NI PCI 1588

[Elvita]

NI PCI 1588 guarantees a clock accuracy upto +/- 230ns.I have two questions regarding the same. 1.Does it compensate for clock drift.If not ,how can it achieve an accuracy in the ns range? 2.When used with Ethernet as the underlying medium,again an accuracy of 230ns seems far-fetched due to the delay introduced by the random nature of the CSMA/CD protocol.[ (2^k -1)*51.2 us for 10 Mbps Ethernet]

[JoshuaP]

Those are good questions. The PCI-1588 boards uses a very precise 10MHz (1.5 ppm) TCXO oscillator to start with to minimize drift between synchronization updates. The board then uses the 1588 protocol to perform the synchronization updates and adjust the 1588 time to ensure that the slave clocks do not drift from the master clock over time.

When the 1588 devices are added to the network they will communicate with the other 1588 devices to determine which device is the best master and the rest of the 1588 devices will become slaves.

The 1588 protocol takes care of accounting for network latencies by measuring the time it takes to send packets between the master and the slaves. This of course will work very well on networks that have minimal jitter. Straight cables as well as hubs have the best performance and least amount of jitter. Devices such as switch will add additional jitter to the system, because of there non-deterministic behavior. There are switch (transparent switches) that are specifically designed for 1588 enabled networks that can also be used to minimize the jitter.

For more information here is a good document that may provide more information.
Introduction to Distributed Clock Synchronization and the IEEE 1588 Precision Time Protocol

-Josh

[Elvita]

Thanks Josh,

From your reply ,I conclude, the NI PCI 1588 makes no "external, physical" compensation for clock drift.
It just relies on its highly precise TCXO .

The datasheet mentions

TCXO Characteristics
Frequency ............................................ 10 MHz
Initial accuracy.................................... ±1.5 ppm

So in a second the oscillator drift is one microsecond.How can it then achieve accuracy in the nanosecond range,with no compensation for drift?

And even if that is somehow possible,I still do not get understand how it can guarantee accuracy upto nanoseconds on Ethernet,when the random delay introduced by the medium protocol is in the microsecond range.

I am sorry but the hyperlink you sent me does not seem to work.

Thanks for your reply once again!

[f.grasdepot@laposte.net]

Hi,

the only way to achieve a microsecond accuracy over Ethernet ( more
accurately a microseconds repeatabilty ) is to load the full program as
a batch and to have a batch of commands executed locally. Otherwise,
you cannot guarantee the timing over the network. This is explained in
my LabTab software manual, among other documents.

FG

[Chris_R.]

Hi Elvita,

The link above has an extra quotation mark.  Just remove that in the link or click here for the corrected link.

[Elvita]

Thanks Chris,
I googled the link and found it really helpful!

FG,can you send me the link to the LabTab software manual which contains the Ethernet working description

Thanks again,

Elvita.

[JoshuaP]

Here is a link from the 1588 site that may help.

http://ieee1588.nist.gov/z_RA_1_FreqCompClk.pdf

[Dennis_Knutson]

Forget about it. FG has made a lot of postings here to (I think) advertise his LabTab product and as far as I can tell by looking at the manual, it has absolutely nothing to do with Ethernet. I'm not sure it has much to do with LabVIEW either.