SC-04-08 Richard Cabot posted an update in the group SC-04-08 3 weeks ago No folders found. Please create and select folder. Documents Folder Title Following special characters are not supported: \ / ? % * : | " < > Privacy Public All Members My Connections Only Me Cancel Create 000151.html 9 KB HTML - Click to view Options Copy Download Link [SC-04-08] WG meeting report New York 2013-10 pre { white-space: pre-wrap; /* css-2.1, curent FF, Opera, Safari */ } [SC-04-08] WG meeting report New York 2013-10 Ben Kok ben at benkok.com Wed Jan 15 07:42:45 MST 2014 Next message: [SC-04-08] WG meeting report New York 2013-10 Messages sorted by: [ date ] [ thread ] [ subject ] [ author ] All, Please find the report of the Work Group meeting in New York below Best regards, Ben Kok SC-04-08 workgroup chair ============================================================================ Report of the meeting of the SC-04-08 Working Group on Measurement of sound systems in rooms of the SC-04 Subcommittee on Acoustics, held in New York, 2013-10-18 The meeting was convened by chair B. Kok. The chair welcomed those present at this first meeting of SC-04-08, both physically and remotely using the Internet. The meeting included representatives from relevant SMPTE engineering groups. The agenda were accepted as written. There were no comments on the notes of the previous meeting in Rome, 2013-05-06 Development Projects AES-X215 Liaison with SMPTE project to codify current procedures to calibrate the Cinema B-Chain Scope: "This project will liaise with SMPTE TC-25CSS in their project to Codify current procedures as used by skilled theater technicians to calibrate the Cinema B-Chain. The project will consider comments as appropriate for communication to SMPTE". Brian Vessa presented on SMPTE activity and cinema context. This inluded recording requirements, the influence of screens and acoustics, current SMPTE standards, X-curve, SPLL settings, variations in EQ, etc. The main goal is to achieve consistent results, taking into account ear/brain compensations and headroom requirements. >From the SMPTE perspective it essentially is a 3-step plan: 1) Optimize current methods 2) Move forward, using contempary (measurement techniques and insights 3) Anticipate to, or initiate, future developments AES-X216 Liaison with SMPTE project on Calibration Pink-Noise Standard and Test File Scope: "This project will liaise with SMPTE TC-25CSS in their project to "Create a pink-noise standard, and a reference pink-noise file available in a digital audio file and DCP formats that conform to the description of the pink noise in ST 202:2010 and RP 200-2012 and for which the time domain (mean, standard deviation, skew, kurtosis, etc.) and frequency domain (energy per bandwidth, bandwidth, etc.) characteristics are defined. Specify the algorithm used to generate the pink-noise file." The project will consider comments as appropriate for communication to SMPTE. Pete Soper discussed the SMPTE need for a pink noise test signal. Already in wide use to support existing measurement methods, but poorly specified so that variations in different pink-noise sources were a source of error in practical measurements. Future measurement methods are not constrained to use pink noise, but for optimization of current methods it seems essential to have a well-defined signal. A replacement was proposed by using a cyclic signal based on a repeated Pseudo Random Noise sequence containing all frequencies, f.i. MLS with pink shaping. This will be easier to specify as a more stable and reproducible source. If the sequence length is chosen to be similar to the settling time of the analyser (typical 1 s for setting 'slow'), this would ensure a stable reading as temporal irregularities in the measured response will repeat with the sequence length. This would substitute for less-well-defined PN generators in existing measurements using RTAs. An intensive discussion was on the required length of the repeated sequence, aspects to consider are: - Frequency resolution (typical 1/sequence length, thus 1 Hz for a 1 s length) The bandwidth of the 20 Hz 1/3-octave band is app 4,5 Hz is 4.5 point