Threshold FET Nine/e preamplifier
The first thing I noticed about the FET nine/e is the outboard power supply: connected to the preamp with a flexible umbilical cord, this is quite a substantial piece of equipment itself, making the power supply of the original FET nine seem as if it should only have been used with a calculator. (The power supply is identical to that used with the FET ten/e, except the FET ten/e has two such power supplies, one for the phono and one for the line section.) Unlike many such devices, the power supply's transformer emits no audible mechanical hum. An outboard power supply eliminates the possibility of magnetic, thermal, and vibrational interfering effects of the transformer on the preamp's circuitry; the only real disadvantage of this design seems to be cost.
The FET nine/e's phono section has switchable high or low gain, the high-gain setting suitable for low-output moving-coils, and with switch-selectable settings for impedance and capacitance. The tape outputs have 100 ohm resistors in series, providing a small degree of buffering. Volume control is set with a stepped 32-position attenuator; the steps are 3dB at the upper and lower ends of the range, 1.5dB in the center.
In their literature, Threshold makes a point of emphasizing the commitment to a minimalist design approach, with as few components in the signal path as possible but with extensive attention paid to bias and power-supply regulation. As the model designation suggests, the Threshold nine/e uses J-FETs and MOSFETs (no, not nine of them), the J-FETs in a parallel-pair configuration to reduce noise. Construction quality is first-rate, with extensive use of metal-film resistors and military-grade glass-epoxy circuit boards.
The Threshold FET nine/e arrived without a manual, and I had no way of determining what the phono stage gain, impedance, and capacitance switches had been set at, so my initial listening was only to the line stage. The sound I heard was clean, clear, crisp (I'm starting to run out of "c"-words), with fine dynamics, good bass extension, andI'm sorry to sayjust a bit on the hard, "transistory" side. In my opinion, the Achilles' heel of digital recording/playback is how it handles complex orchestral and choral sounds; the FET nine/e revealedand, I think, exaggeratedthese problems to an extent that made some recordings difficult to listen to.
For example, played through the FET nine/e, the orchestral tutti on the Mehta Swan Lake (London 410 551-2) had a hardness that was distinctly more annoying than via the Conrad-Johnson PV11 or the Audio Research SP9 Mk.II. With the Muse 100 amplifier rather than the Luxman MQ 68C in the system, this hardness was exacerbated, the residual hardness of the amp adding to that of the preamp.
Listening to a CD transfer of some early Fritz Wunderlich material (Sonic CD 74505), I was struck by how poor the recording quality was, with considerable spittiness evident on sibilants. With the PV11 or the SP9 Mk.II, although I could still hear recording deficiencies if I concentrated on them, I was more aware of the beauty of the voice and the music itself. With recordings that had less of that hard quality to begin with, the FET nine/e did not draw unfavorable attention to itself, but comparison with the PV11 and the SP9 Mk.II showed the latter two to be a touch more natural-sounding.
Similarly, the sense of depth was more convincing through the other two preamps, especially the PV11. The situation did not change appreciably when using the phono stage, except that, with the switches set for high gain and a 100 ohm load (I did get the manual eventually), the noise level was just tolerable, significantly higher than with the SP9 Mk.II used in the same manner (footnote 1). Putting these criticisms in context, it must be noted that the magnitude of all these differences was small (or, in audiophile language, "huge"), and that the original Quads, while themselves anything but bright or hard-sounding, are extremely revealing of any such tendencies upstream.
Yes, but which one is most accurate?
My favorite definition of accuracy comes from J. Gordon Holt's The Audio Glossary: "the ultimate objective of an ideal sound system, which everyone claims they want but nobody likes when he has it." I think Gordon was not writing entirely tongue-in-cheek; a component with a very high degree of technical accuracy, but in which the remaining inaccuracy is of the conspicuously "electronic" sort, may be less preferred than a component that is technically not as accurate, but whose inaccuracy is the sort that sounds like it could have been part of the original sound rather than an electronic artifact.
In routine listening, we don't know what the original sound was like, so our judgment of "accuracy" is actually a measure of plausibility: how much does the sound resemble what we immediately recognize as real (produced by musical instruments, including the human voice) rather than reproduction? (footnote 2) Of course, when the listener was present at the original recording session, this provides another sonic reference, but here the comparison involves a memory rather than a readily available actual sound, and we know that memory is prone to a lot of error/distortion.
A bypass test, where, for example, a line-level input signal to a preamp is compared to the signal after it's gone through the line-level stage, is a useful here-and-now comparison (actually, it's only short-term rather than long-term memory). However, I would still be reluctant to take the results of this test at face value if they contradicted the conclusions of the more conventional listening test; eg, if the product that ranked first in the bypass test turned out to be the one that rendered most records unlistenable.
With some skepticism, then, I conducted bypass tests on all three preamps under consideration. This was done by a) feeding the signal from the output of the Aragon D2A to the CD input of the PV11, b) taking the signal from the tape-out jacks of the PV11 to the input of the Stax CRM-T1 headphone amplifier (footnote 3), and c) comparing the sound of this relatively direct connection (the signal at the tape-out jacks has gone through the input selection switch but no amplification or buffering) with the signal taken from the main outputs, the volume control of the preamp set to provide approximately unity gain (footnote 4).
As it turned out, the results of the bypass tests were not that dissimilar from the results of long-term listening tests. (Whew!) Through the FET nine/e's line stage, voices lost some of their natural warmth, and sibilants tended to acquire a bit of roughness. The SP9 Mk.II line stage was close to neutral in tonal balance, with some (inevitable) loss in transparency and a slight lightening of the sound.
The Threshold FET nine/e, despite its many admirable characteristics, was less able to sustain this illusion; I was more aware of the medium than the message. The Audio Research SP9 Mk.II, in its latest incarnation (Chinese tubes and all), proved to be generally competitive with the PV11, offering a similarly "musical" sound allied with a somewhat lighter tonal balance. If forced to choose, I would go with the Conrad-Johnson PV11 because of its strong "rhythmic" character and its magical way with voices, but I would not deride anyone whose preference was for the SP9 Mk.II. Both preamps are solid Class B contenders.
Footnote 1: The phono-stage noise level was high enough that I wondered whether the unit might be defective, but a check with Threshold revealed that their automatic quality-control equipment had passed the unit as being within spec. A cartridge with a higher output than the AQ 7000 would be a better match.
Footnote 2: Roaming the halls of the New York Penta at the 1990 Stereophile show, I heard a faint sound emanating from one of the rooms in the distance. "That's it! Those are the speakers I want!" I thought as I rushed down the corridor. I threw open the door of the room where the sound system of my dreams was apparently playing. It was Igor Kipnis, practicing on the harpsichord.
Footnote 3: It might seem that it would have been more reasonable to take the output of the D2A straight to the CRM-T1, but this would have left uncontrolled the possible effect of the additional cable and contacts.
Footnote 4: Absolute polarity was maintained by using the polarity-reversal switch on the D2A.