忘記密碼
 立即註冊
樓主: alphatm.tw

Group Delay與minimum phase測量

[複製鏈接]

164

主題

846

回帖

0

積分

新手上路

三星級會員

舊站積分
143
舊站發表文章數
812
 樓主| 發表於 2012-9-1 14:33:48 | 顯示全部樓層
我測試出來的這個是一個設計好的8吋同軸喇叭,就像是tannoy system 800單體那種類型的

我測試出來的結果是,如果我把訊號長度家常從16K加到64K 第一個是分析的頻率可以更低,在低頻的穩定性也會來的好一些,可是用MLS來分析低頻重覆性沒有sinusoidal來的穩定

所以你的說法是指說,如果我用MLS Logchirp分析的時候使用Logchirp會比MLS分析低頻的時候來的穩定一些嘛?有空的話我來測試看看


原發表者是 soongsc
[B]
先確認一下,您測的是單體,而且是無障板?

我用不同訊號測結果的平整度是不同的,我認為這是合理的結果。Logchirp透過即時平均可以降低反射波的影響,低頻還蠻管用的。 [/B]

20

主題

4368

回帖

0

積分

新手上路

五星級會員

舊站積分
508
舊站發表文章數
4372
發表於 2012-9-1 16:57:31 | 顯示全部樓層
通常同軸喇叭至少是兩音路,盆內也有散射交替影響,這是這種設計的天限。

我覺得你把MLS訊號本身長度與MLS量測後分析其impulse的許樣長度搞混了。增加MLS訊號長度是提升量測頻率解析度,在低頻本來就更現現效果。我用SOUNDEASY量通常設定在256K。

取自SOUNDEASY使用手冊:
True Frequency Resolution
The system transfer function is computed by applying an FFT to the impulse response or a portion of it as
in the case of windowed, anechoic loudspeaker measurements. The displayed frequency resolution is given as the
sample rate divided by the FFT size (i.e. 48000/262144=0.183Hz). Displayed frequency resolution is just the FFT
bin  separation.  However,  the  true  frequency  resolution  of  a  measurement  is  at  best  equal  to  1/T  where  T  is  the
length of the impulse response segment used in the FFT calculation (i.e. the width of a rectangular time window).
Thus, it is possible to have a very fine displayed resolution but a rather coarse true resolution. For example,
you could apply a 32768-point FFT to a 512-point segment of impulse response. In that case, the large FFT only acts
to interpolate the true frequency resolution to form a much finer displayed resolution.
[url="http://augene.blogspot.com"]Hear The REAL Thing[/url] [url="http://www.augene-audio.com.tw"]感受藝人的心靈[/url]

164

主題

846

回帖

0

積分

新手上路

三星級會員

舊站積分
143
舊站發表文章數
812
 樓主| 發表於 2012-9-1 22:16:35 | 顯示全部樓層
感謝你的指正,你說的那個測試項目,我有翻一下說明書,他的說明如下
MLS & LOG CHIRP SIZE
The MLS & LOG CHIRP Size is the number of points that defines the MLS & LOG
CHIRP sequence. In terms of generating a digital signal these are the number of
samples before the signals is started again from the beginning. CLIO handles MLS &
LOG CHIRP sizes from 4k to 256k. These terms are inherited by the computer world
and are somewhat imprecise. The real size is the nearest power of 2, the 4k being
4096 points and the 256k 262144 points long (one less! for MLS, exactly this for
LOG CHIRP). The first important consequence of setting the MLS & LOG CHIRP size
is the length of the Impulse Response recovered which is always as long as the MLS
& LOG CHIRP itself. From the users point of view what matters is how long this
impulse is in terms of seconds. This in turns also depends on the Sampling
Frequency set. It is easily calculated as the size divided by the sampling frequency.
The default is 16k and 48000 Hz, that is 16384/48000=0.341 seconds. We will deal
with this in more detail later. For now it is enough to say that CLIO also uses this
120 10 MLS & LOG CHIRP
size for the FFT. This is important as the frequency resolution you get is calculated
as the sampling frequency divided by the FFT size. Again for the default settings
this is 48000/16384=2.929 Hz. This is already a high resolution. However thinking
in terms of octave or fractions of an octave, which are the terms of a logarithmic
frequency axis, 2.929Hz is around 1/2218 of an octave at 10kHz while is around
1/3 of an octave at 10Hz.

參閱圖片,參數值在右下角 @_@
所以48000Hz/32768=1.46Hz
那這個畫面中出現的FreqLO指的就是,點到點的解析度是1.46Hz對吧!!

所以32768/48000Hz=682.65ms
這個計算出來的也就是impulse response 長度



您需要登錄後才可以回帖 登錄 | 立即註冊

本版積分規則

Archiver|MyAV視聽商情網 test

GMT+8, 2026-11-19 12:04 , Processed in 0.090578 second(s), 27 queries .

Powered by MyAV!

快速回覆 返回頂部 返回列表