Digital Signal Processing: Why It’s Important
BackgroundSurrounding the category of equipment known as the system DSP or system processor is a great deal of confusion and speculation. This article will seek to seek to speak with clarity regarding this topic. Because this is a technical topic, the article is technical in nature.
For the past 30 years or so, sound systems designed to reproduce music at high sound pressure levels have typically used active “crossovers”. That is to say that separate amplifier channels would be used to power the high, midrange, and low frequency drivers in the loudspeaker systems.
Before digital technology was reliable and affordable, the left and right buses of the mixing console would be routed through a chain of analog signal processing devices (usually mounted in a rack that sat adjacent to the console, termed the “Drive Rack”) before the signal hit the inputs of the various amplifiers. This would typically include a graphic equalizer, a limiter of some sort, and a frequency dividing or “crossover” network, which would divide each input channel into 2, 3, or 4 filtered outputs.
Some of the frequency dividing networks that were used were provided by the loudspeaker manufacturers. These devices often had no knobs on the front panel, as they contained filters that were configured specifically for the loudspeaker systems with which they were used. Others were more generic and contained adjustments like input level, output level, crossover point, and polarity toggle. Seldom, however, did the generic units include provisions for adjusting filter type/slope, or introducing driver alignment delay, which will be discussed later on in this article.
As an aside, because these adjustments are critical in getting the various loudspeaker components to play well together as a system, a particular model of loudspeaker might sound and behave completely differently when used with one frequency dividing network as compared to another. This sometimes gross inconsistency had a negative impact on the loudspeaker manufacturer and is one of the reasons that several prominent manufacturers are selling boxes with built-in processing and amplification.
BSS Audio was one of the first companies to embrace digital technology to combine some of the functions of a typical concert system “drive rack” into a single box. The BSS Omnidrive was one of the first “System DSP’s.” It resembled the earlier analog crossover units in that the I/O routing was fixed (2 inputs by 8 outputs, essentially a 4-way crossover), but added limiting, driver EQ, delay, phase adjustment, and multiple filter types and slopes to choose from. It additionally had provisions for storing presets, password protection, and a humidity/temperature “auto correction system” which created quite the stir in the live audio community – all accessible via the front panel GUI.
Over the next few years, DSP technology found its way into hardware aimed at other markets, such as hotel and convention center “room combining systems”, business conference room systems, courtroom systems, systems for transit facilities (i.e. airports), and systems for sophisticated distributed applications such as casinos and theme parks. Two loose categories of system processor emerged: those that were built primarily for concert audio and those that were built for permanently installed systems. Though the line is now somewhat blurry, the distinctions can still be seen.
System processors geared toward the concert audio market have traditionally been designed around a fairly rigid signal flow paradigm (2-4 line level inputs from a mixing console [i.e. for left, right, sub, fills] by 4-8 outputs [i.e. L-R high, mid, low, sub]), and are typically “fixed architecture,” meaning that there is no provision for passing multiple channels of audio from unit to unit in the digital domain (though this is beginning to change). Furthermore, while it is has, in recent years, become commonplace to see computers used in the process of setting up sound systems, most of the parameters in the DSP’s built for live audio can be accessed through a front panel GUI. Some of the frontrunners in this market are XTA, BSS, Klark Teknik, Dolby Lake (now owned by Lab Gruppen), Meyer Sound, dbx, and Sabine.
By contrast, systems built for installed applications are commonly extremely flexible – both in terms of signal flow and I/O configuration. These systems are typically configured by way of a computer program and have limited front panel controls. The system designer can basically choose from among a variety of processing objects including mixers, matrices, various kinds of dynamics processors, equalizers/filters, and other objects like duckers and ambient noise compensators. The number of objects that can be used is limited only by the processing power of the hardware. The processing objects are brought into the design via a drag and drop interface, then the signal flow is drawn from physical input through the processing blocks to physical output. The hardware I/O is often modular, and these types of systems are typically equipped with the facilities to be linked to other units via multi-channel digital audio bus. Several manufacturers adopted the CobraNet standard; however, because of its high latency (time delay) and limited channel count, many of the newest platforms employ proprietary protocols. One emerging standard is Dante.
Inputs in these types of system processors are sometimes not derived only from console mix busses but also directly from signal sources, such as microphones, playback devices, and safety systems. In a church context, this is helpful in that the system may set up to provide for basic functionality (i.e. one microphone plus music playback) through the use of a very basic interface like a wall panel, without the mixing console having to even be turned on. Outputs in these systems may be routed to main clusters, delay lines, distributed systems, and recording devices. In addition to the audio I/O, many of these systems also provide for logic functionality through GPIO (General Purpose Input / Output) and other such interfaces. This allows for functions in the DSP to trigger and be triggered by non-audio events. For example, a fire alarm could mute the P.A. via contact closure if required. Some of the leading manufacturers in this category of equipment are Biamp (Audia/Nexia), BSS (Soundweb London), EV (Netmax), and QSC (QSControl and the beautiful new QSys platform).
System DSP’s are incredible tools, but unfortunately are not utilized effectively in many small to medium sized sound systems. Here are a few rules of thumb pertaining to how system processors should be used:
Portable Sound Reinforcement ContextWith small- to medium-sized sound systems, it is best to think of the DSP’s processing capabilities in two distinct sectors, each having a different function. (The use of DSP’s with large arrays is another matter, and is outside of the scope of this article). We will refer to the two sections as the ‘input processing’ and ‘output processing’ sides. First and foremost, you should look at the documentation that came with your hardware and fully understand the signal flow and routing.
