Jump to content

Talk:Frequency response

Page contents not supported in other languages.
From Wikipedia, the free encyclopedia

Audio frequency response

[edit]

Frequency response is often used to describe the performance of audio-related devices, however, it is also often used to describe the performance of, for example, coaxial cables, category cables, video switchers, wireless communications. All of those examples operate well into the gigahertz range. Subsonic examples could include earthquakes, electroencephalography (brain waves). This article should touch on all major disciplines where a frequency response measurement is often used. Snottywong 14:37, 13 September 2007 (UTC)[reply]

I agree with you. Perhaps you'll want to add to or reorganize my changes. Binksternet 16:17, 13 September 2007 (UTC)[reply]

Phase Response

[edit]

This article implies that the phase response is part of the frequency response. While they are often seen next to each other in specs and whatnot, I think the frequency response and the phase response are two different things. Phase shouldn't be mentioned in this article.Snottywong 14:22, 13 September 2007 (UTC)[reply]

I don't agree. Frequency response and Bode plot are for me the same thing: Magnitude AND phase plot together. If you want to be more specific, you could say magnitude/amplitude or phase plot. User:Nillerdk (talk) 09:06, 27 July 2008 (UTC)[reply]
Right. Phase response and frequency response go hand in hand. Binksternet (talk) 18:12, 27 July 2008 (UTC)[reply]
Right. The output test sinusoid of an LTI system is changed by the way the LTI system responds to a given frequency. And it does so in only two ways: Gain and Phase. That is why they are both included when mentioning "frequency response". And in specific situations, one can refer to "gain vs. frequency", and "phase versus frequency". Ohgddfp (talk) 05:33, 17 October 2020 (UTC)[reply]

Comments from 2004-2005

[edit]

Shouldn't this page be merged with transfer function? Jorge Stolfi 03:20, 25 Mar 2004 (UTC)

No. The two are not the same thing at all. Graham 05:08, 25 Mar 2004 (UTC)
Ok, but then the definition of "frequency response" needs to be made more precise.
As it is, one could argue that the two are synonymous, and that the phrase "X has a frequency response of 20Hz - 20,000Hz ±1dB" is only an informal way of saying "the frequency response (=transfer function) of X has constant modulus, ±1dB, between 20Hz and 20,000Hz".
So, assuming that the quoted example is indeed pretty much the definition, what about this rewording:
"The frequency response of a signal processing system is the range of frequencies over which the system's gain is constant, within a prescribed tolerance. For example, a high-fidelity audio amplifier may be said to have a frequency response of 20Hz - 20,000Hz ±1dB, which tells you that the system responds equally to all frequencies within that range and within the limits quoted.
 
It is commonly used in connection with electronic amplifiers and similar systems, particularly in relation to audio signals. As such it is not a measure that is very useful in terms of the quality of reproduction, only that it fulfils the basic requirements needed for it.
Jorge Stolfi 00:17, 26 Mar 2004 (UTC)
About "The frequency response of a signal processing system is the rant of frequencies over which the gain is constant, ..." No. Sorry. One can say the frequency response of a device is within a certain tolerance, such as 20 -20,000 Hz +/- 1dB, but frequency response is not itself a tolerance. Ohgddfp (talk) 05:59, 17 October 2020 (UTC)[reply]

I accept the rewording is something of an improvement, so feel free to amend the article. However, you are confused, I think, as to what transfer function means. The rewording doesn't mention this, so it doesn't matter in this context. Transfer function has a much wider meaning than frequency response, and can be applied to almost any system that has an input and an output. In respect of an amplifier, the transfer function is more to do with its linearity (i.e. distortion) than frequency response, though I suppose the case could be made for talking about the transfer function as it varies with frequency. Knowing this, one could extract the frequency response from it. TF is a complex, multi-dimensional aspect of a system, the FR is merely one limited "view" of it, which ignores many other parameters. Hope this helps! Graham 01:57, 26 Mar 2004 (UTC)

someone needs to add how the frequency response is related to the Fourier transform, eigenfunctions, and LTI system theory.

