[hpsdr] coherent dual mercury receivers

[Joe Martin K5SO]

Hi Phil,

I wrestled with this same situation earlier with HPSDR for the timing  
for pulsar data acquisition, as you may recall from some of our  
private communications a while back regarding my use of Atlas bus line  
A19 for "external sync".   Of course, A19 is now designated for use   
by Cyclops but the use of a different bus line could be implemented  
for a1 pps GPS sync signal very easily.

The timing situation can be made better than your 1 in 62 sample  
uncertainty that arises from use of the C&C data by using, say, one of  
the bits in the 16-bit MIC data from Ozy/Magister/Metis to the PC  
instead.  Doing this would improve the real time uncertainty by a  
factor of 62 for you for starters.  No time stamping is necessary as  
long as the comm path between HPSDR and the PC doesn't drop data  
packets.

As you know, the timing uncertainty is limited by the minimum time  
step between samples that are passed to the PC and that limitation is  
1/192,000 sec at the moment if the highest sampling rate is selected;  
a bit more than 5 uSec between samples.  If time resolution better  
than that is required it seems to me that we would need to run at a  
higher final sampling rate than 192,000 samples/sec, i.e., with less  
total decimation.   Metis should allow for that, I believe, right?

In any case, your Griffin project requirements are not quite the same  
as the issue that Georg brings up with regard to coherent operation of  
dual Mercury boards, I think.  Coherent operation is possible without  
time stamping in Georg's case, unless I've missed the mark completely,  
hihi!

73,  Joe K5SO


On Mar 13, 2011, at 7:16 PM, Phil Harman wrote:

> I do have a requirement to time stamp samples from Mercury.
>
> This is for the Griffin project:
>
> http://openhpsdr.org/wiki/index.php?title=GRIFFIN
>
> I'm developing a Chirp beacon that can be used as a beacon (with an  
> ERP
> much higher than its actual power output) and a radar for measuring  
> real
> time propagation on the HF bands and in particular 6m.
>
> The Chirp beacon linearly sweeps a carrier over 1kHz in one second  
> (as an
> example). The start of the sweep is triggered by the 1PPS signal  
> from a
> GPS receiver.
>
> On receive we can feed 1PPS from a GPS receiver to the Atlas bus and  
> use
> this to time stamp the sample that coincides with the start of the  
> Chirp.
>
> If we time stamp at the output of the ADC then there is a high  
> probability
> that the sample will be removed during decimation.
>
> If we time stamp after decimation and filtering then we can  
> currently only
> indicate which USB frame contains the sample that coincides with the  
> 1PPS.
> That assumes that we just set a bit in the C&C data. In which case we
> could be out by a 62 samples.
>
> Anyone see an elegant solution?
>
> 73 Phil...VK6APH
>


 1300109816.0