DAC-equipped devices that went through a major independent reviewer’s test bench between August 31 and October 30, 2026 measured anywhere from 61 dB to 122 dB SINAD, a 61 dB gap that equals about 1,122 times more noise and distortion voltage at the bottom of the range (10^(61/20) = 1,122) (Audio Science Review measurement database, 2026). On paper the chips inside these products look far closer to each other: the ES9039PRO flagship is rated at -122 dB THD+N per channel by its maker (ESS Technology announcement, 2022, vendor claim), and the AK4499EX at -124 dB THD (AKM product page, vendor claim). Meanwhile, a 34-listener AES study found no significant link between THD and distortion ratings (r = -0.42, p = 0.06) (Lee and Geddes, AES Convention Paper 5891, 2003). We aggregated data from that database, the Archimago’s Musings measurement blog, Audio Precision, Analog Devices, ESS Technology, AKM, Texas Instruments, and research published by the Audio Engineering Society and in Frontiers in Psychology. For sales and market size of this gear, see our headphone DAC and amplifier market statistics; this page sticks to what the converters actually measure.
TL;DR
- DAC-equipped devices among the 20 latest database entries (August 31 to October 30, 2026) measured 61 dB to 122 dB SINAD (Audio Science Review measurement database 2026).
- The ES9039PRO chip is rated at -122 dB THD+N and +132 dB dynamic range per channel (ESS Technology announcement 2022, vendor claim).
- The AK4499EX chip is rated at -124 dB THD and 135 dB S/N in stereo (AKM product page, vendor claim).
- A perfect 16-bit converter tops out at 98.08 dB, a perfect 24-bit converter at 146.24 dB (calculation with the Analog Devices MT-003 formula).
- An ES9038Q2M dongle hit about -120 dB THD+N, as advertised, but rose to -96 dB (0.0016%) above -0.5 dBFS (Archimago’s Musings 2022).
- The ES9039Q2M datasheet’s own THD+N slides from -120 dB to -108 dB as the band widens from 20 kHz to 160 kHz (ESS datasheet v0.2.2, vendor claim).
- At 48 kHz sampling, THD+N is measured through a 22 Hz high-pass and a 20 kHz AES17 low-pass filter (Texas Instruments TIDUEV2 2025).
- A flagship lab analyzer is specified at -117 dB residual THD+N plus 1.0 microvolt, typically better than -120 dB (Audio Precision APx555 specifications).
- THD did not significantly predict distortion ratings from 34 listeners: r = -0.42, p = 0.06 (Lee and Geddes, AES 2003).
- A 16-listener study used 15.3% THD as its ‘mild’ saturation level for speech (Biberger and Ewert, Frontiers in Psychology 2022).
- A meta-analysis of 18 studies, 450 participants and over 12,000 trials found trained listeners told hi-res from standard audio apart about 60% of the time (Reiss, JAES 2016).
1. What SINAD and THD+N Mean, in Plain Numbers
SINAD (signal to noise and distortion) answers one question: how loud is the wanted tone compared with all the rubbish the device adds to it, hiss and harmonics together? Audio Precision defines SINAD as the reciprocal of THD+N (total harmonic distortion plus noise), so the two are the same ratio seen from opposite ends: 120 dB SINAD is the same as -120 dB THD+N, or 0.0001% (100 x 10^(-120/20)). A bigger SINAD is better; a more negative THD+N is better (What Is SINAD? explainer).
The useful mental model is bits. Every 6.02 dB of SINAD is worth roughly one bit of clean resolution, so a 98 dB device behaves like a perfect 16-bit converter and a 122 dB device like a perfect 20-bit one. That is why the CD-era ceiling of about 98 dB keeps reappearing in the measurements in sections 3 and 4 (MT-003 converter tutorial).
