I posted a response about this on FB. Basically, the discussion often misses the mark because we’re talking past each other: it’s 'scientific measurements' versus 'subjective hearing and feeling.' In my view, the best approach is to flip the script and use science to challenge those same arguments. Imagine two people watching cars pass by on a highway. One might scientifically conclude that they are all traveling at 120 km/h with a variance of less than 0.5%, meaning there is no 'scientific' difference between them. Meanwhile, the other person waxes lyrical about the smooth, quiet ride of a high-end Mercedes or BMW, or the instantaneous throttle response of a Ferrari.
The only thing they might agree on is that an old Fiat Panda is clearly struggling to maintain that 120 km/h: it’s probably running hot, burning a bit of oil, and screaming at high RPMs. But their consensus ends there: for one, anything from a Volkswagen Golf upwards is identical because they all hit 120 km/h... period. Right, now let’s translate that to the world of audio, and opamp rolling in the Fosi... but no clear for utterly clear (Sparkos 3602, Burson Vivids,...)
Discrete op-amps like Sparkos and Vivids are in the same league of "crisp and clear." I’d rather try a completely different approach, as the LMs in the Fosi units (like my V3 Monos) also belong to that same clean, clinical school of thought. So, just for the sake of it, I ordered a set of OPA2604s for very little money to try out in my Fosi ZD3.
The ZD3 feeds my ZP3 through both XLR and RCA, creating a sort of "hybrid op-rolled" preamp setup. This allows me to switch between op-amps on the fly—rolling the XLR path while keeping the RCA stock, and then vice versa. In my experience, op-amps seem to have a much greater impact in preamps than they do in power amps.
While I do believe in measurements and agree that distortions below -70dB are generally inaudible in a vacuum, I don't believe a single-tone waveform can compete with the complex signal-splitting job our brains perform. You cannot ignore the fact that while a single 2kHz tone may look identical on two different graphs, actual music will show slight variations. Specifically, how a component handles the timing and integration of harmonics under load is where the real differences lie.
The core of the debate is that we are often looking at the wrong domains. A SINAD figure derived from a constant 1 kHz sine wave tells us about baseline quality, but almost nothing about how a component reacts to the complex, non-periodic signals of music. Here is why the standard FFT might be missing the mark:
The Failure of Static FFT with Transients (Settling Time) Standard FFT measurements are steady-state. Music is a succession of impulses. An op-amp with a narrow Phase Margin or a sub-optimal feedback loop exhibits "ringing" and overshoot following a steep transient. You won't see this on a static FFT because the error is in the time domain and often ultrasonic. In a Null Test, these op-amps reveal a distinct spike in the residual signal at every transient, affecting timing precision and soundstage focus—elements to which our hearing is extremely sensitive via the Precedence (Haas) Effect.
Thermal Tail (Thermal Memory) Transistors on the silicon die change temperature with every current spike, modifying bias and gain-factors in real-time. An Audio Precision sweep is too slow to capture this; by the time the sweep registers, the chip has reached thermal equilibrium. This results in signal-dependent modulation perceived as a lack of micro-dynamics—a dynamic non-linearity that static THD measurements simply mask.
RFI/EMI Rectification and Input-Stage Behavior Op-amp input stages (JFET vs. Bipolar) react differently to high-frequency interference from DAC clocks and switching power supplies. Some op-amps "rectify" or demodulate this ultrasonic garbage into the audible band. This doesn't appear as a distinct peak on an FFT but rather as a subtle elevation of the noise floor during playback, impacting the "black background" and perceived depth.
IMD in the Ultrasonic Domain and Load Interaction Standard analysis stops at 20 kHz, but DACs produce significant artifacts above that. An op-amp with a lower Slew Rate or higher high-frequency distortion can modulate these ultrasonic tones back into the audible range via Intermodulation Distortion (IMD). Furthermore, real-world loads (cables, traces) are capacitive and inductive, not just resistive. The stability of the feedback loop under these dynamic loads is crucial for transparency.
Low distortion is a prerequisite for good audio, but not a guarantee of transparency. Relying solely on the FFT is like looking at a still photograph of a race to conclude who has the fastest sprint. The truth lies in the time domain and dynamic stability.
