Serial port splitter s/w
From: Chris Albertson
[]
Except as an exercise there is almost no point any more using GPS to
drive a local NTP server if you have a very good Internet connection.
My fiber Internet connection is as good as Ethernet as the Pool
Servers work well for me. Those using slower or a widely shared
connection might not get as good a result.
Chris Albertson
Redondo Beach, California
---==========
The Raspberry Pi with GPS/PPS certainly beats any internet connection I've
ever had delivered to this house, though:
http://www.satsignal.eu/ntp/Raspberry-Pi-NTP.html
http://www.satsignal.eu/mrtg/performance_ntp.php
Cheers,
David
SatSignal Software - Quality software written to your requirements
Web: http://www.satsignal.eu
Email: david-taylor@blueyonder.co.uk
On Wed, Feb 26, 2014 at 11:15 AM, David J Taylor <
david-taylor@blueyonder.co.uk> wrote:
The Raspberry Pi with GPS/PPS certainly beats any internet connection I've
ever had delivered to this house, though:
Yes it will. But the point of the server is to deliver time to other
computers. The the measure is "goodness" is not the accuracy of your local
NTP server but the accuracy of the client computers that use it. So if
I point my desktop computer at a set of 5 to 7 Internet pool servers or it
I point the same desktop to my local NTP server I don't se a lot of
difference.
With the notebook computer using WiFi I don't see any difference.
This is what I moment by saying it is hardly worth it if you happen to have
Ethernet-like connectivity all the way back to your ISP. Of douse many
people don't and have to use DSL or Cable TV lines.
The ARM has some real potential for time keeping. Inside the chip in
addition to the ARM CPU are two "RPU" processors. Thee run at exactly
200MHz one instruction per cycle with very deterministic timing. They have
access to all of the ARM memory and interrupts. They are in effect two
micro controllers that are in addition to however many ARM cores there are.
A really good project would be to move NTP's PPS timing to a PRU. But
as I said the weak link is getting time OUT of the NTP server, not into it.
But again the ARM is great for real time because it has both the ARM
cores runnig Linux and the PRUs that you can program in assembly with
deterministic timing.
But as of today I don't think the PRUs are used for this but see below,
lots of potential uses for building things like TICs
http://processors.wiki.ti.com/index.php/Programmable_Realtime_Unit_Subsystem
http://www.satsignal.eu/ntp/Raspberry-Pi-NTP.html
http://www.satsignal.eu/mrtg/performance_ntp.php
Cheers,
David
SatSignal Software - Quality software written to your requirements
Web: http://www.satsignal.eu
Email: david-taylor@blueyonder.co.uk
time-nuts mailing list -- time-nuts@febo.com
To unsubscribe, go to
https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.
--
Chris Albertson
Redondo Beach, California
From: Chris Albertson
Sent: Wednesday, February 26, 2014 7:32 PM
To: Discussion of precise time and frequency measurement
Subject: Re: [time-nuts] Serial port splitter s/w
On Wed, Feb 26, 2014 at 11:15 AM, David J Taylor <
david-taylor@blueyonder.co.uk> wrote:
The Raspberry Pi with GPS/PPS certainly beats any internet connection I've
ever had delivered to this house, though:
Yes it will. But the point of the server is to deliver time to other
computers. The the measure is "goodness" is not the accuracy of your local
NTP server but the accuracy of the client computers that use it. So if
I point my desktop computer at a set of 5 to 7 Internet pool servers or it
I point the same desktop to my local NTP server I don't se a lot of
difference.
With the notebook computer using WiFi I don't see any difference.
This is what I moment by saying it is hardly worth it if you happen to have
Ethernet-like connectivity all the way back to your ISP. Of douse many
people don't and have to use DSL or Cable TV lines.
The ARM has some real potential for time keeping. Inside the chip in
addition to the ARM CPU are two "RPU" processors. Thee run at exactly
200MHz one instruction per cycle with very deterministic timing. They have
access to all of the ARM memory and interrupts. They are in effect two
micro controllers that are in addition to however many ARM cores there are.
A really good project would be to move NTP's PPS timing to a PRU. But
as I said the weak link is getting time OUT of the NTP server, not into it.
But again the ARM is great for real time because it has both the ARM
cores runnig Linux and the PRUs that you can program in assembly with
deterministic timing.
But as of today I don't think the PRUs are used for this but see below,
lots of potential uses for building things like TICs
http://processors.wiki.ti.com/index.php/Programmable_Realtime_Unit_Subsystem
---========
Chris, even with Wi-Fi connected computers, mostly running Windows, there is
a huge difference between talking to a stratum-1 server on my LAN compared
to running just Internet servers. Our experiences differ, as I am on a
cable modem connection from the UK's Virgin Media. Folks need to measure
what performance they are getting and choose their own best path, otherwise
it's guesswork.