| Metric | Value | Source |
|---|---|---|
| Relationship between SINAD and THD+N | SINAD is the reciprocal of THD+N; it can never be less than 1, so it is always positive in dB | Audio Precision, What Is SINAD? |
| Test tone frequency usually used for a SINAD measurement | 400 Hz or 1 kHz | Audio Precision, What Is SINAD? |
| Theoretical signal-to-noise ratio of a perfect N-bit converter | 6.02N + 1.76 dB (74 dB for a 12-bit converter) | Analog Devices, MT-003 Rev. A 2008 |
| Converter bit depth implied by a SINAD figure (effective number of bits, ENOB) | ENOB = (SINAD - 1.76 dB) / 6.02 dB | Analog Devices, MT-003 Rev. A 2008 |
| Harmonics that usually dominate a THD figure | The first 5 harmonics | Analog Devices, MT-003 Rev. A 2008 |
| Condition under which THD+N equals SINAD | Noise measured over the full band from DC to half the sample rate | Analog Devices, MT-003 Rev. A 2008 |
| THD+N of -120 dB written as a percentage | 0.0001% (100 x 10^(-120/20)) | Calculation from the definition above |
Most recent available data: MT-003, revision A (2008). The math has not changed; the definitions still govern every datasheet and review cited below.
2. Chip Datasheets vs Finished Devices: Vendor Claims Under Independent Test
A datasheet figure describes a bare chip on the maker’s evaluation board, under the maker’s chosen conditions. A finished product adds a USB receiver, power supply, output stage and volume control, any of which can become the bottleneck. The independent measurements below show that a well-built product can hit its advertised -120 dB, and that an old chip sets a hard limit no matter how good the rest of the box is. The SMSL R100’s AD1861 chip was specified at 0.003% THD+N, which is -90.4 dB, and the finished DAC measured exactly 90.4 dB SINAD.
Treat the three vendor rows as what they are: the chipmakers’ own claims, quoted under their own test conditions (ESS Technology announcement, May 16, 2022; AK4499EXEQ product page). The independent rows come from bench tests of shipping products (E1DA #9038D6K dongle review).
| Metric | Value | Source |
|---|---|---|
| ES9039PRO chip, THD+N and dynamic range per channel | -122 dB THD+N; +132 dB dynamic range (+140 dB in mono mode) | ESS Technology announcement, May 16, 2022 (vendor’s own claim) |
| ES9039Q2M chip, THD+N and THD at full scale (0 dBFS), 20 Hz to 20 kHz band, 48 kHz sampling | -120 dB THD+N; -126 dB THD; 130 dB A-weighted dynamic range | ES9039Q2M datasheet v0.2.2 by ESS Technology (vendor’s own claim) |
| AK4499EX chip, distortion and signal-to-noise ratio | -124 dB THD (listed as THD, not THD+N); S/N 135 dB in stereo, 138 dB in mono | AKM AK4499EXEQ product page (vendor’s own claim) |
| E1DA #9038D6K dongle built on the ES9038Q2M chip, measured THD+N at -0.7 dBFS (2.6 Vrms) | About -120 dB, matching the dongle maker’s advertised figure | Archimago’s Musings, review of December 6, 2022 |
| Same E1DA #9038D6K dongle, measured unweighted dynamic range | Close to 124 dB | Archimago’s Musings, review of December 6, 2022 |
| Topping DX9 built on the AK4499EQ chip, third harmonic of a 10 kHz tone at 0 dBFS | About -118 dB | Archimago’s Musings, Topping DX9 follow-up, June 14, 2025 |
| Measured SINAD of two desktop DACs reviewed in September and October 2026: D&A Alpha (399 dollars) and SMSL R100 (289 dollars, AD1861 chip released 36 years ago) | 122 dB and 90.4 dB | Audio Science Review reviews, September 7 and October 15, 2026 |
Most recent available data: the ES9039PRO launch figures date from 2022, and the E1DA result from December 2022; a newer dongle appears in section 4.
3. Measured SINAD of Desktop DACs, Interfaces and AV Gear in 2026
Price does not sort these results. A 183 dollar USB-to-balanced box measured 61 dB SINAD while a 190 dollar desktop DAC reviewed almost eight weeks later measured 103 dB, a 42 dB gap between products that cost almost the same. At the top end, a 3,500 dollar AV processor (116 dB) and a keyboard with a built-in audio interface (117 dB) sit 1 dB apart. The question for buyers is no longer whether 120 dB is possible, it is whether a given product was engineered to get there.
