USRP users mailing list -- Universal Software Radio Peripheral by Ettus Research ()

headers Birhane Alemayoh | 9 Jul 2012 14:40 · /examples/

_______________________________________________
USRP-users mailing list
USRP-users@...
http://lists.ettus.com/mailman/listinfo/usrp-users_lists.ettus.com

On 07/09/2012 05:40 AM, Birhane Alemayoh wrote:
> dear All,
> I want implement the frequency hopping in GSM on real time bases using N200
> and DBSRX2 doughterboard. That means I want write an application that
> follows the hopping and capture the signal accordingly from the host PC.
> However, I need to know the tuning latency of the DBSRX2 doughterboard if
> this is possible. The dwell time of  GSM signal is 4.615 msec.
> 
> I could n't get any document that tells the tuning latency of the DBSRX2
> doughterboard.
> So could you please tell me the tuning latency of DBSRX2?
> Can I achieve my requirement using  DBRSRX2 doughterboard  plus considering
> the OS scheduling, IP-stack, Ethernet transit, USRP command processing
> issues too.
> Thank you
> Birhane
> 
> 

Basically you cant retune the LO fast enough. The DBSRX2 is I2C based,
and the I2C control transactions are just too slow. On the other hand,
the SPI based daughterboards like SBX and WBX have 300us tune and lock
time and can be programmed just as fast.

However, you can leave the LO fixed and just shift digitally in the
cordic. This assumes that your frequencies of interest are within the
daughterboard's LF bandwidth. This will enable very fast frequency
hopping for you. Probably about 100us latency due to ethernet.

And even faster, if your bandwidth of interest fits into the maximum
host sample rate. You dont have to tune anything, you can just select a
desired chunk of the available spectrum.

-josh

> 
> _______________________________________________
> USRP-users mailing list
> USRP-users@...
> http://lists.ettus.com/mailman/listinfo/usrp-users_lists.ettus.com
> 

Hello Josh,

I am using WBX over USRPE110,
I want to test the Tuning latency using usrp->set_center_freq(freq).
For that, I surrounded the command with tic-toc. I got the latency around 5000 
microsecond.
- Is it a good way to test the over all latency?
- I think its huge latency, as I read in previous threads, you mentioned its 
around 300 microsecond?

Kind regards,
Sam

On 10/04/2012 05:16 AM, sam wrote:
> 
> 
> Hello Josh,
> 
> I am using WBX over USRPE110,

FYI... This subject is DBSRX2, which in this case, DBSRX2 I2C based
control, and I2C needs a lot of 2-way control and ACKing. So my advice
is for WBX/SBX...

> I want to test the Tuning latency using usrp->set_center_freq(freq).
> For that, I surrounded the command with tic-toc. I got the latency around 5000 
> microsecond.

I highly recommend you try the master branch of UHD if you havent. The
E100 now has the same fifo control interface as the N210s.

What this means is not only timed commands, but things like tuning the
WBX should queue up a bunch of commands ASAP w/o needing to ack.

> - Is it a good way to test the over all latency?

There is a example app in host/examples/ to measure turn around time. We
should also have a nice wiki page on this soon on account of the GR Conf
2012 presentation.

> - I think its huge latency, as I read in previous threads, you mentioned its 
> around 300 microsecond?

The worst case lock time is 300us, after the commands arrive. Thats not
latency.

