You are here

Roland DS90 & V-series

Digital Monitors & Speaker Modelling By Hugh Robjohns
Published November 1999

Roland DS90 & V-series

Roland have moved into the active monitor market with a product which offers digital inputs and a 'Speaker Modelling Function' courtesy of their V‑series recorders and mixers. Technology overkill or a studio necessity? Hugh Robjohns finds out...

Although not particularly well known for high‑quality studio monitoring systems in the UK, Roland have extensive experience in all of the various aspects of the component technologies, so their brand‑new DS90 'reference monitor loudspeaker' is an interesting development. While they're perfectly usable on their own merit as a conventional loudspeaker complete with the added bonus of an S/PDIF digital input, the DS90s are also the intended recipient of the 'Speaker Modelling' functions available in the latest V‑Expanded versions of Roland's V‑series digital workstations and mixers. Consequently, this review is really in two parts, firstly considering the speaker in its own right, and secondly taking account of the speaker modelling technology.


The DS90 is a fairly conventional bass‑reflex cabinet design measuring 372 x 229 x 330mm (hwd). The two speakers are identical, rather than left and right‑handed, and weigh 13.5kg. Two ports vent to the front of the cabinet at the bottom of the front panel, which is constructed from one‑inch‑thick MDF. Both loudspeaker drivers are magnetically shielded; the bass‑mid unit features a 160mm polypropylene cone which crosses over to a 25mm soft‑dome tweeter at 2.6kHz. The frequency response of the complete system is given as spanning 48Hz to 20kHz (at the ‑3dB points), a good result for a relatively compact box, although the bass end drops away rapidly below the corner frequency at around 24dB/octave — an inherent consequence of the bass‑reflex design.

The low‑level active crossover is a fourth‑order (24dB/octave) design which should allow very smooth integration of the two drive units, although the frequency‑response curves published in the manual suggest a few 'woodles' in the crossover region. In fact, the response shows a mild suck‑out in the 400Hz area, with a pronounced peak at 800Hz (perhaps amounting to 6dB between the two amplitude extremes). There is another suck‑out at 2kHz, running into a slightly turbulent area before a gentle lift above 10kHz. All of these variations fall within the +/‑3dB specification limits, but the relative amplitude changes in adjacent bands can prove audible and contribute to the 'sonic character' of the speaker.

First Listening

The DS90 is easy to interface with a wide variety of equipment and is quite capable of delivering loud and punchy music. Unfortunately, though, I have to say that it does not really excel as a reference monitor — at least not in the sense I was anticipating when I started this review! The problem is that the DS90 simply does not seem able to present the necessary level of information from a complex audio track to really warrant being called a reference monitor in the accepted sense of the term. However, as a general‑purpose music recording loudspeaker the DS90 performs quite adequately.

As active loudspeakers go, and especially given its digital input facility, the DS90 is priced very attractively. However, with monitoring speakers you do tend to get what you pay for, and a few hundred pounds more will buy you a far greater degree of resolution and fidelity from numerous other well known and respected speaker manufacturers. The DS90 does have its good points, though — it integrates very well with the VS and VM series of workstations and digital mixers, and the inclusion of both analogue and digital inputs makes this a very versatile and flexible product.

Speaker Modelling

One of the key selling points of the DS90 is that the 'Speaker Modelling' facilities now available in Roland's VS‑ and VM‑series digital products (with the V‑Expanded option) have been optimised for this specific speaker system. For anyone not familiar with the technology, Roland have provided the option to use some of the workstation or mixer's DSP resources (utilising the VS8F2 Effects DSP card) to modify an audio signal in such a way as to 'model' various kinds of common 'reference' loudspeakers. The idea is to allow the user to audition a mix on a variety of common loudspeaker types, thereby optimising the mix for the best uniformity, without actually having to switch between numerous real, physical speakers in the control room

The VS8F2 Effects DSP card, which adds the Speaker Modelling algorithms to Roland's VS1680 hard disk multitracker.The VS8F2 Effects DSP card, which adds the Speaker Modelling algorithms to Roland's VS1680 hard disk multitracker.As a concept this is quite neat — especially if it could be extended to simulate the quality of a £25,000 professional monitoring system from a £900 powered mini‑monitor! Needless to say, the reality is not quite that wonderful...