in this octave band sufficient for a good measurement? - Longer sequences offer better frequency resolution, but that would require the use of Leq measurements, it is not clear of the equipment used for the current methods support this. - The sequence length/frequency resolution is mainly a concern at lower frequencies and when narrower band analysis that 1/3-octave are used; maybe it is an option to define a dedicated low frequency test signal with a longer sequence, to be used only during tuning the system and not for regular maintenance. - There was much discussion about an acceptable settling time. 10sec felt to be high. 30-sec v difficult. 3 to 4 sec would be acceptable to a technician for quick checks in a cinema. - The repetitive character might be (subjectively) disturbing for some technicians, as they have always learned that pink noise should not be repetitive. Some further research will be needed to evaluate this concept. JMW preferred a pseudo-random generator using a shift register as a simpler implementation. A practical pinking filter will need a sensible specification. Needs to produce a fairly rapid settling time - so can't spend minutes assembling a spectrum. The test signal needs some bandwidth limits to avoid effects caused by out-of-band components. Proposed is to develop a tolerance window with minimum and maximum level per frequency, similar to frequency response tolerances for microphones, filters, etc.; this then also defines the allowable ripple in the pass-band. Amplitude stability - the rms level of pink noise should be referenced to -20dB FS Brian Vessa would like versions of pink noise at 48kHz and 96kHz sampling frequencies. The strawman crest-factor specification was constrained between 11.5 and 12dB. No real significance to this number - just seems to be close to current generators. Pete Soper was aware that CF will be modified by downstream elements (amplifiers, transformers, filters), but feels this should not change the PN spec - the distortion is all part of the B-chain. Others felt that a lower CF would be better. The group decided to consider the 12dB to be a maximum CF and not the upper limit of a 0.5dB window. AES-X218 Measurement and calibration of sound systems in rooms Scope: "to document relevant parameters affecting the audience experience of a sound system in a generalized room, and identify appropriate objective methods of measurements to quantify them. The intent is to identify tools and procedures for a technician to measure the performance of a loudspeaker system in a room, and then support accurate and reliable calibration of this system to a specified performance." Eddy's Nordic document (NT ACOU 108, available for download from the group's document site) is interesting as a starting point for this project. Formed a pilot group to create a list of topics: Glen Leembruggen David Murphy Peter Mapp Eddy Brixen Sunil Bharitka Felipe Tavera John Geraghty Weislaw Woszczyk Ben Kok AES-X219 Method of measurement for frequency and impulse response of sound systems in auditoria Scope:" to specify a method of measurement for frequency and impulse response of sound systems in auditoria". A scope for this project awaits refinement in project AES-X218. Liaisons No liaison matters arose. New Projects No new projects were proposed. New Business There was no new business. Date of next meeting The next meeting will be scheduled in conjunction with the AES 136th Convention in Berlin, 2014-04 --- Dit e-mailbericht bevat geen virussen en malware omdat avast! Antivirus-bescherming actief is. http://www.avast.com Next message: [SC-04-08] WG meeting report New York 2013-10 Messages sorted by: [ date ] [ thread ] [ subject ] [ author ] More information about the SC-04-08 mailing list Expand 000152.html 3 KB HTML - Click to view Options Copy Download Link [SC-04-08] WG meeting report New York 2013-10 pre { white-space: pre-wrap; /* css-2.1, curent FF, Opera, Safari */ } [SC-04-08] WG meeting report New York 2013-10 Andreas Silzle andreas.silzle at iis.fraunhofer.de Wed Jan 15 10:13:12 MST 2014 Previous message: [SC-04-08] WG meeting report New York 2013-10 Next message: [SC-04-08] WG