Abstract

This article explores the differences between using digital signal processing (DSP) and fully analog systems in loudspeaker system design. While traditional analog systems are valued for their simplicity with no analog-to-digital converter (ADC) or digital-to-analog converter (DAC) stages, DSP offers precise control and potential improvements in sound quality in a cost-effective way. The article presents a detailed methodology and test setup to compare the performance of DSP and analog systems, highlighting the benefits and trade-offs of each approach. Measurements and analysis aim to provide a transparent, data-driven comparison to help manufacturers and system integrators make informed decisions.

Introduction

Many factors come into play when balancing the pros and cons of a loudspeaker system that uses digital signal processing (DSP) vs. fully analog. This has made DSP use in loudspeaker system design a polarizing topic in recent years.

In favor of the analog approach, traditional passive analog crossover networks in two-way systems are well understood, with no analog-to-digital conversion, minimal group delay, and near zero latency. Some manufacturers consider all-analog designs as a point of differentiation, and some consumers believe that DSP degrades sound quality.

Yet, in favor of DSP, manufacturers and system integrators are beginning to recognize its potential for targeted design improvements. For example, in high end recording studios, DSP can be a critical and highly accurate method to achieve a tuned monitor system in a treated room.

This article aims to quantify some of the benefits and trade-offs of designing loudspeaker systems with DSP. Measurements and analysis are offered in hopes of providing a transparent and data-driven summary of the benefits of DSP-based implementations compared to the classical analog approach.

Methodology

For this article, high quality components were sourced to evaluate whether DSP can improve measured performance compared to a traditional analog crossover implementation. The digital crossover was designed to mimic the topology of an analog bi-amped system with per channel equalization. The main objectives were to reduce the standard deviation of the frequency response and verify that DSP didn’t degrade other measured attributes of the system.

Figure 1 shows the completed signal chain topology.

Topology of the digital crossover in SigmaStudio^®:

1. Defect correction: Fixes narrow-band issues in each individual speaker system.
2. Stereo crossover block: Many crossover types are available for designers to choose from.
3. Stereo equalizer: Controls the equalization (EQ) of both channels of high and low outputs from the crossover.
4. Gain control: Enables level matching for each of the crossover outputs individually.
5. Time alignment block: Allows for very fine delay parameters to be set for matched in-phase response.
6. Look-ahead limiter: Provides driver protection as a safeguard. This adds additional latency, so applications such as recording studios may want to avoid it.

Test Setup

The test setup (Figure 2) used Acoustic Elegance TD15H-4s for woofers. For mid/high frequencies, the ESS Heil Air Motion Transformer™, known for its linear response, low crossover point, and wide dispersion pattern, was used. These were coupled with a high performance passive crossover (Figure 3) and powered with a Behringer NX1000 amplifier, which can output 300 W per channel at 4 Ω with a THD of 0.05%.

For DSP system measurements, the Analog Devices EVAL-ADAU1467Z was used with SigmaStudio, a free programming environment for SigmaDSP^® products. SigmaStudio is a block-based IDE graphical user interface that includes such features as EQ, crossovers, routing, delays, metering, and limiting. The output of this system consisted of separate high-and low-pass line level analog audio signals. The high-pass outputs were fed to an ICEpower 1200AS while the Behringer powered the woofers.

The testing room was semi-treated and approximately 5.7 m × 6.4 m. Speaker location and room were consistent across the tests.

Results: Room Response

The first test was to compare the performance of a digital crossover to the analog passive crossover network. When measuring the resulting listening position response of both systems, notice that the smoothed frequency response of the DSP system had less standard deviation from an ideal flat frequency response (Figure 4).

In free field, the analog system had a standard deviation in the woofer (20 Hz to 800 Hz) of 4.2 dB, while the digital system had a deviation of 2.9 dB. For the tweeter region (800 Hz to 20 kHz), the standard deviation was within the measurement error of a high-end type 1 sound level meter for both analog and DSP.

