Less added.
Evaluate noise, distortion and unwanted changes to the signal.
SZABOWORKS / AUDIO RESEARCH PROPOSED STANDARD
SZW Extreme Fidelity
Not a sound.
A standard.
A clearer way to know what your audio equipment can do. Measured from the first input to the final output.
EF.EXTREME FIDELITY
SIGNAL INTEGRITY, BY DESIGN
Capture exactly. Process intentionally. Reproduce exactly.CONCEPT PREVIEW · OCTOBER 2026
Great audio.
With the evidence
to back it up.
A premium price is not a specification.
Extreme Fidelity is our proposal for a public, measurement-first audio standard. The goal is simple: help you choose equipment that preserves the signal, behaves predictably and performs consistently.
Its limits, test conditions and results would be published together. So you could see what was tested, how it was tested, and what actually passed.
More understood.
Know how equipment behaves at different levels, temperatures and loads.
Proof included.
Connect each future mark to a public report and the exact version tested.
ONE IDEA. TWO PERFORMANCE TIERS.A mark that
A mark that
should mean something.
Two proposed tiers. A clear category. And evidence behind every future claim.
EF.CERTIFIEDPROPOSED MARK
EF Certified
Would identify equipment meeting all applicable core requirements for measured performance, stability and consistency.
EF.REFERENCEPROPOSED MARK
EF Reference
Would identify equipment meeting stricter limits across the same applicable tests. A complete path would earn the System designation separately.
Mark concepts shown for the proposed program. Not licensed certification badges.
Good parts are a start.
The whole system matters.
A DAC, processor and amplifier can each perform well on their own. EF System would test how they work together, from the input you use to the output you hear.
Digital input→DSP→DAC→Amplifier→Speaker output
ONE END-TO-END TEST · EF REFERENCE SYSTEM, IF QUALIFIED
Integration brings its own questions: shared power supplies, interference, channel separation, gain, clipping and processing. An approved module would never automatically certify the finished product.
The same principle applies to a microphone input feeding digital audio, an analog input feeding speakers, or a digital input feeding headphones. Every claimed path would be identified in the report.
BEYOND A SINGLE NUMBERDifferent speakers.
Different speakers.
The same intention.
The proposed EF Real Load Test would measure how an amplifier’s output changes as its electrical load changes.
Resistive and simulated speaker loads would be evaluated across frequency and level. The aim: understand behavior beyond a single 1 kHz result.
Load networks, phase angles, stress limits and pass criteria remain to be validated. Tests would stay within the product’s declared operating envelope.
PROPOSED LOAD FAMILIES
AEveryday listening8 Ω speaker-equivalent concept
BA more demanding system4 Ω reactive-load concept
CHigh-current capability2 Ω stress concept · where supported
DCapacitive behaviorSpecialist-load concept · where supported
FOR THOSE WHO WANT THE DETAILSOpen numbers.
Download proposal (.txt) ↓Open numbers.
Open questions.
Six proposed starting categories. Explore the candidate targets below.
Draft 0.1 · October 2026. These values are engineering proposals, not validated audibility thresholds, measured product results or final pass/fail rules.
PROPOSED CATEGORY / DAC
DRAFT 0.1Digital in. Analog out.
The complete analog output, including its power supply, filter and output stage.
| Measurement | EF Certified | EF Reference |
|---|---|---|
| THD+N · 1 kHz | ≤ −110 dB | ≤ −118 dB |
| Dynamic range · A-weighted | ≥ 120 dB | ≥ 128 dB |
| Response · 20 Hz–20 kHz | ±0.10 dB | ±0.03 dB |
| Channel mismatch | ≤ 0.10 dB | ≤ 0.03 dB |
| Crosstalk · 1 kHz isolation | ≥ 110 dB | ≥ 120 dB |
| SMPTE IMD | ≤ −100 dB | ≤ −110 dB |
| Idle artifacts | < −120 dBFS | < −130 dBFS |
| Output impedance | < 100 Ω | < 20 Ω |
Beyond the numbers. Level and frequency sweeps, low-level linearity, multitone, reconstruction filters, source-jitter susceptibility, balanced output symmetry and distortion versus load.
PROPOSED CATEGORY / ADC
DRAFT 0.1Analog in. Every detail accounted for.
