Technical Analysis of the Matisse Fantasy Mk II Tube Preamplifier

Published by IWISTAO

Introduction

The Matisse Fantasy Mk II is a high-end stereo tube line preamplifier revered for its blend of classic circuit design and refined execution. The circuit itself is relatively straightforward – it uses a pair of dual-triode tubes (two 12AX7/ECC83 and two 12AT7/ECC81 in total for stereo) configured as two cascaded gain stages per channel. As the original designer noted, the Mk II “breaks no new ground” in topology, instead achieving excellence through quality components, careful layout, and power-supply engineering to minimize noise and RFI interference. The result is a preamplifier with approximately 26 dB of voltage gain, very wide bandwidth, and low distortion, yet one that preserves the sonic nuance of vacuum tubes. In this article, we provide a detailed technical look at the Fantasy Mk II’s design: its two-stage amplifier topology, key component values (and how they influence performance), power supply architecture, signal flow, output impedance characteristics, and noise-handling strategies. We’ll also compare the Matisse design to other typical tube preamp topologies and highlight what makes it unique. The discussion is rigorous but aimed at experienced electronic engineers, so expect proper terminology and a focus on how design choices impact sonic performance.

Circuit Topology: Two-Stage Cascaded Amplifier with Feedback

At its core, each channel of the Matisse Fantasy Mk II is a two-stage common-cathode tube amplifier using one 12AX7 and one 12AT7 triode in series. The first stage (12AX7) is a high-gain voltage amplifier, capacitively coupled into the second stage (12AT7), which serves as a voltage driver and output stage. Notably, no cathode follower stage is used for buffering; the output is taken from the plate of the second triode via a coupling capacitor. Instead of an extra buffer tube, the design employs a global negative feedback loop (series-shunt feedback) from the output of the second stage back to the input of the first stage. This topology both stabilizes the gain and lowers the output impedance and distortion of the overall preamp. The closed-loop gain is set to about 20× (26 dB), significantly less than the raw open-loop gain of the two triodes, indicating a moderate amount of negative feedback is applied. The feedback is applied in a series manner at the first stage cathode, meaning the output of the 12AT7 (after the output coupling capacitor) is fed via a resistor network into the 12AX7’s cathode, thereby degenerating the first stage based on the second stage’s output. This global feedback linearizes the frequency response (the Fantasy boasts a nearly 1 MHz bandwidth in some implementations and ~10 Hz–90 kHz ±1 dB in published specs and reduces distortion (total harmonic distortion <0.05% is reported. It also dramatically lowers the effective output impedance seen at the output jacks, from what would be tens of kiloohms down to a few kiloohms – all achieved without a dedicated cathode-follower tube.

Figure: Simplified schematic of one channel of a Matisse Fantasy Mk II style preamplifier (Analog Metric reference design). The first stage 12AX7 (left) and second stage 12AT7 (right) are coupled via C1 and utilize a series-shunt feedback network (e.g. resistor Rf ~200 kΩ from the 12AT7 plate output back to the 12AX7 cathode). Key components include the input grid stopper (≈500 Ω), grid leak bias resistors (e.g. 600 kΩ at the 12AX7 grid, 470 kΩ at the 12AT7 grid), cathode resistors (with bypass capacitors), plate load resistors (220 kΩ on 12AX7, 100 kΩ on 12AT7 in this design), and the coupling capacitors (C1 inter-stage, C3 output). The B+ high-voltage supply (≈420 V) feeds the plates through load resistors, and the heaters (not shown) are powered by 12.6 VDC. (Circuit reference: Analog Metric “MAT Fantasy” design)( manualzz.com, diyaudio.com).

Each stage is biased in common-cathode configuration with the usual supporting components: a cathode resistor for bias (bypassed by a capacitor for AC gain in at least the first stage), a high-value resistor from grid to ground (grid leak) to set the DC operating point and input impedance, and an anode (plate) load resistor to the B+ supply. The first stage 12AX7, being a high-mu (μ≈100) triode, provides the majority of the gain. The second stage 12AT7 (μ≈60, with higher transconductance and lower plate resistance than a 12AX7) adds additional gain but is also chosen for its ability to drive the output load at lower impedance. Both stages contribute to the overall amplification, after which the output signal is capacitively coupled to the next component (power amplifier or other load). It’s worth noting that the Fantasy Mk II is a line-level preamplifier (not a phono stage), so it handles signals on the order of a few hundred millivolts up to a couple of volts. With ~26 dB of gain, the Fantasy can output several volts RMS; in fact, one clone’s specs cite an output up to 8.5 V RMS before clipping, indicating substantial headroom thanks to the high-voltage B+ supply.

Feedback Loop: The negative feedback (NFB) network in this design is a hallmark of the Matisse topology. A resistor (on the order of 180–220 kΩ in the clone schematics) feeds a fraction of the 12AT7’s output voltage back to the 12AX7’s cathode (which is the inverting input point for a common-cathode stage). Often a second resistor (or the cathode bias resistor itself) forms a divider with the feedback resistor to set the exact feedback fraction. The use of series-shunt feedback means the output voltage is fed back in series with the input signal (at the cathode of the input stage) and sampled as a shunt from the output node (the plate of the output stage) – a classic voltage feedback topology for single-ended amplifiers. This approach reduces gain, output impedance, and distortion, while extending bandwidth. Indeed, the Analog Metric documentation for a Matisse-derived kit lists 26 dB gain and an output impedance of ~4 kΩ, significantly improved from an estimated >10 kΩ output impedance without feedback (one builder measured around 14 kΩ open-loop output Z in a variant of the circuit) (diyaudio.com). The feedback also flattens the frequency response and improves linearity: the Fantasy Mk II’s open-loop gain is high enough that even swapping the tubes for lower-mu types barely changes the closed-loop gain, as reported by users (i.e. the gain is “locked in” by the feedback network, not the intrinsic tube gain). This indicates a strong feedback factor, although not so excessive as to drive the amp into instability or completely suppress the tube’s sonic character.

To maintain stability with such wide bandwidth, a small compensation capacitor is often incorporated in the feedback network. In some builds, a capacitance on the order of 10–30 pF is placed across the feedback resistor to roll off high-frequency loop gain and prevent overshoot or ringing (diyaudio.com). In fact, DIY experimenters found that paralleling ~28 pF across the feedback resistor tamed a slight high-frequency ringing observed with a 1 kHz square wave (diyaudio.com). This minor compensation ensures rock-solid square-wave response and no ultrasonic oscillations, given the very high gain-bandwidth of the amplifier (which otherwise extends into the MHz). The inclusion of a grid stopper resistor at each input (discussed below) also contributes to stability by damping potential VHF/UHF resonances. Overall, the topology of the Matisse Fantasy Mk II is that of a high-gain two-stage amplifier with moderate global negative feedback, a design that achieves low distortion and low output impedance while still using a minimal number of active stages – a deliberate choice to keep the signal path short.

Key Components and Design Elements

Let’s examine the critical components in the Matisse Fantasy Mk II circuit and how each contributes to performance and sonic characteristics:

• Input Grid Stopper Resistors: A small resistor (typically around 500 Ω) is installed in series with the grid of each tube (e.g. between the incoming signal lead and the 12AX7’s grid, and similarly at the 12AT7’s grid). In the Fantasy clones, a 510 Ω carbon composition resistor is often used in this role. The grid stopper serves to dampen any high-frequency parasitic oscillations by isolating the tube’s input capacitance and the wiring inductance from high-speed transients or RF interference. Essentially, it forms a low-pass RC filter with the tube’s Miller capacitance, filtering out frequencies in the MHz range that could drive the triode into unwanted oscillation. Importantly, a grid stopper of this size has negligible effect on the audio band (for example, 500 Ω against a Miller capacitance of ~100 pF yields a cutoff around 3 MHz). Thus, it stabilizes the circuit without audibly affecting frequency response. Sonically, the presence of grid stoppers helps ensure the preamp sounds clean and stable under all operating conditions – they prevent ultrasonics or RF from sneaking in and causing distortion or interference. Some builders looking to maximize bandwidth have tried reducing these resistors (e.g. to 300 Ω or even removing them)(diyaudio.com), but such tweaks should be approached with caution: while a slightly lower value may further extend the already-wide HF response, the risk of oscillation or RFI susceptibility could increase if the grid stopper value is too low or omitted. The stock design’s values are a safe compromise that prioritizes stability and sonic purity.

