You are here

Apogee PSX100 & Rosetta

High-resolution Digital Converters By Hugh Robjohns
Published November 1999

Apogee PSX100 & Rosetta

Apogee have a long association with state‑of‑the‑art digital converters. Hugh Robjohns investigates two of their latest 'affordable' models, the fully equipped PSX100 A‑D and D‑A converter and its simpler derivative, the Rosetta A‑D.

Apogee are virtually synonymous with high‑quality A‑D converters, having been involved in filter and converter design practically since the dawn of the digital age. The PSX100 is derived, in part, from their advanced modular AD8000 eight‑channel A‑D converter, and shares its styling features of a distinctive purple front panel with silver buttons. However, the PSX100 is Apogee's first high‑resolution, high‑sample‑rate converter with semi‑independent 24‑bit A‑D and D‑A converters capable of operating at 44.1, 48, 88.2 or 96kHz. Word‑length reduction is handled by the latest UV22HR algorithms and built‑in bit‑splitting options allow a full‑resolution signal to be recorded on ADAT or DTRS multitracks. This new machine has been squeezed into a compact 1U rack case with a chunky power‑supply heat sink along the left edge, and has become the flagship model in the Apogee range.

The Rosetta (see box on page 208) is, effectively, half of a PSX100, providing just the A‑D stage as a cost‑effective high‑quality front end for for any digital system. The 88.2/96kHz sampling options are optional but the full range of digital outputs is provided, including ADAT and TDIF ports.


The Rosetta, like its big brother, will convert analogue signals to any digital format, though it lacks the PSX's D‑A stage.The Rosetta, like its big brother, will convert analogue signals to any digital format, though it lacks the PSX's D‑A stage.

The front panel of the latest Apogee converter contains many more buttons and LED indicators than the average digital converter system, implying a great deal more flexibility, or complexity, depending on your point of view! Unfortunately, though, many of its useful facilities are hidden away as alternative button functions, none of which are marked on the front panel!

Perhaps the first thing to note about the unit is its ability to operate in three distinct modes. The most usual one is 'Confidence Mode' where the A‑D and D‑A remain completely separate from each other. This is intended for use with a digital mastering recorder, the Apogee unit providing its analogue interfacing and allowing independent conversion of its output. The second mode is the 'Analogue Monitor' option, where the D‑A is connected internally to decode the A‑D output directly, replicating the monitoring arrangements of most digital recorders (four‑head DAT machines with true‑confidence replay facilities excepted). The third mode is 'Digital Copy' which routes a selected digital input to all of the available outputs, both analogue and digital (in all the relevant interface formats) simultaneously. The A‑D stage becomes redundant, but the UV22 bit‑reduction signal processing can be used to reformat the digital signal if required.

The PSX100's digital interconnectivity is extremely comprehensive, with dual AES‑EBU, S/PDIF (co‑ax and optical), ADAT and TDIF ports all available simultaneously (although the ADAT and optical S/PDIF share the same connector). The analogue interfacing is also surprisingly configurable — inputs and outputs can accommodate balanced or unbalanced signals, on XLRs, and be set independently for +4dBu or ‑10dBV nominal levels. Even signal polarity is adjustable! Both analogue and digital connectors are located on the rear panel, while four front‑panel 'tweakers' allow the precise relationship between analogue and digital levels to be fine‑tuned over a 20dB range for each channel of the A‑D and D‑A stages. A word clock reference input is also available (also accepting a video sync if an optional board is installed), and a word clock output provides for the synchronisation of other equipment.

The front panel controls are laid out quite logically, with the A‑D facilities on the left‑hand side and the D‑A features to the right, a stereo LED bar‑graph meter separating the two sections. The 'A‑D Sync' selector cycles through a long list of options, starting with internal 44.1 and 48kHz clocking options before moving on to the external word clock, the digital input to the D‑A stage, and (if the optional card is installed) a video reference input. The PSX100 employs 'smart synchronisation' as part of the very low‑jitter clocking system. This measures the incoming sample rate and consequently, locking to external digital sources can take several seconds.

Accessing the high sample‑rate options is one of the 'hidden' functions, and is achieved by holding the 'A‑D Sync' button depressed for two seconds. Both the A‑D and D‑A stages have to operate at the same rate in this mode, which is probably an acceptable compromise to the machine's flexibility. Not all the digital outputs are usable in this mode — see the box overleaf for a description of the facilities — and there is no provision for any kind of sample‑rate conversion.

