alright.
The squash limiter works by subtracting a threshold from the ABS level
sample by sample
This resultant is negative clipped at 0, so that only peak events >
threshold produce any positive and non zero numbers
This value is fed (sample by sample again) into a short FIR low pass
filter with the output of the FIR driving a gain control of the source
being controlled. maybe 5 taps max.
Of course, the signal needs to be delayed before it gets to the gain
control, because for a 5 tap basic LPF FIR, there is a 3 tap group delay
in getting the answers out of the filter.
The LPF bandwidth controls the control bandwidth of course. The squash
limiter for OFDM will punch 'holes' in the ensemble , and may drive the
receiver equaliser mad.
David, can you provide some comment on what the equaliser /tracking will
do and likely maximum duration/slew rate suggested ?
A guess is fine. My guess is the effect on DBPSK wont be as harsh as,
for example, 16QAM where amplitude matters......
Ideally, for single sided spectrum (like being fed into the microphone
input) the sample rate at which this is done should be , ideally ,
higher than the applied signal bandwidth. This permits catching and
controlling all events with fidelity.
NOW ! how this works in practice for OFDM I dunno ! since it iwll
generate momentary squashes of the entire
-glen
_______________________________________________
Freetel-codec2 mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