meeting report New York 2013-10 Messages sorted by: [ date ] [ thread ] [ subject ] [ author ] Hi, Please find attached a paper about the difficulties of MLS measurements. The cyclic error is one problem, the high sensitivity to create distortion in the reproduction chain by the rectangular form of the signal is the next. With the sine sweep method you can separate the response from the distortion and select by a window the response alone. In my experience you can gain 10-15dB in S/N with a sine sweep measurement compared to a MLS measurement, with the same hardware setup. And this is a lot for room transfer function measurements. Best regards, Andreas -- Dr.-Ing. Andreas Silzle Senior Scientist Fraunhofer IIS Laboratories phone: +49 9131 776-6248 Am Wolfsmantel 33 fax: +49 9131 776-6099 D-91058 Erlangen, Germany andreas.silzle at iis.fraunhofer.de Previous message: [SC-04-08] WG meeting report New York 2013-10 Next message: [SC-04-08] WG meeting report New York 2013-10 Messages sorted by: [ date ] [ thread ] [ subject ] [ author ] More information about the SC-04-08 mailing list Expand 000153.html 4 KB HTML - Click to view Options Copy Download Link [SC-04-08] WG meeting report New York 2013-10 pre { white-space: pre-wrap; /* css-2.1, curent FF, Opera, Safari */ } [SC-04-08] WG meeting report New York 2013-10 John Woodgate jmw at jmwa.demon.co.uk Wed Jan 15 10:28:00 MST 2014 Previous message: [SC-04-08] WG meeting report New York 2013-10 Next message: [SC-04-08] WG meeting report New York 2013-10, MLS Messages sorted by: [ date ] [ thread ] [ subject ] [ author ] In message <52D6C1A8.103 at iis.fraunhofer.de>, dated Wed, 15 Jan 2014, Andreas Silzle <andreas.silzle at iis.fraunhofer.de> writes: >Please find attached a paper about the difficulties of MLS >measurements. Attachments are not permitted. The proposal is to derive a predictable pink-noise signal from an MLS, not to use an MLS directly. >The cyclic error is one problem, What is a 'cyclic error'? >the high sensitivity to create distortion in the reproduction chain by >the rectangular form of the signal is the next. After the signal has been through the 'pinking' filter, it no longer has a rectangular form. >With the sine sweep method you can separate the response from the >distortion and select by a window the response alone. There should not be significant distortion during the measurement. If there is, it is being done at too high a level. -- OOO - Own Opinions Only. With best wishes. See www.jmwa.demon.co.uk Nondum ex silvis sumus John Woodgate, J M Woodgate and Associates, Rayleigh, Essex UK Previous message: [SC-04-08] WG meeting report New York 2013-10 Next message: [SC-04-08] WG meeting report New York 2013-10, MLS Messages sorted by: [ date ] [ thread ] [ subject ] [ author ] More information about the SC-04-08 mailing list Expand 000154.html 5 KB HTML - Click to view Options Copy Download Link [SC-04-08] WG meeting report New York 2013-10, MLS pre { white-space: pre-wrap; /* css-2.1, curent FF, Opera, Safari */ } [SC-04-08] WG meeting report New York 2013-10, MLS Andreas Silzle andreas.silzle at iis.fraunhofer.de Wed Jan 15 11:09:55 MST 2014 Previous message: [SC-04-08] WG meeting report New York 2013-10 Next message: [SC-04-08] WG meeting report New York 2013-10, MLS Messages sorted by: [ date ] [ thread ] [ subject ] [ author ] Hi, Am 15.01.2014 18:28, schrieb John Woodgate: > In message <52D6C1A8.103 at iis.fraunhofer.de>, dated Wed, 15 Jan 2014, > Andreas Silzle <andreas.silzle at iis.fraunhofer.de> writes: > >> Please find attached a paper about the difficulties of MLS measurements. > > Attachments are not permitted. Sorry, this I have not known. Here the reference: Krarup Olesen, S., et al. An improved MLS measurement system for acquiring room impulse responses. in NORSIG 2000, IEEE Nordic Signal Processing Symposium. 