• Grid Leak (Bias) Resistors: Each triode’s grid is referenced to ground through a high-value resistor that provides a return path for grid bias current and sets the input impedance. In the Fantasy Mk II, the first stage grid leak is on the order of 600 kΩ, which, combined with any input potentiometer or source impedance, defines a very high input impedance for the preamp. This ultra-high input impedance ensures minimal loading on source components, preserving the full frequency spectrum and dynamics of the source signal. (600 kΩ is far above typical line source output impedances, meaning virtually no current is drawn from the source.) The trade-off with such a large grid resistor is a potential for slightly higher hiss (thermal noise of a 600 kΩ resistor) and the possibility of grid bias wandering if the tube has any grid current or leakage. However, using a high-quality resistor here (e.g. metal film or a low-noise carbon film) mitigates noise, and the 12AX7’s grid current is negligible at proper bias, so the design benefits from the high impedance without significant drawbacks. The second stage 12AT7’s grid leak resistor was intended to be about 470 kΩ, a value confirmed by experienced builders of Matisse-derived kits. Interestingly, some clone PCB documents erroneously showed a 47 kΩ grid resistor for the 12AT7, which would drastically lower the impedance and alter the biasing – indeed a builder found that substituting the correct 470 kΩ value improved the sound and adhered to the intended design. With ~470 kΩ on the second stage grid, the inter-stage coupling forms a high-pass filter with the coupling capacitor (discussed next) that is well below the audio band. In summary, the grid leak resistors in the Fantasy are deliberately large to maximize input impedance and minimize low-frequency phase shift, contributing to an open and effortless sound. They set the bias point in combination with the cathode resistors, ensuring the tubes operate in their linear region.

• Cathode Resistors and Bypass Capacitors: The bias for each triode is established with a cathode resistor (forming self-bias via cathode degeneration). For the 12AX7 first stage, a resistor on the order of 1.5 kΩ is used, and for the 12AT7 second stage around 1 kΩ (exact values chosen to set a suitable idle current and plate voltage). These resistors serve two purposes: they establish the DC operating point (bias) by allowing a certain cathode current to develop a voltage (Vk) that biases the grid at about –Vk relative to cathode, and they also provide local negative feedback (degeneration) for AC signals if not bypassed by a capacitor. In the Matisse Fantasy Mk II, the first stage cathode resistor is bypassed with a fairly large capacitor (e.g. 100 µF) (diyaudio.com), effectively removing degeneration at audio frequencies and allowing the 12AX7 to achieve its full gain. Bypassing the cathode resistor maximizes open-loop gain (which is beneficial since the design then applies global feedback to reach the target gain), and it tends to increase the second harmonic distortion (since local feedback is removed) – a trait that can add a touch of warmth or richness to the sound. The large value of the bypass cap ensures the bypass is effective down to very low frequencies; a 100 µF cap with a 1.5 kΩ resistor yields a pole around 1 Hz, so essentially all audio frequencies are amplified without local cathode degeneration. The second stage cathode resistor in some implementations is partially or fully unbypassed, which means the 12AT7 has some local feedback to tame its gain and output impedance. The original design documentation does not explicitly state if the 12AT7 cathode was bypassed, but builders have experimented both ways. Leaving the second stage cathode unbypassed will reduce the stage gain and distortion, improving linearity at the cost of a bit less overall loop gain; bypassing it would maximize open-loop gain (hence allowing more feedback or more overall gain). Given the Fantasy’s high specified bandwidth and gain, it’s likely that both cathodes were bypassed in the stock design (to get maximum open-loop gain), with the global feedback then handling linearization. Some DIY enthusiasts have tried omitting the 12AT7’s bypass cap to see if it “sweetens” the sound by introducing a bit more second-harmonic character at that stage. In any case, the cathode resistor values and bypass choices affect the trade-off between gain and linearity. Sonically, a fully bypassed design (with heavy global feedback as in the Fantasy) yields a clean, dynamic sound with lots of gain on tap, while a partially degenerated design might sound slightly more relaxed or “vintage” at the expense of higher output impedance and less gain. The Matisse’s approach clearly favored achieving strong gain and low output impedance via feedback, leveraging bypass caps accordingly.

• Plate Load Resistors: High-value anode resistors connect each triode’s plate to the high-voltage B+ supply. These resistors convert the tube’s plate current variations into voltage swings (by Ohm’s law, V = I × R). In the 12AX7 first stage, the plate load is typically on the order of 220 kΩ–240 kΩ, and for the 12AT7 second stage it’s lower (around 100 kΩ). The exact values are chosen based on the tube’s characteristics and the available B+ voltage. With a B+ around 420 V, the designer can afford relatively large plate resistances while still leaving plenty of voltage headroom for the signal swing. A larger plate resistor yields higher gain (closer to the tube’s μ) and higher output impedance from that stage, whereas a lower value plate resistor would reduce gain but allow more current (and potentially swing into lower load impedances). The 12AX7, having a large internal plate resistance (~62 kΩ) and low idle current, often uses a large plate resistor to get high gain – 220 kΩ is a common value and appears to be used in this design. The 12AT7, with a lower plate resistance (~11–12 kΩ) and able to run at higher current, is served well by something like a 100 kΩ plate resistor, which balances a decent gain (perhaps 20–30 from that stage) with the ability to drive the next stage (which in this case is just the output load via a capacitor). Using a 100 kΩ load on the 12AT7 at a few milliamps would drop about 200–250 V, meaning the plate might sit around 150–200 V DC – a reasonable operating point under a 420 V supply. The high supply voltage is a luxury that allows these large swings without clipping. From a sonic perspective, the plate resistors (especially large carbon composition types often found in vintage tube gear) can impart subtle coloration (e.g. slight compression at high signal currents due to resistor non-linearity or thermal effects). However, many modern builds use precision metal film resistors here for lowest noise and distortion. Some builders have opted for boutique resistors (like Riken or Allen-Bradley carbon comps) in pursuit of a particular sonic flavor (diyaudio.com), as these can add a touch of the classic “tube sound” (minor second-harmonic warmth) to an otherwise very neutral circuit. Technically, the plate resistor values in the Fantasy are chosen to yield high gain and wide voltage swing, which contribute to the preamp’s ability to deliver dynamics effortlessly. The large plate loads also mean the output impedance of each stage is high (e.g. the 12AX7’s output Z ≈ Rp ‖ R_load ~ 60 k‖220 k ≈ 47 kΩ, the 12AT7’s maybe ~10 k‖100 k ≈ 9 kΩ). These high impedances are mitigated by the feedback loop, which makes the effective output much lower impedance (as seen by the load). But it’s also a reason why adding a cathode follower was considered unnecessary – with feedback, the 4 kΩ output impedance was deemed low enough for line-level use. In contrast, a design without feedback might have needed a follower to drive long cables or low-Z loads.

• Coupling Capacitors: The Fantasy Mk II uses capacitive coupling between stages and at the output to block DC. After the 12AX7’s plate, a coupling capacitor (let’s call it C_couple1) carries the AC signal to the 12AT7’s grid, and after the 12AT7’s plate, another coupling capacitor (C_couple2) feeds the signal to the output jack (and onward to the power amplifier). In a quality preamp, these capacitors are typically high-quality film capacitors chosen to be large enough to pass audio frequencies down to 20 Hz or below without significant attenuation. In the Matisse (and clones), one finds values around 0.22 µF to 0.47 µF for the inter-stage coupling cap and ~4.7 µF for the output coupling caps (diyaudio.com). For example, the Analog Metric kit schematic uses 0.22 µF between stages and 4.7 µF at the output, with film/foil polypropylene capacitors. These values create high-pass filters with the following corner frequencies (assuming 470 kΩ grid leak on the second stage and a typical 50 kΩ–100 kΩ power amp input on the output):

  • Inter-stage: f_cutoff ≈ 1 / (2π * 470kΩ * 0.22µF) ≈ 1.5 Hz. Even using 0.1 µF (as some modders experimented) yields ~3 Hz cutoff with 470 kΩ – still well below the audible range, meaning no audible bass attenuation but a slightly faster recovery from transients and less subsonic transmission. The choice of 0.22–0.47 µF ensures the coupling is essentially flat at 20 Hz (with only a tiny fraction of a dB rolloff).

