FreeCalypso > hg > themwi-rtp-lib

comparison src/twjit.c @ 41:bda6b24385f7

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
twjit config: revisit default settings The only actual change here is bumping the high water mark setting from 2 to 3; the rest of this patch is comment changes.
author Mychaela Falconia <falcon@freecalypso.org>
date Fri, 20 Dec 2024 22:13:01 +0000
parents 1485211d4492
children 334d883b96ba
comparison
equal deleted inserted replaced
40:d73b6ec27ae6 41:bda6b24385f7
14 #include <themwi/rtp/twjit.h> 14 #include <themwi/rtp/twjit.h>
15 15
16 void twrtp_jibuf_init_defaults(struct twrtp_jibuf_config *config) 16 void twrtp_jibuf_init_defaults(struct twrtp_jibuf_config *config)
17 { 17 {
18 memset(config, 0, sizeof(struct twrtp_jibuf_config)); 18 memset(config, 0, sizeof(struct twrtp_jibuf_config));
19 config->bd_start = 2; /* smallest allowed */ 19
20 config->bd_hiwat = 3; /* Nstart+1 is practically-useful minimum */ 20 /* While the theoretical minimum starting fill level is 1, the
21 config->thinning_int = 17; /* prime number, usually 340 ms */ 21 * practically useful minimum (achieving lowest latency, but not
22 config->max_future_sec = 10; /* 10 s is a long time for voice */ 22 * incurring underruns in normal healthy operation) is 2 for typical
23 * network configurations that combine elements with "perfect" 20 ms
24 * timing (T1/E1 interfaces, external IP-PSTN links, software
25 * transcoders timed by system clock etc) and GSM-to-IP OsmoBTS
26 * whose 20 ms timing contains the small inherent jitter of TDMA. */
27 config->bd_start = 2;
28
29 /* The high water mark setting determines when the standing queue
30 * thinning mechanism kicks in. A standing queue that is longer
31 * than the starting fill level will occur when the flow starts
32 * during a network latency spike, but then the network latency
33 * goes down. If this setting is too high, deep standing queues
34 * will persist, adding needless latency to speech or CSD.
35 * If this setting is too low, the thinning mechanism will be
36 * too invasive, needlessly and perhaps frequently deleting a quantum
37 * of speech or data from the stream and incurring a phase shift.
38 * Starting fill level plus 2 seems like a good default. */
39 config->bd_hiwat = 4;
40
41 /* When the standing queue thinning mechanism does kick in,
42 * it drops every Nth packet, where N is the thinning interval.
43 * Given that this mechanism forcibly deletes a quantum of speech
44 * or data from the stream, these induced disruptions should be
45 * spaced out, and the managing operator should also keep in mind
46 * that the incurred phase shift may be a problem for some
47 * applications, particularly CSD. Our current default is
48 * a prime number, reducing the probability that the thinning
49 * mechanism will interfere badly with intrinsic features of the
50 * stream being thinned. 17 quantum units at 20 ms per quantum
51 * is 340 ms, which should be sufficiently long spacing to make
52 * speech quantum deletions tolerable. */
53 config->thinning_int = 17;
54
55 /* With RTP timestamps being 32 bits and with the usual RTP
56 * clock rate of 8000 timestamp units per second, a packet may
57 * arrive that claims to be as far as 3 days into the future.
58 * Such aberrant RTP packets are jocularly referred to as
59 * time travelers. Assuming that actual time travel either
60 * does not exist at all or at least does not happen in the
61 * present context, we reason that when such "time traveler" RTP
62 * packets do arrive, we must be dealing with the effect of a
63 * software bug or misdesign or misconfiguration in whatever
64 * foreign network element is sending us RTP. In any case,
65 * irrespective of the cause, we must be prepared for the
66 * possibility of seeming "time travel" in the incoming RTP stream.
67 * We implement an arbitrary threshold: if the received RTP ts
68 * is too far into the future, we treat that packet as the
69 * beginning of a new stream, same as SSRC change or non-quantum
70 * ts increment. This threshold has 1 s granularity, which is
71 * sufficient for its intended purpose of catching gross errors.
72 * The minimum setting of this threshold is 1 s, but let's
73 * default to 10 s, being generous to networks with really bad
74 * latency. */
75 config->max_future_sec = 10;
23 } 76 }
24 77
25 /* create and destroy functions */ 78 /* create and destroy functions */
26 79
27 struct twrtp_jibuf_inst * 80 struct twrtp_jibuf_inst *