2000. Kolmården, Sweden. > > The proposal is to derive a predictable pink-noise signal from an MLS, > not to use an MLS directly. > >> The cyclic error is one problem, > > What is a 'cyclic error'? When you repeat the same MLS sequence several times, you get in the measurement result spikes in the time domain above the normal noise floor, when you have the slightest distortion of e.g. the loudspeaker in the processing chain. You can reduce this by using different MLS sequences of the same length, see paper. > >> the high sensitivity to create distortion in the reproduction chain >> by the rectangular form of the signal is the next. > > After the signal has been through the 'pinking' filter, it no longer > has a rectangular form. Ok, then it is maybe less critical. It should be tested and still compared to the sine sweep method. > >> With the sine sweep method you can separate the response from the >> distortion and select by a window the response alone. > > There should not be significant distortion during the measurement. If > there is, it is being done at too high a level. There is the question, what SNR do you need for your measurement results. For room impulse responses for convolution you need something greater than 70dB and this is difficult to achieve with MLS. To get a frequency response to setup an EQ less is enough. Best regards Andreas -- Dr.-Ing. Andreas Silzle Senior Scientist Fraunhofer IIS Laboratories phone: +49 9131 776-6248 Am Wolfsmantel 33 fax: +49 9131 776-6099 D-91058 Erlangen, Germany andreas.silzle at iis.fraunhofer.de Previous message: [SC-04-08] WG meeting report New York 2013-10 Next message: [SC-04-08] WG meeting report New York 2013-10, MLS Messages sorted by: [ date ] [ thread ] [ subject ] [ author ] More information about the SC-04-08 mailing list Expand 000155.html 7 KB HTML - Click to view Options Copy Download Link [SC-04-08] WG meeting report New York 2013-10, MLS pre { white-space: pre-wrap; /* css-2.1, curent FF, Opera, Safari */ } [SC-04-08] WG meeting report New York 2013-10, MLS Smithers, Michael MZS at dolby.com Wed Jan 15 14:25:31 MST 2014 Previous message: [SC-04-08] WG meeting report New York 2013-10, MLS Next message: [SC-04-08] WG meeting report New York 2013-10, MLS Messages sorted by: [ date ] [ thread ] [ subject ] [ author ] Hi, If the MLS sequence is going through a shaping filter to make it more pink, then it begs the question, why start with MLS at all? If there are requirements for a spectral shape (say "pink"), a maximum peak-to-RMS and a particular length in samples, one can simply start with the desired magnitude response and random phase, then iterate between time and frequency domains - FFT/IFFT and at each cycle nudging the magnitude frequency response and maximum sample - until the requirements are met (within some tolerances). This results in a cyclic time signal with a fairly low or well controlled peak-to-rms level (and thus good SNR) and with the desired spectrum. My apologies if I am missing other constraints with regards the proposal. Best regards, Michael Smithers Senior Staff Engineer, Research Dolby Australia, Pty. Ltd. Level 3, 35-51 Mitchell St. McMahons Point, NSW 2060, Australia T +61-2-9101-7981 F +61-2-9101-7800 M +61-402-349-774 www.dolby.com | michael.smithers at dolby.com -----Original Message----- From: sc-04-08-bounces at standards.aes.org [mailto:sc-04-08-bounces at standards.aes.org] On Behalf Of Andreas Silzle Sent: Wednesday, January 15, 2014 10:10 AM To: Working group on Measurement and equalization of sound systems in rooms Subject: Re: [SC-04-08] WG meeting report New York 2013-10, MLS Hi, Am 15.01.2014 18:28, schrieb John Woodgate: > In message <52D6C1A8.103 at iis.fraunhofer.de>, dated Wed, 15 Jan 2014, > Andreas Silzle <andreas.silzle at iis.fraunhofer.de> writes: > >> Please find attached a paper about the difficulties of MLS measurements. > > Attachments are not permitted. Sorry, this I have not known. Here the reference: Krarup Olesen, S., et al. An improved MLS measurement system for acquiring room impulse responses. in NORSIG 2000, IEEE Nordic Signal Processing Symposium. 