Buzz on channels 2 and 20: diagnosis

Rendered from the working notes 03-diagnosis.md on 2026-10-11.

Diagnosis: buzz on channels 2 and 20, TRC, 2026-09-25

Written by the pane after reading 01-session-reconstruction.md in full, the GX4816 photos, the Radial Catapult manual (~/trc-studio-manuals/radial-engineering-catapult-rx4__24.pdf) and the electrical project findings. Confidence labels: high / moderate / low.

A. What the evidence pins down (fact, from the transcript)

1. The buzz was heard on the Trident monitor path (amps next to the Trident, "all monitors, solo master"). So "channel 2" in tonight's tests is Trident channel 2, fed by the mix-room Catapult split. The same source line also feeds GX4816 input 2 through the mic-closet Catapult.
2. The buzz needs THREE things present at once: (a) the Cat6 link between the mix-room Catapult and the mic-closet Catapult, (b) the XLR from the mic-closet Catapult output 2 into GX4816 input 2, (c) nothing else. It does NOT need GX power, the SLink to the SQ-7, or the SQ-7 on. (T2, T5, T6, T13, T25)
3. Replacing the XLR patch (T8) and replacing the whole Cat6 run (T30 to T32) did not remove it. The original Cat6 passed a wiremap (T18).
4. A source plugged straight into GX input 2 with the Catapult bypassed is clean (P7, T28). Trident channel 2 with the mic-closet side disconnected is clean (P1, P3 to P5).
5. Channels 1 and 16 are clean on the very same path. Channel 20 behaves like channel 2, except pulling its XLR leaves a faint residue (T24).
6. Unplugging GX aux outputs (to PAs, in-ears, monitors) changes the buzz inconsistently but never removes it (T9). Greg's body position near the rack makes it louder (T21). Catapult ground lift: no change (P10).
7. Mic-closet Catapult output straight into a QSC powered monitor, no A&H in the path, also buzzed and stopped when the Cat6 was pulled (P15). Channel used in that test not recorded.

B. The mechanism that fits every observation (moderate-high confidence)

Common-mode noise converted to differential noise on two specific channels.

• Two rooms, two ground potentials. The mix room (Trident, mix-room Catapult, its rack) sits at one ground potential. The mic closet rack (GX4816, SQ-7, and every PA, in-ear and monitor amp hanging off the GX aux outputs) sits at another. The electrical project found the isolated-ground (IG) greens reach no electrode, so those grounds float relative to the service ground, and the difference between rooms carries mains noise (60 Hz and dimmer/switch-mode harmonics, which is what "buzz" rather than "hum" implies).
• The Catapult is passive. A Catapult channel is one twisted pair, pin 2 and pin 3, wired straight through from the mix-room XLR to the mic-closet XLR. Pin 1 rides on the Cat6 shield. The ground-lift switch only lifts pin 1; it does nothing to pins 2 and 3.
• A GX4816 input is never "open". Even powered off, each input pin 2 and pin 3 connect to the GX chassis through the phantom feed resistors (6.8 kΩ each, to the 48 V rail which sits at 0 V when off), RF capacitors and clamp diodes. That is why plugging the XLR into a dead GX still completes a path: GX chassis → 6.8 k → pin 2 and pin 3 → twisted pair back to the mix room → Trident input → Trident ground. The noise voltage between the two grounds now drives a common-mode current down the pair.
• On a healthy channel both legs carry the same current and the Trident's balanced input rejects it (channels 1 and 16). On channels 2 and 20 one leg differs from the other, so part of the common-mode noise becomes a differential signal and the Trident amplifies it as buzz. The imbalance can live in only three places: GX4816 inputs 2 and 20 (a drifted or damaged phantom resistor, RF cap or clamp on one leg), the mic-closet Catapult channel positions 2 and 20 (a solder joint, a jack contact, or the RJ45 jack pin behind it), or the mix-room Catapult / Trident-side wiring of channels 2 and 20 (less likely since the mix-room-only test is clean, but an imbalance only shows when a second ground is attached, so "clean alone" does not clear it).

Why each observation fits:

Greg's "that nullifies my hypothesis" at 04:34Z was premature. The direct-into-GX test only shows the GX is quiet with one ground reference; it says nothing about how GX input 2 behaves with two.

B2. What the studio records add (from 02-system-model.md, all cited there)

• Every orange isolated-ground outlet tested in the building reads open ground, with neutral-to-ground floating 30 to 60 V (Elijah's 2026-08-03 handoff; Greg's 2026-01-22 chat). A chassis on an open-ground outlet floats to tens of volts AC through its own mains-filter capacitors. That is the noise voltage in section B, and it is documented, not assumed.
• The GX4816 outputs feed the Crown amp on the Furman, the PSM900 in-ear transmitters and the subs on their own wall outlets. Any of those keeps the GX chassis tied to some ground even with its own cord pulled, which is why "power out entirely" changed nothing.
• Channels 2, 20 and 22 were already logged as "down / crossing" around 2026-03-16, channel 2 for "a couple months" before that and 20 and 22 "since New Year". Bypassing the Catapult cleared channel 2 then too. The fault is persistent and per-channel, not tonight's news.
• The August handoff recorded "loud buzz when phantom power (48 V) is engaged, most notably on live-room mic lines". Phantom pushed through a passive split into a second console is the DC ground-loop path in C4 below.
• The install crew "pressed a bunch of the ground switches" on the Catapults and the buzz "magically worked" for a while. Nobody has recorded the lift positions since.
• Lights and audio share a panel; Lutron dimmers are on the control-room and iso-room track lighting; the stage-right snake runs under the LED wall and had noise in June.
• Trident 88 has a Star Ground Link on its PSU; its state is unrecorded. Channel 26 on the Trident was reported buzzing on 2026-09-12 and was not retested tonight.

