<html>
<head>
<meta http-equiv="Content-Type" content="text/html; charset=UTF-8">
</head>
<body text="#000000" bgcolor="#FFFFFF">
On 8/12/18 17:47, Paul Theodoropoulos wrote:
<blockquote type="cite"
cite="mid:2130d352-63d2-0b06-9894-cb413bf8e8f1@anastrophe.com"> be
able to learn more after the USB-TTL adapter I ordered arrives, so
I can run the ublox uCenter software which is only available on
Windows. I'm not going to try taking the lid off.<br>
<span>
<div> <br>
There's no evidence that the 'battery' (or more likely 'super'
cap) on the package actually is connected to anything. It
seems to lose everything on a power-cycle.</div>
</span></blockquote>
Followup. Much probing in uCenter, no configuration will 'take' -
set stationary mode, survey-in, blah blah - all lost at the next
powercycle, no evidence that almanac/ephemeris data is saved, so
it's a fifteen minute slog at each startup. Blech.<br>
<br>
Was getting pretty poor performance in the original build on the Pi
Zero W, so switched the board to an existing RPI 3B+, dramatic
improvement:<br>
<br>
<a class="moz-txt-link-freetext" href="https://imgur.com/a/H0t8D9n">https://imgur.com/a/H0t8D9n</a><br>
<br>
So, pretty clearly the Pi Zero W, as attractive and tiny and
inexpensive as it is, is not a great platform for NTP. Well, in
absolute terms sure, it's an order of magnitude better than relying
on getting time only over the WAN. But with a base difference in
cost of only $25 for an RPi 3 B+, it's not worth bothering. Use the
Zero W for some other task.<br>
<br>
Note, I did apply some of my personal favorite fine-tunings - set up
realtime partitioning, pin gpsd and ntpd to their own cpu's, but
they only give tiny, incremental improvements - nothing even visible
in these graphs.<br>
<br>
<br>
<pre class="moz-signature" cols="72">--
Paul Theodoropoulos
<a class="moz-txt-link-abbreviated" href="http://www.anastrophe.com">www.anastrophe.com</a></pre>
</body>
</html>