Commits on Source (24)
-
Trevor N authored53b64748
-
Trevor N authored
The buffer length check was incorrect, so large packets were overwriting the clock type stored in the control structure. Expanded the buffer size to accommodate the largest TSIP packet listed in the receiver manuals.
38c37cc0 -
Trevor N authored
removing file-level state variable allows for multiple receivers to be used simultaneously
bfddacca -
Trevor N authored
When not using event capture, offset can vary by hundreds of ms between ntpd startups because the poll timer's random relationship to the top of the second causes the timestamp of a random packet to be used, but only the first packet at the top of the second contains the timecode (for the top of the second).
6f113661 -
Trevor N authored
TSIP does not contain checksums, but most receivers have parity transmission enabled by default so check it when available.
52cc8485 -
Trevor N authored
Remove the ability to disable event polling because packets 8f-ad and 8f-0b are needed for the rollover workaround (#461) but the Palisade and Acutime receivers transmit only packet 8f-ad if their event inputs aren't used. Running with event polling disabled is normally done for troubleshooting purposes, but it's also used for some Palisades with firmware that doesn't allow event inputs. A no-cost firmware upgrade is available which enables event inputs for those receivers.
2b50a20b -
Trevor N authored
* remove partially-implemented Acutime Gold 'Port B' mode -- gpsd can be used when connected to port B. * remove decoding of packets that have information also contained in the timing superpackets * remove decoding of packets that are not helpful for troubleshooting
564b187d -
Trevor N authored5b70a2ea
-
Trevor N authored
Testing of the leap detection logic with a Spirent GSS6100 GPS signal generator revealed that the logic for Palisade and Acutime receivers was incorrect. The leap-second-flag bits (packet 8f-ad byte 19) are the same for both insertion and deletion, so assume insertion. To simplify polling: don't count packets, instead finish the poll with the last packet in the sequence sent after an event or auto-report.
3bdde6fe -
Trevor N authored21317d50
-
Trevor N authored
feature request listed as bug #1765 in the reference implementation
1ef5c918 -
Trevor N authored8b3559df
-
Trevor N authoredf4fea7fe
-
Trevor N authored
The simplest solution of waiting a few microseconds between setting and clearing RTS causes problems with the Acutime 2000 and Praecis. The Acutime 2000 may become unresponsive if the pulse time is between 5us and 600ms, and the Praecis will occasionally produce event timestamps that are incorrect by a few us when the pulse width is between ~5us and 100us. Extending the RTS set time to the entire polling interval works for all supported receivers, but a sleep for at least 12ms is needed between clearing and re-setting RTS for the receivers to re-arm their event inputs.
4ee142d4 -
Trevor N authored
fixed displaying when time transfer skipped now works with Thunderbolt
93ea72cb -
Trevor N authored82320a5f
-
Trevor N authoreda01dfba0
-
Trevor N authored64d5a2bb
-
Trevor N authored85c8e246
-
Trevor N authored
tested for proper startup with host time set to 1970 and 2037
28e0476f -
Trevor N authored9513772c
-
Trevor N authoredfd077e04
-
Trevor N authored34b30741
-
Trevor N authored2d5caf47
This diff is collapsed.
This diff is collapsed.