  • Output: f_cutoff ≈ 1 / (2π * 50kΩ * 4.7µF) ≈ 0.68 Hz (with 50 kΩ load). Even into a lower 20 kΩ load, f_cutoff would be ~1.7 Hz. Thus the output coupling cap is generously sized to drive any realistic line input impedance with full bass extension. The large 4.7 µF polypropylene caps (visible as the big red rectangular caps on the PCB.

    
    

    contribute to the Fantasy’s renowned bass reproduction – there’s essentially no low-frequency phase shift or loss in the audible band.

  
  
  
  
  

  Figure: Physical implementation (DIY build) of a Matisse Fantasy preamp circuit. The large red film capacitors (4.7 µF, 630 V polypropylene) are the output coupling caps, responsible for passing audio to the next stage while blocking high DC (≈200 V) from reaching the output. High-quality parts are used throughout: note the discrete resistors at each tube socket, likely grid stoppers (blue resistors near tube grids) and metal film resistors for plate loads and cathodes. The 12AX7 and 12AT7 tubes (dual triodes) are the four glass valves shown. Using such premium components and a clean layout contributes to the Fantasy’s low noise floor and sonic transparency.

  The quality of the coupling capacitors is often discussed in tube audio circles because these caps lie directly in the signal path. The Matisse Fantasy’s design and later clones typically use high-grade film caps (polypropylene or polypropylene in oil) instead of electrolytics, even for the relatively large 4.7 µF value, to ensure minimal dielectric absorption and linear, low-distortion performance. Some DIY builders have experimented with exotic coupling caps (like paper-in-oil types such as Jensen or Audio Note) to tweak the sonic flavor. These can impart a subtle warmth or “vintage” character, albeit sometimes at the expense of a bit of clarity. The stock design aims for transparency: polypropylene film caps are very neutral, preserving the crisp detail and wide bandwidth that the circuit is capable of. In terms of sonic impact, coupling caps can affect the bass tightness and overall clarity – a sufficiently large and high-quality cap will keep bass punchy and avoid any midrange veiling. The Fantasy’s use of large film capacitors helps give it a full-bodied yet controlled bass reproduction and a very clean transient response. Additionally, by setting the low-frequency cutoff below ~2 Hz, the preamp avoids phase distortion in the audible bass range and ensures that subsonic components (like turntable rumble, if any input is vinyl via an external phono stage) are largely passed through. Some designers might deliberately introduce a higher high-pass corner (say ~5–10 Hz) to filter out subsonic noise, but the Matisse philosophy appears to favor maximum signal integrity, leaving any filtering to other components if needed.

• Negative Feedback Network Components: We touched on the global feedback loop conceptually in the topology section, but in practical terms it consists of a few key parts: the feedback resistor (connecting output to the first stage cathode) and possibly a shunt resistor from the cathode to ground, and sometimes a small feedback capacitor. For example, the design may use a resistor around 200 kΩ from the 12AT7 plate output back to the 12AX7 cathode, and the 12AX7 cathode resistor (bypassed by a cap for AC) might be part of a voltage divider for feedback. In some layouts, a dedicated resistor (say 20 kΩ) from cathode to ground is used in parallel with the cathode bypass cap to set the exact feedback ratio (this would form a divider with the 200 kΩ feedback resistor). The feedback capacitor (as mentioned, ~20–30 pF) across the feedback resistor shapes high-frequency response and stability. These NFB components determine the closed-loop gain (~26 dB), the phase margin, and the reduction in output impedance. Because of the global feedback, the effective output impedance is roughly the original output Z divided by the loop gain. If the 12AT7 stage had ~10–15 kΩ output Z open-loop, and the loop gain is on the order of 3–4 (i.e. about 12 dB of feedback, given open-loop gain perhaps ~40 dB and closed-loop 26 dB), we’d expect output Z around a few kilohms – consistent with the cited 4 kΩ. This lower output impedance means the preamp can better drive cable capacitances and typical amplifier inputs without high-frequency roll-off or signal loss. It’s still advised to use reasonably short, low-capacitance interconnects (as any ~4 kΩ source would), but in practice 4 kΩ is low enough to handle 2–3 meters of cable and a 50 kΩ amp input with no issue. The feedback network also reduces distortion dramatically – that <0.05% THD figure is achieved by trading off the excess gain for linearity. The harmonic spectrum with feedback tends to be predominantly low-order (the tubes’ inherent 2nd harmonic is reduced but not eliminated, and higher orders are minimized). Sonic impressions of this configuration are that the Fantasy Mk II sounds clean, fast, and detailed, with a relatively neutral tonal balance compared to zero-feedback tube preamps. Listeners often note that while it uses tubes, it is not overtly “tubey” or lush; instead it offers a lively yet controlled presentation. This is the direct result of the design choices in the NFB loop – enough feedback to tame the tube nonlinearities, but not so much as to cause transient intermodulation or a sterile character. The use of a compensation capacitor in the feedback loop, as found necessary by builders, further ensures that the preamp remains unconditionally stable even with challenging loads (capacitive loads or the volume control at various settings). In summary, the feedback components are carefully chosen so that the Fantasy Mk II achieves the best of both worlds: the low distortion and flat response of a well-engineered amplifier, and the subtle harmonic “sweetness” of triodes operating in a comfortable, linear range.

Power Supply Architecture and Noise Handling

One often-overlooked aspect of preamplifier design is the power supply, yet it is crucial for low noise and consistent sonic performance. The Matisse Fantasy Mk II features a robust power supply design that provides both a high-voltage B+ for the plates and a low-voltage supply for tube heaters. This preamp was often built with an external power supply unit (PSU) to keep any stray hum fields away from the sensitive audio circuitry. The main requirements are: approximately 420 V DC B+ and 12.6 V for the heaters. The high voltage rail is relatively uncommon (many tube line stages run 200–300 V), but the 420 V allows the use of large-value plate resistors and yields high headroom, as discussed. In practice, builders have found the circuit will run on as low as ~260 V, but the original Mk II used the full 420 V for optimum performance.

B+ Supply: Providing 420 V DC in a quiet, stable manner requires either a well-filtered rectifier supply or a regulator (or both). The original Matisse design documentation is not public, but anecdotal evidence suggests it may have used a series-pass regulator or at least heavy decoupling. Some DIY versions integrate a high-voltage regulator using an LM317 (configured for high voltage with appropriate components) to trim the B+ to the desired 380–420 V range. Others have employed discrete or tube-based regulators (for example, using gas regulator tubes or a pentode pass element) in pursuit of ultra-stable voltage. The current draw of the preamp is modest (a couple of milliamps per triode, perhaps ~8–10 mA per channel), so the supply is more about voltage stability and low ripple than raw power. Ample reservoir capacitance and RC filtering are typically used: one builder noted using multiple C-L-C (capacitor-inductor-capacitor) sections (a choke-input filter) to reduce ripple to negligible levels. In fact, that builder experimented with two PSU configurations: one with a tube-regulated supply and one with a purely passive C-L-C-L-C filtered supply, switchable for comparison. Interestingly, he observed that the simple passive filtered supply gave a “very nice and smooth sound,” whereas the regulated supply was presumably more tight and analytical. This illustrates how the power supply can subtly influence sonic character: a stiff, regulated B+ might enhance bass control and transient precision, while a gently sagging, purely capacitive/choke filter can introduce a slight gentleness or “bloom” that some find musically pleasing. The Matisse Fantasy Mk II’s designers likely optimized the PSU for low noise and impedance. The presence of “two large reservoir decoupling capacitors” in the design specs implies that after initial rectification and filtering, additional big capacitors are placed near the circuit to decouple the B+ for each channel or each stage. Indeed, in the analog metric PCB there are positions for large electrolytic caps per channel. Moreover, each stage is usually fed from B+ via a dropping resistor (RC decoupling) to isolate the first stage from any second stage fluctuations. This would mean the 12AX7 gets its B+ from a node after an additional RC filter, providing further isolation and stability (and preventing the output stage’s current draw from modulating the supply of the input stage, which could introduce coupling or instability especially with feedback). In summary, the B+ architecture is a high-voltage, low-ripple supply with likely under 1 mV of ripple, which is critical to achieving the Fantasy’s >85 dB signal-to-noise ratio. All high-voltage supply nodes are decoupled and bypassed (often a large electrolytic || a small film cap for high-frequency bypass) to ensure a low noise floor and stable operation.

