|
|
Posted by Roberto Divia on 05/22/07 12:21
Rick Merrill wrote:
> Roberto Divia wrote:
>> In any case, for DV cameras the "throughput" issue does not exist (but
>> the
>> "CPU occupancy" does...).
>
> Could you expand on that terminology?
The majority of DV cameras are tape-based and transfer video at tape speed.
Video goes via the isochronous channel of FireWire (unused for normal traffic)
and the data flow is naturally limited by the bitrate of the DV-AVI stream
(~3.6 MB/s sustained). Therefore peak throughput is not an issue and neither
is maximum sustained throughput. All the link (FireWire or USB) has to do
is to keep up with the data stream. During the process, however, the CPU
must be able to follow the peripheral, the data link and to control both.
USB is totally under control of the CPU (setup of each individual transfer,
completion and control of each individual transfer, control of the link)
while FireWire - which is more slave-driven - uses isochronous traffic:
this requires zero control from the CPU and no arbitration from the
peripheral, reducing considerably the CPU load during the transfer and
the probability of a TC break (lost frames). Lost frames are mainly due to
the inability of the CPU to keep up with the data stream. By lowering the
overall CPU load and by giving to the peripheral a dedicated (and guaranteed)
data link, we reduce considerably the possibility of lost frames, which are
the most common error that can happen during synchronous data transfers.
If instead we speak of asynchronous data transfers (e.g. from a HDD-based
video camera), things are totally different...
Ciao,
--
Roberto Divia` Love at first sight is one of the greatest
Dep:PH Bat:53 Mailbox:C02110 labour-saving devices the world has ever seen
Route de Meyrin 385 ---------------------------------------------
Case Postale Phone: +41-22-767-4994
CH-1211 Geneve 23 CERN Fax: +41-22-767-9585
Switzerland E-Mail: Roberto.Divia@cern.ch
Navigation:
[Reply to this message]
|