The Melissa Amplifier

The Designer’s Notes on Goals, Workings, and Operations

 

Carl Berger and Jukka Makinen

 

21JUL06

Design Goals

The Melissa Amplifier design project began and was chronicled on the AX84.com website bulletin board. Two designers (the authors) familiar with “dc-drive” preamp stage techniques (more on that to follow) soon settled in on design goals for this amplifier:

1.      Gig-able without P.A. assist for small to medium gigs.

2.      Reasonably small – shouldn’t be a bear to haul.

3.      Exceptionally good distortion sounds – from the preamp only, or preamp and power-amp.

4.      Exceptionally versatile tones from a single channel amp – anything from glassy clean to Vox-ish warmth to Tweed Deluxe raw to very heavy distortion (in other words, beyond “classic rock” levels).

5.      Able to get the immediacy and punch of a pentode push-pull output stage.

6.      Minimize the number of tubes, and minimize complexity.

 

Workings

To achieve the reasonably small yet gig-able goals, a small push-pull design was settled upon. With an efficient speaker setup, a 15 watt tube amp would suffice – for anything shy of keeping up with a “AS LOUD AS I CAN BE DUDE, WHY?” drummer.

 

The bulk of the versatility would come from two circuit elements that the authors had used in their prior respective builds: “dc-drive,” and a mids-control better described as a “character-control.” Other elements would be tried to further increase the ranges of feel and tonality: ultra-linear output vs. traditional output modes, and High B+ vs. Low B+ modes. In the Low B+ mode, the output tube were to be biased hotter to better get the “Tweed” resistive-cathode biased output stage feel.

 

A few words on the more novel aspects of this amp follow, dc-drive and the Middle Control. In an effort to keep this text as short and simple as practical, the more traditional circuit elements of this amplifier (i.e. ultra-linear output stage, lower-voltage/higher-current output stages vs. higher-voltage/lower-current output stages, preamp master volumes) should be web-searched by the reader. Many great descriptions of these “path more taken” circuit features are available via web-searches.

 

DC-Drive

The term “dc-drive” refers to the coupling between two tube stages and the biasing of the latter of the two stages. It has appeared in the “Taz Tweed,” “Raffler,” “Blue,” and now “Melissa” designs. The technique came about as the results of bench-top experiments to minimize blocking distortion in guitar preamps. Blocking distortion is widely held by players of classic-rock to heavy-metal to nu-metal as an unmusical distortion that must be minimized.

 

This amplification approach has the following features:

 

  1. It is blocking-distortion resistant. In this approach, the traditional AC coupling-path from a plate to a grid is augmented with a DC coupling path as well. The DC path is formed by a resistive voltage divider from the plate to the grid. The result is that the temporary bias-offset brought about with grid-conduction charging the coupling-cap is greatly reduced. Two things bring this about:

 

    1. Because the grid is at a higher dc bias voltage, a much larger cathode resistor is used to keep the grid to cathode voltages at a good bias level. This larger cathode resistor greatly reduces the coupling-cap charging, and so the temporary bias-offset is greatly reduced.
    2. Because the effective resistance that governs the discharge of the coupling cap is reduced, the discharge time is greatly reduced – and so the temporary bias-offset is even further reduced.

 

  1. It creates the potential for very large gain variation from the dc-driven stage. Because of its large cathode resistance, the gain of the dc-driven stage can be reduced typically to 1.3 by simply removing the cathode-bypass capacitor commonly found in these stages. (Web-search “cathode degenerative feedback” for explanations on this.) With the cathode-bypass cap restored, the full “open-loop” gain of the stage is achieved, typically anywhere from 40 to 60 for 12AX7 triodes. As such, a simple potentiometer in series with the cathode-bypass cap effects a very wide-range gain control.

 

  1. Set for high-gain, it can produce smooth, deep distortion. Set for low gain, it produces a warm, slightly compressive response for large input signals.

 

    1. For high-gain operation of the dc-driven triode (i.e., cathode resistance being capacitor bypassed), the slight compression from the large cathode resistor and minimized blocking distortion can create a remarkably smooth, deep distortion from a single stage. There are limits, when pushed too-hard, such a stage sounds as if it “loses focus” – not so much sounding like a traditional blocking effect as a loss of clarity.
    2. For low-gain operation of the dc-driven triode (i.e., cathode resistance without a bypass capacitor), the negative feedback from cathode to grid partially linearizes/cleans-up the output on the plate from a large input signal. For the larger input signals, the stage is temporarily biased slightly colder. The negative feedback from cathode to grid, and the dc-drive characteristics described above reduce this temporary cold-biasing effect, so that a non-cutoff, yet reduced gain operation region exists for a band of high-level input signals.. As such, the result is a slightly compressed response when driven by the right input levels.

