[time-nuts] Re: Austron Oscillator Products - GSM

[Magnus Danielson]

Hi,

To add to this, is GSM and some of the others one also uses Timing 
Advance (TA) to steer the phase of the handset so that the TDM frame 
arrives in the middle of the TDM slot, as experiences by the 
basestation. As you realize, this is an indirect range indication. In 
well functioning systems, the TA is set to reasonably close values as it 
handsover to another cell, and by collecting the TA corrections needed 
as it arrives, the recommendations set can be improved. Doing this 
improve the hand-over times and frees capacity quicker in the old cell, 
allowing more calls there. With 4G and 5G in TDD mode (for 5G it's 
needed to deliver all promises) the phase is managed on network where as 
2G, 3G, and 4G (FDD) avoids that, and only required frequencie to be 
within +/- 50 ppb. Toss in a raising awareness that GNSS may not be 
there all the time, there is a bit of a challenge, which is taken with 
various degrees of seriousness depending on which country you're in. 
Let's just say it keeps me busy on my daytime job.

Cheers,
Magnus

On 2024-03-02 12:27, john.haine--- via time-nuts wrote:
> TVB in an email was asking about the numerology of GSM.
>
> The channel spacing in GSM is 200kHz = 13MHz/65, and all subsequent
> standards have channel centre frequencies that are on the same 200kHz raster
> though the channels are now much wider.  200kHz can also be divided by 25 to
> 8kHz and locked to 2048kHz divided by 256 hence lock to the E1 rate.
>
> The channel symbol rate is 13MHz/48 = 270.8333.. kbit/s.
>
> From
> https://www.rfcafe.com/references/electrical/gsm-specs.htm#:~:text=A%20GSM%2
> 0Multiframe%20is%20the,Bit%20being%203.69231%20%CE%BCs%20long.:
>
> "A GSM Multiframe is the basic unit, and is 120 ms long. There are 26 Frames
> in each Multiframe, with each Frame being 4.61538 ms long (120 ms/26).
> Within each Frame are 8 Timeslots at 576.92 μs per Timeslot (577 μs in round
> numbers). Finally, there are 156.25 Bits per Timeslot, each Bit being
> 3.69231 μs long."
>
> So every handset needed a 13MHz oscillator that had to be within 1ppm before
> lock and was then pulled in to lock using feedback from an "f burst" in each
> downlink timeslot.  VCTCXOs were used initially so low cost very small
> components became available based on hundreds of millions of devices being
> made.  Eventually digitally locked oscillators took over I believe.  The
> specs for 4G and 5G are considerably tighter I think.
>
>
>
> -----Original Message-----
> From: John via time-nuts <time-nuts at lists.febo.com>
> Sent: Saturday, March 2, 2024 7:27 AM
> To: time-nuts at lists.febo.com
> Cc: John <reid.john at hotmail.com>
> Subject: [time-nuts] Re: Austron Oscillator Products - GSM
>
> Ericsson GSM products did lock their master oscillator to the 2.048 mbps
> stream.
>
> The master oscillator was 13 MHz, and I understand the RF was locked to
> that. As well as lots of timing bibs and bobs..
>
> John
>
>
> From: <john.haine at haine-online.net>
>
>> Subject: [time-nuts] Re: Austron Oscillator Products
>> To: "'Discussion of precise time and frequency measurement'"
>> 	<time-nuts at lists.febo.com>
>> Message-ID: <011201da6afe$e7bea780$b73bf680$@haine-online.net>
>> Content-Type: text/plain;       charset="utf-8"
>>
>> GSM standards development started in 1982 though there was prior work in
> several labs.  Basic phy layer parameters were probably fixed by 1985 as the
> development cycle was quite long drawn out (especially for handsets as the
> "long pole" went through the tester!).  Networks were launched in 1991 so
> devices must have been in procurement several years before then.  When were
> these devices made?  I think they might have locked BTS clocks to the
> incoming PCM 2.048 Mbit/s links, not sure about the RF, though given the way
> all the numbers are interrelated they might have had to.
> _______________________________________________
> time-nuts mailing list -- time-nuts at lists.febo.com To unsubscribe send an
> email to time-nuts-leave at lists.febo.com
> _______________________________________________
> time-nuts mailing list -- time-nuts at lists.febo.com
> To unsubscribe send an email to time-nuts-leave at lists.febo.com