driver46.html revision 290001
1<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN"> 2<html><head> 3 <meta http-equiv="Content-Type" 4 content="text/html;charset=iso-8859-1"><title>GPSD-NG client driver</title> 5 6 <link href="scripts/style.css" type="text/css" rel="stylesheet"> 7 <style type="text/css"> 8 table.dlstable { font-size:85%; } 9 td.ttf{ font-family:Courier; font-weight:bold; } 10 </style></head> 11 12 13 14 <body> 15 <h3>GPSD NG client driver</h3> 16<p>Last update: 17 <!-- #BeginDate format:En2m -->30-Apr-2015 05:53<!-- #EndDate --> 18 UTC</p> 19 <hr> 20 <h4>Synopsis</h4> 21 22 <p> 23 Address: 127.127.46.<i>u</i><br> 24 Reference ID: <tt>GPSD</tt><br> 25 Driver ID: <tt>GPSD_JSON</tt><br> 26 Serial Port: <tt>/dev/gps<i>u</i></tt> as symlink to the true 27 device (not used directly; see below)<br> 28 Features: <tt></tt> 29 </p> 30 31 <!-- --------------------------------------------------------- --> 32 33 <br><h4>Description</h4> 34 <p> 35 This driver is a client driver to the <i>GPSD</i> daemon, which 36 over the time became increasingly popular for UN*Xish 37 platforms. <i>GPSD</i> can manage several devices in parallel, 38 aggregate information, and acts as a data hub for client 39 applications. <i>GPSD</i> can also auto-detect and handle PPS 40 hardware signals on serial ports. Have a look 41 at <a href="http://www.catb.org/gpsd/">the 42 <i>GPSD</i> project page</a>. 43 </p> 44 <p> 45 <b>It is important to understand that this driver works best 46 using a GPS device with PPS support.</b> 47 </p> 48 <p> 49 The GPSD-NG protocol is text based, using JSON notation to 50 transfer records in form of JSON objects. The driver uses a 51 TCP/IP connection to <tt>localhost:gpsd</tt> to connect to the 52 daemon and then requests the GPS 53 device <tt>/dev/gps<i>u</i></tt> to be watched. (Different clock 54 units use different devices, and 55 <i>GPSD</i> is able to give only the relevant information to a clock 56 instance.) 57 </p> 58 <p> 59 This driver does not expect <i>GPSD</i> to be running or the 60 clock device to be present <i>a priori</i>; it will try to 61 re-establish a lost or hitherto unsuccessful connection and will 62 wait for device to come up in <i>GPSD.</i> There is an initial 63 10 seconds delay between a connection loss or failed attempt and 64 the next reconnect attempt; this makes sure that there is no 65 thrashing on the network layer. If the connection fails again, 66 an exponential back off is used with an upper limit of 67 approximately 10 minutes. 68 </p> 69 <p> 70 The overall accuracy depends on the receiver used. The driver 71 uses the error estimations (95% probability limits) provided by 72 <i>GPSD</i> to set the clock precision dynamically according to 73 these readings. 74 </p> 75 <p> 76 The driver needs the VERSION, TPV, PPS, WATCH and TOFF objects 77 of the <i>GPSD</i> protocol. (Others are quietly ignored.) The 78 driver can operate without the TOFF objects, which are available 79 with the <i>protocol</i> version 3.10 and above. (Not to be 80 confused with the <i>release</i> version of <i>GPSD</i>!) 81 Running without TOFF objects has a negative impact on the jitter 82 and offset of the serial timing information; if possible, a 83 version of <i>GPSD</i> with support for TOFF objects should be 84 used. 85 </p> 86 <p>The acronym <u>STI</u> is used here as a synonym for <i>serial 87 time information</i> from the data channel of the receiver, no 88 matter what objects were used to obtain it. 89 </p> 90 91 <!