C. Alternative causes, ranked

1. Imbalance inside GX4816 inputs 2 and 20 (moderate). A&H inputs use matched phantom resistors; a hot-patch with phantom on or a static event takes one out. The phantom LEDs in tonight's photo look lit on some inputs in rows 1 and 2; if phantom is on at 2 and 20 and off at 1 and 16, that alone is a difference to test.
2. Imbalance in the mic-closet Catapult channel 2 / 20 (moderate). Same symptom, different box. Cheap to discriminate (test 1 below).
3. Ground potential difference is the driver in every case (high that it exists, given the IG finding). Even a perfect channel gets some buzz if the difference is large; channels 2 and 20 are just the worst. Fixing the bond (IG bus to service ground) lowers the whole floor.
4. Phantom power from two consoles on one line (low-moderate). If the Trident or the SQ-7 has phantom engaged on 2 and 20 only, 48 V is being pushed through the passive split into the other console's input network, which is a DC ground-loop path on exactly those channels. Untested (T26 result missing).
5. Split pair in the Cat6 termination (low). Would explain a single channel but not survival of a full cable swap; ruled out unless the silver cable was also hand-crimped.
6. Noise source is a specific load (unknown). LED drivers, dimmers, Lorex camera PoE switch, the PA amps on the GX auxes, HVAC. The planned all-off test (T33) never ran.

Ruled out by the tests: the XLR patch cable, the Cat6 cable as a cable, the SLink, the SQ-7, GX power, the DPA snake, the wall panel, the mic and mic cable, the Trident alone.

D. Tests that decide it, in order (each under 10 minutes)

1. Swap test. Move the mic-closet Catapult output 2 XLR into GX input 1, and Catapult output 1 into GX input 2. Buzz follows to the Trident channel of whichever line is in GX input 2 → the GX input is the asymmetric part (replacement or A&H repair is justified). Buzz stays on Trident channel 2 → the asymmetry is upstream in the Catapult path. Repeat with 20 and 16.
2. Second grounded device. Plug mic-closet Catapult output 2 into a known-good grounded input (QSC, or a spare preamp) instead of the GX. Buzz present → Catapult/pair side. Absent → GX input.
3. Meter the grounds. AC volts between the GX chassis and the Trident chassis (a long wire to the DMM), and between the ground pins of the two racks' outlets. Anything above about 50 mV AC is a ground difference worth killing; note the reading with lights and HVAC on and off.
4. Temporary bond. Run a heavy wire (12 AWG) between the mic-closet rack and the Trident chassis. Buzz drops → ground potential difference confirmed; the permanent fix is the IG bus bond already recommended in the electrical project.
5. Phantom audit. Read phantom state on the Trident, the SQ-7 and the GX LEDs for 1, 2, 16, 20. Turn all four off and listen.
6. All-off test (T33). Original Cat6 in, buzz audible, kill circuits one at a time at the panels: lighting, HVAC, the PA amps, the camera/network switch. The circuit that silences it names the noise source.
7. Isolate the split. Put a mic-level 1:1 isolation transformer on channel 2 between the Catapult output and the GX input (Radial TX4M/RX4M modules or a single inline iso). Buzz gone → permanent fix for every channel: replace the mic-closet RX modules with transformer-isolated ones, which is exactly what Radial builds them for.

E. Questions for Greg and Elijah

1. Which console had phantom on for channels 2 and 20 tonight, and is it off on 1 and 16? (Trident switches, SQ-7 scene, GX PP LEDs.)
2. In P15 (Catapult straight to the QSC), which channel number, and was anything else connected to that Catapult at the time?
3. In P13 ("speaker directly on the A&H channel"), where exactly was the speaker connected?
4. In T15 (Ethernet moved to other Catapult ports), did the buzz move to other channel numbers, stay on 2 and 20, or vanish?
5. Which Catapult module and bank carries channel 20 on each end, and are 2 and 20 on the same Cat6 run or different runs? Elijah's "16 and 20 on the same route" does not match 4-channel banks.
6. Which outlets feed the mic-closet rack and the mix-room rack, and are they IG (orange) outlets?
7. Is the Cat6 between rooms shielded, with shielded plugs at both ends?
8. Does the buzz change when the studio lights are dimmed, when HVAC cycles, or at a different time of day?
9. Did the mains all-off test happen after 22:05 PDT, and what happened?
10. Where is Elijah's paper test sheet?