In the analog system, the shaping network was adjusted slightly for better subjective listening response, which explains gain differences between the mids and highs in the graph. The woofer low-pass output of the crossover had no shaping network.

Results: Crossover Response

Next, the response of the crossovers was measured electrically using analog probes into an Audio Precision APx555. As expected, the crossover through the DSP was smooth and had no variation between left and right channels. A fourth-order 24 dB/octave Linkwitz-Riley filter centered at 800 Hz was used, a specification that is difficult to achieve in an analog system without expensive components.

The analog system, despite having low tolerance and premium components, exhibited a variance in response between the left and right channels (Figure 5). This highlights the inevitable variation in speaker components when mass manufacturing speaker systems.

In analog systems, variation in speaker components can only be compensated by increasing the complexity of the crossover network, matching the network to the drivers, or having tighter tolerances on the speaker components. All these solutions increase the cost to achieve market quality.

However, in a digital crossover system, component variations can be fixed more easily. If the voicing needs to be adjusted because the woofer doesn’t roll off where expected, that’s a software change, not a hardware change. This flexibility allows manufacturers to accept drivers with wider tolerances while still preserving quality and reducing defect rates. Quick correction of part variation also allows designers more time to fine-tune the overall voicing consistency of each system.

Results: Latency

Latency through DSP is sometimes identified as a challenge compared to near zero latency through an analog crossover and amplifier. To quantify this, the digital crossover (analog in to analog out) on the APx555 was measured and a broadband system latency of 3.4 ms was found regardless of EQ correction. This delay should be considered insignificant in all but the most time-critical environments, such as professional recording settings. Bluetooth® Classic can often exceed 100 ms of latency, for example.

Results: EQ Response

Finally, the EQ response was tuned at the listening position in the room. This can be achieved easily through a DSP, offering real-time control and adjustments that are difficult to match in an analog system. This allows for further optimization of the system, including lower observed peaks (in some cases, due to room effects), frequency response extension, and gain matching the tweeters and woofers.

DSP: A Holistic Voicing Approach

Analog crossover designs require filter banks, where each section is matched based on specific design parameters. This approach is well suited to divide and conquer acoustic and electrical domain problems. However, building the perfect filter bank is inconsequential if the speakers are mismatched because the final voicing heard by the listener is the composite acoustic and electrical response.

Using DSP allows for a holistic voicing approach. Speaker bandwidth and sensitivity can be corrected in software. There is no need for resistive networks to gain match between channels; this is simply a slider in SigmaStudio. If the speaker rolls off earlier than expected, the crossover frequency can be adjusted up or down to correct this without changing component values or redesigning the network.

When EQ correction was applied based on the listening position measurements, overall system frequency response was flattened compared to the analog response (Figure 6). High frequencies were extended with a high shelf filter, and bass frequencies were boosted as well. Specific room modes can also be addressed and smoothed, given known listening positions to tune for.

Alignment Flexibility with DSP

Another design advantage of integrating DSP is the ability to fine-tune time alignment and correct mismatch between the woofer and the tweeter. In a traditional analog design, physical components must be carefully aligned to avoid phase and frequency response problems. This limits industrial design choices and could require multiple prototype builds to test alignment properties.

With DSP, the designer gains significant flexibility to create differentiated products. Any resulting misalignment can be easily identified and corrected by reversing the polarity of one of the transducers in SigmaStudio and measuring the frequency response. A sharp null will be observed at the crossover point with a perfectly aligned response. This can be accomplished quickly in a preproduction state.

Filter Design Optimization

In system voicing, one of the most straightforward approaches for filter design is to use predefined filter types (low-pass, high-pass, etc.) and filter class types (Butterworth, Chebyshev, elliptic, and Bessel). Modern filter design generally uses method-constrained optimization approaches such as Parks-McClellan and Yule-Walker.

