Re: [Openbts-discuss] Best USRP version for OpenBTS

[Matt E. <ma...@et...>]

On Tue, Oct 18, 2011 at 10:49 AM, Alexander Chemeris <
ale...@gm...> wrote:

> Hi Matt,
>
> On Tue, Oct 18, 2011 at 21:32, Matt Ettus <ma...@et...> wrote:
> >
> >
> > On Tue, Oct 18, 2011 at 10:23 AM, Alexander Chemeris
> > <ale...@gm...> wrote:
> >>
> >> B100 may be a nice solution, but it has only single daughter board
> >> slot and while B100 clock is configurable, its tunning step is too
> >> rough to meet GSM requirements. So it will work with some phones and
> >> will have all kind of issues with others.
> >
> > Alexander,
> > That is absolutely false.  We have never found a single phone which fails
> to
> > work with any USRP as long as it has a TCXO with sufficient accuracy.
>  There
> > is simply no need for 1 Hz reference tuning steps.  The tuning is done
> > digitally, and can be done with arbitrary precision.
> > Real world radios don't step their references in 1 Hz steps open loop.
>  It
> > just isn't how radios work.  If you need finer frequency precision then
> you
> > lock to a better reference.
>
> Thank you for the correction. Sure, you can use external 10MHz
> reference with E100 to get required frequency precision.
>
> Regarding digital tuning - as it was discussed a couple of times, it
> introduces clock skew which causes problems with some phones. I guess
> you can compensate this skew in FPGA before downsampling, but AFAIK
> current UHD image doesn't support this.
>
>
NO.  You don't understand, Alexander, and you keep making these incorrect
authoritative statements disparaging our products.

There is no clock skew problem.  ALL radios, even if they have the world's
most accurate clock reference need to  perform timing recovery.  The timing
recovery block takes care of what you are calling clock skew.  All timing
recovery can handle sample rate offset.

Think of it this way -- you have the worlds most accurate oscillator.  Then
you move the phone closer or further from the base station.  The demod still
needs to adjust for the "skew" as you call it.  You can easily move fast
enough to have significant doppler shift, and all radios need to work with
moving users.

If the main oscillator is within frequency spec then there is no need for 1
Hz tunability.

Matt