[phoney]

In hooking up and measuring the audio signal at a VoIP phone, I am seeing about a 2.5dB loss in the audio path from the analog line side across to the RTP side and out a VoIP phone. I inject a known value tone at the LINE jack of the SPA-3k, then measure it at the first possible analog measurement point of my VoIP phone, and I am losing about 2.5dB somewhere in the trip. Of course it could be something in the VoIP phone that haven't accommodated for, and I am pursuing that in parallel to this. But can anyone with Sipura comment on any audio losses through the line to RTP conversion? I have set the FXS port impedance to properly match the source. I have set the FXS Gain to 0dB. Will there be any loss through the conversion from the line side through the hybrid, and onto the RTP side that I do not know about, or are not accommodating for?

[phoney]

I have not yet been able to solve my 2.5dB loss, seemingly through the SPA3000. I don't know for sure that is where I am losing signal, but that is what I am trying to either prove or disprove. I have moved my measurement point up to the RTP stream, and have managed to extract audio data off the ethernet. Measurements so far are worse than measuring at a known analog point in my VoIP phone. So I am no further ahead than I was. Does anyone have information on undocumented audio loss (or documented that I have missed somewhere) from the FXS port through to the Ethernet side? I have ensured that FXS Gain is set to 0dB.

[mberlant]

It seems that you may be chasing a problem that doesn't exist. From what I can discern from your post, you are performing an analog-to-digital transformation using one device and then trying to extract that signal back to analog on a device that does not provide the same electrical interface.

The first thing you would need to do is to extract the signal on a device that reconstitutes the same analog electrical interface as the injection point. Putting an oscilloscope on some IF point inside a VoIP phone is just not equivalent.

Having found an equivalent extraction device, you still may not get the output dBm to match the input dBm. Good luck.

[phoney]

You are correct, in that where I am measuring is a long ways from where I am injecting, and there a number of gain/loss stages along the way. But I have compensated for those, and do not expect to get out exactly what I put in. But I do expect to get out exactly what I put in, plus all known gains added and subtracted along the way, to come out to calculated value when I measure. So I calculate exactly what I should measure at a particular spot, and then measure with a scope to compare. As far as I know, the gains/losses (and this is the rub because I think everything on my end is OK, but I have to start somewhere to find the error) inside the phone I am measuring are compensated for correctly.

I also tried using EtherReal to capture sample audio and extract the audio data from it and do a measurement with PC software. This is a direct extraction from the ethernet link, which eliminates the phone completely from the measurement. But that measurement turned out even worse than measuring it inside a phone. But I am still pursuing that angle as well.

Any thoughts?

[mberlant]

The only thing I can think of is that you try harder for an apples-to-apples comparison. It's just not a reasonable comparison to inject a 600 ohm 20mA signal at one point and measure the output at an interface with any other electrical characteristic at another point.