High noise level reading GSM spectrum


#1

Hi all,

We were reading GSM channels with Tvoid software (
http://wiki.thc.org/gsm/tvoid) and at the beginning it worked fine.
Suddenly, we do not how, it seems that noise level has increased in 30
dB
approx. and only a few GSM channels can be now identified (but they are
all
only 10/15dB above noise power, when at the beginning there was at least
40dB) but this channels can be read so we think the front end is not
broken,
it seems the RF front end is not working properly or its configuration
is
wrong, but we do not know how we have made it. We think that the LP
filter
in our dbsrx daughterboard may be working poorly. I attach you a
screenshot
of the best received GSM channel (it is only 15dB above noise power).

I expect your help
Thank you for your time.

Jose


#2

José Carlos R. wrote:

that the LP filter in our dbsrx daughterboard may be working poorly. I
attach you a screenshot of the best received GSM channel (it is only
15dB above noise power).

There is no low pass filter. Assuming that you are setting the gain
properly, then most likely problem is that you blew out the LNA in the
front, U4. Static or strong signals can do that. You could try
replacing U4, which is an MGA82563.

Matt


#3

Matt E. wrote:

There is no low pass filter. Assuming that you are setting the gain
properly, then most likely problem is that you blew out the LNA in the
front, U4. Static or strong signals can do that. You could try
replacing U4, which is an MGA82563.

Matt
Matt, I thought that there was an LPF on the I and Q channels on the
MAX2118, which is what set_bw() is all about.
The MAX2118 has programmable lowpass filters on the I and Q outputs,
does it not?


Marcus L.
Principal Investigator, Shirleys Bay Radio Astronomy Consortium
http://www.sbrac.org


#4

Marcus D. Leech wrote:

does it not?

Yes, it does have programmable lowpass filters, but he seems to have a
problem on the front end. If a baseband filter had a problem he would
get terrible opposite sideband rejection, which doesn’t appear to be the
case from the picture.

Matt