Talk:Signal-to-noise ratio

From Wikipedia, the free encyclopedia

This article is within the scope of WikiProject Telecommunications, an attempt to build a comprehensive and detailed guide to telecommunications on Wikipedia. If you would like to participate, you can edit the article attached to this page, or visit the project page, where you can join the project as a "full time member" and/or contribute to the discussion.

	Portal This article is within the scope of WikiProject Physics, which collaborates on articles related to physics.
???	This article has not yet received a rating on the assessment scale. [FAQ]
???	This article has not yet received an importance rating within physics.
Help with this template Please rate this article, and then leave comments to explain the ratings and/or to identify its strengths and weaknesses.

Information theory has been around since 1948; telecommunications much longer, and usenet only since the '80s. So "signal-to-noise ratio" cannot have originated in usenet. 131.183.81.100 01:58 Mar 4, 2003 (UTC)

1 What does this mean?
2 What's better?
3 Associated Page may crash your PC
4 Digital SNR derivation
- 4.1 A million different equivalent equations
5 SNR = 8 sigma ... what does the sigma mean?
6 SNR
7 Different meanings of SNR
8 SNR in image processing
9 "unreferenced driveby edit"
10 NPR

[edit] What does this mean?

In the article, this was placed without context:

In Audio Coding there are terms like
- NMT
- TMN
- SMR

These links don't really go anywhere or appear to be much use, so I removed them. --Twinxor 04:07, 12 May 2005 (UTC)

[edit] What's better?

What's better? A high or a low SNR? Maybe someoine can give the formula for calculating SNR? Thanks, --Abdull 10:01, 21 May 2005 (UTC)

Generally, a high SNR is desirable since the signal is defined to be something of interest. I'm hard pressed to think of an example when one would desire a very low SNR (no signal, only noise).

- This is just the case I am working on. In smoothing algorithms "signal" is a second derivative jump due to data, it could be transferred to "noise" component. The trade-off between "noise" and "signal" in that case requies understanding of "what is level of non-recongnizable signal" (193.167.195.60 11:06, 14 March 2007 (UTC)).

(The best I can do is if you interpret the channel as an encrypted one where the signal is the unencrypted data....then you would want the "viewer" to see as much "noise" as possible so they can't "see" the unencrypted data.)  But, generally, the higher the SNR the better.

As for the formula, it can vary slightly. Generally:

$SNR = \frac{P_{signal}}{P_{noise}}$

which is the ratio of powers of the signal and the noise. Cburnett 20:42, May 24, 2005 (UTC)

[edit] Associated Page may crash your PC

Modern PC, plenty of memory using Internet Explorer 6.0. PC crashed about where the second formula is.

Can edit the page OK - but don't know enough to work out why I got a blue screen of death. This occurred on a colleagues PC, so is repeatable (older PC but same vsn of IE).

Help?

[edit] Digital SNR derivation

As far as I know, 6*n dB is an approximation. Also, the formula listed contradicts itself:

$\mathrm{SNR} = 10(2n-1) \cdot \log_{10}(2)+10 \cdot 3 = 6.02 \cdot n + 26.989$

not $6.02 \cdot n + 1.763$

This PDF has a different derivation:

$\mathrm{SNR} = 2^n ( {\sqrt {3} / \sqrt{2}} )$

$\mathrm{SNR_{dB}} = 20 \log_{10} \left ( 2^n ( {\sqrt {3} / \sqrt{2}} ) \right )$

$\mathrm{SNR_{dB}} = 20 \log_{10} ( 2) \cdot n + 20 \log_{10} ( \sqrt {3} / \sqrt{2})$

SNR dB = 6.021 n + 1.763

— Omegatron 15:13, August 30, 2005 (UTC)

Thanks for fixing the formula, User:212.119.9.186!

With more precision is 6.020599 n + 1.760913, if it matters. — Omegatron 15:49, 10 October 2005 (UTC)

Also, $10(2n-1) \cdot \log_{10}(2)+10\log_{10}(3) = 20 \log_{10} ( 2) \cdot n + 20 \log_{10} ( \sqrt {3} / \sqrt{2})$ — Omegatron 01:20, 11 October 2005 (UTC)

Ok, the 6.02n formula corresponds to a noise level caused by a uniform fluctuation between two quantization levels. I think another way of saying this is that the sampled value is assumed to be uniformly distributed across the possible range of values.

The 6.02n+1.761 formula maybe only applies to the ratio of a full-scale sine wave to the error signal, as hinted at in the article? So the signal is assumed to be a sine wave instead of being unknown but uniformly distributed. I will try to do the math later.