The bit-depth view from section 1 makes the spread concrete. Products at the bottom of the range resolve less than 10 effective bits, the 1,849 dollar AV receiver lands just above the CD-quality ceiling, and the best desktop units approach 20 bits (measurement database dashboard).
| Metric | Value | Source |
|---|---|---|
| Measured SINAD of two budget dedicated DACs: SMSL SU-3 (190 dollars, October 24, 2026) vs Interspace BalBox USB-to-balanced box (183 dollars, August 31, 2026) | 103 dB vs 61 dB | Audio Science Review measurement database, accessed November 7, 2026 |
| Measured SINAD of audio interfaces, AV processors and a disc player reviewed September to October 2026 | Topping KB22 keyboard with audio interface 117 dB; Topping M62 portable interface 113 dB; miniDSP Tide 16 AV processor (3,500 dollars) 116 dB; Denon AVR-X3900H AV receiver (1,849 dollars) 102 dB; Sony UBP-X700 disc player 99 dB | Audio Science Review measurement database, accessed November 7, 2026 |
| Effective number of bits implied by 122 dB SINAD | About 20 bits ((122 - 1.76) / 6.02 = 19.97) | Calculation with the ENOB formula in section 1 |
| Effective number of bits implied by 90.4 dB SINAD | About 14.7 bits ((90.4 - 1.76) / 6.02 = 14.72) | Calculation with the ENOB formula in section 1 |
| Effective number of bits implied by 61 dB SINAD | About 9.8 bits ((61 - 1.76) / 6.02 = 9.84) | Calculation with the ENOB formula in section 1 |
| SINAD of a perfect 16-bit converter, the CD-quality ceiling | 98.08 dB (6.02 x 16 + 1.76) | Calculation with the ENOB formula in section 1 |
Context: the database stores SINAD as whole numbers, so a device reviewed at 90.4 dB appears there as 90 dB. Each value is the headline SINAD of one review, at the reviewer’s chosen output level, so a 1 to 2 dB difference between two products is within the range that test conditions alone can produce (see section 5).
4. Dongles, Phones and Laptops: The Portable End of the Spread
Portable converters have closed most of the gap with desktop gear. A 100 dollar USB-C dongle measured 116 dB SINAD in September 2026, the same figure as the 3,500 dollar AV processor in section 3. The weak spot is not the class but the individual design: in the same triple-tone distortion plus noise (TD+N) test, a measurement that plays three tones at once, one portable DAC scored -114 dB while another managed only -56 dB.
Phones and laptops are the least measured part of the market. The public dashboard of the same measurement database lists 42 dongle headphone amps and 39 USB dongles, but only 3 cellphones and 1 laptop, out of 1,452 electronic devices tested. For a phone without a headphone jack, the wired path runs through a dongle like the ones below, and the wireless path through a Bluetooth codec, a separate problem covered in our Bluetooth codec real-world bitrate statistics. The one recent laptop result is a forum member’s test of the 2024 MacBook Air (measurement thread, May 30, 2025).
| Metric | Value | Source |
|---|---|---|
| Measured SINAD of the Fosi DS3 USB-C dongle (100 dollars), reviewed September 9, 2026 | 116 dB | Audio Science Review measurement database, accessed November 7, 2026 |
| 2024 MacBook Air headphone jack, best measured SINAD (full scale into a 30 ohm load) | 97.9 dB; output 1.25 V into 30 ohms and 2.97 V into 300 ohms; crosstalk about -50 dB | Audio Science Review forum member measurement (Cosmos ADCiso analyzer), May 30, 2025 |
| Device counts on the public measurement dashboard: dongle headphone amps and USB dongles vs cellphones and laptops | 42 and 39 vs 3 and 1, out of 1,452 electronic devices tested | Audio Science Review dashboard, accessed November 7, 2026 |
| Effective number of bits implied by the MacBook Air result | About 16 bits ((97.9 - 1.76) / 6.02 = 15.97) | Calculation with the ENOB formula in section 1 |
| Triple-tone distortion plus noise (TD+N) of three portable DACs in the same test: E1DA #9038D6K vs Chord Mojo vs Cayin RU6 | -114 dB vs about -104.5 dB vs -56 dB | Archimago’s Musings, review of December 6, 2022 |
Data note: the MacBook Air figure comes from a single user test with a resistive dummy load. The author later withdrew a second, low-level result from the same thread, which is why only the full-scale figure is used here.