Yeah, right now the article doesn't address frequency response as I learned it at all. Namely if you've got some system with frequency response H(ω), input x(t) and output y(t) then you know that
where the hats indicate Fourier transforms, and therefore the phase of H(ω) is important, contrary to what the article currently says.
It does say that you can find the frequency response by using a Dirac delta function, which is the only reason I didn't doubt the terminology I learned in my very few engineering classes. So if x(t) = δ(t)
--Laura Scudder | Talk 22:08, 16 Apr 2005 (UTC)
Just keep in mind that all of that is only for LTI systems. Cburnett 22:35, Apr 16, 2005 (UTC)
and "frequency response" refers to LTI systems. Ohgddfp (talk) 05:51, 17 October 2020 (UTC)[reply]
About: "No. The two are not the same thing at all." in the above. They are exactly the same thing. That's because an LTI system does not have significant non-linear issues (creates new frequencies and noise at the output that don't exist at the input). What's left is the LTI characteristics. And that is frequency response. For an LTI system, deriving the frequency response from the device's transfer function (for LTIs) is straight forward. And while the transfer function is likely to be an easier way to analyze a device, the transfer function and frequency response represent the exact same information about the device. Ohgddfp (talk) 05:51, 17 October 2020 (UTC)[reply]

There is a merge banner suggesting full frequency response is merged with this page. I'd suggest simply deleting FFR - it's a very subjective definition (what's the full frequency response of an RF amplifier?), largely meaningless. Unless I'm missing something? GyroMagician (talk) 18:14, 3 June 2009 (UTC)[reply]

H(jω) vs H(ω)

[edit]

I don't understand how you can have a function of jω (like what would y(2x) = 5x mean?), but this is often written this way. What's the difference? 71.167.58.9 (talk) 21:59, 21 January 2014 (UTC)[reply]

Applications

[edit]

In the current version, this section opens with "in electronics this stimulus would be an input signal." - this seems a bit lost and maybe belonged to older text. Ok to remove? Vapor57 (talk) 00:12, 18 December 2019 (UTC)[reply]

Technical issues in the intro

[edit]
  • The information about LTI systems isn't focused specifically on frequency, ie the frequency response of an LTI system is LTI, but that's only because the Fourier/Laplace transform is a linear transformation
  • Sinusoidal fidelity is a property of linear systems, not time invariant systems. Time invariance deals with how delays in the input affect the output
  • 1st paragraph last sentence, "purely imaginary" should be "complex" (you can't have magnitude and phase without both real and imaginary components). Also, both the frequency response and the transfer function describe the system independently of the input. The frequency response is a just the special case of taking H(jω) (ie, σ = 0 in the Laplace variable s). They are distinct concepts but very closely related.
  • The 2nd paragraph should be moved to the "Applications" section or made to be more general, it's too specific for the intro (I will do this later)

I've already made some changes to address the first 3 points 30103db (talk) 20:58, 24 March 2022 (UTC)[reply]

Issues in Measurements and plotting
  • 1st method: this is the impulse response, not the frequency response. You have to take the Fourier transform to get frequency
  • 2nd method: a frequency sweep needs to be slow enough to allow the system to reach steady-state. Otherwise you could get a different response depending on which direction you sweep
  • 3rd method: why is deconvolution happening in the frequency domain? Deconvolving the time-domain output would get the impulse response, not frequency, but the reason you move to the frequency domain in the first place is to only have to deal with algebraic operations instead of deconvolution
  • OFDM and N-OFDM are modulation schemes for data transmission, they aren't used for finding frequency response. All 3 sources linked only discuss scheme details, transmission efficient, etc. I don't even know why they were ever included.
  • White noise doesn't have to be filtered to pink noise. In theory either could work but in practice white noise is used because it's easier to generate and flat across all frequencies. It should also be made clear that this is a statistical as opposed to a deterministic method, and that simply using the PSD will result in loss of phase response (cross power spectral density should be used if phase response is required) 30103db (talk) 17:46, 25 March 2022 (UTC)[reply]