DP83640: DP83640

Dan Ringger

Part Number: DP83640

I have a question regarding timestamping accuracy/resolution of the part. What resolution can I expect to get when reading the timestamp via the PTP_CTL PTP_RD_CLK method or via the gpio event timestamping functionality? We have tried both of these methods triggered via a GPS pulse per second output that is accurate to (at worst) 100 ns of UTC. When corrected for frequency, we are only seeing a resolution of +/- roughly 65 us with spikes in these values into the hundreds of microseconds/low milliseconds.

over 7 years ago

0 Ross Pimentel over 7 years ago

TI__Mastermind 23675 points

Hi Daniel,

Have you had a look at this application note?
www.ti.com/.../snla100a.pdf

Kind regards,
Ross

0 Dan Ringger over 7 years ago in reply to Ross Pimentel

TI__Intellectual 1790 points

Hi Ross - Yes. Unfortunately while it discusses the accuracy of the dp83640 (especially as a slave clock) it does not discuss the initial calibration of the PHY to ensure the phase of the PTP clock is in sync with actual UTC time. My question regarding the sampling of the dp83640 clock was also not mentioned in the app note. Thank you

0 Ross Pimentel over 7 years ago in reply to Dan Ringger

TI__Mastermind 23675 points

Hi Daniel,

You will need to use the PTP Rate Control Registers and PTP Temporary Rate Duration Control Registers.
These registers are discussed more in detail here:www.ti.com/.../snla099d.pdf

Beginning on bottom of Page #2 you will find a discussion on rate correction.

Kind regards,
Ross

0 Dan Ringger over 7 years ago in reply to Ross Pimentel

TI__Intellectual 1790 points

The app note does provide good insight into a method to phase correct the PHYs 1588 clock. However, my question is regarding querying timestamps from the dp83640 via the PTP_RD_CLK Method and the event timestamping method via a GPIO. I am seeing a 65 us precision on these returned timestamps. For example if I were to configure a gpio to interrupt on a rising edge of a signal, then query the time stamp of the event from software between many pps the difference between the data points would look something like the following:

1000055000 ns

999990000 ns

0 Ross Pimentel over 7 years ago in reply to Dan Ringger

TI__Mastermind 23675 points

Hi Daniel,

To query timestamps, the general method is to use the PTP_EVNT (0x15h) register followed by the PTP_RXTS (0x1Dh).
Within the datasheet are steps to follow to get the correct values out of PTP_RXTS. If you want to view transmit timestamp, you will need to view the PTP_TXTS (0x1Ch).

Kind regards,
Ross

Interface

Interface forum

DP83640: DP83640