Cheers,
David
SatSignal Software - Quality software written to your requirements
Web: http://www.satsignal.eu
Email: david-taylor@blueyonder.co.uk
Chris, even with Wi-Fi connected computers, mostly running Windows, there
is a huge difference between talking to a stratum-1 server on my LAN
compared to running just Internet servers. Our experiences differ, as I am
on a cable modem connection from the UK's Virgin Media. Folks need to
measure what performance they are getting and choose their own best path,
otherwise it's guesswork.
Yes, exactly. It depends entirely on your internet connection. As soon as
you get even "slow" 10Mb/s fiber the distinction between LAN and Internet
starts to melt away. The problem with Cable TV is that it is a shared
connection with who knows how many others. Shared media have collisions
with back off and retries and are not deterministic.
I do have a local NTP server. I had one back in the days of dial-up phone
modems and still have one with my current fiber connection. I was just
pointing out that the local NTP server is less useful as the Internet
connections get better.
Every few years I get motivated to improve the NTP server. The next step
is for certain going to be to use the ARM's PRUs somehow. That would
remove the effect of the OS from timing which is now the largest source of
error.
Cheers,
David
SatSignal Software - Quality software written to your requirements
Web: http://www.satsignal.eu
Email: david-taylor@blueyonder.co.uk
time-nuts mailing list -- time-nuts@febo.com
To unsubscribe, go to https://www.febo.com/cgi-bin/
mailman/listinfo/time-nuts
and follow the instructions there.
--
Chris Albertson
Redondo Beach, California
On Wed, Feb 26, 2014 at 2:05 PM, Chris Albertson
albertson.chris@gmail.comwrote:
Chris, even with Wi-Fi connected computers, mostly running Windows, there
is a huge difference between talking to a stratum-1 server on my LAN
compared to running just Internet servers. Our experiences differ, as I
am
on a cable modem connection from the UK's Virgin Media. Folks need to
measure what performance they are getting and choose their own best path,
otherwise it's guesswork.
Yes, exactly. It depends entirely on your internet connection. As soon as
you get even "slow" 10Mb/s fiber the distinction between LAN and Internet
starts to melt away. The problem with Cable TV is that it is a shared
connection with who knows how many others. Shared media have collisions
with back off and retries and are not deterministic.
Not generally with broadband. (True broadband that is, such as
Internet-over-cable. I do find it annoying that all higher-speed, i.e.
non-dial-up, access is referred to as "broadband".) There are no collisions
coming downstream because there is only one source -- the head end. There
you just have variable queueing delays if a burst of traffic exceeding
downstream capacity arrives at once. Collisions are possible on the
upstream but less likely.
I do have a local NTP server. I had one back in the days of dial-up phone
modems and still have one with my current fiber connection. I was just
pointing out that the local NTP server is less useful as the Internet
connections get better.
True, but far more deterministic. As you suggest, a local NTP server is
independent of upstream traffic, queueing, and backoff/retry delays.
Personally I would like to have a local stratum-1 source for that reason.
--
Brian Lloyd, WB6RQN/J79BPL
706 Flightline Drive
Spring Branch, TX 78070
brian@lloyd.com
+1.916.877.5067
You guys have me persuaded - I'll get a Raspberry Pi ...
Regards,
David Partridge
-----Original Message-----
From: time-nuts-bounces@febo.com [mailto:time-nuts-bounces@febo.com] On
Behalf Of David J Taylor
Sent: 26 February 2014 19:15
To: Discussion of precise time and frequency measurement
Subject: Re: [time-nuts] Serial port splitter s/w
The Raspberry Pi with GPS/PPS certainly beats any internet connection I've
ever had delivered to this house, though:
http://www.satsignal.eu/ntp/Raspberry-Pi-NTP.html
http://www.satsignal.eu/mrtg/performance_ntp.php
Cheers,
David
SatSignal Software - Quality software written to your requirements
Web: http://www.satsignal.eu
Email: david-taylor@blueyonder.co.uk
On Wed, Feb 26, 2014 at 1:30 PM, David C. Partridge <
david.partridge@perdrix.co.uk> wrote:
You guys have me persuaded - I'll get a Raspberry Pi ...
You might want to think a little more about which to get. The Pi is OK and
will work fine but look at "BeagleBone Black". It's a little nicer for
hardware hackers. Cost is $5 different. You can read specs and the
respective forums.
One thing to watch out for is that all ARM boards run on 3.3 volts. The
serial ports are 3.3 volts. Don't connect a 5 volt signal and certainly
NOT a real RS232 level signal
--
Chris Albertson
Redondo Beach, California
From: Chris Albertson
You might want to think a little more about which to get. The Pi is OK and
will work fine but look at "BeagleBone Black". It's a little nicer for
hardware hackers. Cost is $5 different. You can read specs and the
respective forums.
