Benchmark ADC1

Benchmark are set to further strengthen their excellent reputation with this eagerly anticipated new A-D converter.

The Benchmark ADC1, the long-awaited partner to the highly-regarded DAC1 D-A converter (reviewed in SOS July 2005), has finally arrived, more than two years after it was first announced. This unusually long gestation period is partly due to the company diverting their efforts into developing other products, but it's also because they decided that the original design approach would not achieve the exceedingly high standards established by other Benchmark products and now expected by their customers. Consequently, the original design was largely scrapped and the R&D team started again with a blank piece of paper. Initially, it was suggested that this new stereo A-D converter would also incorporate mic preamp facilities, but this has turned out not to be the case. The ADC1 is thus a relatively straightforward line-level stereo converter, but I understand that a separate product combining a high-quality preamp and A-D converter is in development.

Overview

The ADC1 has been designed to match the DAC1, both technically and aesthetically. It is housed in a similar 1U half-rack case with a built-in linear mains power supply, and it supports all standard sample rates up to 192kHz with 24-bit word lengths.

The rear panel features a pair of electronically balanced XLR inputs for line-level analogue signals, four digital outputs, and reference clock input and output facilities. The three 'main' digital outputs are standard AES3 on an XLR, S/PDIF or ADAT on an optical Toslink connector and AES3-id (the unbalanced, 75Ω form of AES3) on a BNC connector. An adaptor is provided to convert the BNC to a phono socket for easier connection with S/PDIF equipment.

All three digital ouputs carry identical 24-bit data at the selected sample rate, and if the ADAT output mode is selected the appropriate SMux-2 or SMux-4 format is applied automatically. This allows stereo signals at up to 192kHz sample rates to be transported via ADAT, spreading the data across the available channels to keep the individual track data rates standard. At base sample rates (44.1 and 48kHz) only channels one and two are used, the rest being left silent. Double-rate sampling audio (88.2 and 96kHz) occupies the first four ADAT channels using SMux-2, and the quad rates (176.4 and 192kHz) fill all eight channels using SMux-4.

The fourth digital output is called the 'Aux' output and the idea of this facility is to provide a suitable recording feed for DAT or CD-R machines, for example. It is provided on another BNC socket which can be switched to carry exactly the same (bit-accurate) signal as the main outputs, or a 16-bit version. The latter can be generated at either 44.1 or 48kHz sample rates completely independently of the sample rate selected for the main outputs. The word length truncation is linearised automatically with standard TPDF (Triangular Probability Distribution Function) dither, giving a flat noise spectrum.

The BNC word-clock input can accept any common digital reference — straight word clock (including x256 Super Clock), AES3-id or S/PDIF — while the word-clock output carries a buffered clock running at the converter rate, whether that is derived from an internal or external clock.

The front panel is pretty busy compared to the DAC1, but that is an unavoidable aspect of A-D conversion — there are many more things to control and configure than in a D-A converter. However, the limited space on the front panel has forced Benchmark to combine several functions onto a single switch and I feel this makes the unit a little less intuitive to use.

The right-hand side of the panel is concerned with setting the analogue input gain, and to that end each channel is equipped with a three-position toggle switch providing 0, +10 or +20dB of primary input gain. A second set of toggle switches is used to switch between the Calibrated and Variable controls for each channel's second gain stage. The Calibrated control is a 10-position trimmer, accessed through a hole in the front panel, which allows a preset, calibrated level to be established. The Variable rotary knob is used to set the input level manually and it has the same kind of lightly detented action as the DAC1's output-level knob.

Both Variable and Calibrated controls have a range of about 24dB, but there is some confusion as to how that is applied. The handbook says the adjustment range spans -1.3 to +22 dB, while the front panel markings around the manual level knobs suggest a range of -5 to +19 dB, and I'm inclined to believe the latter. Regardless of this anomaly, the complete gain range allows a 0dBFS peak-level signal to be achieved with analogue input levels anywhere between -14 and +29 dBu. As supplied, the SMPTE standard alignment of -20dBFS is achieved with a +4dBu input if all the controls are set to their 0dB gain markings — and the calibrated trimmers are set to that level at the factory.

To the left of the input-level controls is a stereo horizontal bar-graph meter, with nine LEDs per channel. The display resolution can be set using the adjacent three-position toggle switch, selecting either a 48dB range, with the LEDs spread in 6dB steps, or a 20dB range, with the top six LEDs being in 1dB increments. The first seven LEDs are green, with the penultimate being yellow and the last — which lights when the quantiser has been overloaded — being red. There is no soft-limiting function available in the ADC1, so you really don't want to see any red lights! A peak-hold function to extend the illumination time of brief transients is available if the 20dB range option is selected.

The most complicated and least intuitive part of the converter is the last control section to the left of the meter. A square of nine green LEDs is used to indicate the sample rate, internal or external clock source, S/PDIF or ADAT optical output, and the format and sample rate of the Aux output — and it's all configured through a single three-position toggle switch! The use of different-coloured LEDs might have made interpreting the display a little easier, and some markings around the switch would have helped the occasional user.

The toggle (Mode) switch has momentary (non-latching) up and down positions. Pushing the switch up repeatedly cyles through the clocking options for the main outputs, and the block of four LEDs in the top left corner of the array reflects the current setting. The vertical pair on the left illuminates to show 44.1 or 48kHz sample rates, while the second vertical pair illuminates to indicate x2 or x4 modes — the combination enabling all standard rates between 44.1 and 192kHz to be indicated.

In addition, the last option before the internal rates cyle around again is to sync to the external clock input, indicated by the LED in the bottom left corner of the array. If an acceptable external clock is selected, the bottom LED comes on and the appropriate sample-rate LEDs illuminate to reflect the measured incoming rate. If no external clock is present (or it is unstable), the external clock LED flashes to indicate a problem.