As anyone who has ever tried to design their own loudspeakers will know, they are very complex machines in their own right. Most manufacturers provide only an on‑axis frequency reponse plot of their loudspeaker systems, and these almost always show complex variations in amplitude response, particularly around the crossover region, where the integration of drivers is less than perfect. A few manufacturers might even include the phase response of the system to show how the drive units interact in time — often highlighting the practical difficulties of matching units accurately in the time domain. The most confident designers will also provide additional traces showing how the frequency response varies in the vertical and horizontal planes, and these often give the best clue to how a speaker will sound in a real room.

The cabinet shape, size and construction, the way the drive units are mounted, and the shapes of the cones or domes all play a significant part in the frequency response and spectral energy of the on‑axis sound, as well as in defining the off‑axis contribution from the speaker. However, it is this indirect sound field which excites most room reflections in the listening environment, making an enormous contribution to our perception of the 'quality' and 'balance' of the speaker.

V‑Series Modelling

The Speaker Modelling incorporated in Roland's V‑series digital products is a process by which the on‑axis frequency response of a desired speaker system is replicated using DSP equalisation processes, in an attempt to make an entirely different loudspeaker (in this case, the DS90) sound like the target system. However, if any degree of accuracy is to be acheived the acoustic starting point must be known — which is why Roland have produced the DS90 as their one and only standard reference point — and, equally importantly, the actual monitoring loudspeaker has to be capable of responding adequately to the demands of the modelling. After all, it would be hopeless if the modelling called for more bass extension than the loudspeaker was physically capable of producing, for example!

However, no matter how good the model is in its definition of the desired on‑axis response, the physical properties of any real loudspeaker — such as HF diffraction, cabinet resonances, radiation patterns, and so on — cannot be replicated. Furthermore, those very characteristics in the generating loudspeaker will inevitably impose their own sonic signature on the sound output.

Although accurate speaker modelling has severe practical limitations, being able to hear how a mix would sound on a very small radio loudspeaker can be instructive. As long as the technology is taken with a large pinch of salt, the speaker modelling provided in the V‑Studio and V‑Mixer systems can be fun and educational.

The Roland system currently offers 12 different algorithms, starting with a 'Thru' mode (effectively a straight bypass allowing comparison of any effect with the unprocessed version). The second algorithm, called 'Super Flat', is intended to optimise the frequency response of the DS90, and, claim Roland, produces "an even flatter sound with a wider range"!

The next six algorithm options are thinly disguised models of some of the most popular and common active 'reference' monitors in the home or small recording studio. Names like 'Powered GenBlk', 'Powered E‑bas', 'Powered Mack' suggest fairly obvious candidates, and the graphical images presented on‑screen with each selection make the links even more evident! Equally, the 'Small Cube', 'White Cone' and (oops, what a giveaway!) 'White Cone + tissue' algorithms are not difficult to attribute to well‑established reference speakers.

The remaining four program options provide generic lo‑fi references such as 'Small Radio', 'Small TV', 'BoomBox', and 'BoomBox LoBoost' — the last two intended to replicate typical ghetto‑blaster systems. In addition to the 12 core speaker modelling choices, the DSP program also incorporates a phase reversal facility, a switchable and tunable bass‑cut filter, LF and HF trimmers (with 12dB of gain swing) and an adjustable peak limiter system.

Switching between the various acoustic models certainly revealed dramatic sonic changes in the output from the loudspeaker, and broadly along the lines that might be expected of the various simulations too. Unfortunately, I did not have any of the 'alleged' reference monitors to hand during the review period for a direct comparison, although previous experience with most of these commercial designs (along with past comparisons between them and my own PMC reference monitors) leads me to the belief that the Roland modelling does a fairly reasonable job of simulating the overall on‑axis frequency response of the chosen models.

...the 'Small Cube', 'White Cone' and 'WhiteCone + tissue' algorithms are not difficult to attribute to well‑established reference speakers.