Using DSP and SigmaStudio, the original topology can be collapsed into four filters and four limiters. The frequency flatness, phase response, time alignment, and cutoff regions can all be used as constraints in constrained optimization. Combining the finite and infinite impulse responses (FIR and IIR) of digital filters allows for more optimization options.

Digital speaker voicing also allows for much more platform reuse as many products have different driver combinations but similar speaker power requirements. Using a DSP enables a single board that could be used for multiple products. This capability is unavailable in an analog crossover design, where adjustability and topology are fixed at the initial design. In digital crossover design, topology and adjustability are merely variables that can be replaced at will.

Testing Free-Field Response

A final test was run to understand the free-field response of the speaker in open space (in this case, the roof of our lab) to avoid any reflections (Figure 7). The free-field response was an important test to verify whether DSP introduces ringing artifacts or group delay.

Upon reviewing the spectrogram of analog and digital systems (Figure 8), no additional ringing was seen in the digital system. This confirms that DSP crossovers do not add any negative time-domain effects to playback. In fact, the analog system has additional resonances at 300 Hz and 500 Hz. The air motion tweeter has relatively consistent performance across digital and analog crossovers.

The dashed line on the spectrograms in Figure 8 denotes the peak amplitude of the spectrum. The time in milliseconds on the plot is taken from the peak amplitude, not the beginning of the measurement, which is why some millisecond values are negative on the plot. The speakers were placed on a table to elevate them above the railing as a source of reflections. However, raising the speaker added a ground reflection that created a notch at 600 Hz.

Conclusion

The analog and digital crossovers tested have similar performance, but a smoother response from the ADAU1467 DSP signal path was observed, despite implementing a higher order filter. This defies the conventional wisdom that analog crossovers are superior to their digital counterparts.

On a practical note: the passive system as tested had a bill of materials (BOM) costing approximately $137 in mid 2024, while the digital system’s BOM was $28 (quoted at 10 to 100 pieces). Notably, this BOM does not include a requirement of the digital crossover system, and that is the need to bi-amp the system. However, a lower power amplifier can be used to drive the high frequency transducer.

Digital voicing can be done much more simply and at a lower cost than with an analog system. Any sort of in-room speaker voicing can be done inside the DSP easily, and manufacturers implementing DSP have sometimes given this control to end consumers in the form of apps and digital room correction.

Although good analog design will continue to be a staple for audio engineers in years to come, DSP is gaining much more acceptance in the industry to improve products, lower costs, speed time to market, and make end-of-line optimizations that are not possible in the analog domain.

In addition, there are hundreds of additional functions and algorithms available to product designers looking for marketable differentiation and custom performance. Many products in the SigmaDSP line integrate asynchronous sample rate converters (ASRCs), which allow multiple digital inputs with different clock domains to run together, creating flexibility for different use cases and sources.

Other algorithms such as equal loudness compensation, tone generation, speaker management/diagnostics, mixing/muxing, dynamics processing, and GPIO conditioning are available to users of this software at no additional cost.

While there are other measurements that we plan to undertake for a future article, this first attempt at quantifying performance using DSP has shown that the benefits are abundant and clear.

About the Authors

Phenix Nunlee

Phenix Nunlee is an audio product applications engineer at Analog Devices. He holds a bachelor’s degree in engineer science with a specialization in electrical engineering with minors in electrical engineering and music te...

Ryan Boyle

Ryan Boyle is an audio marketing manager for Analog Devices consumer business unit. Ryan holds an electrical engineering degree from UMass Lowell with a minor in sound recording technology. Prior to joining ADI, he worked ...

Matthew Tyler

Matthew Tyler is the managing director of Wearable & Prosumer Solutions at Analog Devices, Inc. (ADI). He is a passionate innovator with a lifelong love of audio and music technologies. He graduated from the University of ...

David M. Thibodeau

David Thibodeau started his audio electronics training in 1977 with the military. He then went on to work at several recording studios including Middle Tennessee State University, where he held the position of chief engine...