-- --------------------------------------------------------- --> 92 93 <br><h4>Naming a Device</h4> 94 <p> 95 The <i>GPSD</i> driver uses the same device name as the NMEA 96 driver, namely <tt>/dev/gps<i>u</i></tt>. There is a simple 97 reason for that: While the NMEA driver and the <i>GPSD</i> 98 driver can be active at the same time <b>for different 99 devices</b>, they cannot access the same device at a 100 time. Having the same name helps on that. It also eases 101 migration from using NMEA directly to using <i>GPSD</i>, as no 102 new links etc need to be created. 103 </p> 104 <p> 105 <i>GPSD</i> is normally started with the device name to access; 106 it can also be instructed by hot-plug scripts to add or remove 107 devices from its device pool. Luckily, the symlinks used by the 108 NMEA driver are happily accepted and used by <i>GPSD</i>; this 109 makes it possible to use the symlink names as device 110 identification. This makes the migration from the built-in NMEA 111 driver a bit easier. 112 </p> 113 <p><b>Note:</b> <i>GPSD</i> (as of version 3.10) cannot use kernel 114 mode PPS on devices that are hot-plugged. This would require to 115 attach the PPS line discipline to the character special file, 116 which is not possible when running with root privileges already 117 dropped. This is not likely to change in the future. 118 </p> 119 120 <!-- --------------------------------------------------------- --> 121 122 <br><h4>The 'mode' word</h4> 123 <p> 124 A few operation modes can be selected with the mode word. 125 </p> 126 <p> 127 <table border="1" frame="box" rules="all"> 128 <th colspan="3">The Mode Word</th> 129 <tr> <td>Bits</td><td>Value</td><td>Description</td> 130 </tr> 131 <tr> <td rowspan="4"align="center">0..1</td> 132 <td align="center">0</td> 133 <td>STI only operation. This mode is affected by the timing 134 stability of whatever protocol is used between the GPS 135 device and GPSD. 136 <br> 137 Running on STI only is not recommended in general. Possible 138 use cases include: 139 <ul> 140 <li>The receiver does not provide a PPS signal. 141 <li>The receiver <i>does</i> provide a PPS signal and 142 the secondary PPS unit is used. 143 <li>The receiver has a stable serial timing and a proper 144 fudge can be established. 145 <li>You have other time sources available and want to 146 establish a useful fudge value for <tt>time2</tt>. 147 </ul> 148 </td> 149 </tr> 150 <tr> 151 <td align="center">1</td> 152 <td>Strict operation. This mode needs a valid PPS and a 153 valid STI to combine the absolute time from the STI with 154 the time stamp from the PPS record. Does not feed clock 155 samples if no valid PPS+STI pair is available. 156 <br><br> 157 This type of operation results in an ordinary clock with a 158 very low jitter as long as the PPS data is available, but 159 the clock fails once PPS drops out. This mode is a 160 possible choice for receivers that provide a PPS signal 161 most of the time but have an unstable serial timing that 162 cannot be fudge-compensated. 163 </td> 164 </tr> 165 <tr><td align="center">2</td> 166 <td>Automatic mode. Tries to operate in strict mode unless 167 it fails to process valid samples for some time, currently 168 120s. Then it reverts to STI-only operation until the PPS 169 is stable again for 40s, when strict mode is engaged 170 again. 171 <br><br><b>Important Notice: This is an expiremental 172 feature!</b><br> Switching between strict and STI-only 173 mode will cause changes in offset and jitter. Use this 174 mode only if STI-only works fairly well with your setup, 175 or if you expect longer dropouts of the PPS signal and 176 prefer to use STI alone over not getting synchronised at 177 all.</td> 178 </tr> 179 <tr> 180 <td align="center">3</td> 181 <td><i>(reserved for future extension, do not use)</i></td> 182 </tr> 183 <tr> 184 <td align="center">2..31</td> 185 <td colspan="2"><i>(reserved for future extension, do not 186 use)</i></td> 187 </tr> 188 </table> 189 </p> 190 191 <!