Re: [Discuss-gnuradio] Estimating SNR

[Marcus Müller]

Yes, there are limits. When you do an uhd_usrp_probe the RX and TX gain ranges will be displayed.

Notice that the same gain setting doesn't mean the same power output at all frequencies – all gains are relative to 0dB gain, in the respective linear gain regions, which again, for all devices and all frequencies, are different. The USRPs are not calibrated measurement equipment!

However, we've got measurement runs for some daughterboards under [1]; look at the TX power over gain pages with fixed frequency.

Best regards,
Marcus
[1]http://files.ettus.com/performance_data/

On 18.05.2016 09:07, Raja Muneeba wrote:

Hi,

I am working on finding power on receiver side of USRP. And changing gain value at the transmitter helps. I want to know what is the maximum gain value that I can select for USRP sink if centre freq is 2.4Ghz(if that matters). Are there any limits for that?

Thanks!

From: Marcus Müller <address@hidden>
Date: Thursday 12 May 2016 14:11
To: Raja Muneeba <address@hidden>, GNURadio Discussion List <address@hidden>
Subject: Re: Estimating SNR

Hi Muneeba,

please stay on-list with your replies!
So,

On 12.05.2016 12:48, Raja Muneeba wrote:

Thankyou for your answers. Im trying to solve this. I am now using fft Sink at receiver to see the power(db) vs frequency plot. I turn off the transmitter and see the power at fft, and consider it noise. Then I turn on the transmitter and get the value at fft. I conside that signal strength+noise. So (signal strength+noise)-noise=signal strength. And SNR could be signal strength/noise. Am I going in the right direction?

I think so! You should start to model your noise. Is your noise model "white"? If it is, what does it say about the power of noise contained in different, equally wide bandwidths?

What is the power(db) on fft actually? Is it dbm?

No,

If its db, then in comparison to what?

Full scale! This is DSP, there's no physical units to the numbers you handle. So, if you feed in a constant 1 into your fft plot, you'll see a specific value.

One more basic thing as I am extremely new to signal processing. What frequency of signal source ,sampling rate and gain should I choose for transmitting over centre frequency of 2.48GHZ.

That doesn't have to do anything with each other:
Sampling rate is just a number that the signal source uses together with the signal frequency to calculate how many *samples* long a full period will be – there's no "real world" time in this kind of DSP, just samples. So a signal source with f_sample=1 and f_signal = 0.2 will have exactly the same output as a signal source with f_sample = 5e6 and f_signal = 1e6.

Now, you don't sample at 2.48GHz  directly – your USRP is an upconverter that allows you to get complex baseband equivalent to what happens in a real passband at RF frequencies.
The idea of complex baseband is very important here – GNU Radio has a suggested reading wiki page[1], but seeing that you're probably a student at the Aalto university, I'd just ask for the scriptum or recommended book of the basic digital communications lecture[2] in electrical engineering, and read up till you reach the baseband receiver. Such a lecture was where I learned those principles!

So, now the sampling rate you chose for your USRP sink or source does have a real-world effect: it actually sets the rate at which samples are converted to/from analog values. And hence, the sampling rate you use with the hardware directly sets the Nyquist rate, and hence, your signals' bandwidth.

What would be the reason for choosing that. Right now I have 1M sampling rate and 1k frequency of signal source. My USRP is N210 with SBX daughter board.

That are in any case values that the SBX and the N210 support.

Best regards,
Marcus

Can you suggest some documentation which can help me deciding these factors?

[1] http://gnuradio.org/redmine/projects/gnuradio/wiki/SuggestedReading
[2] https://mycourses.aalto.fi/course/view.php?id=5109

From: Marcus Müller <address@hidden>
Date: Wednesday 11 May 2016 13:13
To: Raja Muneeba <address@hidden>, "address@hidden" <address@hidden>
Subject: Estimating SNR (was: Re: query)

SNR is signal strength/noise

Yeah, obviously, but what does that mean? how can you *describe* signal strength with a mathematical formula based on the digital samples you have in GNU Radio?

The point is that you'll just need an estimator for the signal energy, and an estimator for the noise power.

There's a lot of estimators for both; for a single tone, Welch's method might be pretty usable. Another approach would be a narrow filter, and a signal-to-magnitude-squared converter.
For noise power, you could just calculate the overall receive signal power, and subtract the signal power estimate.

Best regards,
Marcus

On 11.05.2016 11:56, Raja Muneeba wrote:

SNR is signal strength/noise

From: Marcus Müller <address@hidden>
Date: Wednesday 11 May 2016 12:48
To: Raja Muneeba <address@hidden>, "address@hidden" <address@hidden>
Subject: Re: query

Hi Muneeba,

so the point is: What *is* SNR? Can you define it mathematically?

Best regards,
Marcus

On 11.05.2016 11:47, Raja Muneeba wrote:

HI,

I am using usrp N210 with db SBX, and I do not have any message wave (or source file). I simply transmit and receive a sin wave via USRPs (find my graphs attached). I am not doing any modulation/de modulations.

I need to know the SNR value at the receiver end. I also want to adjust the signal strength to change SNR. SNR is signal strength/noise in db. How can I do this?

I see solutions with modulation schemes, but since I don’t use any modulation, how can it be done without that?

Regards,

Muneeba R.