Ayre KX-R line preamplifier
"The reason nobody ever talks about this issue," said Hansen, when he previewed the preamplifier to his dealers (and me) just before the CEDIA Expo in September 2007, "is that it's like the airit's omnipresent, so nobody even thinks about its existence. Almost every preamplifier on the market uses an attenuator [ahead of the output stage] to reduce the gain. There are two problems with this: 1) once you get beyond 10k or 20k ohms [series resistance], you begin to affect frequency response along with the volume; and 2) the active circuitry outputs a constant noise voltageincrease the volume level and you increase the signal/noise ratio. The maximum SNR is therefore at full output, which nobody listens at. With VGT, the SNR is constant, regardless of volume setting."
The KX-R is a seriously gorgeous piece of audio eye candy, so I implored Hansen to let me review it for Stereophile.
"No problem," he told me, "but we've learned from experience. First we need to finalize the design, then we'll satisfy pre-orders and stock our dealers. After that, we'll let the press listen."
Those eight months were long.
"Before I talk about the KX-R," Charles Hansen said to me, "I have to explain that it was designed with what I can only describe as an extreme degree of madness. That's to say that, in addition to working the kinks out of VGT, we examined everything, even stuff like the panel display.
"Pretty much 99.99% of all front-panel displays are 'multiplexed,' which means that not all of the dots are on all the timethey flash, much like the pixels on an old CRT monitorand turning those dots on and off rapidly creates electrical noise that can radiate into the rest of the circuitry, degrading the sound. We found a special display that is not multiplexedthe segments are always on, and no electrical noise is generated. The catch, of course, is price: We take about a tenfold price hit for that."
On to VGT. The active device in the KX-R is a FET (a transconductance device), meaning that the input voltage controls the output current. The current needs to be changed back to an output voltage later in the circuit, which in a conventional design is typically accomplished with a fixed resistor. This is called a "transimpedance" device because impedance is the opposite of conductance.
"In the KX-R, the idea is very simple," Hansen said. "We simply use a variable resistance instead of a fixed resistance. Changing the value of the resistance changes the gain of the circuit. With a variable-gain circuit there is no input attenuator, as is normally found in a preamplifier. This brings about several advantages: The input impedance is not restricted by the value of the volume control (input attenuator). I mentioned that with conventional attenuators, the frequency response can vary with volume. In contrast, the KX-R has an input impedance of 1M ohm per phase, and the frequency response goes out past 250kHz regardless of volume setting. Another benefit is that the signal path is simplified, and there are fewer switches in the way.
"With VGT, the signal/noise ratio is constant regardless of the volume setting. A typical listening level might be 20dB below full gain, which translates into an effective increase in S/N ratio of 20dB. This is on top of an already quiet circuit, and improves the resolution of low-level detail audibly."
If VGT were simple, Hansen observed, everybody would be doing it. "I'm not saying we invented this. PS Audio's VCA technology sounds like they might be doing something along these lines, and I suspect CEC might be doing something in the same ballpark. I don't know, because nobody's talking about the real intricacies of their designsand neither am I."
Key to VGT was Ayre's development of what they call EquiLock circuitry for the MX-R monoblock. "In a conventional circuit, the gain transistor has a load, usually either a resistor or a current source," Hansen said. "When the current through the gain transistor changes, then the voltage across the load also changes, which, in turn, means that the voltage across the gain transistor is changing. In fact, all of the parameters (transconductance, capacitance, etc.) vary when the voltage across the transistor varies.
"The EquiLock circuit adds another transistor between the gain transistor and the load. (In our case, the load is actually a current mirror.) This extra transistor holds the voltage of the gain transistor at a fixed level while still transmitting the changes in current to the load (the current mirror). By stabilizing the voltage across the gain transistor, all of the parameters of the gain transistor are also stabilized. The circuit is very similar to a cascode circuit, which has been used by other manufacturers, but EquiLock is an improvement over a conventional cascode circuit."
And there are elements that make VGT difficult to pull off. The circuit Hansen developed works only with zero-feedback designs, which just happen to be what Ayre specializes in. "It might take a while for others to catch up," Hansen cackled.