Klaus Heinz: To DSP or not to DSP – that is the question
As the use of DSP – Digital Signal Processors – in Studio Monitors gets more and more common, some people have started to consider DSPs to be the incarnation and/or the proof of progress in speaker-technology, others suspect sonic degradation and do not want to see (or even better, hear) them in the signal path.
So let us have a closer look at the benefits and limitations of DSP in monitors, based on fact, not opinion. Signals can be sent to the speakers either digitally or analog. If DSPs are used, additional conversion is needed one time for digital signals (D/A) or even two times in the case of analog signal processing (A/D and D/A). For arriving analog signals it means two extra conversions, in the case of digital signals an internal D/A converter replaces some external gear.
»Good monitors normally will work in a studio situation where excellent converters are around, and even with a careful design in the converter stage of the monitor, it is likely that the signal quality is lowered compared to sophisticated A/D or D/A equipment.«
On the other hand: What can DSPs do that analog circuitry cannot, and most importantly – can they deliver the better sound? The main job of a monitors input section is to form a frequency dividing network to separate low, mid and high frequencies, to apply equalizers that boost or lower certain areas within the frequency spectrum, to insert limiters and to provide different voltage gains to the individual bands and to the overall amplification. In rare cases a delay is applied to the tweeter to compensate the earlier start of the high frequencies compared to the woofer output, thus introducing virtually equal distances for the drivers to optimize the time behavior of the emanating sound.
The kind of DSP filters that can do all this are called IIR or Infinite Impulse Response filters. Important: Their characteristics are the same as to be found in analog circuitry. A practical advantage of DSPs lies in the convenience of defining and typing in the filter- or EQ-values as well as the possibility to use very steep filters with up to 96 dB/octave (16th order). However, experience shows that a 24 dB/oct (or 4th order) filter is the maximum of what is needed for no compromise results, and that there is no real benefit from deeper filter steepness, as exciting as it may sound at a first glance.
HEDD Type 30 is a Midfield Studio Monitor, based completely on analog circuitry. In contrast to DSP-loudspeakers, various digital connections can be established via the HEDD Bridge, a modular system of input cards.
An additional and theoretically teasing approach however is the use of FIR filters, the so called Finite Impulse Response filters that enable linear phase behavior for the complete monitor independent from its frequency response. Phase response describes the timely relationship between the different frequencies whilst travelling through the unit, the lower frequencies need more time to pass the speaker than middle or high frequencies. A loudspeaker causes timely errors for the signal, no avoiding that – for the moment.
It is now time to discuss the character of the so called Fourier Transformation, a ground-taking equation in Physics that delivers the fundamental relationship between the time and the frequency domain of waves or vibrations of any kind. Fourier Transformation allows calculating the interdependencies between the time and the frequency domain of a signal. The record of an impulse (in the time domain) can be captured and its correspondent frequency response can be calculated. Doing this mathematically correct is a very time consuming process, it needed the Cooley–Tukey FFT algorithm to »invent« the FFT, the Fast Fourier Transformation that does the job in a fraction of the time. Today FFT plays an outstanding role in many technical fields, among them frequency/time responses for loudspeakers and modal analysis of cones or cabinets.
As both domains – frequency and time domain – are part of the Fourier equation they influence each other inevitably. High frequencies are short in time and low frequencies are long, it is a kind of reciprocal relationship. The frequency response of speakers (or other electro-acoustical devices) cannot be changed without affecting its phase response and vice versa. To deliver linear frequency and phase response at a time is not possible in the analog world, that is why phase behavior and its influence on sound quality are an “eternal” issue in speaker discussions and many scientific papers.
»A new trick became reality with modern DSP and FIR (Finite Impulse Response) filters. It became possible to store parts of the signal in the DSP and wait, until everything would be in the right timely order before being transmitted further. By doing so phase can be influenced independently from frequency response, a linear phase loudspeaker finally is possible.«
The price to be paid for the linear phase response is a delay in the overall signal behaviour, practical numbers are between 1 ms and 500 ms, mainly depending on how detailed and how low in frequency the filter is applied. This is no problem in a pure reproduction chain, but of some danger in a recording situation.
Another and quite different question is the connectivity to digital sources, something of increasing relevance these days. We implemented a modular input system for this purpose, the so called HEDD Bridge, that allows to use different input cards that adapt to different digital signal sources and protocols. A whole new dimension in connectivity is established by offering cards that connect the monitors to Audio over IP (AoIP) networks like Dante / AES 67). One can fairly expect that these networks in the future will be replacing the majority of studio connections for a fraction of the price, adding administration of sources and destinations, something not possible before.
- Quick and comfortable filter and equaliser setting-up
- Piece-to-piece consistency for the electrical filter values
- Possible user interface
- FIR filters, that allow linear phase behavior of the complete monitor (on cost of signal delay). FIR´s get additional importance the moment one looks out for a room correction system – to be discussed in a later essay.
- Room correction as an extension (although only valid for limited listening positions)
- Steeper filter characteristics up to 96 dB/oct. (analog: typically not more than 24 dB/oct.). The effect of filters > 24 dB/oct is not relevant for designing filter networks within a speaker design
- At least one, in most cases two additional A/D or D/A conversions, sound quality is affected
- Small remaining minuses in distortion and noise compared to excellent analog designs
- Linear Phase filters (FIR) improve audible sound quality to a small degree but introduce time delays between 2 and 500 ms typically, not bearable in live recording situations
- There are excellent sounding analogue monitors, and there are poor sounding DSP-based speakers.
- The list of additional aspects and considerations is long and would blast a blog article.
- Judgements of sound quality in the end resides in the users opinion.
- HEDD monitors are designed 100% analog, firstly because we prioritize audio quality over everything else, and secondly because of the lack of a compelling reason to use DSP. It is not a 100:0 conviction but a clear preference after dealing with the case for many years.
Do not decide or judge on a monitor due to its approach (digital or analog) alone, take the time to listen and to compare – there are many different ways to either spoil or perfect either an analog or a digital monitor design.