Tests reveal how supposedly flawless DACs add noise bursts that look invisible on a spec sheet.
Twenty years ago, John Siau was testing a sample rate converter (SRC) at Benchmark Media when something strange happened. The measurements looked flawless, with a 130 to 140 dB signal-to-noise ratio, textbook performance.
But he could hear something. With the chip in the signal path, the music sounded brighter. The specs said the chip shouldn’t be doing anything audible. His ears disagreed.
As he traced the cause, he uncovered a mechanism called intersample clipping that, as we’ll see, affects nearly every DAC.
What Causes Thousands of Clips Per Song
The issue is called intersample clipping, and it can happen hundreds, and sometimes over a thousand, times in a single song
PCM digital audio captures amplitude at fixed sample points, like measuring ocean waves only at predetermined intervals. But the actual peaks of the waveform can fall between those samples. When mastering engineers normalize audio to 0 dBFS (the digital maximum), they’re measuring those sample points, not the true peaks.
As your DAC reconstructs the analog waveform through interpolation, those between-sample peaks emerge. They can exceed 0 dBFS by up to 3.01 dB. If the DAC has no headroom to accommodate them, it clips.
Those clipped bursts create intermodulation products manifest as added brightness rather than obvious crackling, which is precisely why it went undetected for decades.
But this distortion isn’t rare. Benchmark’s analysis of Steely Dan’s “Gaslighting Abbie” found 1,129 intersample overs in five minutes, roughly 3.7 per second. The company’s testing revealed something damning.
No chip manufacturer, they noted, has adequately addressed the issue.
Why Your DAC Chip Doesn’t Matter
Audiophiles spend endless forum threads debating ESS versus AKM chips as if the silicon alone determines sound quality. Siau thinks this is misguided.
“A DAC is not simply about the chipset,” he said. “It’s like saying a car is only determined by the engine that’s in it. Well, that’s not the case.”
The numbers bear this out. About 90% of the components in a digital-to-analog converter are analog: I/V converters, voltage references, resistor matching, and power supply regulation. All of it affects the final sound. The same ESS chip wired in voltage mode (lowest parts count, lowest performance) versus current mode (best performance) yields dramatically different results.
When comparing DACs with identical chips, they found 90% of measurable performance differences trace to analog implementation.
Benchmark uses ESS chips in its DACs. It also bypasses their built-in digital filters entirely.
Filters That Fundamentally Fail
If chip selection is oversold, what about all those filter options that DAC manufacturers advertise? Minimum phase, slow rolloff, NOS?
Siau is blunt. They fundamentally fail.
Take the “no oversampling” or NOS approach, marketed as delivering “better transients” by eliminating the ringing associated with digital filters. The problem is what happens without proper filtering. Aliasing.
The marketing demos don’t help either. DAC manufacturers show pristine square waves with no ringing, proving their alternative filters work.
Benchmark tests multiple generations of analog-to-digital-to-analog conversion. With linear phase filters, the waveform remains identical through four generations. With alternative filters, he noted, “it changes radically as we go down the signal chain.”
If your goal is coloring the sound, alternative filters can do that. But accuracy?
“I can put on pink glasses when I’m watching a movie,” Siau offered. “But it’s not going to be the most accurate representation.”
The Fix Nobody Ships
Benchmark’s fix is straightforward. Build around 3.5 dB of digital headroom before the D/A chip. This allows the interpolation filter to reconstruct those between-sample peaks without clipping.
Its trade-off is real, as you lose roughly 3 dB on the spec sheet. But independent measurements confirmed that Benchmark’s DAC3 handles +3 dB intersample overs without distortion.
However, the broader industry has quietly acknowledged the problem, even if DAC manufacturers haven’t.
All major streaming services now recommend -1 dBTP true-peak headroom: Spotify, Apple Music, YouTube.
For consumers, the takeaway is simple. Look for DACs that specify intersample headroom. Or just use digital volume control. Even -3 dB provides protection against the clipping Siau identified two decades ago.
The industry, meanwhile, continues selling you chip wars and filter options while ignoring what Benchmark has called “the elephant in the room.”
A ‘phile and his money are soon parted.
How can we identify which DACs suffer from this problem?