Cheers,
-josh

Josh Blum <josh <at> ...> writes:

> 
> 
> On 10/04/2012 05:16 AM, sam wrote:
> > 
> > 
> > Hello Josh,
> > 
> > I am using WBX over USRPE110,
> 
> FYI... This subject is DBSRX2, which in this case, DBSRX2 I2C based
> control, and I2C needs a lot of 2-way control and ACKing. So my advice
> is for WBX/SBX...
> 
> > I want to test the Tuning latency using usrp->set_center_freq(freq).
> > For that, I surrounded the command with tic-toc. I got the latency around 
5000 
> > microsecond.
> 
> I highly recommend you try the master branch of UHD if you havent. The
> E100 now has the same fifo control interface as the N210s.
> 
> What this means is not only timed commands, but things like tuning the
> WBX should queue up a bunch of commands ASAP w/o needing to ack.
> 
> > - Is it a good way to test the over all latency?
> 
> There is a example app in host/examples/ to measure turn around time. We
> should also have a nice wiki page on this soon on account of the GR Conf
> 2012 presentation.
> 
> > - I think its huge latency, as I read in previous threads, you mentioned its 
> > around 300 microsecond?
> 
> The worst case lock time is 300us, after the commands arrive. Thats not
> latency.
> 
> Cheers,
> -josh
> 

Thanks Josh for your reply, I got it now, that timing the command is timing 
bunch of commands all together, and the time I got is the delay from all these 
commands.

I am trying retune WBX Rx frequency and continue streaming with max delay of 
1500 us, and 45 MHz separation (mainly listening to downlink and uplink GSM 
channels). I read that WBX has single Rx front-end chain, so the only solution 
would be to keep hopping between these channels (with 45 Mhz separation).

My question is, does the streaming continue directly after the lock time (300 
us) ? or there are other dependencies to start streaming ?

> 
> Thanks Josh for your reply, I got it now, that timing the command is timing 
> bunch of commands all together, and the time I got is the delay from all these 
> commands.
> 

Basically, you can time when the tuning happens, but you have to give
yourself some window of padding. This link is for the SBX/WBX phase
sync, but should stand has a code example for using a timed command to
tune:
http://files.ettus.com/uhd_docs/manual/html/sync.html#align-los-in-the-front-end-sbx-wbx-n-series

> I am trying retune WBX Rx frequency and continue streaming with max delay of 
> 1500 us, and 45 MHz separation (mainly listening to downlink and uplink GSM 
> channels). I read that WBX has single Rx front-end chain, so the only solution 
> would be to keep hopping between these channels (with 45 Mhz separation).
> 
> My question is, does the streaming continue directly after the lock time (300 
> us) ? or there are other dependencies to start streaming ?
> 

If you are using continuous streaming, then you will see an
uninterrupted flow of samples. Basically, sampling doesn't care about
what is occurring on the frontend.

So what you need to do issue the tune command at a specified time, say
x. Then you should ignore samples (on the host) that are overlapping
between x and x + delta.

-josh

> 
> 
> _______________________________________________
> USRP-users mailing list
> USRP-users@...
> http://lists.ettus.com/mailman/listinfo/usrp-users_lists.ettus.com
> 

Hello Josh,

Thanks for your reply. As a suggestion from the "Deepsec" conference guys, 
is to listen separately to uplink and downlink channels (using two USRPs N with 
gigabit ethernet connection), 
and not to keep hopping. As later I need to statistically 
analyze all uplink/downlink signalling, without even missing x time of samples.

Before taking the decision for this approach, I would like to ask your 
suggestions of the followings:

- Is there SINGLE USRP/Daughterboards combination that allow us to listen 
simultaneously to uplink/downlink GSM (900 MHz)/UMTS (1900 MHz) (with 45MHz 
separation).
- On Ettus website, there are a new series of Daughterboards coming soon, do you 
think any can support previous requirements (I can wait a little for them?).
- I am searching for portable Batteries for my USRPs (outdoor testing), any 
suggestions ?

Best regards,
Sam

Tuning latency of DBSRX2 for frequency hopping

Re: Tuning latency of DBSRX2 for frequency hopping

Re: Tuning latency of DBSRX2 for frequency hopping

Re: Tuning latency of DBSRX2 for frequency hopping

Re: Tuning latency of DBSRX2 for frequency hopping

Re: Tuning latency of DBSRX2 for frequency hopping

Re: Tuning latency of DBSRX2 for frequency hopping