The input stage and conversion path measured together.
| Measurement | EF Certified | EF Reference |
|---|---|---|
| Dynamic range | ≥ 118 dB | ≥ 125 dB |
| THD+N · −1 dBFS | ≤ −105 dB | ≤ −115 dB |
| Response · 20 Hz–20 kHz | ±0.10 dB | ±0.03 dB |
| Crosstalk isolation | ≥ 105 dB | ≥ 115 dB |
| Channel gain mismatch | ≤ 0.10 dB | ≤ 0.03 dB |
| Input noise | Declared + verified | Declared + verified |
Beyond the numbers. Anti-aliasing, overload recovery, low-level linearity, clock behavior and processing disclosure. Noise reduction, AGC and other optional processing would be disabled in Reference Mode.
PROPOSED CATEGORY / Microphone input
DRAFT 0.1Start with a cleaner capture.
The microphone preamp and, where present, its ADC as one usable path.
| Measurement | EF Certified | EF Reference |
|---|---|---|
| Equivalent input noise | ≤ −128 dBu | ≤ −130 dBu |
| Maximum gain | ≥ 60 dB | ≥ 70 dB |
| CMRR · 1 kHz | ≥ 80 dB | ≥ 100 dB |
| THD+N · 20 dB gain | ≤ −100 dB | ≤ −110 dB |
| Gain accuracy | ±0.25 dB | ±0.10 dB |
Beyond the numbers. Source impedance, bandwidth and weighting must accompany EIN. Gain sweeps, overload, input symmetry, RF susceptibility and phantom-power voltage, current, noise and transients would be evaluated.
PROPOSED CATEGORY / Power amplifier
DRAFT 0.1Stay faithful. Even under load.
Performance at the speaker terminals, across supported loads and output levels.
| Measurement | EF Certified | EF Reference |
|---|---|---|
| THD+N · 5 W / 4 Ω | < −100 dB | < −110 dB |
| THD+N · 100 W / 4 Ω | < −95 dB | < −105 dB |
| SNR · A-weighted | > 115 dB | > 125 dB |
| Output noise | < 20 µV | < 10 µV |
| Response · 20 Hz–20 kHz | ±0.10 dB | ±0.03 dB |
| Output impedance | < 20 mΩ | < 5 mΩ |
| Crosstalk isolation | > 100 dB | > 115 dB |
| DC offset | < 10 mV | < 3 mV |
Beyond the numbers. Frequency and power sweeps, continuous and burst output, reactive loads, thermal drift, clipping recovery, channel loading and protection behavior. The 100 W point applies only to equipment rated to support it; lower-power qualification rules are still to be defined. No forced 2 Ω rating.
PROPOSED CATEGORY / Headphone amp
DRAFT 0.1Quiet enough for the small details.
Noise and control across sensitive earphones and demanding headphones.
| Measurement | EF Certified | EF Reference |
|---|---|---|
| Supported resistive loads | 16 / 32 / 80 / 150 / 300 / 600 Ω | Same supported-load matrix |
| Output impedance | Limit under development | Candidate target < 0.5 Ω |
| Noise, distortion and power | Limits under development | Limits under development |
| Volume tracking and crosstalk | Limits under development | Limits under development |
Beyond the numbers. Sensitive-IEM noise, voltage and current capability, low-volume channel balance and reactive-load stability. Test only the loads claimed by the manufacturer; publish exclusions.
PROPOSED CATEGORY / Complete system
DRAFT 0.1The entire path earns the mark.
A module-level result would never substitute for end-to-end measurements.
| Measurement | EF Certified | EF Reference |
|---|---|---|
| Signal path | Every claimed input-to-output mode | Every claimed Reference path |
| Noise, distortion and headroom | Path-specific limits pending | Path-specific limits pending |
| DSP and sample-rate handling | Declared and evaluated | Reference Mode evaluated |
| Hardware and firmware | Report exact revisions | Report exact revisions |
Beyond the numbers. Digital to speaker, analog to speaker, microphone to digital, digital to headphone and analog to digital. Combined-stage limits must be established independently of the component tables.
What makes a number comparable?
A released procedure would have to state signal level, gain, sample rate, measurement bandwidth, weighting, load, channels driven, ambient temperature and warm-up period. It would also define fixture tolerances, analyzer residuals, calibration, measurement uncertainty and the decision rule at each limit.
High-frequency distortion tests need an explicit analysis bandwidth and harmonic count. A 20 kHz test with a 20 kHz low-pass measurement cannot describe distortion above that bandwidth. Noise requires RMS/peak and weighting definitions; SNR requires an output reference; microphone EIN requires a defined source impedance.
Continuous output power would be stated with load, frequency, distortion limit, duration and channels driven. Burst output would be reported separately. Results below analyzer capability would be bounded by the measurement limit rather than given invented precision.