One thing to watch out for is that all ARM boards run on 3.3 volts. The
serial ports are 3.3 volts. Don't connect a 5 volt signal and certainly
NOT a real RS232 level signal
---==
It would be interesting to see a performance comparison between the RPi and
the BBB. when running NTP from a PPS signal. The RPi is much more popular
and has a much wider user base, though.
I've used mine for a 1.09 GHz ADS-B receiver (aircraft tracking), logging
room temperature, as a wall-clock synced from NTP, of course, and for
various other applications. The GPS units you can get these days support
3.3V operation so that's not a difficulty in practice. I've used these
units for both "with soldering" and "no soldering needed" choices:
With soldering:
http://www.adafruit.com/products/746
"No soldering required"
http://ava.upuaut.net/store/index.php?route=product/product&path=59_60&product_id=95
Certainly, for external true RS-232 devices an interface is needed, but
these can again be bought for the Raspberry Pi (data only):
http://www.davidhunt.ie/?p=3091
http://www.abelectronics.co.uk/products/3/Raspberry-Pi/29/Serial-Pi-RS232-Interface
Cheers,
David
SatSignal Software - Quality software written to your requirements
Web: http://www.satsignal.eu
Email: david-taylor@blueyonder.co.uk
A number of people have reported issues with the Raspberry Pi Ethernet hardware when used for NTP, as it is actually a USB-Ethernet bridge and the drivers may not be all they could be. I have not had problems myself but I do not run NTP on it.
The Beaglebone Black (BBB) is supposed to be better in that regard since the Ethernet MAC is directly on the CPU. The BBB is also entirely open source, unlike the proprietary Broadcom CPU on the RPi.
Didier KO4BB
On February 26, 2014 11:32:36 AM CST, Chris Albertson albertson.chris@gmail.com wrote:
It's not going to work.
If the purpose of running the Thunderbolt is only to drive NTP then
you don't need LH. NTP's only tags the pulses to the nearest
microsecond, nano sec on accuracy is lost on NTP. I'd even say the
TB is the wrong GPS for NTP. It costs to much and uses to much power.
But if you are also, or mainly, using the Thunderbolt for it's 10Mhz
and NTP is a secondary function then a TB makes sense.
Yes you NEED the PPS on the serial cable. Thunderbolts do not send
NMEA. Thunderbolts send their own data format that is unique to
Trimble. Don't modify the GPS receiver. Make a special cable
adapter. When you do this pay attention to polarity of the PPS
signal. It is easy to get it backwards. You want the raising edge of
the TB pulse to interrupt the computer. It you invert the signal the
wrong number of times the time will be "off" by the ouse length and I
don't know if the pulse length is controlled to the level the leading
edge is. Remember RS232 uses negative and positive voltage, data
lines use negative logic, control lines positive. The TB's PPS is
TTL level. Many rs232 ports do accept t/l level if you get the
polariy correct.
Again don't even bother to run an NTP server without PPS. You may as
well just get time from some internet time servers.
You can NOT control a GPS from two ports. Both NTP and LH will try to
send commands to the GPS.
Likely, almost certainly you need to build a small circuit board the
has two connectors that face the TB (PPS and serial) and one that
faces the computer. The little perfboard makes a neat way to or
connect cables but you could solder up a y--cable
The best thing to do is get a cheaper GPS, and one that uses less
power to drive the NTP server. The old Motorola Oncore series are
cheap and the new breed of very small GPSes are good too. DOn't spend
more than $40 or $50 on a GPS to drive NTP as ,again NTP record
microseconds.
You could free up that Windows PC too. It is not the best platform
for NTP. Asmall ARM based system (even the Rasbery Pi) will
outperform a Windows based NTP server. and use a LOT less power
(Power cost for a NTP server is more than you think, it came to about
$300 a year for me if I used a standard PC and a thunderbolt.
Switching to a very tiny ARM based system and a smaller GPS gave as
good performance and power savings paid for the hardware in 1/2 a
year. $.21/KWH about 170W and 8760 hours per year comes to a $300
power bill. My current system is powered by a 1000 mw plug-in power
cube and does not need a cooling fan.
On Wed, Feb 26, 2014 at 7:53 AM, David C. Partridge
david.partridge@perdrix.co.uk wrote:
I'm running Meinberg NTP on the Windows 7 x64 machine to which my
Thunderbolt is attached.
I'd like to be able to share the serial port between LH and NTP so
that I can run the machine as an NTP Stratum 1 server locked to the TB,
and also be able to use LH to check things.
I looked around the with Google, and saw *numerous" serial port
splitters. Which is recommended?
Also what's the best way how to configure NTP to lock the the TB on a
serial port? Do I need to modify the TB to deliver the PPS down one of
the serial data lines or will NTP work well by parsing the NMEA time
messages?
and follow the instructions there.
--
Chris Albertson
Redondo Beach, California
time-nuts mailing list -- time-nuts@febo.com
To unsubscribe, go to
https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.
--
Sent from my Motorola Droid Razr 4G LTE wireless tracker while I do other things.