Sorry but the analogue frequencies of telephone calls are not directly related to the digital transmission of data by VOIP. So your analogy is incorrect.

I vaguely recall, from about a million years ago, that the digital sample rate of an analogue source needs to be twice the frequency in order to record it accurately. See Nyquist's theorem. So if the maximum frequency is (say) 4.4KHz them a sampling rate of 8.8Kbps is needed. Add in a bit extra and round it up for better sound quality and 10Kbps ought to do the trick.

Edited by JHo1 (Sat 27-Aug-22 11:24:47)

In reply to a post by Realalemadrid:

Sorry but the analogue frequencies of telephone calls are not directly related to the digital transmission of data by VOIP. So your analogy is incorrect.

Why is he wrong?

The wider the frequency range, the greater the data to be sent, analogue or digital and as such the more bandwidth in this case would be required.

The analogue signal is converted to digital and then sent, exactly the same applies to music and video too. Granted you then start to get into compression ratios and codecs etc however generally the more you put in, the bigger the output, unless you know of some magic that changes this?

In reply to a post by JHo1:

I vaguely recall, from about a million years ago, that the digital sample rate of an analogue source needs to be twice the frequency in order to record it accurately. See Nyquist's theorem. So if the maximum frequency is (say) 4.4KHz them a sampling rate of 8.8Kbps is needed. Add in a bit extra and round it up for better sound quality and 10Kbps ought to do the trick.

You were doing well up until the last sentence

Each sample will require a number of bits to indicate its magnitude, usually a minimum of 8 for reasonable quality, so a sample rate of 10khz needs a bit rate of at least 80kbps.

Higher sampling rates are better but not necessary. A CD of uncompressed music uses 44.1kHz though it may have been down sampled from a higher frequency recording. You're right though in that our sample is not a single bit but rather, well how many? I think we may be furiously agreeing here.

Either way I think we can agree that, even without compression, we're talking in the low kbit range and the impact of a voip conversation on a domestic multi megabit line is going to be very hard to notice.

I found this https://www.nextiva.com/blog/voip-codecs.html which describes three voip codecs, two at 64kbps and one at.... eight! I'm sure there are others

Edited to add nextiva website

Edited by JHo1 (Sat 27-Aug-22 12:22:38)

I have VoIP in its own queue on pfSense so can monitor exactly how much the bit rate is.

Using codec G711, the codec encodes at 64Kbps and the data rate seen in pfSense queue is around 92Kbps. This is a bit higher than what is usually shown online for total expected rates for this codec when overheads are added, which I think in my case is because SIP packets are also included in that queue inflating it slightly. This is in one direction, so it is 92Kbps down and 92Kbps up. With other codecs common to VoIP it can be considerably less than this, for example G729 is around 27Kbps with overheads.

One of the issues with VoIP is the packets are quite small, so the transfer of audio packets are fairly inefficient, but not usually a problem in the home for a couple of calls. A large call center using VoIP would need pretty fast kit to cope with the high number of packets it has to process.

Overall VoIP is very low bandwidth, and given our super fast connections in comparison it is fairly hard to completely max the line out so much that VoIP packets can not find a way through.

The typical codecs used in VoIP are not new. There are better sounding codecs for the same or lesser data rate. For example G711 is from 1972 but still widely used today, one reason it is still used is that the patent has expired so it is free, and pretty much supported by every bit of kit.

Edited by E300 (Sat 27-Aug-22 13:45:43)

Either way I think we can agree that, even without compression, we're talking in the low kbit range and the impact of a voip conversation on a domestic multi megabit line is going to be very hard to notice.

I found this https://www.nextiva.com/blog/voip-codecs.html which describes three voip codecs, two at 64kbps and one at.... eight! I'm sure there are others

You’re correct in that the total bandwidth for a single voice channel is quite small, but the perceived “impact” from the OP question is not the actual consumed bandwidth but the impact other traffic may have on the real time stream - that is would it be affected by other traffic that could delay the sending or receiving of the packets for the voice connection - factors we know as round trip delay (ping) and jitter. The evidence is that unless a given broadband connection is very, very highly used then in reality for a residential connection with one or perhaps a handful of simultaneous voice connections it will generally be OK. So although you could mitigate by partitioning the available data path (virtual LAN) or by prioritising voice traffic by tagging the packet (QoS) it’s not really necessary on a domestic connection, or at least it’s positive impact is probably quite negligible.

On the other hand corporate connections with much higher volumes of VoIP (simultaneous channels) are a different matter and it is good practise by the service provider on something like a leased line WAN connection to partition some of the available bandwidth into a separate VLAN for VoIP/Telephony so that it is not “crowded out” by other less delay sensitive traffic on the main VLAN.

The sound quality of voice connection is typically measured by something called a Mean Opinion Score (MOS). Have a search for it and do some extra background reading whilst you are looking at VoIP codecs. There is a lot of history (and some corporate politics) with certain wideband codecs especially and there has been a lot of evolution in the last few years.

Most standard VoIP interconnections have for decades followed the TDM world of telephony and have defaulted to G.711 a or u-law codecs which are a pretty good trade off for quality vs DSP processing requirements vs bandwidth (the three conflicting holy grails of voice codecs).

Here is some pretty good starter for ten, background on codecs:

https://info.teledynamics.com/blog/demystifying-code...

https://info.teledynamics.com/blog/the-wonderful-wor...

The most common VoIP codecs you’re going to find are:
G.711 A-law
G.712 u-law
G.722
Opus
iLBC
G.729A
GSM

Some of the best quality calls will be proprietary platform such as Apple FaceTime calls which use I believe use the AAC-ELD codec for the voice element of calls.

But is 8 bit actually needed for voice? Then throw in the codec and compression.

We managed to get reasonable voice quality, that was acceptable across 2.4 and 4.8 kbps links although most was at either 4.8 or 9.6 kbps!

This is all interesting in a nerdy way

Apple FaceTime is amazing, when came out of hospital a few years ago I stayed with my sister and brother-in-law for a week and they used an Ipad to chat to my nephew, even on a naff ADSL connection the quality was excellent, both video and audio.
Far better than anything I saw on Skype or other systems at the time.

Thanks for all the responses, most informative and interesting.

I'll probably go with A&A when the time comes as it didn't get used an awful lot and (at the time of writing) £1.20 per month to host a number is better than £7.