[time-nuts] Rohde & Schwarz GPSDO

Martin Burnicki martin.burnicki at burnicki.net
Sat Oct 10 12:58:55 EDT 2015

Magnus Danielson wrote:
> Hi,
> On 10/09/2015 09:35 PM, Rami Vainio wrote:
>> On 7.10.2015 14:45, Arthur Dent wrote:
>>> I believe that like a lot of the Meinberg receivers that
>>> this uses a down converter to give an IF frequency of
>>> 35.4 MHz. If you don't have the converter that apparently
>>> isn't included with the receiver you have a $300 paperweight.
>>> You might want to check with the seller before bidding.
>> Hi, here is quick and dirty way to make your own LNC to Meinberg 166 (I
>> got it with no antenna and broken 10MHz oscillator):
>> https://www.dropbox.com/sh/n4lob51pzs236wb/AABy7DVGxX4EIaazizGaAx5Na?dl=0
> A friend has one of those boxes, as I gave it to him.
> One of the Meinberg brothers did the actual DSP code in it.
> I know of very few vendors (Meinberg and Trimble) of GPSDOs that does
> both the GPS and DO parts, and if you take on the effort to do the GPS
> side, then you get a better system understanding. Most GPSDOs builds on
> OEM modules.
> As I recall, this is a one-channel receiver with analogue Early, Prompt
> and Late correlator integrators, where the Altera produces the chips and
> the DSP does all the nav-stuff.
> Martin might remember more (or just ask).

Yes, the GPS166 was the first GPS receiver developed by Meinberg at the
beginning of the 1990s. It was just a single channel receiver where GPS
reception (correlators etc.) was done by a DSP, and decoding and
evaluation was done by a Siemens 80C166 microcontroller which was a
brandnew, sophisticated 16 bit CPU with a lot of useful peripherals
(counter/timer/compare/capture registers, GPIO) built-in.

Since the design was targeting stationary installations for time
synchronization, the GPS166 was already designed with a down converter
built into the antenna, so you could use very long standard coaxial
cable like RG58 between the antenna and the receiver. For stationary
installations it also didn't matter that the device could only track a
single satellite.

The satellite signals were used to discipline an on-board oscillator
(TCXO or OCXO), and that oscillator was used to drive a counter chain
using one of the 80C166's timers to implement the on-board time.

All output signals were derived from that counter chain, and the
counter/compare units provided by the CPU were used to generate an exact
1 PPS output signal, etc., directly coupled to the on-board time.

The resulting accuracy was better than 1 microsecond, and this was much
better than the accuracy you could achieve with usual radio clocks at
that time.

The subsequent GPS receiver models by Meinberg (GPS167 etc.) used some
special correlator chip which then became available, but the correlators
were fully controlled by our own firmware. Those devices could track up
to 6 satellites in parallel. This was limited by the CPU power of the
80C167 processor used for those devices which still had to do the
decoding of satellite data.

Current GPS receiver models run an ARM processor with correlators
implemented in an FPGA, so thes devices are much more powerful.

Unfortunately the design of the original GPS166 converter/antenna was
not applicable to multichannel receivers like the GPS167 and later
series, so the converter had to be modified.

This is the reason why you can't use a GPS166 antenna with GPS167 and
newer receivers and vice versa, however the GPS167 antenna is still
compatible with the antennae shipped with newer GPS receiver models.


More information about the time-nuts mailing list