Posted at 2009-10-04 16:56 BST by RevK
Update #0: 2009-10-04 16:58 BST

We are doing a lot of work in October, upgrading some key servers,
moving some key servers, adding new servers (e.g. email), and so on.
This will all help improve our services.

We'll post more details during the month.

Obviously we'll aim to minimise disruption to services in the process.[IMAGE]

URL: http://aaisp-planned.blogspot.com/2009/10/major-upgrades-in-october.html

-- 
AAISP Planned Work Blog
http://aaisp-planned.blogspot.com/

RevK wrote:
> adding new servers (e.g. email), and so on.
> This will all help improve our services.
> 

Was ther a plan to add an NNTP server following the shut-down of the 
Gradewll service?   I seem to remember some mention of a server being 
provisioned.   Or is my memory faulty?

-- 
Ron

In article <hab0n6$3h6$1@news.eternal-september.org>, Ron Lowe wrote:
> Was ther a plan to add an NNTP server following the shut-down of the 
> Gradewll service?   I seem to remember some mention of a server being 
> provisioned.

The servers list on the A&A website at 

http://aaisp.net.uk/kb-broadband-serverlist.html

says "News service. Expected to be available soon ...".

I'm certainly hoping that an NNTP server will be one of those that is 
introduced in these new upgrades ... but perhaps we'd better wait until 
Adrian posts "more details during the month", as promised, to find out.

Cheers,
 Daniel.

On 05/10/09 08:38, Daniel James wrote:
> http://aaisp.net.uk/kb-broadband-serverlist.html

"NTP  time.aaisp.net.uk  Use this to set your clock by NTP"

Does A&A have only one NTP server? It is usually recommended to use
three NTP servers, and good ISPs usually provide more than one, which
saves from have to connect to heavily loaded public servers, possibly
with poor latency.

If an ISP hosts its own NTP servers they can provide their customers
with better quality time with reduced latency, reduce the load on public
servers and reduce (slightly) their external bandwidth requirements. I
believe that running your own stratum 1 server off a GPS receiver or
similar is quite inexpensive these days.

Roger

Roger Lynn wrote:
> On 05/10/09 08:38, Daniel James wrote:
>> http://aaisp.net.uk/kb-broadband-serverlist.html
> 
> "NTP  time.aaisp.net.uk  Use this to set your clock by NTP"
> 
> Does A&A have only one NTP server? It is usually recommended to use
> three NTP servers, and good ISPs usually provide more than one, which
> saves from have to connect to heavily loaded public servers, possibly
> with poor latency.
> 
> If an ISP hosts its own NTP servers they can provide their customers
> with better quality time with reduced latency, reduce the load on public
> servers and reduce (slightly) their external bandwidth requirements. I
> believe that running your own stratum 1 server off a GPS receiver or
> similar is quite inexpensive these days.

It would be handy to have three time servers available but for me
the delay for time.aaisp.net.uk is 17ms which is only 2ms from next 
closest.

I'm setting up to use a Conrad module that I've modified with 60kHz 
crystal and currently gives offset mostly better than 200us. That
cost about 12quid for two, when they were doing two for one. With 
internet time the offset is mostly better than 500us but varies a
lot more than MSF, maybe +/-5ms vs +/-1ms.

Garmin GPS18X-LVC cost around 70quid and gives offset mostly 2us
varying to maybe 6us but I'd need a mast above roofline for reliable
reception.

Conrad module tuned to DCF77 suffers from bad reception here (1000km
range) for a couple of periods most evenings and is comparable to 
internet time, but potentially at closer range DCF77 can be
synchronised to the signals phase modulation which should be within a
few us or better.


David

David Lord  writes:

> Roger Lynn wrote:
>> On 05/10/09 08:38, Daniel James wrote:
>>> http://aaisp.net.uk/kb-broadband-serverlist.html
>>
>> "NTP  time.aaisp.net.uk  Use this to set your clock by NTP"
>>
>> Does A&A have only one NTP server? It is usually recommended to use
>> three NTP servers, and good ISPs usually provide more than one, which
>> saves from have to connect to heavily loaded public servers, possibly
>> with poor latency.
>>
>> If an ISP hosts its own NTP servers they can provide their customers
>> with better quality time with reduced latency, reduce the load on public
>> servers and reduce (slightly) their external bandwidth requirements. I
>> believe that running your own stratum 1 server off a GPS receiver or
>> similar is quite inexpensive these days.
>
> It would be handy to have three time servers available but for me
> the delay for time.aaisp.net.uk is 17ms which is only 2ms from next
> closest.
>
> I'm setting up to use a Conrad module that I've modified with 60kHz
> crystal and currently gives offset mostly better than 200us. That
> cost about 12quid for two, when they were doing two for one. With
> internet time the offset is mostly better than 500us but varies a
> lot more than MSF, maybe +/-5ms vs +/-1ms.
>
> Garmin GPS18X-LVC cost around 70quid and gives offset mostly 2us
> varying to maybe 6us but I'd need a mast above roofline for reliable
> reception.

