Forums

I and Q SNR in a IQ mixer

Started by Paul Solomon October 31, 2005
Hi,

This should be a quick one (hopefully), I have been doing some exploritory 
surgery with signal tap and matlab on the FM demod I am working on and I 
have noticed that I and Q channels have different SNR, one might have 30dB 
the other 50dB. Is this normal, or would you expect the SNR of I and Q to 
be the same??

Note: The I and Q that I am measuring is generated simple by mixing the output 
od an ADC with the sin and cos output of the same NCO. Then these are filtered 
and decimated by 16 and I am measuring after that point. the filter is a 
simple 1/16 band FIR filter to allow for the decimation. 


"Paul Solomon" <psolomon@tpg.com.au> wrote in message 
news:34f9711612698c7acf27766fed0@news.tpg.com.au...
> > Hi, > > This should be a quick one (hopefully), I have been doing some exploritory > surgery with signal tap and matlab on the FM demod I am working on and I > have noticed that I and Q channels have different SNR, one might have 30dB > the other 50dB. Is this normal, or would you expect the SNR of I and Q to > be the same?? > > Note: The I and Q that I am measuring is generated simple by mixing the > output od an ADC with the sin and cos output of the same NCO. Then these > are filtered and decimated by 16 and I am measuring after that point. the > filter is a simple 1/16 band FIR filter to allow for the decimation.
Hello Paul, This SNR difference could be pointing to your root cause. I would expect the SNRs to be similar. Clay
"Paul Solomon" <psolomon@tpg.com.au> wrote in message 
news:34f9711612698c7acf27766fed0@news.tpg.com.au...
> > Hi, > > This should be a quick one (hopefully), I have been doing some exploritory > surgery with signal tap and matlab on the FM demod I am working on and I > have noticed that I and Q channels have different SNR, one might have 30dB > the other 50dB. Is this normal, or would you expect the SNR of I and Q to > be the same?? > > Note: The I and Q that I am measuring is generated simple by mixing the > output od an ADC with the sin and cos output of the same NCO. Then these > are filtered and decimated by 16 and I am measuring after that point. the > filter is a simple 1/16 band FIR filter to allow for the decimation.
Hello Paul, This SNR difference could be pointing to your root cause. I would expect the SNRs to be similar. Clay
On 1 Nov 2005 11:56:45 +1000, Paul Solomon <psolomon@tpg.com.au>
wrote:

>Hi, > >This should be a quick one (hopefully), I have been doing some exploritory >surgery with signal tap and matlab on the FM demod I am working on and I >have noticed that I and Q channels have different SNR, one might have 30dB >the other 50dB. Is this normal, or would you expect the SNR of I and Q to >be the same?? > >Note: The I and Q that I am measuring is generated simple by mixing the output >od an ADC with the sin and cos output of the same NCO. Then these are filtered >and decimated by 16 and I am measuring after that point. the filter is a >simple 1/16 band FIR filter to allow for the decimation.
If the SNRs aren't identical in I and Q then the noise is somehow correlated to one of the orthogonal channels. That does indicate a problem somewhere. As Clay mentioned, it shouldn't be too hard to track down since it is channel specific. Eric Jacobsen Minister of Algorithms, Intel Corp. My opinions may not be Intel's opinions. http://www.ericjacobsen.org
Could it not also be a simple gain difference between the channels?

--Randy

On 2 Nov 2005 03:37:38 -0800, "Randy Yates" <yates@ieee.org> wrote:

>Could it not also be a simple gain difference between the channels? > >--Randy
Yeah, I suppose if the transmitter has different gains in I and Q that would also explain it, but he's seeing a 20dB difference. That'd be tough to miss in a transmitter. Eric Jacobsen Minister of Algorithms, Intel Corp. My opinions may not be Intel's opinions. http://www.ericjacobsen.org
Eric Jacobsen <eric.jacobsen@ieee.org> writes:
> > On 2 Nov 2005 03:37:38 -0800, "Randy Yates" <yates@ieee.org> wrote: > >>Could it not also be a simple gain difference between the channels? >> >>--Randy > > Yeah, I suppose if the transmitter has different gains in I and Q that > would also explain it, but he's seeing a 20dB difference. That'd be > tough to miss in a transmitter.
Hi Eric, I was thinking on the receive side. Maybe he's got something strange like overlapping data sections or something weird in the DSP. Could be anywhere - you know how fickle the simplest circuit can sometimes be. -- % Randy Yates % "I met someone who looks alot like you, %% Fuquay-Varina, NC % she does the things you do, %%% 919-577-9882 % but she is an IBM." %%%% <yates@ieee.org> % 'Yours Truly, 2095', *Time*, ELO http://home.earthlink.net/~yatescr
Question for you guys (that are smarter than I am)....

If the (random AWGN)  noise were added near  the front end of the radio
(as is the normal case) and then the signal were hard limited  (as is
the normal case for an FM system)... would the hard limiter impact the
SNR of the I and Q differently?  I'm thinking the limiter removes much
of the AM noise but does not reduce the FM (phase) noise.  But I'm not
clear if this translates into I and Q that way or even if the OP is
hard limiting his signal before demodulation.


Mark

On 2 Nov 2005 19:30:22 -0800, "Mark" <makolber@yahoo.com> wrote:

>Question for you guys (that are smarter than I am)....
Don't think that for a second... ;)
>If the (random AWGN) noise were added near the front end of the radio >(as is the normal case) and then the signal were hard limited (as is >the normal case for an FM system)... would the hard limiter impact the >SNR of the I and Q differently? I'm thinking the limiter removes much >of the AM noise but does not reduce the FM (phase) noise. But I'm not >clear if this translates into I and Q that way or even if the OP is >hard limiting his signal before demodulation.
No, it should affect I and Q the same. Remember, the definitions of which is I and which is Q are more or less arbitrary. It is often up to the receiver to sort out the phase ambiguity by using something embedded in the signal, otherwise there's no way to tell which is I and which is Q. The amps and noise sources can't tell either, until they're actually separated by the mixer. Eric Jacobsen Minister of Algorithms, Intel Corp. My opinions may not be Intel's opinions. http://www.ericjacobsen.org
On Thu, 03 Nov 2005 03:08:39 GMT, Randy Yates <yates@ieee.org> wrote:

>Eric Jacobsen <eric.jacobsen@ieee.org> writes: >> >> On 2 Nov 2005 03:37:38 -0800, "Randy Yates" <yates@ieee.org> wrote: >> >>>Could it not also be a simple gain difference between the channels? >>> >>>--Randy >> >> Yeah, I suppose if the transmitter has different gains in I and Q that >> would also explain it, but he's seeing a 20dB difference. That'd be >> tough to miss in a transmitter. > >Hi Eric, > >I was thinking on the receive side. Maybe he's got something strange like >overlapping data sections or something weird in the DSP. Could be anywhere - you >know how fickle the simplest circuit can sometimes be.
I mentioned the transmitter first since that's the most obvious place where a channel imbalance would result in different SNRs, but in this case the difference is huge. A transmitter with that much channel mismatch could easily be detected with some of the simple and common Tx tests, e.g., single side tone on each side, etc. Since the main noise source in a practical system is the LNA and since many sim models add the noise at IF as well, one would expect that the most likely way to get differing SNRs due to channel imbalance would be at the Tx. It would help to get clarification of where the noise is being added in this particular sim. But I see your point, too, that if the same sections process I and Q via multiplexing weird stuff like this can happen. Eric Jacobsen Minister of Algorithms, Intel Corp. My opinions may not be Intel's opinions. http://www.ericjacobsen.org