2000. Kolmården, Sweden. > > The proposal is to derive a predictable pink-noise signal from an MLS, > not to use an MLS directly. > >> The cyclic error is one problem, > > What is a 'cyclic error'? When you repeat the same MLS sequence several times, you get in the measurement result spikes in the time domain above the normal noise floor, when you have the slightest distortion of e.g. the loudspeaker in the processing chain. You can reduce this by using different MLS sequences of the same length, see paper. > >> the high sensitivity to create distortion in the reproduction chain >> by the rectangular form of the signal is the next. > > After the signal has been through the 'pinking' filter, it no longer > has a rectangular form. Ok, then it is maybe less critical. It should be tested and still compared to the sine sweep method. > >> With the sine sweep method you can separate the response from the >> distortion and select by a window the response alone. > > There should not be significant distortion during the measurement. If > there is, it is being done at too high a level. There is the question, what SNR do you need for your measurement results. For room impulse responses for convolution you need something greater than 70dB and this is difficult to achieve with MLS. To get a frequency response to setup an EQ less is enough. Best regards Andreas -- Dr.-Ing. Andreas Silzle Senior Scientist Fraunhofer IIS Laboratories phone: +49 9131 776-6248 Am Wolfsmantel 33 fax: +49 9131 776-6099 D-91058 Erlangen, Germany andreas.silzle at iis.fraunhofer.de _______________________________________________ SC-04-08 mailing list SC-04-08 at standards.aes.org <http://standards.aes.org/sc.cfm?ID=91> Previous message: [SC-04-08] WG meeting report New York 2013-10, MLS Next message: [SC-04-08] WG meeting report New York 2013-10, MLS Messages sorted by: [ date ] [ thread ] [ subject ] [ author ] More information about the SC-04-08 mailing list Expand 000156.html 4 KB HTML - Click to view Options Copy Download Link [SC-04-08] WG meeting report New York 2013-10, MLS pre { white-space: pre-wrap; /* css-2.1, curent FF, Opera, Safari */ } [SC-04-08] WG meeting report New York 2013-10, MLS John Woodgate jmw at jmwa.demon.co.uk Wed Jan 15 14:44:06 MST 2014 Previous message: [SC-04-08] WG meeting report New York 2013-10, MLS Next message: [SC-04-08] WG meeting report New York 2013-10, MLS Messages sorted by: [ date ] [ thread ] [ subject ] [ author ] In message <6520235A3394884B8601431AC44CF6BE69989DB0 at DLB-XCHPW01.dolby.net>, dated Wed, 15 Jan 2014, "Smithers, Michael" <MZS at dolby.com> writes: >If the MLS sequence is going through a shaping filter to make it more >pink, then it begs the question, why start with MLS at all? Because it's much better controlled in all characteristics than other noise sources. > >If there are requirements for a spectral shape (say "pink"), a maximum >peak-to-RMS and a particular length in samples, one can simply start >with the desired magnitude response and random phase, then iterate >between time and frequency domains - FFT/IFFT and at each cycle nudging >the magnitude frequency response and maximum sample - until the >requirements are met (within some tolerances). This results in a cyclic >time signal with a fairly low or well controlled peak-to-rms level (and >thus good SNR) and with the desired spectrum. Are you claiming that that is easier than passing an MLS through a pinking filter? All (well, almost all) opinion is that the crest factor should be close to 4. For the purpose for which the signal is required, SNR is not critical. Please look at the project documents on the web site. -- OOO - Own Opinions Only. With best wishes. See www.jmwa.demon.co.uk Nondum ex silvis sumus John Woodgate, J M Woodgate and Associates, Rayleigh, Essex UK Previous message: [SC-04-08] WG meeting report New York 2013-10, MLS Next message: [SC-04-08] WG meeting report New York 2013-10, MLS Messages sorted by: [ date ] [ thread ] [ subject ] [ author ] More information about the SC-04-08 mailing list Expand 000157.html 8 KB HTML - Click to view Options Copy Download Link [SC-04-08] WG meeting report New York 2013-10, MLS pre { white-space: pre-wrap; /* css-2.1, curent FF, Opera, Safari */ } [SC-04-08] WG meeting report New York 2013-10, MLS David Josephson dlj at josephson.com Wed Jan 15 15:05:26 MST 2014 Previous message: [SC-04-08] WG meeting report New York 2013-10, MLS Next message: [SC-04-08] Uploads Messages sorted by: [ date ] [ thread ] [ subject ] [ author ] Perhaps the