-- --------------------------------------------------------- --> 192 193 <br><h4>Syslog flood throttle</h4> 194 <p>This driver can create a lot of syslog messages when things go 195 wrong, and cluttering the log files is frowned upon. So we 196 attempt to log persistent or recurring errors only once per 197 hour. On the other hand, when tracking a problem the syslog 198 flood throttle can get into the way.</p> 199 <p>Therefore, fudge <i>flag3</i> can be used to <i>disable</i> the 200 flood throttle at any time; the throttle is engaged by 201 default. Running with the syslog flood throttle disabled for 202 lengthy time is not recommended unless the log files are closely 203 monitored.</p> 204 205 <!-- --------------------------------------------------------- --> 206 207 <br><h4>PPS secondary clock unit</h4> 208 <p>Units with numbers ≥128 act as secondary clock unit for the 209 primary clock unit (u mod 128). A secondary unit processes only 210 the PPS data from <i>GPSD</i> and needs the corresponding master 211 unit to work<a href="#fn1" name="fn1bl"><sup>1</sup></a>. Use 212 the 'noselect' keyword on the primary unit if you are not 213 interested in its data. 214 </p><p>The secondary unit employs the usual precautions before 215 feeding clock samples:</p> 216 <ul> 217 <li>The system must be already in a synchronised state. 218 <li>The system offset must be less than 400ms absolute. 219 <li>The phase adjustment from the PPS signal must also be less 220 than 400ms absolute. 221 </ul> 222 <p>If fudge flag <tt>flag1</tt> is set for the secondary unit, the 223 unit asserts the PPS flag on the clock as long as PPS data is 224 available. This makes the unit eligible as PPS peer and should 225 only be used if the GPS receiver can be trusted for the quality 226 of its PPS signal<a href="fn2" 227 name="fn2bl"><sup>2</sup></a>. The PPS flag gets cleared if no 228 PPS records can be aquired for some time. The unit also flushes 229 the sample buffer at this point to avoid the use of stale PPS 230 data.</p> 231 <p><b>Attention:</b> This unit uses its own PPS fudge value 232 which must be set as fudge <tt>time1</tt>. Only the fudge 233 values <tt>time1</tt> and <tt>flag1</tt> have an impact on secondary 234 units.</p> 235 236 <!-- --------------------------------------------------------- --> 237 238 <br><h4>Clockstats</h4> 239 <p>If flag4 is set when the driver is polled, a clockstats record 240 is written for the primary clock unit. (The secondary PPS unit 241 does not provide clock stats on its own.) The first 3 fields are 242 the normal date, time, and IP address common to all clockstats 243 records. 244 </p><p> 245 <table border="1" frame="box" rules="all"> 246 <th colspan="2">The Clockstats Line</th> 247 <tr> <td>field</td><td>Description</td> </tr> 248 <tr> 249 <td align="center">1</td> 250 <td>Date as day number since NTP epoch.</td> 251 </tr><tr> 252 <td align="center">2</td> 253 <td>Time as seconds since midnight.</td> 254 </tr><tr> 255 <td align="center">3</td> 256 <td>(Pseudo-) IP address of clock unit.</td> 257 </tr><tr> 258 <td align="center">4</td> 259 <td>Number of received known JSON records since last 260 poll. The driver knows about TPV, PPS, TOFF, VERSION and 261 WATCH records; others are silently ignored. 262 </td> 263 </tr><tr> 264 <td align="center">5</td> 265 <td>Bad replies since last poll. A record is considered 266 malformed or a bad reply when it is missing vital fields 267 or the fields contain malformed data that cannot be 268 parsed. 269 </td> 270 </tr><tr> 271 <td align="center">6</td> 272 <td>Number of sample cycles since last poll that were 273 discarded because there was no GPS fix. This is 274 effectively the number of TPV records with a fix value 275 < 2 or without a time stamp. 