Your browser is out-of-date!

Update your browser to view this website correctly. Update my browser now


dCS 904 A/D and 954 D/A Converters

I enjoyed using the dCS 904 and 954 converters so much that their U.S. distributor, Independent Audio, literally had to twist my arm to pry them away from me. This pair of converters is the best-sounding digital equipment I have ever used and - if I could have afforded to purchase them, they would still be in my rack.

I enjoyed using the dCS 904 and 954 converters so much that their U.S. distributor, Independent Audio, literally had to twist my arm to pry them away from me. This pair of converters is the best-sounding digital equipment I have ever used and – if I could have afforded to purchase them, they would still be in my rack. The first choice of several of my international classical music-recording colleagues, these units are without parallel in today’s crowded converter marketplace.
Product PointsApplications: Applications: High-end mastering and monitoring

Key Features: PCM sample rates up to 192 kHz; DSD; various filter and noise-shaping choices and much more

Price: $7,335 each

Contact: dCS/Independent Audio at 207-773-2422;


+ Excellent sound quality

+ Wealth of high-resolution recording options


– Cumbersome synth-type interface with multiple nested menus

– No metering on 904 A/D unit

The Score: If you need 192 kHz and/or DSD conversion, this is the only show in town.

At first glance, their front panels look identical – a bank of six pushbuttons, another bank with three more buttons, two red seven-segment displays; one single digit and one five digit display – and that’s about it. Each of the first five buttons has a single dedicated function; the others have different functions, depending on the results of special button-pushing routines.

The Function menu, for example, which is accessed by holding down the Menu Step button and pressing the Menu Set button, has 25 menu choices on the dCS 954. These are selected by pressing the Menu Step button repeatedly – and most of these have multiple subchoices as well!

The rear panel on the dCS 904 analog-to-digital converter begins with a set of balanced stereo analog inputs on XLRs. There are two small 20-turn trimpots for adjusting the input sensitivity over a +/-6 dB range, but be aware that +14 dBu level is required for 0 dB FS. Thus, users of -10 dBu or even 0 dBu equipment will need to interpose another amplifier to drive the 904 to full level.

The middle section of the rear panel has four XLR digital outputs, as well as separate AES/EBU reference I/O XLR connectors. The unit can operate in high-speed, double-wide, or quadruple-wide mode. All four XLRs are used in “quad AES” at 176.4 or 192 kHz or DSD mode. The next section contains four BNC jacks, used for SDIF-2 and word clock I/O. A pair of DB-15 connectors labeled “remote in/out” is next. A multivoltage IEC connector setup completes the rear panel.

On the model 954 – the digital-to-analog converter – the rear panel arrangement is the same as on the 904, except it contains XLR and RCA phono connectors for its analog outputs, and four digital XLR connectors for digital inputs.

Both units support DSD at 2.822 MHz, and PCM from 192 kHz down to 32 kHz. The data formats supported are AES/EBU, dual and quad AES, SDIF-2 (for PCM and DSD sampling), as well as DSD packed into four AES links (for use on eight-channel 16-bit MDMs, for example). In addition to functioning in master mode, both models can sync to word clock, AES reference, digital input signal or, with an option installed, video.

The noise-shaping truncation DSP functions include first-, third- and ninth-order curves. At 88.2 and 96 kHz, digital connections can be made through high-speed or dual AES (double-wide) protocols; at 176.4 or 192 kHz, dual or quad AES protocols may be observed. A high-quality (120 dB dynamic range) signal generator function with MHz resolution (ditherable in the DAC unit), is included.

The units have last setting recall, can be controlled by optional PC software and can be set to lock out sensitive menu changes. The various menu options are controlled by firmware EPROMs; I swapped out chips several times while the units were installed in my studio. Firmware can also be changed via an RS232 port on the unit’s rear panel.

In use

I did many of my tests with the units connected “back-to-front” and, in fact, that was the only way I could audition their DSD functions. To hear analog input converted by the 904, I simply connected a suitably high-level input source, pressed the dedicated sample rate and output format buttons to set those parameters, and sent a digital output to a 954 input.

It’s nice that all four XLRs are active in all modes so, for instance, with high-speed 96 kHz conversion, the same digital data is duplicated on all four connectors.

Setting the 954 DAC to select the various transmission protocols (high-speed, dual, or quad) wasn’t quite as easy. It involved multiple button presses in a particular sequence. Eventually, I found a Function menu item that automated its digital input selection but still, I often had to manually input how may of the XLRs should be read as comprising part of a dual or quad signal. Its “auto detect” feature frequently seemed a little capricious.

For DSD auditioning, I used both the BNC and the XLR connectors to send data from the 904 to the 954, but when I tried to use the XLRs to send the data to and from an actual MDM recorder (a TASCAM DA78 set to 16-bit operation, with a Spectral Translator Plus doing the format conversion), the resulting extreme high noise accompanying the signal played back by the MDM suggested that my format conversion circuits were not sufficiently sample accurate to be useful for DSD recording. Hard disk or MO drives would be preferable.

Unfortunately, time constraints prevented me from borrowing a suitable Genex machine, so my DSD observations had to be based solely on back-to-front connections between the two units. Still, the fact that one can – in theory, at least – use a standard 16-bit MDM as a stereo DSD recorder (in the same way that one can use a similar MDM with an Apogee PSX-100 to record stereo 96 kHz data) is a definite plus.