Heater Supply: The 12AX7 and 12AT7 heaters can be wired for 12.6 V (each tube’s two 6.3 V heater halves in series). The Fantasy Mk II uses a 12.6 V DC heater supply for all four triodes (likely drawing on the order of 300–600 mA total). Using DC for heaters is a deliberate choice to eliminate heater-induced hum. AC heater drive is simpler but can easily introduce 50/60 Hz hum into sensitive high-gain stages, especially with a 12AX7’s high gain. By rectifying and regulating the heater supply to 12.6 V DC, the Matisse ensures that the filament does not inject any ripple into the cathode or surrounding circuitry. The heater supply in some clone designs is derived from a separate transformer winding feeding a diode bridge and a 7812-type regulator or similar, producing a stable 12 V (some run the heaters slightly under 12.6 V to extend tube life and further reduce noise). Additionally, because the 12AT7 in the second stage might have a cathode potential of a few volts, there’s usually no need to elevate the heater reference in this design (heater-to-cathode voltage difference stays within safe limits even with DC). Nonetheless, some high-end implementations do elevate the heater supply to ~40 V above ground to minimize any heater-cathode leakage effects in the first stage, further reducing hum. In terms of sonic impact, a DC heater supply contributes to the jet-black background and low noise of the preamp – you won’t hear the faint buzz that some AC-heated tube preamps exhibit. It also avoids adding any intermodulation that heater ripple could cause in the audio band.

Grounding and Layout: The Fantasy Mk II pays special attention to grounding scheme and PCB layout to reduce noise. The design advertises a “symmetric layout” and “dedicated ground and power rails” for minimal parasitic coupling. Likely, this means the PCB is laid out in a dual-mono fashion (left and right channels physically mirrored), and the ground is arranged in either a star or low-impedance bus configuration to prevent ground loops. A star ground brings all critical return paths to a single point, minimizing hum. In the photo of a build above, one can see a thick ground trace connecting major points and large ground plane areas around the signal circuitry.

. Engineers will recognize that high impedance nodes (like the 600 kΩ grid resistors) are very susceptible to interference; thus, the wiring around the input stage is kept short, and likely the chassis provides some shielding. The tubes themselves are usually placed such that their input pins are near the input RCA jacks, etc., to keep the paths short. The use of tube sockets on a PCB requires careful consideration of layout to avoid oscillation – hence the grid stopper resistors are placed as close as possible to the grid pins (often directly on the socket). Also, high-voltage capacitors (like coupling and supply caps) are spaced to avoid coupling into low-level lines.

From a noise perspective, the Fantasy Mk II is specified at >85 dB S/N (A-weighted), which for a tube line stage with 26 dB gain is quite good. This implies a low microphonics tube selection (12AX7s can vary in microphonic behavior; likely the design calls for selected low-noise tubes for the first stage) and excellent filtering of the power. Any residual hum at the output would be in the microvolt range, effectively inaudible. The combination of DC heaters, ample filtering, and feedback (which can also reduce low-frequency hum by the loop gain factor) yields a noise floor that is largely a non-issue even in high-sensitivity systems. The open-loop design might have a bit more hum or drift, but once the loop is closed, any B+ noise is attenuated as well (since the feedback will correct low-frequency error to an extent). It’s a testament to the design that despite using high-μ tubes like the 12AX7, known to be prone to noise if not careful, the Fantasy Mk II manages a quiet operation. Proper dressing of input wires, quality shielded input jacks, and possibly input muting or relay soft-start circuits (to avoid pops at turn-on) are also part of the complete product. Some versions include a muting relay delay at startup – this is hinted by mentions of “Delay Protection” modules in related documentation, which would keep the outputs disconnected until the tubes warm up and any DC transients settle.

In summary, the power supply and noise-reduction strategy in the Matisse Fantasy Mk II involve: a high-voltage supply either regulated or heavily filtered to deliver ~420 V with minimal ripple, a DC regulated heater supply for hum-free filament power, extensive decoupling and star grounding to eliminate ground loops and inter-stage coupling, and a thoughtful physical layout to reduce RFI and electromagnetic interference. Each of these choices contributes to the unit’s reputation for background silence and stability, allowing the sonic qualities of the amplifier circuit to shine without intrusion from hum or noise. For the engineer, it’s a reminder that a great circuit isn’t just about the schematic – it’s equally about how the power and ground are handled.

Typical Signal Flow and Operation

To put it all together, let’s walk through the signal flow in one channel of the Matisse Fantasy Mk II, highlighting how the audio signal is processed and how the design elements we’ve discussed come into play:

1. Input Stage (12AX7 common-cathode): The audio signal (from a source or the wiper of a volume potentiometer, if the preamp has one at the input) first sees a high impedance input network. The 12AX7’s grid is tied to ground through ~600 kΩ, defining the input impedance. A small series resistor (~510 Ω) sits right at the grid as a stopper. This stage is typically biased at a few volts cathode-to-grid; for instance, with a 1.5 kΩ cathode resistor and say ~1 mA of current, the bias would be ~1.5 V (keeping the grid at –1.5 V relative to cathode). The plate might idle around 150–200 V. When an AC signal enters, the 12AX7 amplifies it with a gain roughly around 50–60 (without feedback). The actual gain in circuit, with feedback, will be lower; but let’s consider open-loop first: The plate voltage swings in opposition to the grid signal, amplified by the tube’s transconductance and the plate load resistor. The cathode bypass capacitor on this stage ensures that AC gain is high; the cathode stays at nearly AC ground, so the stage operates with maximal gain on AC signals. The output of this stage is the plate of the 12AX7, which is connected to the next stage through the coupling capacitor. Before moving on, note that a portion of the output (from stage 2) is coming back into this stage’s cathode via the feedback resistor. That means the 12AX7’s cathode node carries a small version of the output signal, effectively subtracting from the input (negative feedback). Thus, the 12AX7 actually amplifies the difference between the input signal and the feedback signal. If, for example, the closed-loop gain is set to 20, the feedback network will feed back 1/20 of the output to the cathode, such that the 12AX7 amplifies until the output is 20× the input (so that the fed-back portion equals the input at the cathode and equilibrium is reached). This dynamic is continuously active and is what flattens the frequency response and stabilizes gain. The input stage is therefore a gain stage and error amplifier (in feedback terms) all in one.

2. Inter-stage Coupling: The AC signal from the 12AX7’s plate goes through a coupling capacitor (C1, ~0.22 µF). This cap blocks the ~150 V DC on the 12AX7 plate, but passes the changing AC voltage. On the far side of the cap, at the 12AT7’s grid, a high-value resistor (~470 kΩ) to ground provides a reference so that the grid has a DC bias point (usually at 0 V DC, since the 12AT7 cathode will be biased positive relative to ground). The coupling capacitor and this grid-leak resistor form a high-pass filter, but as noted, it’s set well below the audio band (~1–2 Hz). Thus, audio-frequency signals are transferred effectively, while any DC offset or very slow drift is blocked. By the time the signal reaches the 12AT7’s grid, it may have an amplitude of perhaps a couple of volts (if the input was, say, 100 mV, the 12AX7 might amplify it to a few volts). The feedback from the output is also effectively mixed at this point: the feedback resistor brings a bit of the output back to the 12AX7 cathode, which is effectively in series with the input signal source’s connection to the 12AX7 grid (through the cathode’s influence on grid-to-cathode voltage). In essence, the loop ensures that the 12AT7 grid receives just the right drive such that the output will be the amplified replica of the input (with gain 20). If the output tries to deviate (say at certain frequencies due to tube gain roll-off), the feedback alters the drive at the 12AX7 to compensate. This way, by the time the signal arrives at stage 2’s grid, it already carries the corrections imposed by feedback for any frequency response shaping needed.

