Comtech EF Data CDM-625A - Manual (Page 688)

CDM-625A Advanced Satellite Modem MN-CDM625A
Appendix M Revision 3
M–8
M.6.1 VersaFEC ACM Latency
In an ACM system that has a number of ModCods, each having a different latency, what defines
the overall system latency? The answer is simple – the latency of the worst-case ModCod. (This
may not seem obvious to some, and it is beyond the scope of this appendix to provide a rigorous
defense of this statement. It is, however, a correct statement despite certain believers in non-
causal systems and encoders that possess the magical quality of negative latency…)
Examining the data in Table M-2, latency for each ModCod is shown for the example of
VersaFEC ACM at a fixed 100 ksymbols/second rate. Of particular note is that even though the
ModCods span a 7:1 variation in throughput, the latency is only varying between 25 and 34
milliseconds. A careful analysis will show that this is a consequence of using a constant number
of symbols per block. In the example shown the worst-case latency for this ACM scheme is 34
milliseconds.
Table M-2. VersaFEC Implementation of ACM – 100 ksymbols/sec Example Case
ModCod Modulation Code Rate
Spectral efficiency,
bps/Hz
Bit rate (throughput)
Minimum Latency,
In milliseconds
00 BPSK 0.488 0.49 49 kbps 34
01 QPSK 0.533 1.07 107 kbps 32
02 QPSK 0.631 1.26 126 kbps 30
03 QPSK 0.706 1.41 141 kbps 28
04 QPSK 0.803 1.61 161 kbps 26
05 8-QAM 0.642 1.93 193 kbps 30
06 8-QAM 0.711 2.13 213 kbps 28
07 8-QAM 0.780 2.34 234 kbps 27
08 16-QAM 0.731 2.93 293 kbps 27
09 16-QAM 0.780 3.12 312 kbps 26
10 16-QAM 0.829 3.32 332 kbps 25
11 16-QAM 0.853 3.41 341 kbps 25
OVERALL SYSTEM LATENCY = Worst-case ModCod (ModCod0)
Latency = 34 milliseconds, NOT INCLUDING WAN BUFFER OR SATELLITE PATH
By way of comparison, consider the same 100 ksymbols/second rate, but this time using DVB-S2. It
becomes clear that there is an unintended penalty (besides demodulator complexity) to having a
constant number of bits per block. Each time the ModCod is lowered and the throughput is
reduced, the latency grows accordingly due to the block size being related to data rate, not symbol
rate.
Remembering that, for the ACM case, the system latency is equal to the latency of the worst-case
ModCod, DVB-S2 shows a severe penalty. For 16k short blocks, this calculates to be 329
milliseconds (+ WAN Buffer delay) versus 34 milliseconds (+ WAN Buffer delay) for VersaFEC ACM.
For 64k block DVB-S2, the core latency is 4 times higher. Assuming a WAN Buffer of 20
milliseconds:
• Latency for 64k block DVB-S2 ACM at 100 ksps = 1336 milliseconds
• Latency for 16k block DVB-S2 ACM at 100 ksps = 349 milliseconds
• Latency for VersaFEC ACM at 100 ksps = 54 milliseconds