5. Methodology: Bandwidth, the AES17 Filter and Test Level Move the Number
Two honest measurements of the same converter can differ by more than 10 dB. The ES9039Q2M datasheet itself shows THD+N worsening from -120 dB to -108 dB when the measurement band is widened from 20 kHz to 160 kHz, because a wider window lets in more ultrasonic noise. The AES17 low-pass filter is how measurement practice keeps that noise out: Texas Instruments’ own DAC reference design applies a 20 kHz AES17 filter at 48 kHz sampling and a 40 kHz one at 96 and 192 kHz (Texas Instruments TIDUEV2 reference design, October 2025).
Test level matters just as much. Some converters, like the E1DA dongle in the table, are cleanest a fraction of a dB below full scale, and an amplifier’s SINAD often improves as output rises because noise stays fixed while the signal grows. That is why a published SINAD figure is only comparable when the output voltage and load are stated alongside it (SMSL H400 MK2 review, October 30, 2026). The analyzer itself also has a floor (Audio Precision APx555 specifications).
| Metric | Value | Source |
|---|---|---|
| Input filters for a DAC THD+N test at 48 kHz sampling | 22 Hz high-pass plus 20 kHz AES17 low-pass | Texas Instruments, TIDUEV2 reference design, October 2025 |
| Input filters for dynamic range and signal-to-noise tests at 96 and 192 kHz sampling | 22 Hz high-pass, 40 kHz AES17 low-pass, A-weighted (THD+N is measured without A-weighting) | Texas Instruments, TIDUEV2 reference design, October 2025 |
| ES9039Q2M chip THD+N as the band widens: 20 kHz (48 kHz sampling), 40 kHz (96 kHz), 80 kHz (192 kHz), 160 kHz (384 kHz) | -120 dB, -117 dB, -114 dB, -108 dB | ES9039Q2M datasheet v0.2.2 by ESS Technology (vendor’s own claim) |
| SINAD of one analog headphone amp and preamp (SMSL H400 MK2) at 4 Vrms vs 5 Vrms input | 125 dB vs 127 dB; the DAC feeding it fell to 116 dB SINAD when its output was set 12 dB lower | Audio Science Review review by NTTY, October 30, 2026 |
| THD+N of the E1DA dongle just below full scale vs above -0.5 dBFS | About -120 dB at -0.7 dBFS vs up to -96 dB (0.0016%) above -0.5 dBFS | Archimago’s Musings, review of December 6, 2022 |
| Harmonics counted for a 10 kHz test tone when the 20 kHz low-pass filter is switched off (192 kHz sampling) | Harmonics up to 90 kHz | Archimago’s Musings, Topping DX9 follow-up, June 14, 2025 |
| Residual THD+N of a reference audio analyzer in a 22 kHz band | -117 dB plus 1.0 microvolt specified; typically better than -120 dB at 1 kHz and 2.0 V | Audio Precision APx555 specifications (vendor’s own claim) |
Outlier note: the 125 dB to 127 dB results above exceed the typical residual of the reference analyzer in the last row. They were taken with a different chain, an E1DA Cosmos ADCiso with a Cosmos Scaler front end, which is a reminder that the test equipment is part of every figure at this level.
