Vidar Øierås
Hi-Fi freak
- Ble medlem
- 19.01.2003
- Innlegg
- 3.202
- Antall liker
- 341
Hei,
Jeg har diskutert litt med en kar i fra LinearX - de som leverer LEAP LMS f.eks. Jeg mente det fantes mer informasjon i en impulsrespons enn i en ren frekvensrespons. Han var visst uenig i det:
Sitat:
"Vidar påstår: "An impulse response says much more about a speakers behavior than a pure frequency response"
Chris:
That statement is incorrect. This is a common misconception by people that have limited understanding of digital signal processing. The basis of the Fourier and LaPlace transforms is that the frequency domain and time domain represent the same identical information, just in different ways. If one domain showed you different information than the other, the transforms would not exist.
For example, assume you have a transfer function and the time domain response shows a ringing on a pulse/step. In the frequency domain, it is shown as a peak in the frequency response. This is exactly the same anomaly. It appears in both domains. Anything that shows up in one domain must show up in the other domain. This is basic to all signal processing theory.
However there are most certainly differences in how time doman and frequency domain measurements are made, and how they perform in the presence of adverse signal conditions such as background noise. Acoustic measurements often take place in the presence of high background noise, which is not true for electronic equipment measurements. This often presents a problem for time domain based systems which see the noise as dither. This can destroy the accuracy of any time domain distortion measurements, and in the presence of high noise can even destroy the accuracy of time domain amplitude measurements.
There are many factors involved in time domain measurements such as: the type of signal used, its spectral nature and content, acquistion length, gating time, window functions, etc. The operation of these parameters and their understanding is generally beyond the knowledge of most users. The other fundamental difference between frequency domain and time domain measurements is log vs. linear frequency resolution. The linear frequency resolution in the time domain often creates inadequate resolution problems at low frequencies. As a result, many measurements taken on time domain systems can have poor or inaccurate results.
Så ønsket jeg vannfalls-muligheter i deres nye programvare:
Chris:
Waterfall processing has been widely misused and misunderstood. The value of these curves, their accuracy, and their meaning, is highly questionable. Reduced time windowing of a sampled signal causes major loss of information. The curves that result may have peaks, valleys, ripples, etc. but they are only a reflection of the limited data remaining in each time slice window. This may or may not have any significance or importance, and may lead to completely incorrect conclusions.
For example, assume you do waterfall analysis on the response of an enclosure. With each time slice the diffraction around the enclosure is being removed from the response, so each waterfall time curve certainly looks different. However, the diffraction around the enclosure is part of it's entire response. That is what your ear will hear, the total complete response. Your ear does not do time slicing, it hears a continuous time spectrum. To view the response of the enclosure in only a narrow range of time slices yields a completely inaccurate response, and any conclusions based on that analysis are incorrect and meaningless.
The problems involved with waterfall analysis also become acute when you try to obtain the same analysis from different analyzers/software. None of them are the same, they do not match, the results cannot be compared or verified. In many cases the way the waterfall setup is parameterized in each software is very different, and you cannot set them up the same. Moreover, when you change the parameters even a small amount, the resulting set of curves can look entirely different.
I hope that helps. Time and space does not allow me to cover all of these topics here in detail. That kind of treatment would require a book. Suffice to say, LMS has been the most successful electroacoustic analyzer around the world for nearly 20 years, and it is entirely log frequency domain.
The users rely on and trust the measurements and data they obtain from this system. The results are repeatable and they do not require expert knowledge on the part of the user. The LX500 will continue and expand on this legacy and reliable frequency domain tradition, offering yet another new level of capabilities and precision.
Best Regards,
Chris N. Strahm / Engrg
LinearX Systems Inc.
9500 SW Tualatin-Sherwood Rd.
Tualatin, OR 97062-8586 USA"
Noen som kan kommentere dette. Enig eller uenig?
Jeg har diskutert litt med en kar i fra LinearX - de som leverer LEAP LMS f.eks. Jeg mente det fantes mer informasjon i en impulsrespons enn i en ren frekvensrespons. Han var visst uenig i det:
Sitat:
"Vidar påstår: "An impulse response says much more about a speakers behavior than a pure frequency response"
Chris:
That statement is incorrect. This is a common misconception by people that have limited understanding of digital signal processing. The basis of the Fourier and LaPlace transforms is that the frequency domain and time domain represent the same identical information, just in different ways. If one domain showed you different information than the other, the transforms would not exist.
For example, assume you have a transfer function and the time domain response shows a ringing on a pulse/step. In the frequency domain, it is shown as a peak in the frequency response. This is exactly the same anomaly. It appears in both domains. Anything that shows up in one domain must show up in the other domain. This is basic to all signal processing theory.
However there are most certainly differences in how time doman and frequency domain measurements are made, and how they perform in the presence of adverse signal conditions such as background noise. Acoustic measurements often take place in the presence of high background noise, which is not true for electronic equipment measurements. This often presents a problem for time domain based systems which see the noise as dither. This can destroy the accuracy of any time domain distortion measurements, and in the presence of high noise can even destroy the accuracy of time domain amplitude measurements.
There are many factors involved in time domain measurements such as: the type of signal used, its spectral nature and content, acquistion length, gating time, window functions, etc. The operation of these parameters and their understanding is generally beyond the knowledge of most users. The other fundamental difference between frequency domain and time domain measurements is log vs. linear frequency resolution. The linear frequency resolution in the time domain often creates inadequate resolution problems at low frequencies. As a result, many measurements taken on time domain systems can have poor or inaccurate results.
Så ønsket jeg vannfalls-muligheter i deres nye programvare:
Chris:
Waterfall processing has been widely misused and misunderstood. The value of these curves, their accuracy, and their meaning, is highly questionable. Reduced time windowing of a sampled signal causes major loss of information. The curves that result may have peaks, valleys, ripples, etc. but they are only a reflection of the limited data remaining in each time slice window. This may or may not have any significance or importance, and may lead to completely incorrect conclusions.
For example, assume you do waterfall analysis on the response of an enclosure. With each time slice the diffraction around the enclosure is being removed from the response, so each waterfall time curve certainly looks different. However, the diffraction around the enclosure is part of it's entire response. That is what your ear will hear, the total complete response. Your ear does not do time slicing, it hears a continuous time spectrum. To view the response of the enclosure in only a narrow range of time slices yields a completely inaccurate response, and any conclusions based on that analysis are incorrect and meaningless.
The problems involved with waterfall analysis also become acute when you try to obtain the same analysis from different analyzers/software. None of them are the same, they do not match, the results cannot be compared or verified. In many cases the way the waterfall setup is parameterized in each software is very different, and you cannot set them up the same. Moreover, when you change the parameters even a small amount, the resulting set of curves can look entirely different.
I hope that helps. Time and space does not allow me to cover all of these topics here in detail. That kind of treatment would require a book. Suffice to say, LMS has been the most successful electroacoustic analyzer around the world for nearly 20 years, and it is entirely log frequency domain.
The users rely on and trust the measurements and data they obtain from this system. The results are repeatable and they do not require expert knowledge on the part of the user. The LX500 will continue and expand on this legacy and reliable frequency domain tradition, offering yet another new level of capabilities and precision.
Best Regards,
Chris N. Strahm / Engrg
LinearX Systems Inc.
9500 SW Tualatin-Sherwood Rd.
Tualatin, OR 97062-8586 USA"
Noen som kan kommentere dette. Enig eller uenig?