Actually Marcel, it doesn't. The total time window sets the frequency resolution of the impulse. The gating window has no effect on this.
We need to see the impulse response to determine what the issue is. Measure at as low a level to the transducer that you can get away with, i.e. not so low that noise becomes an issue. Set mic gains etc. such that the signal that you get from the mic is nearly as large as possible. This might take some practice.
I think that you should mean "The shorter the time record, the less detail there is ..."
The resolution of the FFT that is taken on the impulse depends only on the total time record. That resolution is set by the total amount of data, even to the point of padding the data out to whatever length you want. The resolution goes up even though you are not adding in any data per-se.
In an anechoic measurement all data beyond a certain point is already zero so adding in zeros just adds to the resolution.
Gating has no effect on HFs only LFs.
Well, if you gate your measured IR, you basically set a limit on what it may contain frequency response wise. The shorter the gating time, the less detail there can be, no matter how much zero padding you do afterwards.
If the IR fits in the gate (i.e. is not touched by the gate too much), everything is fine. If it doesn't then of course you do some kind of "smoothing" of the frequency response.
If the IR fits in the gate (i.e. is not touched by the gate too much), everything is fine. If it doesn't then of course you do some kind of "smoothing" of the frequency response.
Last edited:
Could you show the IR (showing 0-10 ms) so we can see where the gating should be set? After gating you could EQ (normalize) the listening axis and I'd bet it will look quite wonderful.
Sorry Marcel, that is not correct.
The HFs dissipate very quickly and as such do not enter into the picture for a gated measurement. Gating degrades the response from the bottom (LFs) up, so a very short window will degrade a much higher frequency, but the "resolution" remains constant. The degradation of the response from gating is not the same thing as limiting "frequency resolution."
Look at Holm sometime as you change the gating time. The HFs don't change at all (unless you let in some reflections.) In an anechoic measurement changing the gating length has no effect on the response until you begin to "cut-off" the LF resonance tail.
The HFs dissipate very quickly and as such do not enter into the picture for a gated measurement. Gating degrades the response from the bottom (LFs) up, so a very short window will degrade a much higher frequency, but the "resolution" remains constant. The degradation of the response from gating is not the same thing as limiting "frequency resolution."
Look at Holm sometime as you change the gating time. The HFs don't change at all (unless you let in some reflections.) In an anechoic measurement changing the gating length has no effect on the response until you begin to "cut-off" the LF resonance tail.
That's indeed what I say. That it is important to fit the impulse response in the gate, which in fact sets the limit on the resolution of such features.
- If you have only 3 ms available, there's simply a limited amount of FR "detail" you can measure - especially regarding resonances. If the IR takes 2.5 ms to diminish, you have it all (and it also means that there are no such features). If it doesn't, you just don't know what the truth is, but can be sure it's worse than what you measured and calculated.
Last edited:
All my comments, which you have said are incorrect, specifically said 'frequency resolution', as in frequency response graphs.
You appear to have changed my specific resolution reference to mean 'time resolution', as in an impulse response.
Nobody ever said an impulse's resolution changes with gating, I think we all know it doesn't.
I continue to totally think gating lessens frequency response resolution.
(And admitted my bad, saying that it's not much of an issue for the waveguides in this thread. Apologies again.)
If you want to say it only effects LF and not HF,
it's probably fair you define that split in terms of how much gating, and at the freq that separates LF from HF.
Of course, but that's partly because the fundamental frequency resolution given by the gating time is constant across the whole band, i.e. the lower the frequency the bigger the impact it has. The fundamental resolution is indeed 333 Hz in a case of 3 ms gating time, that's 1/2 octave at 600 Hz and 1/30 octave at 10 kHz - and that's all the resolution you have (and again, it can be enough to measure accurately - IF the device is smooth enough so that its IR diminishes in these 3 ms).
Last edited:
Basically, yes (that's really how it's calculated). But that resolution can be easily increased by the zero padding, i.e. making the bars narrower and the resulting "curve" smoother (in fact as smooth as you want - making virtually a continuous line). But that doesn't add any real resolution, which is limited by the gating time from the start.
By the "real resolution" I mean the ability to show the real Q of a resonance, for example, if there's some, or show it at all.
By the "real resolution" I mean the ability to show the real Q of a resonance, for example, if there's some, or show it at all.
Last edited:
I think that we agree, but maybe we have a slightly different definition of resolution. I mean the textbook delta-frequency of the data, that doesn't change. What you are referring to is a smoothing effect of the gate as a convolution smoothing that will lower the Q of a very high Q result. To me, they are different, to you maybe they are the same thing.
Hello mindsource, i guess this is the reason for your disappointment. REWs was intended ad room response tool, so for loudspeakers you need to take way shorter gating window. Just to see what this does go to "IR windows" and set left window to 5ms and right window to 7 ms and apply to all. (These are just my arbitrary guesses!)
Last edited:
What's the driver again? That would be the first thing I would try - test a different one. I don't think that a proper gating will make a night and day difference here. It will get certainly cleaner but I would expect the overall picture to stay about the same. And I don't think the waveguide is to blame.
Last edited:
Or in other words, the smoothing effect of gating on high Q, that changes it to a lower Q, is simply due to the decreased frequency resolution caused by gating .....which can no longer accurately replicate high Q.
In my simpler way of seeing things, there are only two forms of resolution...time and frequency.
We may rename them, re-analyze them in myriad forms, but in the end it comes down to which one do want to look at in high resolution, and why.
Can't have both, in high resolution, at the same time, huh?
Last edited: