Re: [Discuss-gnuradio] VRT Branch: config_clock / PPS questions

[Josh Blum]

On 02/18/2010 12:27 PM, Tim Pearce wrote:
> Hi,
>
> I've finally had a bit of time to have a go with the VRT branch - I noticed
> with the rx_timed_samples example it's now possible to start rx_streaming at
> a specific time so I've quickly merged my custom usrp2 timestamp source
> block (gr-usrp2 - providing a second stream of timestamps - I dont think the
> vrt_header is available after this block at the moment).
> I think I also need to add the python calls to the start_streaming_at
> functionality.

Cool. Yea, the python functionality is not in there.

If you wanted to pass the vrt header to the next block, then you have to 
figure out how ot use this PMT stuff in gnuradio. Hmm....

> As I havent got a scope around at the moment to check, am I correct in
> thinking that the old config_mimo command to use the 10MHz Ref Input needs
> to be replaced with something like:
>
> u=usrp2.source_32fc()
>
> clock_settings=usrp2.clock_configt_t()
> clock_settings.ref_source=clock_settings.REF_SMA
> u.clock_config(clock_settings)

The config mimo was replaced with a clock config call that takes a 
struct of clock configuration arguments. Setting the PPS source, REF 
source, and PPS polarity.

Im not sure what it looks like in python, but the above seems correct.

> I've also noticed the set_time_at_next_pps() command, I'm assuming I can use
> this to force the usrp2 to set zero fractional seconds and some integer (or
> zero) number of seconds - this seems to be similar to what the
> rx_timed_samples example is doing.

set_time_at_next_pps will latch in a new seconds and ticks when the next 
pps pulse arrives. Because there is a race condition, I recommend using 
this only in conjunction with something that can asynchronously message 
your computer on a pps pulse (like a gps with a serial port).


> Is the PPS used after this to keep the integer seconds increment aligned
> with the PPS?

Nope. The pps call is a one time thing.

Once set, the integers seconds will continue to increment on their own 
as the ticks rollover to zero after 100e6 counts. Your usrp2 should be 
able to keep exact gps time without re-syncing provided that the 10mhz 
ref is attached.

You could re-call set_time_at_next_pps every second. It shouldnt be 
necessary unless you are worried about leap seconds.

-Josh