Re: Synchronization: PPS and Tx signal in a USRP N200

[Glen Langston]

Thanks for your very well described study
of SDRs processing GPS signals and resulting timing accuracy.

Best regards

Glen

> On Dec 11, 2021, at 3:28 AM, jmfriedt <jean-michel.friedt@femto-st.fr> wrote:
> 
> If it can be of any help, we discussed generating a GPS-aligned 1-PPS
> in http://jmfriedt.free.fr/ifcs2021.pdf and the oral presentation at
> http://jmfriedt.free.fr/ifcs2021_presentation_jmfriedt.mp4 prepared
> for the International Frequency Control Symposium (IFCS).
> 
> What we realized when working on this topic is that the only place
> where time is known in an SDR processing sequence is at the ADC sample
> acquisition step, since all subsequent processing and data transfer
> afterwards are asynchronous, including the multitasking non-real time
> operating system GNU Radio is running on. Hence, the control on the
> 1-PPS alignement can only be achieved at the FPGA level clocking the
> ADC: the host clock is irrelevant since we do not know how much time
> was needed to transfer and process data. Luckily, GNU Radio provides a
> timestamp on each sample, assuming no data was lost since streaming
> started, so the control signal can be fed back to the SDR clock if the
> delay between reference and generated 1-PPS is measured using the same
> time reference of the ADC clock.
> 
> Best, JM
> 
>>> On 2021-12-10 17:25, isaac mario tupac davila wrote:  
>>>> Hello everyone!
>>>> 
>>>> My name is Isaac. I have a curious situation here... I've
>>>> generated a pulsed signal per second. I've saved one second
>>>> period in a .h5 file, so that I'm sure I'm having a fixed signal
>>>> per second, and then repeat it in my GRC flowgraph. .
>>>> 
>>>> To use this signal as Tx I'm trying to synchronize a USRP N200 and
>>>> a trimble in GRC. After run my flowgraph, this is what I see:
>>>> 
>>>> imagen.png
>>>> 
>>>> imagen.png
>>>> 
>>>> The yellow signals are the PPS of my trimble and my purple signals
>>>> are my tx signal per second. They are periodic in time but when
>>>> you see it deeper, the difference between the timble PPS and my Tx
>>>> signal is approx 97 ms. I think this difference should be close to
>>>> cero, as my USRP and trimble are synchronized. ¿What is happening
>>>> here? ¿Is this a normal behavior? I'm using unknown PPS to
>>>> configure my UHD:USRP sink in GRC....
>>>> 
>>>> I'll appreciate any help to clarify this behavior
>>>> 
>>>> Regards
>>>> Isaac T.
>>>> 
>>> Your flow-graph made very little sense to me.  If you want to 
>>> synchronize your TX, you have to take explicit measures to ASK the
>>> USRP to schedule your transmits at specific times.
>>>   The 1PPS signal only synchronizes an internal time-stamp clock in
>>> the unit.  It has NO WAY of knowing what the *meaning* of your
>>> samples are, so it can't possibly
>>>   synchronize some arbitrary event in your continuous sample stream
>>> to 1PPS without you explicitly asking for when your bursts need to
>>> be sent.
>>> 
>>> https://kb.ettus.com/Synchronizing_USRP_Events_Using_Timed_Commands_in_UHD
>>> 
>> 
>> 
>> 
> 
> 
> 
> -- 
> JM Friedt, FEMTO-ST Time & Frequency, 26 rue de l'Epitaphe, 25000
> Besancon, France