USRP problems (interpolation, decimation, sampling rate, frequency)

I use two USRPs. One sends a sinusoidal signal, and the other receives
it.

A flow graph and a result are here
flow graph:
http://lh5.ggpht.com/_byQV8nmc5FA/TOOTX9ReSYI/AAAAAAAAAQk/hWJ9Kf5GcAA/s800/스크린샷-1.png
result:
http://lh5.ggpht.com/_byQV8nmc5FA/TOOTYGnBCUI/AAAAAAAAAQo/sQmJlvlJJVU/s800/스크린샷-2.png

It seems fine, but notice that the receive signal goes 10 times faster
than
the original one. (see time-axes in both plots)

And lots of uO (usrp overrun). Don’t think about uO because I run it on
Virtual Machine, then what the hell is going on this flow graph?!

And I really wanto know the ratio of an interpolation and a decimation
I thought that interpolation:decimation = 2:1, because USRP sink
processes
128Msamples/interpolation and USRP source processes
64Msamples/decimation
In order to match up both in same value, it has to be 2:1

In short, I want to know about 2 things.

1. RX goes 10 times faster than TX
2. ratio between interpolation and decimation

On Wed, Nov 17, 2010 at 3:41 AM, Songsong G.
[email protected]wrote:

1. ratio between interpolation and decimation

Your scope sinks aren’t using the same sample rate. One is using 500,
one is
using 5k.
2:1 is correct for the ratio I believe.
Not sure what is causing your overruns.
Keep in mind that your true sample rate is being determined by your
USRPs,

-Steven

Thank you, it works.
(I have done what you told me to, but at that time it didn’t work.
MYSTERIOUS…)

Anyway I’ve got a new problem

RX seems similar with TX, but it… JITTERS in an amplitude or a
frequency

In other words, RX keeps a shape of TX, loosely.
Observing for a fine duration, however, an amplitude an a frequency is
moving around the value of them from TX.
Therefore, it seems to be stretched or shrunk in vertical and
horizontal.

Yes, it doesn’t matter for a human to know that RX keeps a shape of TX
but I’m worried that USRP, GRC, PC may not recognize that.

2010/11/17 Steven C. [email protected]