TEST THE EXPERIENCE BEHIND THE SPECIFICATIONOne sweep
01 / SIGNAL 02 / OPERATION 03 / PROCESSING 04 / CONNECTIONS 05 / PRODUCTION 06 / CONFIDENCE
One sweep
isn’t the whole story.
Across the range.
Frequency and level sweeps, multitone, linearity, phase, noise spectra, hum, idle tones and channel matching.
After warm-up, too.
Cold startup through two-hour operation, ambient variation, thermal drift, sustained power and overload recovery.
Nothing hidden.
Document EQ, gain, resampling and protection. Evaluate bypass integrity, internal headroom, latency and sample-rate handling.
From source to system.
USB and network integrity, clock behavior, multichannel timing, interference and supported format changes.
More than one sample.
Proposed qualification on at least three randomly selected production units, followed by retail surveillance and revision-based retesting.
Repeatable elsewhere.
Published fixtures and methods, traceable calibration and uncertainty. Independent laboratory participation is a future goal.
THE REPORT IS PART OF THE IDEALook beyond
Look beyond
the badge.
Every future certified model should have a report you can actually read.
See the measured value beside its limit, with uncertainty, conditions and pass margin. Identify the standard version, laboratory, hardware, firmware and sample set.
A future model ID or QR link would lead to that evidence. A certification registry and verification API are planned concepts, not live services.
EF / TEST RECORDFORMAT PREVIEW
- Product / revision
- To be tested
- Standard
- Draft 0.1
- Laboratory / samples
- Not assigned
- Measured values
- Not available
- Uncertainty / conditions
- Required in final report
NO CERTIFICATION ISSUED
Report structure only. No fictional measurements or passing product are shown.
TECHNOLOGY-NEUTRAL. EVIDENCE-FIRST.Any architecture.
Any architecture.
The same expectations.
Class A, AB or D. Tubes or transistors. Discrete or integrated. R2R or delta-sigma. The proposed requirements would follow the result, not the technology used.
No exclusive hardware.Using SzaboWorks components would not be a condition of qualification.
No special treatment.SzaboWorks products would face the same methods and limits as other manufacturers.
No pay-to-pass.Any future fees would pay for testing and administration, never a guaranteed result.
Preserve everything.
Add nothing.
And when you choose processing:
change only what you intended to change.
Ambitious by design.
Developed in stages.
- NOW
Define the proposal.
Publish the scope and candidate targets. Gather technical feedback and identify gaps.
- NEXT
Validate the methods.
Benchmark hardware, characterize fixtures and analyzer limits, and establish uncertainty and decision rules.
- PLANNED
Prove reproducibility.
Pilot measurements across units and laboratories. Refine targets and define governance before releasing a standard.
- FUTURE
Launch qualification.
Only after validation: publish the specification, open testing, issue traceable reports and introduce ongoing surveillance.
Explore the wider research scope
Analog
Line inputs and outputs, preamps, volume controls, balanced interfaces and MM/MC phono stages.
Digital
DSP, clocks, USB, HDMI, streamers, network audio, codec disclosure and wireless synchronization.
Systems
AVRs, multichannel operation, active-speaker electronics, subwoofer amplifiers and room-correction processing.
Speaker intelligence
Voltage/current sensing, impedance and thermal estimation, fault detection, adaptive protection and telemetry accuracy.
Engineering tools
EF Bench, test tracks, automated reports, approved modules, a public registry and a verification API.
Future acoustic research
Speakers, microphones and headphones need separate acoustic methods. Studio, cinema and automotive applications remain longer-term concepts.
What would EF tell me—and what wouldn’t it?
EF would describe measured audio performance within a stated scope. It would not certify personal taste, musicality, room acoustics, speaker placement or subjective claims about cables. Electronics qualification would not establish the acoustic performance of a loudspeaker.
Reference Mode would remove optional coloration and disclose any essential processing. Mandatory protection remains necessary. Wireless qualification would identify codecs and losses, not imply that a lossy link becomes lossless.
EF would not replace applicable electrical safety, regulatory compliance or product approvals. It is a proposed audio-performance program.
Can I submit a product or use the badge today?
Not yet. Applications, licensing, approved laboratories and pricing have not launched. The marks on this page are concepts. No equipment, including LIAP Mini or a future SzaboWorks reference amplifier, is currently claimed to be EF certified.
Future reports would remain tied to the standard version and hardware/firmware tested. A new standard or significant design change would require the appropriate revalidation.
SZW EXTREME FIDELITYBetter questions.
Better questions.
Better evidence.
Better audio.
Read the proposal. Explore the thinking.
Help shape what a meaningful audio standard could become.
Draft 0.1 · October 2026 · No certification applications or payments are collected on this page.