Steve Underwood <[email protected]> writes:

> Rune Allnor wrote:
>> On 22 Mar, 08:38, minfitl...@yahoo.co.uk wrote:
>>> What is considered real-time? How much latency would be necessary for
>>> it to be no longer real-time? The application is adaptive filtering.
>>> Somebody told me that real-time is strictly one sample in and one
>>> sample out but most of the processors I have met don't work that way
>>> any more - they work on buffers of data and surely all real-time
>>> filters have some latency or is one sample all that is required?
>> "Real time" means that the data are processed as they come in,
>> and that this processing is done inside an acceptable latency.
>> In a phone system, this "acceptable latency" may be on the
>> order of milliseconds.
>> Where I work, the "acceptable latency" is 24 hours. It is still
>> a real-time system.
>
> The last time I built DSP for the telephone network, the acceptable
> latency was 3 sample times - 375us. One to deserialise, one to
> serialise, and just one left for the processing delay in the
> middle. :-)
>
> A true real time application typically doesn't have an "acceptable"
> latency. It has a hard and well defined latency, it absolutely
> positively must meet under all circumstances. For a control loop that
> hard latency is typically very small for acceptable performance. For a
> satellite TV decompressor it might be quite long. Nonetheless, the
> incoming data will *always* be accepted without ever losing a bit, and
> the output will be provided a very much fixed delay after the input
> data is received.

In my opinion any discussion of latency should be removed from the
definition of "real-time." These are two "orthogonal" concepts.
--
% Randy Yates % "The dreamer, the unwoken fool -
%% Fuquay-Varina, NC % in dreams, no pain will kiss the brow..."
%%% 919-577-9882 %
%%%% <[email protected]> % 'Eldorado Overture', *Eldorado*, ELO
http://home.earthlink.net/~yatescr