Pressing the Mode switch down instead of up cycles through the options for the Aux output. These start with sending the same signal as the main outputs, and then a 16-bit signal at either 44.1 or 48kHz. The column of three LEDs on the right side of the array indicates these three modes.

The only LED in the array not yet mentioned is the one in the centre of the bottow row. This illuminates when the optical output is configured as an ADAT interface. Pressing and holding the Mode switch for three seconds toggles the optical output between S/PDIF and ADAT modes.

Technology

The balanced analogue input signals are received and processed mainly with Analogue Devices AD797 op-amps — which are marketed as ultra-low distortion and ultra-low noise devices — plus a couple of OP27 op-amps for good measure. The conditioned input signals are then fed to the A-D converter stage.

The ADC1 employs the same 'UltraLock' technology as the DAC1 to isolate the converter circuitry from any interface jitter associated with external clocks. The idea is that the A-D conversion is performed with a rock-solid local crystal clock, and the output then transcoded through sample-rate converters to provide the required output sample-rate, externally referenced if required. This approach ensures the highest possible conversion quality with minimal jitter, and the graphs and specs published in the handbook would seem to demonstrate the benefits of this clever technique.

The heart of the ADC1's analogue-to-digital conversion is a 192kHz/24-bit AKM5394 converter chip, which is running at a fixed 221.2kHz sample rate (referenced to a local 28.322MHz crystal). The AKM chip is apparently happy to operate at such an unusually high and non-standard sample rate, and Benchmark claim that this approach has significant benefits in terms of the signal-to-noise ratio and filter response in the subsequent sample-rate conversion.

Separate 22.5792 and 24.576MHz crystals provide clock references for the internal 44.1 and 48kHz output rates (and their multiples), and pass these directly to a Xilinx Spartan FPGA (Field Programmable Gate Array). This device is configured to provide all the operational control logic functions, internal signal routing, output formatting (AES, S/PDIF, ADAT and SMux), word-length reduction and even the metering display ballistics.

External clock inputs are routed to one of two receivers. S/PDIF or AES3-id reference clocks are passed to an AKM4114 chip, while standard word clock or Super Clock references are handled by an RS485 receiver. These particular devices were selected for their ability to recover usable clocks from very low-level and poor-quality signals. The recovered clock input is then passed to a VCXO PLL (a voltage-controlled, crystal oscillator-based, phase-locked loop, using the Analogue Devices ADF4106 and Texas Instruments PLL1706 devices), to regenerate a clean, low-jitter external clock reference, which is passed on to the FPGA.

Depending on the operating mode, the FPGA routes the appropriate internal or external clock signals to either or both of a pair of AD1896 sample-rate converters (the same type used in the DAC1). These accept the 221.2kHz sample rate data from the A-D converter, and calculate the output signals for the required sample rates for the main and Aux outputs. These signals are then formatted as necessary and routed on to the appropriate physical outputs by the FPGA.

In Use

Although the operation of the Mode switch is less than perfectly intuitive, the way it functions becomes apparent after a little fumbling around — and both the handbook and quick start guides explain it all extremely well, if you are the sort of person who reads manuals. The rest of the unit is very clear and simple to set up, and the range of acceptable input levels is unusually wide, making it easy to align the ADC1 with both professional and semi-pro equipment.

I particularly liked the option to switch between calibrated and user-adjustable levels at the flick of a switch, and although I was initially concerned that the detented level controls might make matching levels between channels difficult, this proved not to be the case at all. I was able to match channel gains very precisely at any level setting.

Everything worked exactly as advertised from the point of view of external clocking and output formats, and deliberately trying to upset the ADC1 with low-quality and extremely jittery external clocks had no noticeable effect whatever on the converted audio, which is as it should be. The ADC1 is quite lethargic when syncing to an external clock — it can take anything from about two to five seconds — but it gets there in the end and always reported incoming sample rates accurately during testing. The slow lock-up is an inherent side-effect of the jitter-filtering processes involved.

In terms of absolute sound quality and resolution, the ADC1 impressed me greatly — I evaluated the converter by partnering it with Benchmark's DAC1, and comparing it to my own Apogee PSX100 and the Prism Sound ADA8XR, which was reviewed in SOS April 2006. At elevated sample rates the ADC1 boasts a clean, open sound with extended bandwidth in both directions. It has no discernible character of its own and seems able to capture whatever sonic qualities are present in the source without adding, subtracting or obscuring anything. The bottom end is particularly solid and defined, and I found that stereo images were captured accurately and with a very good sense of depth and spaciousness. My only mild criticism is that there is the faintest hint of congestion at the extreme top end when using standard sample rates, although all the double and quad rates sounded very transparent and neutral, with superb resolution and transient detail. In fairness, I should say that the Prism Sound unit showed a similarly subtle effect at base sample rates, and it was far more obvious on the (now obsolete) Apogee.

Conclusions

The ADC1 is expensive in comparison to typical mid-market converters such as those from RME, Lynx and Apogee (the Apogee Rosetta 200 is similarly priced but includes a stereo D-A stage as well, for example). However, I feel justified in suggesting that the ADC1 performs at a significantly higher level and compares more naturally with serious high-end products from the likes of Lavry, Prism Sound and dCS. In that context, the ADC1 represents substantial value for money, giving only a little away in terms of ultimate resolution.

The Benchmark DAC1 impressed me so much that I felt compelled to own it. I suspect my bank manager will fear the same will apply to the ADC1. The long wait has certainly been worthwhile, as the ADC1 lives up to the excellent reputation established with the DAC1 and the pair form a perfect partnership. The ADC1 is highly recommended for serious applications.