discussion needs to be re-centered once again, since many participants have not been following it since the beginning.This is NOT about MLS, and it is NOT about room characterization where SNR is important except in the most basic way. We are looking for a signal where SNR=0, we are just generating noise! Please refer to http://www.aes.org/standards/meetings/aes135-sc-04-08-report.cfm Under consideration is one item, which pertains to AES-X215 and -X216. We are working with SMPTE to define a standard, preferably with an algorithmic base rather than a preferred circuit diagram, that will permit a uniform pink noise to be generated. We all know the limitations of pink noise as a test signal, but that consideration is not part of this discussion. We have considered an MLS as an easily specified signal that can be generated with a simple algorithm and then "pinked" with a suitable filter, also easily defined with sample code and a sample output file. Any other random signal would be OK too, if we can reach consensus on it. Regards David Josephson Chair, SC-04 > Hi, > > If the MLS sequence is going through a shaping filter to make it more pink, then it begs the question, why start with MLS at all? > > If there are requirements for a spectral shape (say "pink"), a maximum peak-to-RMS and a particular length in samples, one can simply start with the desired magnitude response and random phase, then iterate between time and frequency domains - FFT/IFFT and at each cycle nudging the magnitude frequency response and maximum sample - until the requirements are met (within some tolerances). This results in a cyclic time signal with a fairly low or well controlled peak-to-rms level (and thus good SNR) and with the desired spectrum. > > My apologies if I am missing other constraints with regards the proposal. > > Best regards, > > Michael Smithers > Senior Staff Engineer, Research > Dolby Australia, Pty. Ltd. > Level 3, 35-51 Mitchell St. > McMahons Point, NSW 2060, Australia > T +61-2-9101-7981 F +61-2-9101-7800 M +61-402-349-774 > www.dolby.com | michael.smithers at dolby.com > > > -----Original Message----- > From: sc-04-08-bounces at standards.aes.org [mailto:sc-04-08-bounces at standards.aes.org] On Behalf Of Andreas Silzle > Sent: Wednesday, January 15, 2014 10:10 AM > To: Working group on Measurement and equalization of sound systems in rooms > Subject: Re: [SC-04-08] WG meeting report New York 2013-10, MLS > > Hi, > > Am 15.01.2014 18:28, schrieb John Woodgate: >> In message <52D6C1A8.103 at iis.fraunhofer.de>, dated Wed, 15 Jan 2014, >> Andreas Silzle <andreas.silzle at iis.fraunhofer.de> writes: >> >>> Please find attached a paper about the difficulties of MLS measurements. >> Attachments are not permitted. > Sorry, this I have not known. > Here the reference: > Krarup Olesen, S., et al. An improved MLS measurement system for acquiring room impulse responses. in NORSIG 2000, IEEE Nordic Signal Processing Symposium. 2000. Kolmården, Sweden. >> The proposal is to derive a predictable pink-noise signal from an MLS, >> not to use an MLS directly. >> >>> The cyclic error is one problem, >> What is a 'cyclic error'? > When you repeat the same MLS sequence several times, you get in the measurement result spikes in the time domain above the normal noise floor, when you have the slightest distortion of e.g. the loudspeaker in the processing chain. You can reduce this by using different MLS sequences of the same length, see paper. >>> the high sensitivity to create distortion in the reproduction chain >>> by the rectangular form of the signal is the next. >> After the signal has been through the 'pinking' filter, it no longer >> has a rectangular form. > Ok, then it is maybe less critical. It should be tested and still compared to the sine sweep method. >>> With the sine sweep method you can separate the response from the >>> distortion and select by a window the response alone. >> There should not be significant distortion during the measurement. If >> there is, it is being done at too high a level. > There is the question, what SNR do you need for your measurement results. > For room impulse responses for convolution you need