 

  1. It creates a preamp stage that biases well over a fairly wide range of DC at the grid. Because the grid-to-ground DC voltage is much larger than the grid to cathode DC voltage, large changes in DC voltages at the grid translate into small changes in DC grid-to-cathode voltages. As such, the stage bias changes much less for a given DC voltage change at the grid than it would for a traditional tube preamp stage. The benefit here is that a wide range of DC plate voltages at the prior stage can be accommodated (and so the bias point is relatively resistant to changes in tube types, manufacturer, batch characteristics, etc.), and that the correct bias conditions don’t change much for a relatively wide range of B+ voltages

 

Middle (‘Character’) Control

The Middle Control mixes the response of a mid-range notch filter and a mid-range accentuate circuit. At one end of the control, a wide mid-range “hump” centered at about 1.5kHz dominates. This response is a natural for the “warm” sounds of Vox or Matchless, and when placed before a distorting gain stage, it’s a natural for the tough distortion sounds of a cranked Marshall. At the other end of the control, a 300Hz notch dominates, which is very characteristic of blackface Fender designs (e.g.., Deluxe Reverb, Twin Reverb, etc.). Sweeping the control between these two extremes “morphs” the tone response from the one extreme to the other. At about 2/3 of full-scale, the control helps voice the amp in a fashion similar to a Fender tweed-era amp. This tone shaper might better be called a “Character Control.” In addition to lending a wide variety of responses, it also features very low signal attenuation in its pass-band - which is very helpful, as will be discussed later.

 

This tone shaping control, when coupled with the clean-to-deeply distorted responses available from dc-driven design approaches, makes for an extremely versatile preamp channel. Many of the goals listed at the top of these notes were achieved in the Melissa design by combining these two features into a single preamplifier.

 

Melissa Circuit Topology

To date there are two variants of the Melissa design: the full Melissa and the Melissa Lite. The Melissa Lite is a simplified version of the full Melissa, without output stage bias controls or ultra-linear output stage mode. The Melissa Lite schematic is presented (and discussed) first:

 

 

Figure 1 - The Melissa Lite Schematic

 

Power Supply

The power supply is relatively standard for a tube amplifier, including a tube rectifier for the compressed response they lend at loud levels of play. Of interest are the two B+ modes for the amplifier: Low and High. When the power supply is set to High B+ mode, the most power and punchiest response is achieved for louder levels of play. When set to Low B+ mode, the output power is lowered/quieter, and cranked-up settings become more compressed while getting a warmer characteristic feel. Another output stage cathode resistor is placed in parallel with R29, increasing the output stage bias current for this mode. In both modes, maximum idle plate dissipation for the 6V6GTs (14 watts each) are observed.

 

Preamp Distortion

There are two dc-driven stages in this amplifier: the second stage (V1b), and the phase splitter (formed by V2a and V2b).

 

At the second stage, the Input Level control sets the signal level (AC) getting to the grid, while R5 and R7 form a DC voltage divider which presents a fraction of the first stage’s plate voltage to the second stage grid for biasing purposes. At maximum gain, almost all of the signal at the first stage plate appears at the second stage grid, allowing the second stage to be driven into very deep distortion. DC-drive characteristics described before make any signal attenuation to prevent blocking unnecessary, and provide smooth, deep distortion.

 

At the phase splitter, very large signal levels appear at its input when the Gain and Input Level controls are turned up. These levels overdrive the splitter and DC-drive was used to make any attenuation from the second stage unnecessary - achieving greater distortion from this stage than would normally be possible. The Middle Control features very low attenuation (for maximum distortion at the splitter), and a DC voltage divider running through it (for correct dc-drive biasing of the phase splitter).

 

Bass, Middle, Treble, and Bright Controls

The Bass Control appears where it can do the best job in removing any residual blocking distortion: before the second stage. Its works based upon cathode degenerative feedback of only the bass frequencies at the first stage lowering the gain for only bass frequencies. A small cathode bypass cap (C1) is always present, giving midrange and treble frequencies the maximum gain available from that stage. When the Bass control is lowest, degenerative feedback is maximized for bass frequencies, and the bass at the output of this stage is at its lowest. Alternately when the first stage cathode is fully cap-bypassed (via maxing the Bass Control), the bass frequencies amplification becomes “open-loop,” and is as much as that for the midrange and highs. The use of a current limiting diode (D1) at the cathode of V1a (instead of the normal 1.5k resistor or such) allows for greater Bass cut: D1 acts like a normal cathode resistor for the purposes of biasing the first stage (they both would give the correct cathode current). But because it takes a very large change in voltage across the diode to create just a small change in current through it, for audio (guitar) signals it acts like a resistor in the hundreds of thousands of ohms. This in turn makes the Bass control much more effective than if a cathode resistor were used.

 

The Middle Control is actually a mid-cut notch filter (formed by C9, R9, and R12), and also a band-pass filter, (i.e., mids being passed while highs and lows are cut, being formed by C8, R10, and C13). The output of these two filters are mixed together via P4, the overall output appears at P4’s wiper. So, when the Middle Control is fully counter-clockwise the wiper is connected to the mid-cut notch filter, and when the Middle Control is fully clockwise the wiper is connected to the band-pass filter. Various responses are achieved by sweeping P4’s wiper from CCW to CW. The output of P4 is then mixed in with a bit of highs (C12, R11, and R13) to give the output just the right amount of highs when the Middle control is fully CW.

 

The Treble Control appears where it can do the best job in taming any excessive highs present from heavily distorting preamp stages: after the phase splitter and master volume. It is a simple low-pass filter with its control pot allowing more or less treble frequencies to be filtered out before reaching the output tube grids.

 

The Bright Control is most effective when the Gain control is set low. This was done for a purpose, and happens due to feedback for the lows and mids (as set by C8 and P3) becoming smaller as the gain is increased, finally equaling that of the bright frequencies (as set by C7 with the Bright switch on).