After the A/D (analog to digital) conversion, your signal will likely first hit some sort of equalizer (either graphic or parametric), followed by a delay. This is the input processing side of the DSP. We will come back to this.
Next, your signal likely hits some sort of routing matrix. Some DSP’s have a few routing presets to choose from and others are more flexible, letting you configure your own. You will want to set up the routing matrix based upon where the signal will go once it leaves the various outputs of the DSP. (Is your system bi-amplified? Tri-amplified? Will you be running subs? Delay speakers?)
Finally, your signal will hit a series of processing objects which we will collectively call the output side of the DSP. This will likely include a crossover (consisting of a high pass and low pass filter), EQ, delay, limiter, and polarity switch, usually in that order.
It is critical to understand that the parameters on the output side of your DSP are used to deliver the correct frequency content with the correct time offsets to the various components in your loudspeaker system. These are fixed values that should not change once they are set (with the one possible exception being the high pass filter on the sub output), regardless of the environment in which the system is set up. The reasons for this are numerous and are beyond the scope of this article, but it has to with the phase interaction between the drivers in the crossover range. There are two things you should note about setting up the output side of your DSP.
- Some DSP’s come pre-loaded with presets for specific loudspeaker systems. For example, QSC’s signal processors come with presets already installed for some of their loudspeaker enclosures. Meyer products provide the same benefit. This is a quick and easy way to get your system up and running.
- Though most loudspeaker manufacturers publish the operating parameters for their products, getting the system to operate correctly may not be a simple as plugging the numbers into the processor. Surprisingly, although these devices are operating in the digital domain, two system processors with the parameters set up identically may produce two different transfer functions. That is, even though the settings relate to measurable, quantitative values, different processors will not always produce equal results. For this reason, it is very important to be careful what equipment you buy. Essential Audio would welcome the opportunity to work with you to put together a system perfectly suited to your specific need.
Now, back to the input side of the DSP. These are the parameters you adjust to compensate for the anomalies in the acoustical space in which the system is set up. Sometimes called “tuning the system,” (or worse yet, “tuning the room”) this procedure is more correctly referred to as “setting the system EQ” or “aligning the PA.” Provided the output side of your DSP is set up correctly, amp gains are correct, and you have a well-designed loudspeaker system, the system should produce a relatively flat response in an open space with the input EQ flat. When the system is set up in a room, the input EQ is typically used to attenuate (decrease the system’s response at) the room mode frequencies. (Room modes are the frequencies at which a room resonates and are a function of the room’s geometry. You can easily identify a room’s resonant frequencies on a 31 band graphic equalizer within a matter of seconds by boosting the filters one at a time. Boosting one of the sliders will make the room sound like it is about to launch into orbit – much more so than the others. Chances are you have identified the resonant frequencies in tile bathrooms on numerous occasions while humming or whistling.) This is the main function of the input EQ in small sound systems. In medium- to large long-throw systems, air absorption and other atmospheric factors can necessitate the use of high frequency shelving filters to keep the top end of the spectrum intact.
Install ContextIn the permanent install context, system DSP’s should always be configured by a qualified technician, then password protection should be set up. In this context, the process is referred to as “system alignment.” It is a highly sophisticated process and will involve the use of hardware and software measurement tools such as SMAART. In nearly all permanent installs, the sound equipment is prone to tampering, and changes made to the DSP can not only adversely affect the performance of the sound system, but may also cause serious damage to the equipment. Once the system is set up correctly and performing well, there should be absolutely no reason to access the settings again except in cases where there are architectural changes, major changes to the system, or the integrator provides a custom user interface to initiate preset recalls (i.e. to change the localization of the sound system).
Other NotesAuto-EQ and Feedback Eliminators
It has, in recent years, unfortunately become common “wisdom” that it is possible roll in to a venue with a sound system, throw a measurement microphone up on a stand, push a button, and have a perfectly aligned sound system. It just isn’t so. One manufacturer is particularly liable for perpetuating this myth. Aligning a sound system and establishing an EQ curve is a process that must involve listening. Measurement tools are important but must be interpreted against what the technician hears and observes about the setup. The vast majority of systems I have heard that have been set up using Auto EQ sound atrocious. You are almost always better off setting your input EQ to flat, then dealing with the most obvious problem frequencies.
As for feedback eliminators, they have their place. But their place is not the entire mix bus in a high performance music system.This, likewise is a myth, which the aforementioned manufacturer is again responsible for perpetuating. Feedback issues are symptoms of loudspeaker coverage/aiming issues and/or incorrect settings on the output side of your DSP. It is like applying a band aid to a “flesh wound” like the one referred to in the movie “Monty Python”. If you need the feedback eliminator circuit, it will not solve your problem because the need is a symptom of bigger problems.
Your DSP cannot correct for a fundamentally flawed loudspeaker or array design. There are hundreds of loudspeakers that will sadly never sound good because of factors related to the enclosure, horn, and driver selection/arrangement. It really is a matter of physics and many problems cannot be corrected through processing.
Contrasting Normal Mixing and DSP Setup
|Console Adjustments||DSP Configuration|