I was able to use the TASCAM (returned to 24-bit HR mode) to record stereo 192 kHz data, employing the dCS gear’s four XLR connectors set to quad-wide mode. For me, this was the cat’s meow as a two-channel recording setup and, by far, the best stereo recorder I had ever· uh, built. Were I at an earlier period in my recording life, this setup would have easily become my secret weapon.

I did have a few gripes, however, with the equipment. First, there was no level readout – not even a signal presence LED on the 904, so the only way to know what, if any, signal was being output from the A/D converter was to monitor the receiving device. (There are bar and numerical level meters on the 904 Windows remote control-Ed.) This lack of visual feedback was extremely annoying. In all fairness, however, I think it’s a bit of a European convention. I’ll never forget my experience working at Polish Radio in Katowice in 1983, when I saw – for the first time – a room of very basic-looking Telefunken analog recorders with nary a meter in sight. When I asked about this, I was informed, “Well, they’re all carefully calibrated to our consoles, so we see no need for extra meters.”

A similar seemingly strange metering situation is obtained when outputting DSD. Viewed on an MDM, what one sees is eight channels, all with fully lit LEDs, suggesting something is feeding back. I don’t know which is worse – no meters at all or eight meters with all reading full scale at the same time! Of course, a DSD datastream viewed from a PCM perspective looks like (and sounds like) full-scale noise, which is why the meters on my TASCAM recorder were nailed to their end stops when fed DSD.

My more serious gripe concerns the software aspect of the converters’ user interface. Although I’ve certainly programmed my share of user-unfriendly synths – beginning with the DX7 – I’ve come to increasingly appreciate the value of dedicated switches and buttons, rather than living with the necessity of having to step through multiple nested menus via pushbuttons. Especially on location, I’d be hard-pressed (pun intended) to feel that all the various adjustable parameters were still set where I think I left them, without stepping through the Function menus one more time, just to make sure.

Perhaps I’m being overly paranoid (as there really aren’t too many parameters in the menu affecting recording status that are not obvious from either the ADC’s front panel or the proper performance – or lack thereof – of the connected recorder), but still, I want to feel completely secure with my equipment chain when I have an orchestra of highly paid musicians breathing down my neck.

Additionally, my review units were demos that had been exposed to the rigors shipping several times prior to arriving at my doorstep – one of the nine pushbuttons on the DAC was defective. Unfortunately, it was one of the buttons that does double duty in the function mode, so I became quite frustrated while trying to learn to use the unit – until a transatlantic telephone conversation in which I and a gentleman at dCS in the U.K. pushed similar buttons on identical models until we determined that mine was, indeed, broken.

With all the effort I spent learning to use these units, the listening tests proved rather anticlimactic. They sounded simply superb! The 192 kHz conversion, when replayed directly through the dCS model 954 DAC was, by a not inconsiderable margin, the best PCM digital sound I have ever heard. Open, easy-to-listen-to, liquid, analog like – these were some of the adjectives I wrote down on my notepad. The D/A, used with other sources – especially when they were upsampled to 192 kHz via the dCS model 972 unit (PAR, 6/00, p. 58), beat out my reference Wadia DAC in just about all categories.

I’ll make a brief mention of the various digital filters available in the two units. The 904 offers the option of four anti-alias filter responses. Filters 2, 3 and 4 give progressively more relaxed responses, degrading the alias performance but sharpening the impulse response, thus affecting the stereo image. I observed fascinating listening results from playing around with these; suffice it to say that I preferred different filters for different material.

Similarly, the 954 D/A unit, using a high bandwidth oversampling converter, gives one the opportunity to change filter characteristics in the circuit, removing the Nyquist images in the output signal. As we know from theory, sharp filters have the unavoidable mathematical result of giving a poor transient response, marred by significant ringing; this affects the stereo image by smearing the signal energy. The dCS filter choices trade off the filter rolloff and energy smear parameters by offering increasingly relaxed rolloff curves. My listening tests agreed with dCS’s, and I preferred Filter 2 for my classical masters, but enjoyed the softer curves for listening to jazz and pop selections. Fascinating, indeed!

As for DSD · well, this was my first exposure to it and, believe me, it was an eye-opener. Tom Jung is right; DSD sure sounds different from PCM, especially lower sample rate PCM! I don’t think I’ve even developed the right vocabulary yet to qualify the differences.

I’ve written previously that higher sample rate PCM conversion sounds are easier to listen to than 44.1 kHz. DSD continues this trend, but there’s also a big difference in the midrange, which I can’t quite put my finger on. Maybe I’ve been listening to PCM too long, but in some instances, I actually preferred 192 kHz PCM to DSD. I look forward to other opportunities to hear DSD conversion. Anyone who heard the Sony demos at the New York AES Convention 1999 knows how good DSD can sound. And we know that, in theory at least, DSD should be better.

dCS also sells sets of specially modified 904 converters to do multichannel DSD recordings on 24-bit MDMs, with the possibility of simultaneously sending a PCM signal to an external AES/EBU meter. Ah, brave new world!


If I were 30 years younger, I’d sell most of the equipment I currently own and buy the dCS 904, 954 and 972 equipment, along with a couple of Genex MDM recorders, and start my audio recording life all over. As it stands now, I can only dream about it and hope to inspire younger engineers to take on that challenge.