[AMPS] Pre-emphasis filter missing/not working (affects AMPS deviation/speech volume) #11
jolly
Loading…
Reference in New Issue
No description provided.
Delete Branch "%!s(<nil>)"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
These are two interconnected issues.
AMPS operates in analog networks with the notion of reference volume level, and it can be both louder and quiter than that.
In digital recording it seems that nomimal volume level are considered to be about 0.5 of maximum amplitude (-6 dB full scale, dBFS).
AMPS standard TIA_EIA_553_A on PAGE 12 LINE 17 , section 2.1.3.1.1 says that +- 2.9 kHz peak frequency deviation should happen at expected speech nominal volume level.
What is implemented in osmocom-analog AMPS is 2.9 kHz peak deviation from maximum volume level. Nominal volume level is 50% or less of max possible amplitude for digital recordings.
Also section 2.1.3.1.3 page 13 line 3 sets maximum deviation for voice transmission (instantaneous) to +-12kHz.
So, as far as I understand the standard, +-2.9kHz deviation should be for 1 kHz tone at nominal voice level (50% amplitude for digital audio input), and up to +-12 kHz deviation is possible for short-time voice transmission peaks.
For AMPS to get the volume level right deviation should be adjusted here
3bc3e67abb/src/amps/dsp.c (L104)in such a way that 1kHz tone at nominal voice level (seems to 50% or less amplitude for digital recording) will produce +-2.9 kHz deviation on radio part, but not more that +-12 kHz for audio level peaks (probably). Some testing equipment documentation mentions +-11kHz.
This document describes the procedures of AMPS phone measurement device
https://scdn.rohde-schwarz.com/ur/pws/dl_downloads/dl_application/application_notes/1cm53/1CM53_0e.pdf
On pages 8 and 9 it says:
..."The individual channels are arranged with a channel
spacing of 30 kHz. The maximum deviation of the signal should therefore
not be greater than 11 kHz in order to prevent interference on the adjacent
channel."
..."In the AMPS phone, the amplified audio signal from the microphone first
enters a “compressor”. The reference deviation for AMPS is 2.9 kHz. This
means that the audio signal should generate a deviation of 2.9 kHz at full-
scale level. The relationship between audio signal level and deviation is,
however, not linear. A 6-dB-lower input level attenuates the deviation by
only 3 dB. A 6-dB-higher audio signal amplifies the deviation by only 3 dB.
As is usual with analog transmission systems, higher frequencies are
amplified and lower frequencies are attenuated. This is referred to as a
“pre-emphasis filter”."
Full-scale level is nominal voice level here, input signal for voice can be louder and quiter. Definetly it's not maximum digital amplitude (you can't go louder than that).
Expected frequency response of pre-emphasis filter also is given on page 9.
It's quite possible that it's not deviation that is incorrect, it's pre-emphasis filter that is missing or not working properly. As a result frequency that corresponds to voice have much lower amplitude that lower frequencies (at least the way Motorola decodes it).
Anyway, it's definitely shifted in the lower frequency range.
Please see amplitude-frequency response graphs below. Please keep in mind while connection was established, it was extremely unstable and de-synced almost immediate due to noise on the line (maybe hardware is old, maybe radio is noisy). Most stable link speed was around 4800 bps.
I have acquired "cellular data link" for Motorola MicroTAC, which is effectively a modem port breakout for the phone, and performed some testing with a following setup:
(all elements here are attached to the same pc, to eliminate network delay and jitter)
USR Courier I-modem -> BRI -> Digium BRI PCI card -> asterisk -> osmo-sip-connector -> amps -> LimeSDR -> Motorola MicroTAC -> data link box -> USR Courier v.everything.
Why the complexity? Because USR Courier can plot amplitude-frequency response curve for analog line after the connection was terminated.
So, for direct analog line on my setup the "perfect" line looks like this (mind that hardware is pretty old), which corresponds to V.90 setup (digital modem on one side, analog modem dials in):
Frequency response curves (for direction FROM BRI TO MOBILE, modem plots what it receives) over AMPS look like this for following deviations:
2900Hz (current value):
3900Hz
5000Hz
[AMPS] Wrong FM deviation (speech to quiet), wrong pre-recorded message signal level (too loud)to [AMPS] Pre-emphasis filter missing/not working (affects AMPS deviation/speech volume)