The next button on the front panel sets the output resolution of the A‑D, defaulting to 24‑bits but with 20‑ or 16‑bit options available. The word‑length is reduced through Apogee's latest UV22HR process, although this is not available in the high sample‑rate modes. Holding the 'Output Resolution' button down during power‑up resets the machine's configuration to the factory defaults (except for those facilities determined by the rear‑panel dip switches). Normally, the machine recalls the last used settings on power up.

The optical output port can be switched between S/PDIF and ADAT modes with the next button, and a further pair of buttons mute the main left or right digital outputs. I fail to see a practical use for this feature other than for channel identification, which can usually be performed just as easily from the analogue source, and feel that these buttons might have been better used in providing direct access to some of the more useful 'hidden' facilities. However, when the video clock module is installed these Mute buttons also take on the role of adjusting the sample rate to cope with those weird American frame rates! Once again, these alternative key functions are not indicated on the panel.

<h3>Splitting And Limiting</h3>

Above the two Mute buttons a further pair of buttons select the 'Soft Limiter' and the Apogee Bit‑Splitting ('ABS') function. The Soft Limiter is intended to accommodate high signal levels with a reduced (but not absent) risk of digital overloads, and affects signals above about ‑4dBFS. In one of the more bizarre allocations of a hidden feature, pressing and holding the Soft Limit button engages a calibration mode for the bar‑graph meters. Although this would be easier to find if the button was adjacent to the meter, it is, nevertheless, an extremely useful facility. When the mode is engaged each LED in the meter scale represents a 1dB step between ‑20 and ‑11dBFS and the 'Over' LED illuminates when the signal level falls within 0.1dB of the relevant integer value. In this mode, accurate input level and balance calibration is a breeze.

The ABS function is designed to work in conjunction with an ADAT or DTRS modular digital multitrack (MDM) machine, splitting a 24‑bit stereo signal across four tracks on the MDM (channels 5‑8 carrying duplicate data of channels 1‑4). The data allocation conforms with the established 'Paq‑Rat' format with each 24‑bit signal occupying a pair of adjacent tracks and, by pressing and holding the ABS button (another 'hidden' feature), the two AES‑EBU outputs can also be configured to carry this same bit‑split data across their four channels (left from output 1 and right from output 2). This usefully extends the use of the format to most other multitrack recording systems. In the high‑sample‑rate mode, the ABS function outputs data across all eight tracks of an MDM to record a stereo 88.2 or 96kHz, 24‑bit signal. This facility is obviously not available on the dual AES‑EBU outputs.

The digital bar‑graph meters are scaled down to ‑50dBFS with separate red Over indicators, and can be selected to show either the A‑D output or the D‑A input via a button adjacent to the display. The sensitivity to overloads can be set via the rear‑panel dip switches for between 1 and 4 consecutive peak‑value samples, and the Over light can remain illuminated until manually reset (by holding the 'Meter' button), or will self‑cancel after two seconds — again selectable via the dip switches.

Everything to the right of the Meter switch is concerned with the D‑A section of the machine, starting with a button labelled 'Digital Copy'. This effectively disables the A‑D stage and forces its digital outputs to reproduce the selected digital input signal. Not only does this provide for interface format conversion, but also allows the UV22HR process to be applied to a digital source — for example, when reducing a high‑resolution original recording for a 16‑bit CD master. in In the digital copy mode an output muting system automatically kills the output format corresponding to the selected input to avoid creating clock and audio howlrounds. Holding the Digital Copy button for two seconds disables this feature if required.

The 'D‑A Input' button cycles through the available input sources: the two AES‑EBU inputs, co‑axial or optical S/PDIF, optical ADAT (sharing the S/PDIF optical connector) and TDIF. Pressing and holding the button for two seconds selects the 'Analogue Monitor' mode (indicated by an LED labelled 'A‑D Out') routing the A‑D output directly to the D‑A input. This mode has the useful advantage that the A‑D's 24‑bit output signal becomes available at the Auxiliary AES‑EBU port, whilst a UV22‑reduced 20‑ or 16‑bit signal is dispatched from the other digital interfaces. 'MDM In' selects pairs of tracks from the ADAT or TDIF port; in bit‑splitting mode a set of four tracks can be chosen (or all eight for 88.2/96kHz sample rates).