6. Audibility: What Listening Research Says About Distortion
The listening research points away from single-number rankings. In a 34-listener test of 21 distorted stimuli, THD had no significant relationship with ratings (r = -0.42, p = 0.06), while a metric that weights distortion by how the ear hears it reached r = 0.95 on 18 of the stimuli (Lee and Geddes, AES Convention Paper 5891, 2003). A separate JAES study found moderately strong negative correlations between quality ratings and a physical distortion measure called DS, built as an alternative to conventional figures (Tan, Moore and Zacharov, JAES 2003).
The distortion levels used in perception studies also sit in a different universe from DAC measurements. A 2022 quality study labelled 15.3% THD on speech as only ‘mild’ (Biberger and Ewert, Frontiers in Psychology 2022). That is about -16.3 dB (20 x log10(0.153)), more than 100 dB above the -120 dB that current DAC chips claim. Headphone drivers add distortion of their own, a separate story covered in our planar magnetic vs dynamic driver headphone THD statistics. Even for hi-res formats, which change noise floor and bandwidth rather than distortion, a pooled analysis found trained listeners right only about 60% of the time (Queen Mary University of London, C4DM summary of Reiss 2016).
| Metric | Value | Source |
|---|---|---|
| Size of the Lee and Geddes listening test | 34 normal-hearing listeners, 21 stimuli, 7-point scale from ‘better than reference’ to ‘intolerable’ | Lee and Geddes, AES Convention Paper 5891, 2003 |
| Correlation of conventional distortion figures with listener ratings | THD r = -0.42 (p = 0.06); IMD r = -0.35 (p = 0.13); neither significant | Lee and Geddes, AES Convention Paper 5891, 2003 |
| Correlation of the perception-weighted GedLee (Gm) metric with listener ratings | r = 0.68 (p < 0.001) on all 21 stimuli; r = 0.95 (p < .001) on 18 stimuli | Lee and Geddes, AES Convention Paper 5891, 2003 |
| GedLee metric value below which systems are expected to sound distortion-free | Gm < 1.0 predicts ratings of ‘imperceptible’ | Lee and Geddes, AES Convention Paper 5891, 2003 |
| Relationship between the physical distortion measure DS and perceived quality of speech and music | Moderately strong negative correlations | Tan, Moore and Zacharov, JAES vol. 51 no. 11, 2003 |
| THD of the ‘mild’ and ‘moderate’ saturation stimuli in a 16-listener quality test | Speech 15.3% and 21.4%; guitar 22.1% and 27.1% | Biberger and Ewert, Frontiers in Psychology 2022 |
| Pooled evidence on hearing hi-res vs standard-resolution audio | 18 studies, 450 participants, over 12,000 trials; trained listeners about 60% correct (chance is 50%) | Reiss, JAES June 2016, via Queen Mary University of London C4DM |
Most recent available data: the core distortion-perception papers date from 2003, and the meta-analysis from 2016. We found no newer peer-reviewed study that sets a THD audibility threshold for DAC-level distortion with music.
Summary: DAC Chip SINAD & THD+N by the Numbers
| Metric | Value | Source |
|---|---|---|
| ES9039PRO chip THD+N and dynamic range per channel | -122 dB; +132 dB | ESS Technology announcement 2022 (vendor’s own claim) |
| ES9039Q2M chip THD+N at 0 dBFS in a 20 kHz band vs a 160 kHz band | -120 dB vs -108 dB | ES9039Q2M datasheet v0.2.2 by ESS Technology (vendor’s own claim) |
| AK4499EX chip THD and stereo S/N | -124 dB; 135 dB | AKM AK4499EXEQ product page (vendor’s own claim) |
| Measured THD+N of an ES9038Q2M dongle at -0.7 dBFS | About -120 dB | Archimago’s Musings 2022 |
| Triple-tone distortion plus noise of the weakest portable DAC in the same test | -56 dB | Archimago’s Musings 2022 |
| SINAD ceiling of a perfect 16-bit converter | 98.08 dB | Calculation with the Analog Devices MT-003 formula |
| SINAD ceiling of a perfect 24-bit converter | 146.24 dB (6.02 x 24 + 1.76) | Calculation with the Analog Devices MT-003 formula |
