I am using v 8.0 of the code on win32. I have a strange behavior happening. I am talking to a paging speaker. IE... Normally no audio coming from the device. I also have a push to talk button on a USB mic of the machine with my py_pjsua running that I use to mute / unmute the mixer input for the mic. So, I am essentially operating in half duplex, via a muting and un-muting of a microphone. Now the funny stuff, I have one of two things happen consistently when talking to the paging speaker. Either, I will produce no output on the paging speaker or It will start out ok and quit after I have went through a couple of PTT -> talk ->release ptt sequences. In other words, muting the mic appears to somehow kill the audio. Now, looking at a wireshark trace when this is happening shows that the signals are still there but are at a very low level. I do this by using wiresharks ability to replay, voip calls. The waveforms are shown when you do this and it is obvious the levels are extremely low for the audio pjsip is sending at that time. Second, looking at a log for these bad calls vs good ones (this app works perfect for other devices), I notice entrys like Silence det.c VAD cur_threshold updated 4-->3 signal lo=-1 then 3-6 lo=8 ... 6-7 lo=16. That is the actual data for the call. I do not see anything like this for good calls. Third, the paging speaker plays a tone at first when connected up so you know that in fact you are connected to the paging speaker. I mentioned the two behaviors earlier being: no audio at all OR audio and then degrade after a couple of PTT sequences. Well, If I hold down the PTT on my mic on the machine running pjsip_ua while the call is starting up, the audio will always work right off the bat until I go through a couple of PTT cycles. I can only assume that this is happening because my mic is picking up the tone being played. All of these sound like non-related items, I agree. But it is what it is. The question I guess is, In the absence of a signal from the destination and input from a microphone on the py_pjsua client, does the stack change the VAD threshold. If so, will it have an effect on the transmit level of the signal. Or is this a bug? I know it is strange, but I have verified this multiple times. This is the behavior.