But worth it to get those last-second ebay bids in on time! :)

>
> Conrad module tuned to DCF77 suffers from bad reception here (1000km
> range) for a couple of periods most evenings and is comparable to
> internet time, but potentially at closer range DCF77 can be
> synchronised to the signals phase modulation which should be within a
> few us or better.
>
>
> David

-- 

John Devereux

David Lord wrote:
> I'm setting up to use a Conrad module that I've modified with 60kHz 
> crystal and currently gives offset mostly better than 200us. That
> cost about 12quid for two, when they were doing two for one. With 
> internet time the offset is mostly better than 500us but varies a
> lot more than MSF, maybe +/-5ms vs +/-1ms.

As someone who seems to know a lot about time synchronisation, do you 
know of a solution to the fact that my house seems to block most of the 
UK time signal from Anthorn? I have to carry my MSF clocks from the 
South side of my house to the North side every night in order for them 
to synchronise. Putting standard quartz clocks right once a month would 
be easier.

I've wondered if there is some sort of MSF signal booster that I could 
get, or maybe something like your Conrad module, but which transmits on 
the MSF frequencies? I realise that transmitting on the MSF frequency 
would be illegal, but perhaps a booster could listen for the real MSF 
signals to fall silent, then send out its own in the gaps.

I'm actually well placed for reception, with a clear view to the horizon 
in the Northward direction (barring the trees at the bottom of out 
garden) as I'm on one of the highest points on the South Downs. Perhaps 
Anthorn is on the far side of a mountain?

-- 
Steve Swift
http://www.swiftys.org.uk/swifty.html
http://www.ringers.org.uk

On Tue, 06 Oct 2009 09:30:27 +0100, Swifty wrote:

> I have to carry my MSF clocks from the South side of my house to the 
> North side every night in order for them  to synchronise. Putting 
> standard quartz clocks right once a month would be easier.

So flog your MSF jobbies an get some quartz ones or ones that use DCF
(Germany) instead.  B-)

> but perhaps a booster could listen for the real MSF signals to fall 
> silent, then send out its own in the gaps.

The MSF signal is a carrier that goes off precisely at 0mS each
second and has a variable off time to send the time/data information
at 2
bits/second.

http://www.npl.co.uk/science-technology/time-frequency/time/products-a
nd-services/msf-radio-time-signal

http://tinyurl.com/msf-radio-time-signal

> I'm actually well placed for reception, with a clear view to the horizon 
> in the Northward direction (barring the trees at the bottom of out 
> garden) as I'm on one of the highest points on the South Downs. Perhaps 
> Anthorn is on the far side of a mountain?

How about the Lake District and the Southern Pennines? The "Common
Clock  Problems" page on the NPL site may give you a few clues, it
maybe just the fact you want your clocks on the "wrong" side of the
house but it could be local interference.

-- 
Cheers
Dave.

Swifty wrote:
> David Lord wrote:
>> I'm setting up to use a Conrad module that I've modified with 60kHz 
>> crystal and currently gives offset mostly better than 200us. That
>> cost about 12quid for two, when they were doing two for one. With 
>> internet time the offset is mostly better than 500us but varies a
>> lot more than MSF, maybe +/-5ms vs +/-1ms.
> 
> As someone who seems to know a lot about time synchronisation, do you 
> know of a solution to the fact that my house seems to block most of the 
> UK time signal from Anthorn? I have to carry my MSF clocks from the 
> South side of my house to the North side every night in order for them 
> to synchronise. Putting standard quartz clocks right once a month would 
> be easier.
> 
> I've wondered if there is some sort of MSF signal booster that I could 
> get, or maybe something like your Conrad module, but which transmits on 
> the MSF frequencies? I realise that transmitting on the MSF frequency 
> would be illegal, but perhaps a booster could listen for the real MSF 
> signals to fall silent, then send out its own in the gaps.

I suspect a cable connection from an external aerial would be ok, 
possibly amplified with a pickup coil located near each clocks.
That's messy if you have a few clocks. Next easiest would be to
demodulate MSF at external location and use a short range wireless
module with receiver and regenerated low level MSF at each clock
or locate circuitry and inject in clock between demodulator and
decoder.

There are a few in this NG who can probably answer this better
than I can.

My north facing (towards Anthorn) bedroom seems to be a wavetrap
specially designed for MSF (steel joists, copper central heating
pipes and earth bonding I guess). I made a 23" per side pentagon
frame aerial that pulled in the signal but still had a couple
of periods when signal was lost.

> I'm actually well placed for reception, with a clear view to the horizon 
> in the Northward direction (barring the trees at the bottom of out 
> garden) as I'm on one of the highest points on the South Downs. Perhaps 
> Anthorn is on the far side of a mountain?