276 </td> 277 </tr><tr> 278 <td align="center">7</td> 279 <td>Number of serial time information records (TPV or TOFF, 280 depending on the GPSD version) received since last poll. 281 </td> 282 </tr><tr> 283 <td align="center">8</td> 284 <td>Number of serial time information records used for 285 clock samples since the last poll. 286 </td> 287 </tr><tr> 288 <td align="center">9</td> 289 <td>Number of PPS records received since the last poll.</td> 290 </tr><tr> 291 <td align="center">10</td> 292 <td>Number of PPS records used for clock samples on the 293 secondary channel since the last poll. 294 </td> 295 </tr> 296 </table> 297 </p> 298 299 <!-- --------------------------------------------------------- --> 300 301 <br><h4>Fudge Factors</h4> 302 303 <dl> 304 <dt><tt>time1 <i>time</i></tt></dt> 305 <dd>Specifies the PPS time offset calibration factor, in seconds 306 and fraction, with default 0.0.</dd> 307 <dt><a name="fudgetime2"><tt>time2 <i>time</i></tt></a></dt> 308 <dd><em>[Primary Unit]</em> Specifies the TPV/TIME time offset 309 calibration factor, in seconds and fraction, with default 310 0.0.</dd> 311 <dt><tt>stratum <i>number</i></tt></dt> 312 <dd>Specifies the driver stratum, in decimal from 0 to 15, with 313 default 0.</dd> 314 <dt><tt>refid <i>string</i></tt></dt> 315 <dd>Specifies the driver reference identifier, an ASCII string 316 from one to four characters, with default <tt>GPSD</tt>.</dd> 317 <dt><tt>flag1 0 | 1</tt></dt><dd><em>[<b>Secondary</b> 318 Unit]</em> When set, flags the secondary clock unit as a 319 potential PPS peer as long as good PPS data is available. 320 </dd> 321 <dt><tt>flag2 0 | 1</tt></dt> 322 <dd><em>[Primary Unit]</em> When set, <u>disables</u> the 323 processing of incoming PPS records. Intended as an aide to 324 test the effects of a PPS dropout when using automatic mode 325 (mode 2). 326 </dd> 327 <dt><tt>flag3 0 | 1</tt></dt><dd><em>[Primary Unit]</em> 328 If set, <u>disables</u> the log throttle. Useful when tracking 329 problems in the interaction between <i>GPSD</i> and <i>NTPD</i>, 330 since now all error events are logged. Persistent/recurrent 331 errors can easily fill up the log, so this should only be 332 enabled during bug hunts.</dd> 333 <dt><tt>flag4 0 | 1</tt></dt><dd><em>[Primary Unit]</em> 334 If set, write a clock stats line on every poll cycle. 335 </dd> 336 </dl> 337 338 <!-- -- footnotes -------------------------------------------- --> 339 340 <hr> 341 <p><a name="fn1" href="#fn1bl"><sup>1</sup>) </a>Data transmission 342 an decoding is done only once by the primary unit. The decoded 343 data is then processed independently in both clock units. This 344 avoids double transmission over two sockets and decoding the 345 same data twice, but the primary unit is always needed as a 346 downside of this approach. 347 </p> 348 <p><a name="fn2" href="#fn2bl"><sup>2</sup>) </a>The clock driver 349 suppresses the processing PPS records when the TPV/TIME data 350 indicates the receiver has no fix. It can also deal with 351 situations where the PPS signal is not delivered 352 to <i>GPSD</i>. But once it is available, it is also processed 353 and used to create samples. If a receiver cannot be trusted for 354 the precision of its PPS signal, it should not be used to create 355 a possible PPS peer: These get extra clout and can effectively 356 become the sole source of input for the control loop. You do not 357 want to use sloppy data for that. 358 <hr> 359 <p>Additional Information</p> 360 <p><a href="../refclock.html">Reference Clock Drivers</a></p> 361 <hr> 362 <script type="text/javascript" language="javascript" src="scripts/footer.txt"></script> 363 </body></html> 364