Summary
On the Module Audio (M144, ES8388 codec), the left line-in ADC channel is attenuated with a frequency-dependent (high-frequency roll-off) response, while the right channel is flat. We traced the cause on the official schematic to capacitor C9, and confirmed that removing C9 fully restores balanced stereo. This matches an earlier community report (topic 7966) but adds the schematic-level root cause and before/after measurements, in case it's useful for an errata note or a future board revision.
Setup
- Module Audio (M144), ES8388, ADC input select
LINPUT1_RINPUT1 (stereo line-in, TRS "input" jack), 48 kHz / 16-bit.
- Source verified L = R symmetric (confirmed by feeding the same cable into other equipment).
- Reproduced on two separate modules (and two different host boards), with two different cables — so this is systematic, not a single defective unit.
Symptom (measured left-channel attenuation vs. right)
| Test tone |
Left attenuation vs. right |
| 100 Hz |
≈ −6 dB |
| 1 kHz |
≈ −15 dB |
| 4 kHz |
≈ −15 dB (plateau) |
The right channel is flat across the band; only the left rolls off, worsening with frequency until it plateaus. Because the imbalance is frequency-shaped, a fixed digital gain trim cannot correct it.
Root cause (from the official schematic, M144_sch_moduleaudio_v10)
The left ADC input node (MIC_MONO_IN, between C14 → CODEC_LIN1 and R9 22 kΩ → the jack) has an extra branch that the right channel does not:
MIC_MONO_IN ──(C9 100 nF)──(R6 3.3 kΩ)── AUDIO_VDD
AUDIO_VDD is an AC ground, so C9 + R6 act as a frequency-dependent shunt of the left signal to ground. A capacitor's impedance falls as frequency rises (Zc = 1/2πfC), so C9 diverts progressively more of the left signal to ground at higher frequencies — producing exactly the measured high-frequency roll-off. The plateau at high frequency corresponds to the divider bottoming out at R6/(R9 + R6) ≈ 3.3 k / 25.3 k ≈ −18 dB. The right channel path (CODEC_RIN1 ← C13 ← R8 22 kΩ) has no such branch, so it stays flat.
This branch appears to belong to the microphone-bias/filter circuitry on the shared left/mic node, but it degrades the left line-in because they share MIC_MONO_IN.
Fix / verification
Removing C9 eliminates the shunt. After removal, measured L = R (±0.5 dB) across 100 Hz–4 kHz — the left channel becomes flat and equal to the right (true stereo restored). The only function affected is the headset (TRRS) microphone input, which shares that branch; the stereo line-in and the headphone output are on separate nets and are unaffected.
Request
Could M5Stack:
- Confirm whether this left-channel line-in attenuation is expected behavior or a board issue?
- If it's unintended, consider an errata note in the docs and/or a value/placement change for C9 (or decoupling the mic-bias branch from the line-in node) in a future revision?
For anyone hitting this now: removing C9 restores balanced stereo line-in (at the cost of the headset mic).
Reference
Independent earlier report of the same fix by @Dkuzniar: https://community.m5stack.com/topic/7966/m5stack-audio-module-es8388
Summary
On the Module Audio (M144, ES8388 codec), the left line-in ADC channel is attenuated with a frequency-dependent (high-frequency roll-off) response, while the right channel is flat. We traced the cause on the official schematic to capacitor C9, and confirmed that removing C9 fully restores balanced stereo. This matches an earlier community report (topic 7966) but adds the schematic-level root cause and before/after measurements, in case it's useful for an errata note or a future board revision.
Setup
LINPUT1_RINPUT1(stereo line-in, TRS "input" jack), 48 kHz / 16-bit.Symptom (measured left-channel attenuation vs. right)
The right channel is flat across the band; only the left rolls off, worsening with frequency until it plateaus. Because the imbalance is frequency-shaped, a fixed digital gain trim cannot correct it.
Root cause (from the official schematic, M144_sch_moduleaudio_v10)
The left ADC input node (
MIC_MONO_IN, between C14 →CODEC_LIN1and R9 22 kΩ → the jack) has an extra branch that the right channel does not:MIC_MONO_IN ──(C9 100 nF)──(R6 3.3 kΩ)── AUDIO_VDD
AUDIO_VDDis an AC ground, so C9 + R6 act as a frequency-dependent shunt of the left signal to ground. A capacitor's impedance falls as frequency rises (Zc = 1/2πfC), so C9 diverts progressively more of the left signal to ground at higher frequencies — producing exactly the measured high-frequency roll-off. The plateau at high frequency corresponds to the divider bottoming out at R6/(R9 + R6) ≈ 3.3 k / 25.3 k ≈ −18 dB. The right channel path (CODEC_RIN1← C13 ← R8 22 kΩ) has no such branch, so it stays flat.This branch appears to belong to the microphone-bias/filter circuitry on the shared left/mic node, but it degrades the left line-in because they share
MIC_MONO_IN.Fix / verification
Removing C9 eliminates the shunt. After removal, measured L = R (±0.5 dB) across 100 Hz–4 kHz — the left channel becomes flat and equal to the right (true stereo restored). The only function affected is the headset (TRRS) microphone input, which shares that branch; the stereo line-in and the headphone output are on separate nets and are unaffected.
Request
Could M5Stack:
For anyone hitting this now: removing C9 restores balanced stereo line-in (at the cost of the headset mic).
Reference
Independent earlier report of the same fix by @Dkuzniar: https://community.m5stack.com/topic/7966/m5stack-audio-module-es8388