Overdrive, distortion and fuzz circuits: TS808, Klon-style, RAT and Fuzz Face simulated in ngspice, with sound clips

The three types differ in where and how they clip

Every guitar distortion circuit amplifies the signal hard and then limits it at some voltage. What differs is the element that does the limiting and where it sits.

The figure shows the output of each circuit for a 196 Hz (G), 0.3 V sine wave fed through the same pickup. The three diode circuits clip symmetrically, so even harmonics stay below −80 dB and only odd ones appear. The 3rd harmonic is −15.7 dB relative to the fundamental for the TS808 and −9.7 dB for the RAT. The Fuzz Face clips asymmetrically and has a strong 2nd harmonic at −8.9 dB.

Test conditions: a synthesised guitar, pickup included

The clips are not recordings of real pedals. They are the output of the circuit simulator ngspice (WebAssembly build). The guitar is made with modal synthesis, which computes the string vibration harmonic by harmonic. It includes the triangular shape of a picked string, the slight sharpening of harmonics from string stiffness, faster decay of higher harmonics, and the small difference between the string's two planes of vibration. The phrase is 9.4 seconds: an E5 power chord, A5 and G5 stabs, an E minor pentatonic line, a high note held with vibrato and an open E major chord.

There are two pickups, and you can switch between them under each player. The single coil is one coil 43 mm from the bridge (L = 2.5 H, R = 6.5 kΩ, 250 kΩ pots). The humbucker is the sum of two coils at 40 mm and 58 mm (L = 5 H, R = 10 kΩ, 500 kΩ pots), with about 1.8 times the output. Both include 100 pF of self-capacitance and a 3 m cable (300 pF) before each circuit. Where the pickup sits on the string changes the mix of harmonics, and pickup impedance changes the sound of a fuzz a lot, so neither is left out (see the pickup equivalent circuit).

The output was taken at 88.2 kHz, decimated to 44.1 kHz, and set to the same loudness of −20 dBFS RMS, as if you had matched the pedals' Volume knobs. Clips with large peaks, such as the clean one, were kept below −1 dBFS and are quieter by that amount.

Listening to the raw circuit output, you also hear everything above 5 kHz that a guitar speaker would remove, so it sounds harsher than in real life. With the box under the players ticked, your browser applies a 90 Hz–5 kHz band limit while playing. Untick it for the raw circuit output.

TS808: diodes in the feedback round off the peaks, the bass stays clean

The heart of the Ibanez TS808 (Tube Screamer) is a non-inverting op-amp stage (JRC4558). The feedback resistance is 51 kΩ + Drive (500 kΩ) and the inverting input goes to ground through 4.7 kΩ + 47 nF. The gain is 1 + (51k + Drive) ÷ 4.7k, up to 118 (41 dB) at full Drive.

Because of the 47 nF, that gain only applies above 720 Hz (1 ÷ (2π × 4.7k × 47n)). Below it the gain approaches 1, so the bass hardly distorts. Clipping the fundamentals of the low strings would muddy chords, and these values avoid that.

The two diodes sit in parallel with the feedback resistor. When the voltage across it (between the output and the inverting input) exceeds about 0.6 V, the diodes conduct, the feedback gets deeper and the gain falls smoothly. That is soft clipping. Because the stage is non-inverting, the input itself (the gain-of-one part) always reaches the output, which rounds off the corners.

The tone stage after it is built around a 1 kΩ + 0.22 µF low-pass (723 Hz), with a 20 kΩ tone pot adding or removing treble. Together with the 720 Hz high-pass in front, this gives the familiar mid-hump (Figure 6).

Klon-style: germanium-clipped signal plus the clean signal

The Klon Centaur and its descendants (often called "Klones") add a clipped signal and an unclipped one in a summing amplifier. The circuit here takes only the structure from published analyses (dual-gang Gain pot, germanium diodes, summing) with rounded values. It is not an exact copy of the real pedal.

The gain stage is non-inverting, with a gain of 1 + (2.2k + Gain2) ÷ 3.3k above 482 Hz. Its output goes through 1 kΩ to two germanium diodes (1N34A) to ground. Germanium conducts at about 0.3 V, half of silicon's 0.6 V, so it limits at half the voltage.

The clean path divides the input buffer output with 15 kΩ and Gain1 (to ground) and sends it to the summing node. Both pots turn on one shaft: turning Gain up raises Gain2 (more distortion) and lowers Gain1 (less clean signal). In this circuit the share of clean signal is 0.58 near minimum Gain and 0.06 (−20 dB) at maximum.

