Low repeater audio level using discriminator input

[iu6dgd]

Hi everyone, I would like to share my technical problem with you.

My SvxLink repeater is fed directly from the discriminator output of a Motorola GM900, which provides flat, audio without pre-processing, to ensure proper dynamic range and prevent ADC clipping, I have calibrated both RX and TX paths carefully:

1)RX calibration procedure:
-Audio input is taken directly from the discriminator, with some decoupling passive components.
-ALSA input level has been adjusted so that with squelch open (white noise), the peak level sits at approx. −6 dBFS.
-Audio decoding, CTCSS detection recognition work reliably,with excellent dynamics.

[LocRx:CTCSS]
SQL_DET=CTCSS
CTCSS_FQ=85.4
CTCSS_OPEN_THRESH=20
CTCSS_CLOSE_THRESH=10
CTCSS_MODE=2
CTCSS_DEBUG=0
SQL_HANGTIME=200

[LocRx:SIGLEV]
SQL_DET=SIGLEV
SIGLEV_DET=NOISE
SQL_SIGLEV_RX_NAME=LocRx
SQL_SIGLEV_OPEN_THRESH=0
SQL_SIGLEV_CLOSE_THRESH=0
SQL_HANGTIME=200
SQL_DELAY=0

[LocRx]
TYPE=Local
AUDIO_DEV=alsa:plughw:0
AUDIO_CHANNEL=0
SQL_DET=COMBINE
DEEMPHASIS=1
PREAMP=0
PEAK_METER=1

2)TX calibration procedure

-TX calibration has been performed using devcal tx, generating a 1 kHz tone.
-The transmitted deviation has been adjusted using a Radio Testset and spectrum analyzer.
-With the limiter threshold at −6 dBFS, I calibrated for 2.5 kHz deviation, so TX output is correctly aligned when the reference tone is at full internal level.

[Tx1]
TYPE=Local
AUDIO_DEV=alsa:plughw:0
AUDIO_CHANNEL=0
LIMITER_THRESH=-6
PREEMPHASIS=1
MASTER_GAIN=0
CTCSS_FQ=85.4
CTCSS_LEVEL=-30

The RX and TX are individually calibrated and function correctly, but there is a structural gain imbalance in the RX→TX repeater path.
On the RX side, the discriminator noise reaches -6 dBFS with the squelch open, which is optimal for preventing clipping and ensuring reliable decoding of CTCSS, SIGLEV, and DTMF.
However, normal voice peaks are typically around -18 dBFS, meaning the user's audio is approximately 12 dB lower than the reference noise.
When this audio passes through the repeater logic and is retransmitted, the resulting RF modulation is perceived as significantly lower (~10–15 dB) than expected, despite the TX being properly calibrated.

Increasing the PREAMP on the RX side would increase the level, but this would immediately push the noise floor to 0 dBFS or higher, resulting in distortion and unreliable tone decoding, so this approach is not risky.

Increasing the MASTER_GAIN on the TX side is also not difficult, as it would amplify all audio sent to the transmitter, including EchoLink or reflector traffic, resulting in an unbalanced audio difference between the local RX and the reflector RX.

In the past I solved this by increasing the gains on both sides, finding a balance, but it is always very difficult, and it is a shame not to be able to use and rely on calibration tools which are very efficient and precise.

Thanks Marco IU6DGD