The following graphs are spectral analysis of short segments of 1200 AFSK (Bell 202 modulation) which is carrying alternate mark and spaces at the maximum rate.
They demonstrate the effect of overdriving the transmitter audio on spectral distribution of the transmitted modulated signal.
The resultant nullification of the transmitter pre-emphasis contributes to roll-off of the high frequency end of the pass-band which degrades PLL demodulator performance.
The analysis technique can be used to identify such signals off air by performing spectral analysis of the short sync period at the start of each transmit burst.
This graph below is output from the TNC (sending a test pattern of alternate ones and zeros). This TNC uses an AMD7910 modem chip. Note mark and space frequencies and the sideband about half way between them. I expected that the level of the mark and space frequency should have been similar, but in this case there is 3dB roll off at 2200 compared to 1200.
The graph below is output from a receiver. The roll-off evidenced above is worsened, with about 5db at 2200 compared to 1200. I suspect that the additional roll-off can be attributed to equipment manufacturers tendency to over de-emphasize receivers to improve the SINAD specification.
The graph below is output from a receiver when the transmitter input is increased 10dB. A 10dB increase in input should cause about 10kHz deviation in a linear system, but in this case measured peak deviation is 4.5kHz, demonstrating limiting in the transmitter audio stage. The roll-off evidenced above is worsened further, with about 12db at 2200 compared to 1200. The further degradation is caused by the effect of the audio limiter.
The following graph was captured off air from a signal that I cannot decode successfully. It is a spectral analysis of repeated frame codes. Note the similarity in the spectral distribution to the controlled overload case above.
Figure 5 shows defective AFSK signals from different stations and captured locally on the APRS channel. Defects readily identifiable are:
Figure 5 demonstrates that there is a natural concentration of energy during sync because of the code value (0x7E), and whether it is high or low depends on the state of the encoder when it commences to send the first frame code. It is the data part of a frame, where the bit sequence is more likely to be random, that best indicates amplitude / frequency equalisation.
After monitoring the channel for twenty minutes, I did not observe a model signal. However, the first transmission in Figure 5, apart from the excessive TxDelay and slightly low deviation is otherwise good. The blip at the end of the transmission is noise between end of transmission and reciever squelch closure. Note that the squelch will have clipped the front of each transmission.
Figure 6 shows three different AFSK transmissions in the waterfall log in the lower part of the display. The spectrum graph in the top part of the display corresponds to the signal shown at the very top of the waterfall. The red tick marks at the left of the waterfall are at one second intervals.
A description of the highlighted areas follows:
Not highlighted on this figure was that the "sync" period at the start of some station's transmissions were excessive, wasting channel bandwidth. The sync period can be measured against the tick marks. The lowest transmission's sync period is not all shown on the display, and was excessive at more than double the sync period of the top transmission. The parameter that controls the sync period is often termed TxDelay.
The spectral analyses above were done with Spectrogram and Spectran using the PC sound card line input driven from the receiver speaker output.
A TNC Audio Break-Out-Box was a convenient way to monitor and make measurements on received and transmitted signals.
© Copyright: Owen Duffy 1995, 2021. All rights reserved. Disclaimer.