3. Output Stage (12AT7 common-cathode): The second stage 12AT7 takes the AC signal at its grid and amplifies it further. The 12AT7 is biased typically around 1.5–2 V at its cathode (with ~1 kΩ cathode resistor, that implies ~1.5–2 mA of current). Its plate might sit around 200 V DC with a 100 kΩ plate resistor as earlier surmised. Without any feedback, this stage might have a gain of 20 (the 12AT7’s μ of 60 is loaded by the external resistance and internal plate resistance). With feedback in action, the effective gain of this stage is part of the overall loop and might be a bit less. The important point is that the 12AT7 provides additional voltage gain and power to drive the output. The output is taken from its plate, which has a relatively high impedance ( tens of kΩ as discussed). The AC voltage at the plate goes through the output coupling capacitor (C2, e.g. 4.7 µF) and then to the output jacks (typically RCA connectors). Downstream, the preamp is expected to see a load like a power amp input (commonly 47 kΩ or higher). The coupling cap ensures no DC from the 12AT7 plate (which sits at a couple hundred volts!) reaches the outside world – only the centered audio waveform goes onward. The 12AT7’s cathode resistor may or may not be bypassed: if it is unbypassed, the stage has some local feedback which reduces its gain and output impedance; if bypassed, it yields full gain. Regardless, the global feedback from the 12AT7’s output back to the 12AX7’s cathode is closing the loop: any difference between the actual output and the expected output (based on the input) will cause the 12AX7 to adjust its drive. For instance, if the output is too high (perhaps at a frequency where the 12AX7 had extra gain), the feedback will subtract more from the input drive, leveling it out. Conversely, if the output lags (say at high frequencies due to capacitances), the feedback lessens, effectively boosting the input drive to compensate. The result is that the final output is a very faithful, amplified reproduction of the input. The output signal coming off the capacitor is now ready to feed a power amp. With an output impedance of ~4 kΩ and typical cable capacitance, the high-frequency roll-off is negligible (a 4 kΩ source with 100 pF of cable forms a 400 kHz low-pass, far above audio). However, one should avoid extremely long cables or very low impedance loads – as is normal with tube preamps, the design is intended to drive loads ≥20 kΩ for best performance. In practice, virtually all solid-state and tube power amps have 47 kΩ–100 kΩ input impedance, so this is a non-issue. The preamp can easily deliver the necessary voltage swing (several volts) to drive amplifiers to full power, given its high headroom.

4. Volume Control (if present): The Matisse Fantasy Mk II line stage is typically used with a volume control (unless it’s a fixed-gain line stage in a larger system). Usually, the volume potentiometer is placed at the input, before the 12AX7 grid. For example, a 50 kΩ–100 kΩ logarithmic potentiometer might be connected such that the input jack feeds the pot, and the pot’s wiper goes into the 510 Ω grid stopper and into the 12AX7. The bottom of the pot goes to ground. This arrangement means the 12AX7 always sees a source impedance of at most 1/4 of the pot’s value (at mid-position) and the grid leak 600 kΩ is much larger, so it doesn’t significantly load the pot. With this scheme, when the volume is reduced, the input signal is simply attenuated; the feedback loop still ensures linearity at that lower level. One caveat is that at very low volume settings, the source impedance as seen by the grid can be higher (when the pot wiper is near the bottom, the series resistance from the top half of the pot might be significant). A high source impedance into the 12AX7 will interact with the Miller capacitance to roll off highs. For a 100 kΩ pot at mid, source Z ~25 kΩ; with C_miller ~100 pF, that’s a pole around 64 kHz, which is benign. Even at worst (50 kΩ source), it’s ~32 kHz – still above audible range. So, the design likely chose a pot value in this range to balance noise (higher pot => more Johnson noise) and roll-off. Alternatives could be a passive attenuator between stages, but that would not work well with the global feedback (since feedback spans both stages, you normally put volume either at the very input or at the very output; putting it in the middle would change feedback factor with volume position). Therefore, it’s safe to assume volume control is at the input in typical implementations. This means the high input impedance (600 kΩ) is mostly “swallowed” by the pot value (e.g. a 100 kΩ pot in front yields overall ~100 kΩ input impedance), but the specs in the manual might refer to the grid leak alone when stating 600 kΩ.

5. Behavior with Feedback: It is instructive to consider how the signal flow responds to certain conditions under feedback. For example, if a very low-frequency surge comes through (say a burst of 5 Hz from a turntable rumble), the coupling caps will start to roll it off, and the feedback will also see a phase shift – but the loop is designed with enough stability margin (likely dominant pole somewhere to ensure phase margin) that it won’t oscillate. The small compensating cap across the feedback resistor helps by adding a zero/pole that stabilizes HF phase shift. If a very large output is demanded (close to clipping), the loop will eventually run out of headroom and the output stage will saturate or grid-current limit; when that happens, feedback can no longer correct linearly and the preamp will clip, likely in a fairly benign way (softened by the 12AX7’s compression). Under normal operation, though, everything remains in the linear region, and the feedback just continuously corrects. Transient response is very fast – the rise time for a square wave can be on the order of a microsecond or two given the bandwidth (some measured the output square wave performance and with a small compensation cap, got a clean result with minimal overshoot So signal flow from an AC standpoint is wide-open and unimpeded through the amplifier, aside from the subtle guidance of the feedback loop ensuring it stays accurate.

Overall, the signal flow description underscores that the Matisse Fantasy Mk II operates like a precision amplifier for line-level signals: high input impedance, sufficient gain to bring line signals to a standard level, low output impedance to drive the next stage, and correction mechanisms (feedback) to maintain fidelity. Yet it does so with just two active stages in series, which is relatively minimal. Many other preamps might add buffers or tone control stages, but the Fantasy is a purist line stage with a “less is more” philosophy on the signal path, relying on quality components and design to get the job done with only the essential elements.

Output Impedance and Driving Capability

A critical parameter for any preamp is its output impedance (Z_out), which affects how well it can drive the input of a power amplifier and associated cables. The Matisse Fantasy Mk II’s output impedance is specified around 4 kΩ (with feedback in action). This is lower than many simple tube preamps that lack cathode followers, thanks in large part to the global feedback reducing it from what would natively be roughly the 12AT7’s plate resistance in parallel with its plate load (approximately 10–15 kΩ). While 4 kΩ is not as low as a solid-state preamp (which might be <100 Ω) or a tube preamp with a cathode follower (<600 Ω), it is sufficient to drive typical amplifier inputs (which are usually 47 kΩ or higher) without significant loss. When you connect the Fantasy Mk II to an amplifier with, say, a 47 kΩ input impedance, the preamp sees an easy load – it’s 47k in parallel with its own internal 4k, so the overall load on the 12AT7 is about 3.6 kΩ (which is actually dominated by the internal, hmm, but since feedback is present, we consider closed-loop output Z effectively). There is minimal attenuation (voltage divider effect is 47k/(47k+4k) ≈ 0.92, less than 1 dB loss) and the distortion remains low because the tube is not asked to source high current. The current demanded for a given signal is I = V_out / (load || Z_out). For example, to output 2 V into 47 kΩ, the current is ~0.04 mA – trivial for a 12AT7 that can source a couple of mA.

