Topband: DSP and Latency

[Bill Wichers]

A POTS line *should* have predictable latency *during* any particular call. This is because they operate over circuit-switched networks where the connection/route is established at call initiation and remains the same throughout the duration of the call. Essentially, a defined data path is created through the telephone network for the duration of the call. There is no guarantee that the next call, even to the same destination, will be completed through the exact same path so the latency is not predictable (although will likely be very similar) from call to call.

On the Internet, which is a packet-switched network, each data packet (generally not more than about 1,500 bytes per packet) is routed separately, so each *packet* may take a different path to the destination. Per-packet latency is also affected by network load and buffering in routers. This means that latency *during* an Internet session may vary on a packet-by-packet basis and isn't necessarily predictable within the duration of a "call". In practice things are usually pretty consistent though. Keep in mind also that packets may arrive out of order, and packets will likely take completely different paths (and even different service provider networks) in each direction to/from the destination.

50ms, BTW, is about the time it takes to travel from Chicago to Los Angeles via typical backbone network links. I have a gigabit link at work from Southfield (outside Detroit) to Los Angeles that is around 55ms latency round trip.

One last thing: pretty much ALL phone lines are now digital. This is especially true for links between phone exchanges (and any 100 mile link is sure to be between exchanges). There are very few analog trunks left in service. The telephone network standard is for 8k samples/sec, 8 bit samples. Such circuits are transported as 64kb/s data streams without compression. Some systems use in-band signalling that "robs" one bit from each byte resulting in 7 bit effective samples and a 56kb/s data stream for the analog content.

Jim, I remember the old wide-band phone lines one could special order for remote broadcast setups. A lot of that is done through the Internet now. Crazy new world! :-) Many event venues now provide "house bandwidth" for this purpose. Back when I was working in a studio I remember setting up delay lines to compensate for processing delays on other busses, now the internal DSP in the board does it automatically. It's really amazing what some of the new gear can do without outboard boxes!

   -Bill KB8WYP 

Sent from my iPad

> On Mar 18, 2015, at 2:50 PM, Jim Brown <jim at audiosystemsgroup.com> wrote:
> 
>> On Wed,3/18/2015 11:06 AM, Richard (Rick) Karlquist wrote:
>> There is a very straightforward fix for this latency.
>> Simply use a POTS telephone line with a phone patch.
>> Receive audio comes in, and CW tones go out.  At
>> the remote site, a tone decoder generates key closures.
>> I measured 50 ms latency on a 100 mile phone connection.
>> And you never get audio glitches. 
> 
> Good system, BUT -- don't count on that 50 msec being a constant. Telco's routing may change at any time. And I suspect it may be digital in their system. Why do I say this? Roughly 25 years ago, I had a part time gig doing remotes for WGN Radio for their farm show, which included a live jazz combo. The location rotated around their rather large listening area (50kW on 720 kHz). We did the gig on two dialup POTS lines using  Dutch system that dibandsplitting to put something like 100 - 2700 Hz on one line and 2700-5400 on the other.  It was not  unusual for these lines to have slightly different latencies, and the system had the ability to delay one channel to put them in sync.
> 
> 73, Jim K9YC
> 
> 
> _________________
> Topband Reflector Archives - http://www.contesting.com/_topband