The remaining two inches of panel space are, generously, given over to four screwdriver trimmers to adjust the input sensitivity and output levels of the A‑D and D‑A respectively. These operate over a 20dB range so any house line‑up level can be accommodated.

In Use

The PSX100's rear panel features a comprehensive selection of digital connectors, as well as a selection of dip switches which are used to control 'set‑and‑forget' parameters.The PSX100's rear panel features a comprehensive selection of digital connectors, as well as a selection of dip switches which are used to control 'set‑and‑forget' parameters.

Both analogue and digital specifications for the PSX100 are quite excellent, and the handbook provides far more data than is common for such products. Although I find it difficult to remember which buttons provide access to many of the 'hidden' features, I have grown quite attached to this converter which performs superbly well in all manner of situations. The audio quality is exceptional, easily justifying the PSX100's asking price which, although high compared to some of the so‑called 24‑bit converters available on the market, is an absolute bargain when compared to high‑end products from manufacturers like dCS and Prism. Just comparing the Apogee with such worthy products reflects the competence of this machine — it may not quite reach their lofty heights, but it is not far short of them either!

The UV22HR word‑length reduction algorithm does a fine job of converting high‑resolution data to the 16‑bit format, providing a consistent and inconspicuous noise floor and retaining an immense amount of detail from the 24‑bit original. Apogee claim their system works in an entirely different way to conventional dither‑plus‑noise‑shaping systems and is based around a specific statistical modulation of the lower‑order bits producing an energy 'bulge' at around 22kHz. This high‑frequency 'bias' effectively linearises the quantisation process and is claimed to produce a noise floor which is 4 or 5dB below that of conventional flat dither systems. I was not able to compare the UV22 system directly with other generic or proprietary noise‑shaping systems, but it has become almost the de facto standard for CD mastering in the USA, so it must be good... Furthermore, a few deliberate under‑recording tests proved that a great deal of original audio information which was lost when simply truncating the data at the 16th bit could remain audible below the noise floor with the process engaged.

Although the benefits of 24‑bit resolution are plainly audible, I am in two minds about the advantages of the high‑sampling option. Admittedly, I have heard some very convincing demonstrations suggesting improvements to both stereo imaging and the overall 'naturalness' of recordings, and I managed to persuade myself that the PSX showed similar attributes in its high sample‑rate mode. However, at present there are only a few suitable recording and playback formats available and I remain sceptical — especially with the format war between the 24/96 and Direct Stream Digital systems confusing the issue. Perhaps the commercial success of DVD will eventually force the format upon us, but I seriously doubt many domestic users will ever appreciate the supposed benefits!

The soft‑limiter system worked pretty well, especially with transient‑rich material, giving a kind of analogue tape‑crushing effect. Typical drum‑heavy pop material benefited greatly from the process, allowing the A‑D to be driven with slightly higher levels than otherwise but with negligible side‑effects. However, more subtle and complex material sounded much better with the limiter switched off, as might have been expected. At least the option is provided so you can try it and see.

The bit‑splitting functions work as described, and open up a whole world of convenient high‑resolution recording using conventional equipment. In the high‑sampling mode, a digital tone is recorded on tracks 5‑8 along with the lower eight bits of each 24‑bit sample, to indicate the presence of this data. The system worked admirably with a variety of recorders including a DA88 and an original ADAT machine, and I also tested the AES‑EBU option successfully.

The metering, although useful, is barely adequate for accurate monitoring of recording levels and I found I tended to rely more on the meters of the relevant recorders. Apogee recognise this point, which is why they provided the Auxiliary AES‑EBU output — its primary function is to feed an external digital metering unit.

Overall I found the PSX100 to be an excellent machine providing high‑quality audio conversion with superbly flexible facilities. It is priced to compete directly with the likes of the Audio Design ProBox converters and falls well below the cost of top‑end products such as the dCS or Prism units, whilst still going a long way towards matching their quality. For anyone serious about the quality of their recordings, the A‑D is the next most critical item of equipment after the microphone and mic preamp, so a decent unit is bound to cost something akin to a top‑end microphone. Of course, it is very hard to quantify 'quality' but, for my money, the Apogee PSX100 represents good value, not just for the accuracy and high fidelity of its converters, but also for its exceptional interfacing flexibility and economy of rack space — most of its peers would require 2U to do less! In fact, I was so impressed with the PSX that I have put my money behind my comments and have bought one for my location recording system.

