see Figure 1) may be evidence of this concern (also note that Category 3 circuits should have an inherently steeper light-pulse attack than in Category 1, as suggested in Section 4). If minimizing the striking surge is the main criterion, a SB lamp that is "dark-effect reduced" (using radioactive additive; see below) is the best choice. Of course, every choice is a trade-off, and in this case one sacrifice would be lamp efficiency (lower light intensity per milliamp than HB; this could be compensated to some extent by careful choice of LDR, so that tremolo depth does not suffer). The widely available SB lamp that best fits this criteria is the NE-2V.

Indeed, the NE-2V is used in the 1993 Fender Custom Shop Series "Vibro-King" amp, according to its service manual (marked June 1994 Rev A), which lists this bulb by its ASA number, A2B. It's the first item on the parts list for the PCB assembly. Three schematic versions are included with the service manual PDF; the first one (Drawing Number 041253 Rev A, dated 1993) appears to best agree with the manual's parts list. It shows an optical coupler using two LDRs in parallel, and the parts list notes that they are a Cadmium Selenide (CdSe) type; the intensity pot is 250K in this case. The other two schematic versions (041253 Rev E [1995] and Rev F [1995]) show a single photocell and 5-megohm intensity pot. In all three drawings, the photocell(s)/pot network is implemented in a similar manner as the classic Fender amps (right before the "phase splitter" stage, see Section 1 and Figure 1). The LFO and neon lamp driver stages are also quite similar to the classic amps, and (referring now to notation in Figure 1) use driver feed connection "A", 22 F for C5, and 100K for R9, making it most similar (and probably functionally close) to Category 1 among the classics. The differences between the Vibro-King's LFO and the "canonical" circuit are all minor: "speed" pot R3 is 5M instead of 3M (it can probably do a slower tremolo), R4 is 270K instead of 100K, foot pedal switch SW1 is shunted by a 10-F electrolytic capacitor in parallel with a 270-K resistor (this adds contact "de-bouncing"), and the LFO triode is bypassed plate-to-cathode with a 220-pF capacitor for stability. The Vibro-King's lamp driver stage has two minor differences to the classics (still using the notation in Figure 1): R8 shunts just the lamp instead of the lamp-R10 series combination; and, while the Vibro-King uses a 0.022-F striking-surge suppression capacitor, unlike Category 3's C6, it bypasses the cathode to ground rather than to B+ (these connections are functionally equivalent due to the low source impedance of the power supply).

Since Fender's Vibro-King uses a NE-2V lamp in an optical tremolo circuit that is very similar to those of the classic amps, I think that's good evidence that the classic Fender amps probably also used a SB lamp, not a HB type. My inference that the AES Replacement coupler also uses a SB lamp adds weight to this conclusion, because whoever made that replacement probably has more direct knoledge about the original parts than I do. Presumably, minimizing the current surge when the lamp strikes is a major reason for the SB choice, which in turn suggests that, among SB lamps, ones that are "dark-effect reduced" (i.e., radioactive, like the NE-2V) are best.

From the vintage era (see 1963 and 1966 G.E. manuals) up to the present (see neon lamp app notes and spec sheet from Visual Communication Company, LLC), the industry-standard method for dark effect reduction is including a "radioactive additive" in the lamps. The radiation keeps some small portion of the lamp's neon gas ionized (even in darkness when there are no photoelectons emitted from the cathode), so that current can begin to flow as soon as possible once the breakdown voltage is exceeded. Neon lamp manufacturers specify neither the exact radioactive isotope(s) they use, nor its exact ("small") quantity, in their public literature. (The G.E. manuals do state that the dark effect cannot be completely eliminated, except with "prohibitively" [and I guess, dangerously] large radiation levels.) I thought this should be important information for design engineers building equipment that might be expected to function predictably after several half-lives of the "radioactive additive," for all we know. Most likely, the half-life (T-1/2) is long compared to the 25,000-hour useful life of these lamps; but intermittent and pulsed operation extends this, as mentioned earlier, and radioactive decay is independent of lamp use. And what about these lamps' shelf life? Could radioactive decay have affected the dynamic performance of my new-old-stock NE-2H? Does a NE-2V eventually decay into a NE-2E?

Two government documents, NUREG/CR-1775 published in 1980 and NUREG-1717 published in 2001, say that the radionuclides most widely used in neon lamps are (apparently mainly) krypton-85 (Kr-85; T-1/2 = 10.756 years) and (a bit less commonly) hydrogen-3 (tritium; H-3; T-1/2 = 12.32 years). The 1980 report states that 1 to 5 microcuries (Ci) of these isotopes was used per neon indicator lamp in the 1970s. More specifically, the 2001 report states that each lamp typically contains 0.2 Ci of Kr-85 or 20 Ci of H-3. Table 9.1 of the earlier report lists the total number of "electron tubes" (broadly defined; including the lamps) distributed in the US by year from 1970 to 1978 containing various isotopes (not including those containing thorium, a footnote says); annual distribution of units containing Kr-85 climbed from 34 to 120 million in this period. In the 1970's, when the demand for "electron tubes" in general was decreasing, units containing Kr-85 must be some specialized items, and the sheer quantities suggest that many (if not most) are neon lamps. The next two most commonly used isotopes in 1978 were: H-3, 56 million units distributed; and promethium-147 (Pm-147; T-1/2 = 2.6234 years), 1.5 million units.

One microcurie (Ci) is defined as 37,000 disintegrations per second (https://en.wikipedia.org/wiki/Curie). Thus if a brand-new neon lamp contained 0.2 Ci of Kr-85 as suggsted in NUREG-1717, it has 7400 disintegrations per second. Hypothetically, let's say that lamp was made in 1960 and (after some storage period in inventory) found its way into the tremolo circuit of a Fender guitar amp, as an essential part of its optical coupler. This year, in 2013, which is five Kr-85 half-lives since the lamp was made, one expects the lamp to contain less radioactivity by a factor of 2^-5 = 1/32 (about 231 disintegrations per second). How might this affect performance? I can't be sure, but it seems fair to say that the lamp's dark effect reduction would not be quite as effective as it was when new. This would tend to make its striking surge greater, perhaps causing ticking noise. However, normal amp usage (with "vibrato" engaged) uses up lamp life due to cathode sputtering, having the same undesired effect on the striking surge; radioactive decay probably doesn't matter a big deal compared to that. It's just fun to ponder, by way of conclusion, that these venerable old amps, a product of the "space age" (to use vintage-amp-period lingo), probably in this small way "harnessed the power of the atom."

Focusing on questions of linearity, my investigation of Fender optical tremolo continues here.

Back to Clark Huckaby Main Page (shows my email address)

Forward to Next Article: "How Linear is Fender Optical 'Vibrato' (Really Tremolo)?"

Sidebar Demonstrating Neon Lamp Dark Effect

Sidebar on Correcting Magnetic Skew in Oscilloscope

Sidebar on Difference Between Polarity and Phase

Back to Earlier Fender "Vibrato" Article

Copyright 2013 Clark Huckaby. All rights reserved. Commercial use of this content without first obtaining the expressed permission of Clark Huckaby is prohibited by law. Distribution of copies beyond personal use and/or re-posting of this page is prohibited by law.