With Gain turned down, the clean signal leads and the germanium-clipped part sits thinly on top. That sum is where the reputation for a thicker sound that does not sound obviously distorted comes from. The real pedal runs its op-amps from 18 V (doubled with a charge pump) so the clean path does not clip first. The simulation uses ±9 V for the same swing.

With the mixer below you can change only the amount of clean signal and hear the sum. The two paths were solved separately in ngspice and are added in the page. The summing amp is linear, so the sum is the same as solving the whole circuit.

RAT: huge gain, then diodes to ground flatten the peaks

The ProCo RAT amplifies hard with a non-inverting LM308 stage, then clips with two diodes to ground. The inverting input goes to ground through 47 Ω + 2.2 µF and 560 Ω + 4.7 µF, so above 1.5 kHz the gain is 1 + Distortion ÷ 43 Ω, up to 2300 (67 dB).

The real gain is set by the op-amp. With its external 30 pF compensation the LM308 has a GBW of about 1 MHz, so the gain at 2 kHz tops out around 500. Its slew rate is also slow at 0.3 V/µs, which turns large high-frequency swings into triangles. This is why the LM308 is considered essential to the RAT's sound.

Unlike the TS808, the diodes are outside the gain stage. After 1 kΩ the signal is simply cut at ±0.6 V, giving a flat-topped hard clip. The final Filter is a low-pass of 1.5 kΩ + Filter pot (100 kΩ) and 3.3 nF, with a corner from 32 kHz (pot at 0) down to 475 Hz (pot at max). The clips use 15 kΩ (about 2.9 kHz).

For how 1970s distortion like the RAT differs from 1990s-and-later high gain (Boss MT-2 Metal Zone), see Vintage distortion vs modern high gain.

Fuzz Face: two transistors saturate and cut off

The Fuzz Face uses no diodes. It is a two-transistor direct-coupled amplifier, with 100 kΩ of DC feedback from Q2's emitter to Q1's base setting the operating point. The more the Fuzz pot (1 kΩ) is bypassed by 20 µF, the higher the AC gain.

The gain is so high that Q2 saturates in one direction (collector down to about 0.9 V) and cuts off in the other (up to nearly 9 V). The swing from the 4.4 V operating point is different up and down, so the waveform is an asymmetric square and even harmonics such as the 2nd appear (Figure 1).

The input connects through a capacitor straight to Q1's base, so the input impedance is low. That is why it interacts directly with the pickup coil and the guitar's volume control, and why a buffer in front changes the sound.

The original used germanium PNP transistors. This is the silicon NPN version. In the silicon version Q2's operating point drops, so 8.2 kΩ was changed to 5.6 kΩ to put the collector at 4.4 V (a usual adjustment for silicon builds).

Frequency response: which band is boosted before clipping

Circuits that clip the same way still sound different depending on which band they boost first. The figure shows the small-signal response, with the diodes not conducting, normalised to 0 dB at 1 kHz.

The TS808 has a hump peaking near 800 Hz, −24 dB at 20 Hz and −13 dB at 6.3 kHz. The TS808's reputation for mids comes from this hump becoming the distorted part of the sound. In the Klon-style circuit the clean path carries the bass, so it is only −8 dB at 20 Hz and −2 dB at 6.3 kHz. The RAT peaks near 1.8 kHz and cuts the lows about as much as the TS808.

All clips side by side

The same phrase through nine circuits and settings. Distorting a chord creates components at the sums and differences of its notes' frequencies (intermodulation). For a power chord (root, fifth, octave) most of them land on harmonics of the original notes, but the final E major, which contains a third, produces components that do not, and it gets muddy. Hard clipping produces more of them.

What the simulation leaves out

• The op-amps are simple models with only GBW, slew rate and output swing (JRC4558 3 MHz and 1 V/µs, TL072 3 MHz and 13 V/µs, LM308 1 MHz and 0.3 V/µs). Noise, input offset and output current limiting are not modelled.
• The op-amp circuits replace the real "9 V single supply + 4.5 V bias" with a ±4.5 V split supply (±9 V for the Klon-style). To the signal it is the same.
• Input and output buffers (the TS808's transistor, the RAT's JFET) are ideal voltage buffers.
• Pots are fixed resistors chosen for the knob position. The difference between log and linear tapers is folded into those values.
• No guitar amp or speaker is included. The optional band limit only roughly imitates a speaker's range and is not the response of any particular cabinet.