High Sample‑Rate Options

Apogee PSX100 & Rosetta

When the gurus designed the original interfacing specifications for our digital audio world, they failed to foresee the move towards high sampling rates. Consequently, the existing interface standards have had to be modified to cope. The AES‑EBU format is, by and large, the de facto professional interface format for stereo 16‑ to 24‑bit, 44.1 or 48kHz systems. However, it does not have sufficient data capacity to carry stereo 24‑bit audio at 88.2 or 96kHz, and two interim solutions have appeared (although a definitive AES standard is being drawn up). Some companies have chosen to run their existing AES‑EBU interface at double speed ('double fast') although this places enormous technical demands on both the interface circuitry and the connecting cables. A more elegant, if expensive, approach is the 'double wide' format employing two AES‑EBU interfaces, each carrying just one channel of audio data instead of a stereo pair, but with two samples in the time frame normally allotted to a single sample. Thus, AES‑EBU output 1 carries left‑channel data with 24 bits at 88.2 or 96kHz, and output 2 carries the right channel. Apogee had originally only implemented this mode, but have now made a free EPROM upgrade available which allows the PSX100 and Rosetta to use double‑fast mode as well, thus increasing their flexibility still further. This was not implemented on the review units.

The ADAT and TDIF ports output high sample‑rate data across all eight tracks simultaneously, the last four carrying an identification tone along with the lower eight bits of each sample. To avoid confusing DTRS machines, a dip switch option allows the word clock output socket to generate a half‑speed clock (ie. 44.1 or 48kHz) during high sample‑rate recordings. Similarly, when the TDIF input is selected, the system expects to see a half‑rate clock source.

At present, there is no standard at all for S/PDIF data at the elevated sample rates, and the PSX100 simply drops alternate samples from the high‑sample rate audio to output data at 44.1 or 48kHz. This data has not been re‑filtered to comply with the Nyquist rules and so is likely to suffer from aliasing distortions. The same applies to the auxiliary AES‑EBU output, although it should still provide a signal which is adequate for most digital metering systems, if required. For the same reasons, it is not possible to select the S/PDIF inputs to the D‑A in the high sample‑rate mode. The UV22HR facility is also disabled in the 88.2/96kHz modes, as Apogee believe it is pointless to use shorter word lengths with high sample rates!

Rosetta — Half The Features, All The Quality

The Rosetta is less than half the price of the PSX100, from which it is derived, but offers only half the facilities. Containing just the 24‑bit A‑D stage of its bigger sibling, the Rosetta is intended as a high‑quality front end for any DAT, MDM, or digital workstation, and its specifications are identical to those of the A‑D section of the PSX100.

Unlike the PSX100, there are no hidden button features — what you see is what you get. The standard version offers 44.1 and 48kHz sample rates, while the 'Rosetta 96' model offers the additional rates of 88.2 and 96kHz. A pair of XLR sockets provide the analogue inputs (dip switches selecting +4dBu or ‑10dBV reference levels and the wiring polarity) and the Rosetta retains a full complement of digital outputs. These include two AES‑EBU ports, co‑axial and optical S/PDIF, and a TDIF socket. The optical interface can be reconfigured to the ADAT format via a front‑panel switch if required. A word‑clock output is provided but no input, restricting the flexibility of the unit to some degree, although such a facility would rarely be used in the intended market sector for the Rosetta.

The Rosetta represents a simple, high‑quality converter which is priced very competitively against its obvious direct rivals. Like the PSX100, it is also considerably cheaper than any of the high‑end units whilst still delivering virtually all of the audio quality.


  • A very favourable quality/cost balance from both units.
  • The PSX100 is supremely flexible and adaptable.


  • No labelling of the hidden button functions on the PSX100.
  • No bit‑splitting modes on the Rosetta.


The PSX100 is a high‑quality and superbly flexible 24‑bit A‑D and D‑A combination which justifies its relatively high price. The Rosetta is simply the A‑D stage from the PSX with the same high specifications, and is also good value.