Is there another for a Gaussian distribution? — Omegatron 17:32, 30 November 2005 (UTC)

As I've explained in the article, the noise is signal-dependent, so SNR for a system requires that you come up with a model of what that signal is going to look like first. — Omegatron 20:25, 12 March 2006 (UTC)

RMS signal = $A \over \sqrt{2}$ or $FSR/2 \over \sqrt{2}$ where FSR is the full-scale range of the converter

RMS noise = $q \over \sqrt{12}$ where q is the LSB

[edit] A million different equivalent equations

$10(2n-1) \cdot \log_{10}(2)+10\log_{10}(3)$
$20 \log_{10}(2^{n-1} \sqrt{6})$ [1]
$20 \log_{10} \left ( {2^n \sqrt{12} \over 2 \sqrt{2}} \right )$ [2]
$20 \log_{10} ( 2) \cdot n + 20 \log_{10} ( \sqrt {3} / \sqrt{2})$ [3]
$2n\cdot10\log_{10}(2)+10\log_{10}(3/2)$ Totally forgot we have a derivation over at Talk:Quantization_noise#Thesis — Omegatron 02:35, 13 March 2006 (UTC)

[edit] SNR = 8 sigma ... what does the sigma mean?

Nasa describes the performance of a CCD in a space-based optical telescope as having SNR = 8 Sigma

at this page: http://kepler.nasa.gov/sci/basis/character.html

I'm sure someone can explain to me if this just means SNR is 10^8? Or what does it mean?

-Kevin

I'm not exactly sure, but sigma often refers to standard deviation, so my guess would be that this means that the SNR is equivalent to (1/) 8 standard deviations in a gaussian distribution or 6x10^-16 or 150dB. This is of course unfounded conjecture, use at your own risk (this corresponds to 25bits, which seems dubious)

[edit] SNR

How can I calculate the SNR for video quality? sorry about my spelling

[edit] Different meanings of SNR

The current article focusses on the SNR of sound. However, there are alternative meanings and definitions for the SNR, such as for analytical instruments. For instruments, the SNR (signal intensity / standard deviation of background) is a measure for the quality of the signal: if the SNR for a signal is less dan 3, it is not detected, whereas if it is higher than 10, the signal can be quantified. See also the SNR on the Dutch wikipedia. Therefore, I suggest to adapt the article or write a new completely new one. Annabelleke 13:56, 15 March 2007 (UTC)

Nonsense; there is only one definition of SNR (except that in video and image the definition gets perverted somewhat by common usage), and it has nothing specific to sound; the dB scale is commonly used for all sorts of sound and non-sound signals, but is completely optional. And detection thresholds are not so simple as you say. Dicklyon 23:34, 15 March 2007 (UTC)

Yeah, OK. There's is indeed only one definition (cfr. introduction of the article), but it is expressed in different ways for different disciplines. Therefore, there should be at least a paragraph or notion on the different expressions of definition. In many disciplines, in fact in most disciplines, people don't speak of the power of power ratio. This term comes from fourier transform analysis of a signal.

In addition, in the case of analytical chemistry, e.g. for spectroscopic measurements, detection limit is defined, by making use of the SNR.

Annabelleke 08:02, 16 March 2007 (UTC)

Please do add info about that. Actually, I'm no longer sure of the definition in terms of power ratio, since the SNR is often quantified as an amplitude ratio; either way, one article should cover it all. Need to review some books... Dicklyon 08:18, 16 March 2007 (UTC)

[edit] SNR in image processing

The article says "In image processing, the SNR of an image is usually defined as the ratio of the mean pixel value to the standard deviation of the pixel values" but that sounds confusing to me. I guess it involves time series (thus every pixel in a static image is acquired several times, and its standard deviation can be assigned to noise) or a constant (homogeneous) image; but the way it's written suggests that the standard deviation is calculated over all the pixels in a single frame image, whatever it is. The latter is of course false: the variability of different pixels in an image can be due to the object being imaged, and not all of it is just noise. Default007 13:40, 14 August 2007 (UTC)

[edit] "unreferenced driveby edit"

I, too, usually just revert small uncommented changes to new or anonymous editors. But in this case, the removal of the square from the amplitude ratio was correct. It's taken care of by the 20 instead of 10 multiplying the log of the ratio. Editors, please use edit summaries to tell us what your edit is about, so it can be distinguished from drive-by vandalism. Dicklyon 14:50, 17 October 2007 (UTC)

[edit] NPR

At NPR (disambiguation), someone had added an entry for Noise Power Ratio, though we don't seem to have such a page. It seems to me that this should probably redirect (if at all) to here, Signal-to-noise ratio. Do other editors agree? Or is a separate page needed? References are here:[4][5] Thanks, El on ka 04:02, 16 May 2008 (UTC)

In books, it does seem to be a genuine term and acronym. Better read some of those books and see what it means. Dicklyon (talk) 05:31, 16 May 2008 (UTC)

Categories: WikiProject Telecommunications | Unassessed physics articles