However, if one tried to use this preamp to drive an unusually low impedance, say a pro audio 600 Ω load or headphones directly, it would not be suitable. At 600 Ω, 4 kΩ output Z is far too high – the voltage division would be severe (only ~13% of the voltage would appear across 600 Ω, and the tube would have to source ~3.3 mA per volt, likely straining it). But such a scenario is outside the intended use; in home audio, loads below 10 kΩ are virtually never seen by preamps. For driving long cable runs, the 4 kΩ source impedance means one should be mindful of cable capacitance. A typical coaxial audio cable might have 20–30 pF/ft; a 10 ft cable could be ~200–300 pF. With a 4 kΩ source, that’s a low-pass with f_c ≈ 1/(2π4k250e-12) ≈ 160 kHz – entirely negligible for audio. Even a very long 20 ft run (maybe 500 pF) would still be ~80 kHz cutoff. So in practice, the Fantasy preamp can drive normal cable lengths without high-frequency loss. The relatively higher Z_out compared to cathode-follower designs does mean that the preamp will have a higher sensitivity to capacitive loading (i.e. adding 1000 pF might show a bit of overshoot or peaking if not compensated), but the inclusion of that small feedback capacitor was likely to keep the square wave clean under standard loading conditions.

Another aspect of output impedance is its frequency dependence. The output coupling capacitor (4.7 µF) and the load form a high-pass; at very low frequencies, the effective output impedance rises because the cap’s reactance becomes significant. At 20 Hz into 50 kΩ, the cap’s X_c = 1/(2π204.7e-6) ≈ 1.7 kΩ, which in series with 4 kΩ output Z gives a slight bass attenuation into very low loads – but into 50k, 1.7k is small. Into 20 kΩ, that reactance is ~1.7 kΩ and now the divider with 4 kΩ +1.7 kΩ vs 20 kΩ might cause maybe ~0.8 dB rolloff at 20 Hz – still not bad. Generally, the use of a large output cap mitigates frequency-dependent Z_out issues.

In contrast to the Fantasy’s design, many tube preamps include a cathode follower output stage explicitly to reduce output impedance to a few hundred ohms, making them more bulletproof into any load or cable. The Matisse design eschews the cathode follower, likely to avoid the additional active stage in the signal path (which could introduce its own distortion or sonic signature). Instead, the global feedback partially substitutes for a cathode follower’s effect by lowering Z_out by the loop gain factor. The result is admittedly a higher Z_out than follower-equipped designs, but within acceptable limits for its intended use. For example, Marantz’s classic Model 7 preamp used a cathode follower after a two-stage gain circuit, yielding ~600 Ω Z_out, and designs like the Audio Research SP series often have follower buffers. The Fantasy’s ~4 kΩ is higher, but users have generally not reported issues driving standard amps; it’s part of the design trade-off that avoids the potential “drive hardness” some audiophiles attribute to cathode followers. In subjective terms, some listeners find that eliminating the cathode follower can lead to a more “direct” or transparent midrange, presumably because there’s one less coupling capacitor and one less stage’s distortion in the chain (though a well-designed cathode follower is very linear). The feedback ensures that, despite not having a dedicated buffer, the Fantasy Mk II still behaves in a well-mannered way when connected to typical downstream equipment.

One more consideration is that the output impedance interacts with the feedback loop. Because the feedback is taken from the output (after the coupling cap), if the load impedance is too low, it effectively changes the loop gain (a heavy load will attenuate the output, and the feedback will sense a lower voltage, causing the amp to work harder). Up to a point, the loop will compensate (in fact, feedback will source more current to keep output voltage up, which is good). But if the load is extremely challenging, the headroom may be consumed. In normal use, as described, that’s not an issue, but it’s conceptually why one shouldn’t, for example, directly hook a passive 10 kΩ volume control at the output and then drive a 10 kΩ amp – that would be 5k load combined, the preamp could still manage but it’s less optimal. It’s always best to have the volume pot at the input in this design, preserving a nice high input impedance to the pre and presenting a high stable impedance to the output.

In conclusion, the Fantasy Mk II’s output stage is designed to comfortably drive the high impedance inputs of power amplifiers, delivering the necessary voltage swing with low distortion. The ~4 kΩ output impedance is a deliberate choice that balances simplicity and performance: it’s low enough for intended applications, but avoids adding more circuitry that might detract from purity. Users should treat it like any classic tube preamp – use reasonably short cables and standard amp loads – and it will reward them with coherent, transparent sound and solid bass (thanks to those hefty coupling caps). The design’s cult following in high-end audio circles suggests that in real-world usage, the output drive is perfectly adequate and the sonic benefits of the simpler signal path are well appreciated.

Comparison to Other Tube Preamp Topologies

It’s insightful to compare the Matisse Fantasy Mk II’s design choices against other common tube preamp topologies, as this highlights its unique traits and engineering philosophy:

• Versus a Cathode Follower Output Stage: As mentioned, many tube preamps incorporate a cathode follower after the gain stage to buffer the output. For example, a classic topology is: 12AX7 (gain stage) -> 12AU7 or 12AT7 (cathode follower). This yields a very low output impedance (hundreds of ohms) and high current drive, but the cathode follower itself adds no gain (unity gain stage) and introduces its own distortion (though mostly 2nd harmonic) and can impart a particular sonic character (some describe CF-equipped preamps as a touch “smoother” or, conversely, less “immediate” depending on the implementation). The Matisse Fantasy avoids a follower, instead using feedback to improve output impedance. Sonic trade-off: The Fantasy might sound more open or detailed by skipping the follower, whereas a follower design might handle difficult loads with more ease. However, since in hi-fi the loads are gentle, the Fantasy’s choice prioritizes signal purity over brute-force drive. Interestingly, the global feedback in the Fantasy accomplishes much of what a follower would (low distortion, lower output Z) but while still allowing the second stage to provide gain.

• Versus SRPP / Totem-Pole Circuits: Another popular topology for line stages is the SRPP (Series Regulated Push-Pull) or totem-pole arrangement (as seen in some Audio Research and DIY designs), often using two triodes of a dual tube (like two halves of a 6DJ8) in a stacked configuration. These circuits can provide a decent gain and low-ish output impedance without global feedback by operating the two triodes in a push-pull-like manner for signal swings. For instance, a 6DJ8 SRPP might have an output Z of ~1 kΩ and a gain of ~20 dB, with fairly low distortion if loaded correctly. Compared to the Matisse, an SRPP uses no global feedback (it’s essentially open-loop but has built-in balancing). The Matisse’s use of feedback versus the SRPP’s use of symmetry reflects different approaches: the SRPP relies on device matching and even-order cancellation for linearity, whereas the Matisse uses feedback to linearize. One notable difference is output driving: SRPP can source and sink current actively, whereas the Matisse’s single-ended output can source more easily than sink (the coupling cap provides the path for sourcing current into a load; for sinking, it draws from the load into ground through the cap on the negative swing – either way fine for AC, but the SRPP might drive a heavier load with more ease up to a point). Sonic comparison: A good SRPP can sound very dynamic and punchy, often with a “musical” flavor due to some residual second harmonic, whereas the Fantasy with feedback might sound a bit cleaner and more neutral. The Fantasy’s distortion is likely lower across the board due to feedback, while an SRPP might have a bit more harmonic content (some audiophiles like that for a euphonic touch). The choice of 12AX7/12AT7 in the Fantasy also differs from the low-μ high-Gm tubes often used in SRPP (like 6DJ8, 5687, etc.), again showing the Fantasy’s design lineage is more classic (12A_7 series tubes were used in vintage preamps like the Marantz 7, etc.).

• Versus Fully Passive or Transformer-Coupled Designs: Some modern high-end preamps forego active gain altogether for line level, using autoformers or transformers to attenuate or couple signals, or they might use a single triode with a transformer output. For example, one might compare it to a design that uses a single 6SN7 triode with a transformer to get low output impedance. Those designs typically have no feedback and rely on the inherent linearity of a triode and the bandwidth of a transformer. The Fantasy uses no transformers in the signal path – all coupling is capacitor-based. This avoids any core saturation or bandwidth limit that a transformer might introduce, and it keeps the cost/size down. However, transformer-coupled preamps can have extremely low output impedance (the transformer's secondary can be, say, 600 Ω or less) and block DC inherently. The tonal differences might be that a well-made transformer preamp sounds very smooth and weighty (transformers often add a bit of bloom or “body” in the bass), whereas the Fantasy’s capacitor coupling might sound faster, with crisper transients. Again, these are subjective impressions, but the engineering difference is feedback vs no feedback, and capacitor vs transformer coupling – each with its pros/cons. The Fantasy’s measurable performance (THD, bandwidth) would likely surpass a transformer-coupled single-triode stage, at least on paper, due to the linearizing effect of feedback and the absence of transformer-induced phase shifts at frequency extremes.

