Choosing a DAC: The Chipset behind the Sound
The figures in Part 2, such as 120 dB SNR and 0.0001% THD+N, depend on hardware. They come from the design of the chipset (IC) at the center of the board and the circuitry around it.
Most audio devices use components from a relatively small group of chipmakers. This chapter looks at the main Hi-Fi DAC architectures, the stories behind them, and the sound characteristics associated with each manufacturer.
Understanding DACs for Beginners: A 3-Part Series
- Part 1: What is an Audio DAC? The first thing to buy when your earphones sound dull
- Part 2: How to read DAC spec sheets in 1 minute: 24-bit? SNR?
- Part 3: Choosing a DAC by Understanding Its Chipset
ESS, AKM, and Cirrus Logic: The Top 3 Delta-Sigma Designs
More than 90% of modern DAC chipsets use delta-sigma conversion, which increases resolution through high-frequency oversampling. Manufacturers differ in their decoding algorithms and filters, producing subtle variations in the output waveform. Here are the three major names you’ll encounter.
ESS Technology (Sabre DAC)
– Source: DANAWA –
ESS has the largest share of today’s global Hi-Fi market. Its ES90xx family, including the ES9038PRO and ES9039PRO, appears in flagship desktop gear. The ES92xx family, such as the ES9219, is intended for portable dongles and DAPs.
Measurements by well-known overseas audio communities once exposed what became known as the “ESS Hump,” or IMD Hump. ESS-based devices showed an unexpected increase in intermodulation distortion at particular middle output levels, rather than a uniformly poor result across the range. The finding became a talking point precisely because the behavior appeared in multiple products. An interesting consequence was the pressure it placed on the circuitry around the converter. Manufacturers had to pay closer attention to amplifier design, and the episode helped push that side of the market forward. It is a useful reminder that a chip’s reputation and the results of a finished device are connected through the way that device is built.
ESS uses its proprietary HyperStream algorithm to keep THD low. The Sabre sound is associated with defined edges, clean decay, and precise instrument placement. If you favor an analytical, monitoring-style presentation with little embellishment, an ESS-based device is worth considering.
AKM (Asahi Kasei Microdevices)
– Source: DANAWA –
Japan’s AKM aims to soften the sharp, cold character sometimes associated with digital audio. Its DACs use AK44xx numbering, including AK4490, AK4493, and AK4499.
A major fire at AKM’s Nobeoka semiconductor plant in October 2020 destroyed the facility and disrupted worldwide chip supplies. Many audio manufacturers moved new products to ESS as a result. AKM later returned with a flagship architecture that separates digital processing in the AK4191 from analog output in the AK4499EX.
Its “Velvet Sound” branding reflects the intended character. Suppressing ringing at the end of notes produces a smooth texture and rich lower-midrange harmonics reminiscent of analog tape. Dense vocals and comfortable long listening sessions have made the chips popular in high-end portable devices, including those from Astell&Kern.
Cirrus Logic (CS)
– Source: DANAWA –
Cirrus Logic is a US semiconductor maker focused on efficient designs for mobile and low-power applications. Its DACs use CS43xx numbering, including the CS43131 and CS43198.
It acquired British audio chipmaker Wolfson in 2014, bringing in its engineering experience and warm sonic character.
The CS43131 integrates amplification efficiently to limit heat and battery drain. It’s widely used in dongle DACs and midrange DACs and DAPs, where preserving a phone’s charge matters. The presentation is generally flat and clean, with little emphasis on individual frequency bands and a hint of the natural warmth associated with Wolfson.
R-2R: Converting with a Resistor Ladder
Inside the Gustard R26 R2R + 1Bit DSD DAC
Alongside delta-sigma designs, R-2R ladder networks form another important branch of high-end audio.
Instead of assigning conversion to a single chipset, a ladder design lays out hundreds of physical resistors directly on the circuit board. The parts are arranged in a closely spaced ladder network, and switches open or close in response to the incoming 0 and 1 values. Those changes distribute voltage through the network to produce the output. It is a very tangible way of performing the conversion: the relationships between physical components do the work that the more familiar integrated designs handle inside a chip. That different construction is central to both the attraction and the difficulty of the approach.
A 24-bit implementation needs many resistors per channel, and even a 0.01% error in one component can affect the sound image. The approach is difficult, costly, and physically large. Its appeal is the option of NOS (Non-Oversampling) operation, bypassing the digital filter for natural harmonic characteristics that suit listeners who prefer an analog texture.
Why High-End Designs Use Chips in Parallel
TOPPING D90 III
Even below the most expensive price brackets, a DAC may contain two flagship chips such as the ES9039PRO, each supporting 8-channel processing. Music is normally 2-channel stereo, so why provide 16 audio channels in total (8 channels x 2)?
The same signal is sent through multiple channels, and their outputs are combined. The wanted audio adds together, while random noise from the individual paths is reduced relative to that signal because it doesn’t align in the same way.
Assigning a chip’s 8 paths as 4 for the left channel and the remainder for the right, then combining each set, can theoretically improve SNR by about 6.02dB. Manufacturers use this parallel arrangement to suppress noise at the conversion stage, before the signal reaches the amplifier.
The Chipset Is Only the Starting Point
After comparing chips and architectures, it’s worth remembering a common observation in this hobby: “the chipset doesn’t determine the sound.” Even an excellent flagship component can’t show its full capability if the board’s power supply is poorly designed.
Power is part of that larger picture. A good DAC design uses LDO (Low Drop-Out) regulators inside the device to keep noise arriving through USB power or an adapter away from the audio circuitry. The aim is to supply the individual components with clean DC, so the converter has the conditions it needs to perform well. Conversion isn’t the end of the process, either. The small analog signal leaving the chipset still has to pass through the amplifier stage. How that stage handles and amplifies it contributes to the finished device’s character. The chipset and surrounding circuitry therefore need to be considered together, rather than treating the converter’s name as a complete description of the sound.
Choosing an audio device therefore means looking beyond an expensive chip. The overall engineering matters: how well the design controls digital noise and preserves the analog signal through the rest of the circuit. The quality of that complete design is what I’d weigh against the price, rather than the chipset name alone.
I hope this three-part guide helps you choose equipment whose design, character, and cost make sense for the way you want to listen. Understanding a little of what happens at the boundary between digital data and analog sound can make the hobby easier to enjoy.