In particular, the general sound characteristics such as warmth, brightness, presence and mid‑range emphasis (or the lack of it) are all simulated fairly well and would, therefore, be useful in assessing the balance of a mix under a variety of likely listening conditions. Unfortunately, the underlying (and considerably more complex) electro‑acoustic characteristics stemming from the resonant properties of the cone material (polypropylene sounds very different to paper, for example), the cabinet, and their combined effect on the off‑axis responses cannot be replicated in this simple way.

Consequently, I found that the level of detail and resolution — as well as attributes such as neutrality and accuracy — I would associate with these larger monitors was simply not present in the DS90 representations. This, though, is hardly surprising when you realise that the DS90 costs significantly less than the 'reference monitors' it is trying to emulate. On the other hand, the simulation of the 'Small Cube', and the generic TV and radio speakers, although providing suitably curtailed bandwidths, were almost too good by comparison with the real thing. The colorations and intermodulations expected from a little speaker being asked to do too much were completely absent in the DS90!


The DS90 is a reasonable and very flexible loudspeaker, although I would hesitate to class it as a true reference monitor. However, in concert with the Speaker Modelling facilities of the V‑series digital consoles it provides a range of acoustic simulations which can be useful in assessing the transportability of a mix between typical good hi‑fi and poor 'tranny' radio systems — something which has traditionally required at least a second pair of speakers. I cannot say that the Roland speaker modelling system provides a perfect simulation of other monitors but I'm sure many users will find some benefit in the system, even if only as a means of tweaking the spectral balance of the speakers to suit personal tastes!

Trimming & Interfacing

The DS90 offers comprehensive analogue and digital interfacing. Since the two speakers in a pair are identical, each can be switched between replaying the left and the right channels of a digital signal, or a summed mono version.The DS90 offers comprehensive analogue and digital interfacing. Since the two speakers in a pair are identical, each can be switched between replaying the left and the right channels of a digital signal, or a summed mono version.The DS90 accepts input signals in either digital or analogue form. In the case of good old‑fashioned analogue the nominal signal level is 0dBu, and a combi‑jack socket accepts balanced or unbalanced connections on XLR or quarter‑inch jack. Digital inputs are provided on co‑axial phono and TOSlink optical connections, both expecting S/PDIF‑formatted data at sample rates between 25 and 55kHz. A 'Thru' co‑axial phono connector is also provided to enable the onward passage of a stereo digital signal to a second DS90 speaker. A slide switch on the rear panel selects between the co‑axial phono and optical connectors, and another determines whether the speaker's D‑A responds to the left or right channel data, or a mono sum of both. The built‑in D‑A converter is a 24‑bit design, and its output is presented to a third slide switch which determines whether the analogue or digital input feeds the subsequent amplification stage. All of these facilities tend to be of the 'set and forget' kind, so the use of miniature rear‑panel slide switches is not a problem.

Three rotary controls are also provided on the rear panel: volume, LF trim and HF trim. The two spectral trim controls offer a sensibly restricted amplitude range of just +/‑3dB, centred at 60Hz and 10kHz respectively. These are intended to cater for very gentle response tweaking, to compensate either for room acoustics or personal preferences. After this initial input conditioning, the signal passes through the active crossover for separation into bass and treble elements. The bass signal is then amplified by a 60W power amplifier module to drive the bass‑mid speaker, while the treble is amplified by a separate 30W unit.

The only other facility on the rear panel is the ubiquitous IEC mains socket and associated power switch. A large vertical heatsink area occupies the upper half of the rear panel and gets reasonably warm after prolonged use, but never hot. The handbook advises against using the DS90 flat against walls, or mounted sideways, as both arrangements would impair ventilation of the heatsink. The front panel carries a couple of indicator LEDs mounted just below the bass speaker cone. A green lamp shows when the speaker is powered and an orange LED indicates the presence and selection of an acceptable digital signal.


  • Well built.
  • Comprehensive analogue and digital interfacing.
  • The target for the V‑series 'speaker modelling' algorithms.


  • Not neutral or accurate enough to justify calling it a reference monitor.
  • Relatively expensive.


Essentially a general‑purpose loudspeaker suited for undemanding home‑studio applications. Its particular merit, other than the flexibility of both analogue and digital inputs, is in its role as the recipient of Roland's 'Speaker Modelling' facilities provided by the V‑series digital consoles.