• Versus Multi-Stage High-Gain Preamps (w/ Global Feedback): Some preamps, particularly ones with integrated tone controls or very high gain (like vintage designs that also needed to amplify low-output sources), might use 3 or more triode stages and global feedback around them. For instance, certain vintage preamps or modern ones with op-amp hybrid designs will have lots of gain and then lots of feedback to tame it (some even approaching op-amp like topologies). The Fantasy’s simplicity stands out in that it’s just two stages and one feedback loop. This minimalism likely contributes to its transparent sound – with fewer active stages, there are fewer cumulative phase shifts and potentially a simpler harmonic structure. Multi-stage feedback amps can sometimes suffer from complex distortion spectra or stability concerns if not carefully designed. The Fantasy’s designer kept it simple: two stages are enough to get the desired gain and allow a moderate feedback factor. So while it’s a feedback amp, it’s not a complex one. This approach is more akin to classic hi-fi tube practice (the Dynaco PAS preamp, for example, also used two triodes with feedback, albeit with tone controls in the loop; the Marantz 7 line stage used a two-stage feedback loop as well). One could say the Fantasy is a refined descendant of those classic designs but executed with modern components and higher supply voltage for better performance.

To illustrate the comparison, here’s a summary table contrasting the Matisse Fantasy Mk II with a few typical design approaches:

Aspect	Matisse Fantasy Mk II (12AX7→12AT7 w/ NFB)	Typical Cathode Follower Preamp (e.g. 12AX7→12AU7 CF)	SRPP / No-Feedback Preamp (e.g. 6SN7 SRPP)
Stages & Topology	Two gain stages, global series-shunt feedback applied. No cathode follower; output from plate via cap.	Two stages: one gain, one cathode follower buffer (no global feedback, or sometimes local feedback only).	Two active triodes in SRPP (or similar) config acting as one compound stage. No global feedback (uses internal symmetry).
Tubes Used	12AX7 (high-μ voltage amp), 12AT7 (medium-μ driver) per channel. Both dual triodes used (one half per channel each).	Often 12AX7/12AU7 or 6DJ8/6DJ8 etc. (high-μ for gain, low-μ for follower). Follower tube chosen for current capability.	Usually identical dual triodes (e.g. two halves of 6SN7, 6DJ8, etc.) in a single bottle per channel.
Gain (typical)	~26 dB (x20) closed-loop. Open-loop much higher (~60 dB), but reduced by feedback. Very flat frequency response (±0.1 dB in audio band).	~15–20 dB from gain stage (follower is unity gain). Possibly slight high-frequency roll-off if follower has Miller capacitance loading first stage.	~12–20 dB (depends on tube and config). Response can have slight bass roll-off if output cap smaller, HF extension good if using high-Gm tubes. No feedback-induced gain shaping, so response follows tube’s natural response (often fine to >100 kHz).
Output Impedance	~4 kΩ (with feedback). Without feedback ~15 kΩ (from 12AT7 plate). Can drive ≥20 kΩ loads comfortably.	Very low, ~200–600 Ω typical (cathode follower output impedance ≈ 1/gm of follower tube). Able to drive 5–10 kΩ loads or long cables easily.	Moderate, ~1–2 kΩ typically (SRPP has lower Z_out than single triode but higher than CF). Drives ≥50 kΩ loads well; heavy loads reduce its swing and linearity.
Distortion Profile	THD <0.05%, dominated by low-order harmonics (feedback reduces most distortion). Very low output noise/hum. Sound is clean/neutral with slight tube warmth (due to some 2nd harmonic left).	THD ~0.1–0.3% open-loop (from gain stage), possibly lower overall if local cathode feedback used. Follower adds tiny distortion. Harmonic profile mostly 2nd, some 3rd. Usually very quiet if tubes are good and layout is proper. Sound tends toward classic “tube with a bit of warmth” or very neutral if using lots of local degeneration.	THD ~0.1–1% depending on operating points. Often a richer harmonic profile (2nd dominant, some 3rd) because no feedback to cancel it. Noise floor can be very low (few stages, no global loop to amplify noise). Tends to sound more “lush” or “euphonic,” with less emphasis on absolute neutrality.
Power Supply Approach	High-voltage (~400 V) supply for max headroom. Often regulated or well-filtered; dual-mono decoupling. DC heater supply for low hum. Focus on RFI rejection and stable rails.	Medium B+ (200–300 V) often suffices since only one gain stage needs high voltage. Some use unregulated RC filter, others use regulated B+ for lower noise. Heaters sometimes AC if hum can be canceled, otherwise DC for first stage.	Similar B+ ~250–300 V typically. Usually CRC or CLC filtering (no global feedback to correct PS noise, so need clean supply). AC heaters can be used with careful grounding, or DC if chasing lowest noise.
Unique Traits	- High input impedance (600 kΩ), good for not loading sources.

• Global NFB in a tube preamp (many hi-fi tube preamps avoid NFB; Matisse embraces it to improve linearity).

• Very high bandwidth (reported up to 1 MHz) – indicative of extremely careful design and feedback compensation.

• Uses “classic” audio twin-triode tubes in a refined way, hinting at designs like Marantz 7 lineage but with modern touches. | - Inclusion of cathode follower is common in mainstream designs (e.g., Conrad-Johnson, McIntosh vintage pre’s), prioritizing compatibility with any load.

• No global feedback (in many cases) means harmonic character of tube is more audible, potentially more “tube bloom.”

• Simpler to implement, but possibly measures worse (higher THD, less flat FR if not designed well). | - Minimal stage count and often no feedback yields a very direct signal path (some say more “immediacy” in sound).

• Often lower gain; just enough to cover passive losses and provide some boost.

• Output impedance lower than single-triode but still not as low as CF or solid-state – a compromise design.

• Favored in DIY circles for its simplicity and “vintage” sound. |

This comparison shows that the Matisse Fantasy Mk II is somewhat unique in combining a high-gain tube topology with global negative feedback in a line preamp. By contrast, many later high-end tube preamps (especially those from the late 1990s and 2000s) actually moved away from feedback, aiming for zero-feedback designs, or they employed buffer stages to ensure load drive. Matisse (the designer) took a more engineering-driven approach: use feedback to improve performance, but do it in a simple two-stage loop to preserve the musicality. The result is a preamp often described as having the clarity and low distortion of a more modern design, yet the dimensionality and life of a tube preamp. This balance is not easy to achieve, and it speaks to the thoughtful choices in this circuit.

Another noteworthy trait is the choice of 12AX7 and 12AT7. Many high-end preamps opt for tubes like 6922/E88CC (for their low noise and high transconductance) or 6SN7 (for their linearity and classic tone). The 12AX7 is a high-mu tube typically seen in phono stages or guitar amps; using it in a line stage is somewhat old-school. But in the Fantasy, the 12AX7 is operated at a high voltage and low current, where it can be quite linear, and the feedback further linearizes it. The 12AT7 is also an interesting choice – it has higher Gm and can drive better than a 12AX7, and has more gain than a 12AU7. It’s sometimes criticized for being a bit less linear than a 12AU7 at audio frequencies (it was originally often used in RF circuits), but again, in this design it’s harnessed well: the feedback likely cleans up any minor non-linearity. The combination of these two common tubes means replacements and tube rolling are easy, and indeed some users experiment with different brands or NOS tubes to fine-tune the sound. For instance, one might try a 5751 (a lower gain variant of 12AX7) in the first stage for slightly less gain, or different 12AT7 brands for sonic flavor. However, given the feedback, the differences might be subtle compared to a no-feedback design, because the circuit will correct a lot of the tube’s intrinsic differences. This is another advantage of the feedback design: tube-to-tube variations are minimized, making the unit more consistent and reliable over time (it’s less sensitive to tube aging or swapping, as evidenced by the user who tried a 12AU7 and found gain unchanged.

