Sub-ppm THD and −150 dBV noise floors only mean something if the wiring between the parts of the bench does not inject more noise than the DUT produces. This describes how the generator (DAC), the device under test (DUT) and the capture (ADC) are connected so that external electromagnetic fields induce as little voltage as possible on the signal and signal-ground wires.
There is no separate room-sized cage. The device enclosures and the cable shields together build one continuous conductive envelope around the signal pair — a distributed Faraday cage, all tied through XLR pin 1 and earthed at a single point. Three ideas do the work:
- Best-possible cable shielding, with the right wire used for the right job.
- A continuous shield/chassis envelope — the distributed cage — that excludes external electric fields.
- Single-point grounding: the signal ground is bonded to that shield at exactly one place (the DUT), so shield currents never flow in the signal return.
1. The cable: CAT.8 patch cable, not "audio" coax
The single most important choice is the cable. Use a computer twisted-pair CAT.8 patch cable terminated in XLR connectors.
CAT.8 carries a double shield: an aluminium/PET foil wrapped around each twisted pair, plus an overall foil/braid around all four pairs. A foil is a continuous, hole-free screen — unlike the braided shield of a coaxial cable, which is a mesh full of small openings that leak fields, especially as frequency rises. The twisted pair itself is the second weapon: twisting makes the loop area between signal and return tiny and sign-alternating, so a magnetic field induces nearly equal-and-opposite EMFs in successive twists that cancel.
2. XLR wiring — one pair carries the signal
A standard XLR has three pins: 1 = shield/ground, 2 = hot/positive, 3 = cold/negative. Here the connection is unbalanced, carried on one twisted pair, with the foil/braid used as a separate chassis screen:
| CAT.8 wire | XLR pin | Role |
|---|---|---|
| blue-white | 2 (hot) | Signal |
| blue | 3 (cold) | Signal ground (the unbalanced return) |
| green + green-white + drain | 1 | Cable shield → device chassis |
So the signal (pin 2) and the signal ground (pin 3) are the blue/blue-white twisted pair; the green pair plus the drain wire are all tied to pin 1 and serve only as the shield/chassis screen.
3. Single-point grounding — the key trick
Signal ground (pin 3) is connected to the shield (pin 1) at exactly one place, normally the DUT input or output — test both and keep whichever gives the lower noise. Everywhere else the signal-ground wire is isolated from the shield and from the chassis.
Why this matters: if the signal ground also acted as the shield (as it does in an ordinary single-ended coax), then any current the external field induces in the shield would flow through the signal return and add directly to the signal at the DUT input. By giving the shield its own conductor (pin 1, bonded to every chassis) and bonding the signal ground to it at one point only, the shield/eddy currents have their own return path and do not appear in series with the signal.
Each device has its own shield connected to the XLR shield/ground (pin 1) — not to signal ground (pin 3). The chassis screens form one continuous shield held at cage/earth potential; the signal ground floats on the twisted pair and touches that shield only at the single star point at the DUT.
4. How much noise is actually left? (order-of-magnitude)
The remaining question is whether external fields can still induce a meaningful voltage. Estimate it.
Electric fields — the usual source of mains hum via capacitive coupling — are excluded entirely by the continuous shield/chassis envelope: the foil around the pair and the metal enclosures form one closed conductor around the signal. That coupling path is gone.
Magnetic fields are not blocked by an electrostatic screen; they have to be out-run by minimising loop area. The thin conductive envelope still helps a little via eddy currents, but the dominant defence is simply the tiny loop area of the twisted pair.
Faraday's law gives the induced EMF in the signal loop:
| Quantity | Symbol | Estimate (50 Hz) |
|---|---|---|
| Ambient mains magnetic field in a normal room [1] | Bext | ~0.1 µT (≈1 mG); 0.05–1 µT |
| Residual field inside the cage | Bin | a few dB lower, ~0.05 µT |
| Effective loop area of the twisted pair (pessimistic) | Aeff | ~10 mm² = 1×10⁻⁵ m² |
| Angular frequency | ω | 2π·50 ≈ 314 rad/s |
| Induced EMF | V | 314 · 5×10⁻⁸ · 1×10⁻⁵ ≈ 1.6×10⁻¹⁰ V ≈ 0.16 nV |
| in dBV (re 1 V) | ≈ −196 dBV |
Even taking a deliberately pessimistic case — no envelope attenuation (B = 0.1 µT) and a poorly-twisted run with ten times the loop area (Aeff = 100 mm²):
That is still tens of dB below the converter's own noise floor (a capable ADC sits around −150…−160 dBV per bin). The magnetically-induced mains pickup is negligible.
For contrast, drop the discipline — let the signal-ground conductor double as the shield so the whole cable-to-chassis loop (tens of cm²) is exposed and carries shield current — and the same field induces micro-volts (≈ −120 dBV): audible hum, and a hard floor under every distortion measurement. The twist plus single-point ground buys roughly 50–80 dB.
Example DUTs
The same wiring feeds whatever sits in the cage. Two filters used on this bench:
Summary
- CAT.8 twisted pair + XLR, double foil shield, no braid holes.
- Pin 2 = signal, pin 3 = signal ground, pin 1 = shield/chassis.
- Signal ground bonded to shield at one point only, at the DUT.
- Every chassis on the shield (pin 1), never on signal ground.
- This continuous shield/chassis envelope drives the induced mains pickup to ≈ −170 dBV or lower — well under the noise floor of the measurement itself.
References
- Ambient mains magnetic field. World Health Organization, Electromagnetic fields and public health: exposure to extremely low frequency (ELF) fields (fact sheet, 2007): typical residential power-frequency (50/60 Hz) background magnetic fields average ≈ 0.07 µT (Europe) and ≈ 0.11 µT (North America), rising locally near appliances and wiring. WHO fact sheet.
- Shielding & grounding technique. H. W. Ott, Electromagnetic Compatibility Engineering (Wiley, 2009), ISBN 978-0-470-18930-6 — Chapter 2 "Cabling", esp. §2.12 Coaxial Cable Versus Twisted Pair, §2.13 Braided Shields, and §2.15.2 Grounding of Cable Shields.