I don't understand why the packet is to hight 150-200ms
Can you re-word this sentence? What parameter of the RTP flow has this 150-200 ms value? Because in the
CALLER column of the table, there is
Avg RTP jitter 8.4 (ms),
Max RTP jitter 18 (ms), and
Lost packets 0. With a maximum packet delay of 18 ms, it is not clear to me why the estimated MOS for a 50 ms fixed "de-jittering" buffer should be worse than the one for a 200 ms fixed buffer, unless you use 40 ms packetization time and the buffer size estimation includes the size of the packet being played (so 40 ms of packet being played plus 18 ms would exceed the 50 ms buffer size).
Regardless that, I'd suggest you to sniff the traffic on the Ethernet interface underlying the
pppoe-out1 during the test call with iperf interference (better using port mirroring than sniffing into a file although you can connect an external USB3 drive to hAP ac3), and then let Wireshark show you the RTP flow at that point of the path. You may find out this way that the delay of some RTP packets is caused by queueing outside the Mikrotik.
Also, the packets generated by iperf seem pretty short to me (258 bytes on average), so at a physical 850 kbps line, the transmission time of such packet would be just about 2.5 ms, which is the maximum delay it can cause to an RTP packet that arrives to the queue just after the interfering one started departing. If the physical line is actually faster, it won't be even that much. But if something else is shaping the traffic between the Mikrotik and the SBC, uses a different averaging window, and ignores DSCP, it may cause the "unlucky" UDP packet to have to wait in the queue for the next window since the "interfering" packets just before it have saturated the current averaging window of that shaper. So using larger interfering packets might cause higher PDV of the RTP (a 1500-byte packet will cause about 15 ms).
It cannot be completely excluded that the above happens on the Mikrotik itself, so to test that, reduce the
max-limit for the "
no-mark" queue to 750k to see whether it has a positive impact on the jitter and MOS results.
)
