Rafe Arnott: Your new L2i integrated amplifiers feature an optional L2 DAC module specified with ultra-low distortion, 128dB signal-to-noise ratio dual AKM AK4497EQ chipsets (one per channel) for up to 32-bit/768kHz and DSD 512 playback. This isn’t the first time you’ve tapped AKM’s flagship chip for digital duty (LIO DAC 2.0 in 2017), how many other companies products did you try before settling on this particular implementation? Why no ESS? Why no multibit? Why no R2R?
Vinnie Rossi: The L2 DAC is essentially the same three-PCB stack as the LIO DAC 2.0, but the main voltage regulator feed for the L2 DAC is a Belleson super-regulator. I was very excited when the AKM AK4497 d/a was introduced because it has a non-oversampling, digital filter bypass mode that can be activated in software control mode (an onboard PIC microcontroller sets this when the unit is turned on). With most off-the-shelf d/a chips, you are fixed with using their internal oversampling and using one of their internal digital filters. With the AK4497, it can be run in non-oversamping mode and without having to use any of it’s internal digital filters, so I was able to implement an analog-domain filter in the Class A JFET output stage that follows the d/a conversion – one AK4497 per channel. The AK4497 also supports PCM up to 32-bit / 768kHz, and up to DSD512 (8x DSD).
RA: What made you want to adopt a modular approach to the DAC section in your designs over the past several years to begin with? What are the main benefits from a design and manufacturing standpoint and for the consumer?
VR: With LIO, as well as with the new L2 Signature Preamplifier and upcoming L2 Integrated Amplifiers, the goal is to have the DAC be a modular option for our customers (the same is true with our L2 Phonostage option). It mainly has to do with flexibility. Some customers already have an external DAC that they are happy with, so they simply don’t need one or don’t want to spend the money for our DAC module, or can always add it later. Others love the all-in-one aspect of having the L2 DAC module in the L2 Preamp or L2 Integrated Amp. This eliminates another box, another power cable and set of interconnects. The L2 DAC gets it own dedicated Belleson super-regulated power feed inside the L2 component used, so it is not sharing power from other audio stages.
The other nice thing about the modular approach is the ability to upgrade at a later time if/when a new design comes out. With LIO, we offered the original LIO DSD/PCM DAC module (“DAC 1.0”) for a few years and then later developed the DAC 2.0. Close to 75% of our LIO customers upgraded to the DAC 2.0 and appreciated this level of flexibility.
RA: Inquiring minds are beguiled by acronyms and their meanings, so armchair engineers like myself wonder… you implement JFET (Junction Field-Effect Transistor) semiconductor analog outputs running discretely in Class A for the DAC section, is this because the JFET output characteristic is flatter (more critical for digital) than that of MOSFET (Metal Oxide Semiconductor Field-Effect Transistor)?
VR: JFETs and MOSFETs are fairly similar (both are field effect transistors, both have a high-input impedance and are voltage-driven devices), but yes, for line-level signals, the JFET typically has the advantages of flatter output characteristic curves and better noise performance.
RA: The L2 DAC module is specified with an FPGA (Field Programmable Gate Array – another type of semiconductor) buffer and femto-reclocking circuitry to minimize the effects of jitter. Can you talk to us about your implementation of this design?
VR: For all three digital inputs of the L2 DAC, we electrically isolate from the source to minimize the effects of any electrical noise from the source. But in order to negate the effect of the source’s jitter (timing variations of its digital output signal), we first send the incoming data to an FPGA buffer to store the data, and then reclock the data coming out using our on-board, low-jitter oscillators. Timing becomes more and more critical in the d/a conversion process as the sampling frequency of the data in increased (e.g. a 44.1kHz stream vs. a 384kHz stream, or a DSD64 stream vs. a DSD512 stream). It is similar to driving a car at 60km/h on a road with a speed limit of 60km/h. You can go even faster on that road, but it becomes more difficult to actually stay on the road at 120km/h or 240km/h (less room for error). The car is like the digital data, and jitter is like the car driving off of the road when the d/a required you to be on the road to properly do its conversion.
With so many of our customers using Roon and their sample-rate conversion (for example, you can have Roon covert all your files to DSD512 or PCM 768kHz if you have enough processing power to not bog down your machine), the FPGA buffer and reclock circuitry are helpful when doing d/a conversion at these higher speeds.
RA: In your considerable experience, how much of an impact does the Belleson Super-regulated voltage feeding alll the linear regulators affect the noise floor of the digital module in particular? We often hear how critical a low noise floor is for phono stages, but I think it makes an even bigger difference to digital-audio circuit paths than analog ones – would you agree and why?
VR: If we are talking about a DAC’s analog output stage vs. a phonostage, I believe low-power supply noise is more important for the phonostage because the phonostage is amplifying the signal 100x (40dB, MM input), 1000x (60dB, MC input), or even more depending on the max. gain of the MC input. Even the noise of the amplifying devices themselves comes into play more than with a DAC’s analog output stage, which is usually outputting around 2Vrms (“line-level” signal).
But when you are trying to come as close as possible to realizing the theoretical SNR of 24-bit playback, the analog output stage of the DAC also requires a very low-noise power supply. The Belleson super-regulators provide exceptionally low noise and low-output impedance over a wide frequency range, along with high load regulation. Their website has a section that details their performance attributes, and I used them in both our L2 DAC and L2 Phonostage.