Doesn't seem to be from map, not close anyway and signal should
be able to clear those in Lakes. You may be better with DCF77 from
your location. I'm 20ml north of Manchester on Pennines and about
100ml from Anthorn vs 600ml from Frankfurt. I've only tried DCF77
using the small ferrite of Conrad module and there were only a
couple of nights where signal wasn't lost (switchover from
groundwave to skywave).


David

Roger Lynn wrote:
> On 05/10/09 08:38, Daniel James wrote:
>> http://aaisp.net.uk/kb-broadband-serverlist.html
> 
> "NTP  time.aaisp.net.uk  Use this to set your clock by NTP"
> 
> Does A&A have only one NTP server? It is usually recommended to use
> three NTP servers, and good ISPs usually provide more than one, which
> saves from have to connect to heavily loaded public servers, possibly
> with poor latency.
> 
> If an ISP hosts its own NTP servers they can provide their customers
> with better quality time with reduced latency, reduce the load on public
> servers and reduce (slightly) their external bandwidth requirements. I
> believe that running your own stratum 1 server off a GPS receiver or
> similar is quite inexpensive these days.

I always assumed that the NTP protocol contained some sort of correction for 
latency, otherwise Internet time synchronisation would be very flaky. Was I 
mistaken about this?

Alfred E Neuman wrote:
> Roger Lynn wrote:
>> On 05/10/09 08:38, Daniel James wrote:
>>> http://aaisp.net.uk/kb-broadband-serverlist.html
>>
>> "NTP  time.aaisp.net.uk  Use this to set your clock by NTP"
>>
>> Does A&A have only one NTP server? It is usually recommended to use
>> three NTP servers, and good ISPs usually provide more than one, which
>> saves from have to connect to heavily loaded public servers, possibly
>> with poor latency.
>>
>> If an ISP hosts its own NTP servers they can provide their customers
>> with better quality time with reduced latency, reduce the load on public
>> servers and reduce (slightly) their external bandwidth requirements. I
>> believe that running your own stratum 1 server off a GPS receiver or
>> similar is quite inexpensive these days.
> 
> I always assumed that the NTP protocol contained some sort of correction 
> for latency, otherwise Internet time synchronisation would be very 
> flaky. Was I mistaken about this?

It does but can be better on dialup than adsl due to assymetric
path using adsl. The correction can't allow for difference between
up and down latency so having low latency is more important on
adsl than the line speed.


David

David Lord wrote:
> Alfred E Neuman wrote:
>>
>> I always assumed that the NTP protocol contained some sort of 
>> correction for latency, otherwise Internet time synchronisation would 
>> be very flaky. Was I mistaken about this?
> 
> It does but can be better on dialup than adsl due to assymetric
> path using adsl. The correction can't allow for difference between
> up and down latency so having low latency is more important on
> adsl than the line speed.

Ah, right, I always wondered how they managed it. I assume they measure latency 
in and out and then assume the correct time is half way between the two.

Dave Liquorice wrote:

> get some quartz ones or ones that use DCF (Germany) instead

Had to return my Siemens clock; wouldn't synch anywhere in my house.

> How about the Lake District and the Southern Pennines? The "Common 
> Clock  Problems" page on the NPL site may give you a few clues, it 
> maybe just the fact you want your clocks on the "wrong" side of the 
> house but it could be local interference.

It has always seemed odd to me that the DCF transmitter is sited at a 
location shielded from the bulk of the population by the Lake District 
mountains, and that they took the opportunity to reduce the power from 
60Kw to 17Kw during the move.

I'd blame the O2 Cellmast about 70m from my house, but it's north of the
house, so on the side which works better.

I still think that my house was built out of surplus nuclear shielding;
almost no wireless communications work through the walls.

-- 
Steve Swift
http://www.swiftys.org.uk/swifty.html
http://www.ringers.org.uk

Alfred E Neuman wrote:
> Ah, right, I always wondered how they managed it. I assume they measure 
> latency in and out and then assume the correct time is half way between 
> the two.

I'm sure something like that is going on. If the time from sending the 
request to getting the answer back is less than a second, and you assume 
that the answer you get back was "stale" by half of the elapsed time, 
then you're likely to be less than ½ a second out, which for my Windows 
box is good enough.
It will certainly get my "last second" ebay bids in exactly on time.

-- 
Steve Swift
http://www.swiftys.org.uk/swifty.html
http://www.ringers.org.uk

On 06/10/09 11:36, Alfred E Neuman wrote:
> I always assumed that the NTP protocol contained some sort of correction for 
> latency, otherwise Internet time synchronisation would be very flaky. Was I 
> mistaken about this?

No, but another aspect is that public time servers can often be badly
overloaded. An ISP providing its own servers can reduce the load on the
public ones and provide a better service to its customers.

Roger