Sonic Impact of Design Choices

Ultimately, how do these technical decisions translate to sonic performance? Experienced engineers and audiophiles often observe the following about the Matisse Fantasy Mk II (and its well-made clones):

• Neutrality with a Hint of Tubelike Warmth: Thanks to the global negative feedback, the Fantasy Mk II measures very clean and sounds generally neutral across the spectrum. The frequency response is ruler-flat, and distortion is very low, so it does not impart obvious colorations like frequency tilts or high-order distortion. However, it still uses vacuum tubes in a relatively simple circuit, so there remains a touch of 2nd harmonic warmth and a naturalness to the timbre of instruments. The use of a fully bypassed first stage and those specific tube types means even-order harmonics are likely the main residue (which human ears perceive as warmth or fullness). Feedback will have reduced them, but not banished them entirely. Many users praise the Fantasy’s ability to sound musical yet accurate, avoiding the analytical or sterile presentation that some heavily feedback or solid-state designs can have. This is a direct outcome of striking a balance in feedback amount and preserving a simple signal path.

• Soundstage and Detail: High-quality components (like the polypropylene coupling caps and metal film resistors) and the low noise floor contribute to an excellent retrieval of detail. The soundstage portrayed by the Fantasy is often described as holographic, with solid imaging of vocals and instruments. This likely comes from the low distortion and low noise allowing microdetails and spatial cues to come through. The symmetrical layout and dedicated grounds mean crosstalk between channels is minimal, preserving stereo separation. The timing (transient response) is very fast – as indicated by that 1 MHz bandwidth spec – meaning transients are handled with ease and there’s no slew-induced distortion. Drums and percussive sounds have snap, and complex musical passages remain clear. If any aspect of the design softens the sound, it might be the output coupling caps if a poorer quality cap were used, but in the original they likely used top-notch caps, and indeed clones using film caps maintain that clarity.

• Bass and Dynamics: The use of a high-voltage supply and large coupling caps gives the Fantasy authoritative bass. Users have noted that despite being a tube preamp, the bass is tight and extended (some tube preamps with small coupling caps or undersized power supplies can sound rolled-off or loose in the bass – not the case here). The dynamic range is high (owing to >85 dB S/N and the ability to swing several volts cleanly), so the preamp can handle both very soft and very loud passages without injecting noise or distortion. The headroom from the 420 V B+ means that even high dynamic range recordings won’t cause the preamp to clip or compress; it will typically be the power amp or speakers that reach limits before the preamp does. The feedback also plays a role in bass reproduction: by lowering output impedance, it increases damping of any interaction with the power amp’s input. Bass transients are conveyed with low coloration. If one were to compare to a no-feedback tube preamp, the Fantasy might have tighter, more controlled bass whereas the zero-NFB design might have a bit more rounded, “bloomy” bass. That matches reports that when the Fantasy’s PSU was run unregulated (with more sag), the sound was “smooth” (which might imply slightly softer bass), whereas the regulated/stiff supply gave more “impact”. Many audiophiles equate that impact and precision with “speed” in the bass – an ability to start and stop notes quickly – which the Fantasy seems to deliver.

• Comparison to Solid-State Preamps: Although the question is framed among tube topologies, it’s worth noting that the Fantasy Mk II’s performance encroaches on what solid-state preamps achieve (low THD, wide bandwidth, low output impedance), but with the benefit of tube sonics. It doesn’t have the extremely low (<0.001%) THD or sub-100 Ω output of some op-amp based designs, but musically it likely has a more engaging sound. Engineers might say it has enough feedback to correct gross errors, but not so much as to erase the device’s character. In contrast, a typical op-amp preamp uses tons of feedback and can measure as a straight wire, but some feel they sound too clinical. The Fantasy’s roughly 12 dB or so of feedback (estimating) is a moderate amount – a conscious design choice to reach a target THD of 0.05%, which is well below audibility for most content, but not to chase 0.0005% which might require an op-amp or more stages. This moderate feedback leaves a little bit of benign distortion which many find actually contributes positively to listening enjoyment (the “second harmonic enrichment” theory of tube sound). So in a sense, the Fantasy optimizes listener perceived quality rather than absolute textbook perfection.

• Stability and Consistency: Sonically, the Fantasy Mk II is known to be consistent in different setups – it doesn’t misbehave with certain amplifiers or cables, as long as those are within normal parameters. This is an advantage over some exotic designs which might oscillate or have frequency response interactions with certain loads. The Matisse design’s stability margin (with that compensation cap in place and grid stoppers) is enough that one doesn’t hear variations or instabilities (like ringing on transients). The preamp is also reportedly quiet at idle, so in use one doesn’t hear hum or hiss even on high-efficiency speakers, which can’t be said of all tube preamps (some with AC heaters or no feedback might have a slight hiss/hum that’s audible up close to speakers). This quietness means all you hear is the music. From an engineering perspective, the noise and stability performance are as important to “sonic performance” as the distortion, because any hint of hum or oscillation can really detract from the musical experience. The Fantasy’s attention to these details (through the power supply design and feedback) yields a solid black background, which enhances perceived contrast and detail in the music – quiet nuances are not masked.

• Sonic Adjustability: While not a feature per se, the Fantasy’s simple circuit allows some level of tuning by changing component types, which some users have done. For instance, swapping the cathode bypass capacitor type (electrolytic vs film) or value can subtly change the sound (unbypassed cathode = more local feedback = a bit leaner/cleaner sound; fully bypassed = fuller sound). Changing coupling caps to different brands can impart subtle tonal shifts (some film caps sound very neutral, others slightly warm). Rolling tubes can adjust the micro-details: a Mullard 12AX7 might add a warmer mid, a Telefunken might be more open and extended, etc. The core design is solid, so these tweaks are like seasoning on a well-cooked meal – not necessary, but for those chasing perfection it’s possible. The feedback ratio itself could be adjusted by changing one resistor (the feedback resistor), effectively altering gain and damping. Decreasing feedback (for higher gain) would increase output Z and distortion, potentially giving a more “vintage” tube sound (some might even try this to subjectively see how it sounds). Increasing feedback further would lower gain (not usually desired in a preamp beyond a point) and could risk stability if not re-compensated. It seems the designers chose an optimal point where the sonic gains from feedback were maximized without side effects, so likely best to leave it as-is. Nonetheless, this design invites understanding and potentially customization by skilled users, which is partly why it’s popular among DIYers.

In conclusion, the Matisse Fantasy Mk II preamplifier stands as a technically well-engineered yet sonically rewarding design. It doesn’t pursue novelty for its own sake, but rather refines proven ideas (common-cathode gain stages, negative feedback, quality passive components) to achieve a result that checks off both objective and subjective criteria. Compared to typical tube preamps, it might not be the simplest (because of the added feedback network) nor the most complex (no regulators on board in some versions, only two active stages), but it finds a “sweet spot” that many designers aim for: the musical engagement of tubes with the accuracy of solid engineering. For an electronic engineer, it’s a beautiful example of how careful component choices (like those grid stoppers, bias points, and caps) and a bit of loop feedback can create a high-fidelity piece of equipment. For the experienced listener, it proves that tubes can be very precise and quiet without losing their soul. The Fantasy Mk II has thus earned its place in audio lore, often being the circuit behind various boutique preamps and respected DIY builds, and its topology continues to be a reference for designing modern tube line stages that aspire to the same level of performance and musicality.

References (Design & Discussion)

• Matisse Fantasy Mk II Specifications and Manual (Analog Metric user manual)--manualzz.com

• diyAudio Builder Reports on Matisse Fantasy Circuit (discussion of output impedance, feedback tuning, resistor values)--diyaudio.com

• SoundStage Network Review of Matisse Fantasy (noting classic design approach and component quality focus)--manualzz.com

• Clone Preamp User Experiences (Xindak XA3250 clone of Matisse, performance specs)--stereonet.com

• Forum Discussions on Component Upgrades (grid stopper removal, coupling cap changes, PSU filtering vs regulation)--diyaudio.com