Problem
A user reported a loud 120 Hz hum on Tulip's audio out when connecting a Kurzweil's MIDI OUT to Tulip's MIDI IN. Severing the ground/shield conductor in the MIDI cable eliminated the hum. Using an Alpha Juno as the controller (or patching the Kurzweil through the Juno's THRU) also avoided it.
Root cause: both Tulip and AMYboard tie the MIDI IN TRS jack's sleeve to GND, which violates the MIDI electrical spec and creates a ground loop with spec-compliant source devices.
- Tulip r11:
MIDI_IN_CONN2 (19269 TRS jack) sleeve → GND, before the 220R → H11L1SR2M optocoupler (docs/pcbs/tulip4_r11/tulipv4_r11.pdf)
- AMYboard v1.4:
MIDI_IN jack sleeve → GND, before the polarity bridge → H11L1SR2M (docs/pcbs/amyboard/amyboard-1.4-sch.png)
Why this is our bug, not the source device's
The MIDI 1.0 electrical specification (and the 3.3 V update, CA-033) connects pin 2 (shield) to ground only at MIDI OUT and THRU. At MIDI IN, pin 2 is a no-connect. The opto-isolated current loop runs entirely on pins 4/5 (tip/ring on TRS-A) — ground is never used for data, and grounding the shield at the receiver defeats the galvanic isolation the optocoupler exists to provide.
The Kurzweil grounding its shield at MIDI OUT is exactly per spec. Our grounded IN sleeve bonds the two chassis grounds via the cable shield; with audio cables also connecting the devices, ground-loop current flows and 120 Hz (full-wave-rectified 60 Hz) hum couples into the audio path.
TRS adapters (MMA RP-054, "TRS-A") carry the sleeve↔pin 2 mapping because the same adapter is used for MIDI OUT, where the shield connection is required. The connect-or-don't decision belongs to the device jack — the receiver is the side that must lift it.
Fix (future board revisions, both Tulip and AMYboard)
Disconnect the MIDI IN jack sleeve from GND. Options:
- Fully floating — by-the-book MIDI spec behavior.
- Hybrid lift (recommended) — sleeve → GND through ~100 nF (optionally ∥ ~1 MΩ). Open at DC/60 Hz (breaks the loop) but still an RF shield / ESD path at high frequency. Common practice in modern MIDI gear.
One-trace change; the downstream circuit (220R, 1N4148, H11L1) is already correct. MIDI OUT jacks must keep their grounded sleeve — that side is required by spec.
Workaround for existing boards
A ground-lifted TRS cable/adapter, or a DIN cable with the shield cut at one end, eliminates the hum with no loss of MIDI function (the current loop doesn't use ground). Worth a note in docs/FAQ.
Problem
A user reported a loud 120 Hz hum on Tulip's audio out when connecting a Kurzweil's MIDI OUT to Tulip's MIDI IN. Severing the ground/shield conductor in the MIDI cable eliminated the hum. Using an Alpha Juno as the controller (or patching the Kurzweil through the Juno's THRU) also avoided it.
Root cause: both Tulip and AMYboard tie the MIDI IN TRS jack's sleeve to GND, which violates the MIDI electrical spec and creates a ground loop with spec-compliant source devices.
MIDI_IN_CONN2(19269 TRS jack) sleeve → GND, before the 220R → H11L1SR2M optocoupler (docs/pcbs/tulip4_r11/tulipv4_r11.pdf)MIDI_INjack sleeve → GND, before the polarity bridge → H11L1SR2M (docs/pcbs/amyboard/amyboard-1.4-sch.png)Why this is our bug, not the source device's
The MIDI 1.0 electrical specification (and the 3.3 V update, CA-033) connects pin 2 (shield) to ground only at MIDI OUT and THRU. At MIDI IN, pin 2 is a no-connect. The opto-isolated current loop runs entirely on pins 4/5 (tip/ring on TRS-A) — ground is never used for data, and grounding the shield at the receiver defeats the galvanic isolation the optocoupler exists to provide.
The Kurzweil grounding its shield at MIDI OUT is exactly per spec. Our grounded IN sleeve bonds the two chassis grounds via the cable shield; with audio cables also connecting the devices, ground-loop current flows and 120 Hz (full-wave-rectified 60 Hz) hum couples into the audio path.
TRS adapters (MMA RP-054, "TRS-A") carry the sleeve↔pin 2 mapping because the same adapter is used for MIDI OUT, where the shield connection is required. The connect-or-don't decision belongs to the device jack — the receiver is the side that must lift it.
Fix (future board revisions, both Tulip and AMYboard)
Disconnect the MIDI IN jack sleeve from GND. Options:
One-trace change; the downstream circuit (220R, 1N4148, H11L1) is already correct. MIDI OUT jacks must keep their grounded sleeve — that side is required by spec.
Workaround for existing boards
A ground-lifted TRS cable/adapter, or a DIN cable with the shield cut at one end, eliminates the hum with no loss of MIDI function (the current loop doesn't use ground). Worth a note in docs/FAQ.