| Effective number of bits implied by 122 dB SINAD | About 20 bits | Calculation, section 3 |
| Relationship between SINAD and THD+N | Reciprocal of each other | Audio Precision, What Is SINAD? |
| AES17 low-pass filter for THD+N at 48 kHz sampling | 20 kHz, with a 22 Hz high-pass | Texas Instruments TIDUEV2 2025 |
| Correlation of the physical distortion measure DS with quality ratings of music and speech | Moderately strong negative correlations | Tan, Moore and Zacharov, JAES 2003 |
| Reference analyzer residual THD+N, 22 kHz band | Typically better than -120 dB | Audio Precision APx555 specifications (vendor’s own claim) |
| Correlation of THD with listener distortion ratings | r = -0.42, p = 0.06 | Lee and Geddes, AES 2003 |
| Correlation of the GedLee metric with ratings (18 stimuli) | r = 0.95 | Lee and Geddes, AES 2003 |
| THD of the ‘mild’ speech saturation stimulus | 15.3% | Biberger and Ewert, Frontiers in Psychology 2022 |
| Trained listeners distinguishing hi-res from standard audio | About 60% | Reiss, JAES 2016 |
| Studies and participants pooled in that meta-analysis | 18 studies, 450 participants | Reiss, JAES 2016 |
Methodology and Sources
- Audio Science Review: measurement database dashboard and electronics list (accessed November 7, 2026), SMSL H400 MK2 review by NTTY (October 30, 2026), 2024 MacBook Air headphone output measurements (May 30, 2025).
- Archimago’s Musings: E1DA #9038D6K dongle review (December 6, 2022) and Topping DX9 follow-up tests (June 14, 2025).
- Audio Precision: What Is SINAD? and APx555 specifications.
- Analog Devices: MT-003, Understand SINAD, ENOB, SNR, THD, THD + N, and SFDR (Walt Kester, Rev. A, 2008).
- ESS Technology: ES9039PRO and ES9027PRO announcement (May 16, 2022) and ES9039Q2M datasheet v0.2.2. Vendor claims.
- AKM (Asahi Kasei Microdevices): AK4499EXEQ product page. Vendor claims.
- Texas Instruments: TIDUEV2 Precision DAC for Premium Audio reference design (October 2025).
- Lidia W. Lee and Earl R. Geddes: Auditory Perception of Nonlinear Distortion, AES Convention Paper 5891 (115th AES Convention, October 2003).
- Chin-Tuan Tan, Brian C. J. Moore and Nick Zacharov: The Effect of Nonlinear Distortion on the Perceived Quality of Music and Speech Signals (JAES vol. 51 no. 11, 2003).
- Thomas Biberger and Stephan D. Ewert: Binaural detection thresholds and audio quality of speech and music signals in complex acoustic environments (Frontiers in Psychology, 2022).
- Joshua Reiss: A Meta-Analysis of High Resolution Audio Perceptual Evaluation (JAES, June 2016), as summarized by Queen Mary University of London, Centre for Digital Music.
- Data watch: SINAD figures are not comparable unless output voltage, load and measurement bandwidth match, and the reviews above use different analyzers (Cosmos ADCiso chains and lab analyzers). The database rounds SINAD to whole dB (90.4 dB appears as 90 dB) and does not label the column; the column was matched to SINAD by checking it against review text (125 dB for the H400 MK2, 90.4 dB for the R100). The ES9039 announcement prints the mono figure as ‘+140B DNR’, an evident typo for +140 dB. The AK4499EX figure of -124 dB is listed as THD, not THD+N, so it is not directly comparable with the ES9039 THD+N claims. Lee and Geddes reach r = 0.95 only after excluding 3 of 21 stimuli (r = 0.68 on all 21). The Reiss meta-analysis figures come from the university’s summary; press coverage gives slightly different participant and trial counts. Phone and laptop data are thin: only one recent laptop measurement was found, and we found no 2025 or 2026 peer-reviewed survey of phone or laptop headphone outputs.
- Last updated: November 7, 2026. We update this roundup quarterly, and the next refresh is expected when the AES Europe Spring 2027 Convention papers are published.