something greater than 70dB and this is difficult to achieve with MLS. > To get a frequency response to setup an EQ less is enough. > > Best regards > > Andreas > Previous message: [SC-04-08] WG meeting report New York 2013-10, MLS Next message: [SC-04-08] Uploads Messages sorted by: [ date ] [ thread ] [ subject ] [ author ] More information about the SC-04-08 mailing list Expand 000158.html 7 KB HTML - Click to view Options Copy Download Link [SC-04-08] WG meeting report New York 2013-10, MLS pre { white-space: pre-wrap; /* css-2.1, curent FF, Opera, Safari */ } [SC-04-08] WG meeting report New York 2013-10, MLS Thomas Lagö thomaswowwe at gmail.com Wed Jan 15 11:25:00 MST 2014 Previous message: [SC-04-08] WG meeting report New York 2013-10, MLS Next message: [SC-04-08] WG meeting report New York 2013-10, MLS Messages sorted by: [ date ] [ thread ] [ subject ] [ author ] Dear All, It seems that this discussion never ends. I developed a unique MLS-based RT60 system already in 1979 and many of the comments made are based in the wrong synchronisation methods in between source and data collection. I used a real-time convolution. Also, the discussion is going in “circles” since “non-linear” is mixed with a number of other parameters that should be used carefully. When things are brought out of their context, we quickly end up with with opinions and comments that are discussing different topics simultaneously. How can we make this stop and come to some common understanding on what it is all about? Best regards, Dr. Thomas L. Lagö QirraSound Technologies Europe AB Chief Scientist & Chairman thomaslagoqirra at gmail.com On 15 Jan 2014, at 19:09, Andreas Silzle <andreas.silzle at iis.fraunhofer.de> wrote: > Hi, > > Am 15.01.2014 18:28, schrieb John Woodgate: >> In message <52D6C1A8.103 at iis.fraunhofer.de>, dated Wed, 15 Jan 2014, Andreas Silzle <andreas.silzle at iis.fraunhofer.de> writes: >> >>> Please find attached a paper about the difficulties of MLS measurements. >> >> Attachments are not permitted. > Sorry, this I have not known. > Here the reference: > Krarup Olesen, S., et al. An improved MLS measurement system for acquiring room impulse responses. in NORSIG 2000, IEEE Nordic Signal Processing Symposium. 2000. Kolmården, Sweden. >> >> The proposal is to derive a predictable pink-noise signal from an MLS, not to use an MLS directly. >> >>> The cyclic error is one problem, >> >> What is a 'cyclic error'? > When you repeat the same MLS sequence several times, you get in the measurement result spikes in the time domain above the normal noise floor, when you have the slightest distortion of e.g. the loudspeaker in the processing chain. You can reduce this by using different MLS sequences of the same length, see paper. >> >>> the high sensitivity to create distortion in the reproduction chain by the rectangular form of the signal is the next. >> >> After the signal has been through the 'pinking' filter, it no longer has a rectangular form. > Ok, then it is maybe less critical. It should be tested and still compared to the sine sweep method. >> >>> With the sine sweep method you can separate the response from the distortion and select by a window the response alone. >> >> There should not be significant distortion during the measurement. If there is, it is being done at too high a level. > There is the question, what SNR do you need for your measurement results. > For room impulse responses for convolution you need something greater than 70dB and this is difficult to achieve with MLS. > To get a frequency response to setup an EQ less is enough. > > Best regards > > Andreas > > -- > Dr.-Ing. Andreas Silzle Senior Scientist > Fraunhofer IIS Laboratories phone: +49 9131 776-6248 > Am Wolfsmantel 33 fax: +49 9131 776-6099 > D-91058 Erlangen, Germany andreas.silzle at iis.fraunhofer.de > > _______________________________________________ > SC-04-08 mailing list > SC-04-08 at standards.aes.org > <http://standards.aes.org/sc.cfm?ID=91> > Previous message: [SC-04-08] WG meeting report New York 2013-10, MLS Next message: [SC-04-08] WG meeting report New York 2013-10, MLS Messages sorted by: [ date ] [ thread ] [ subject ] [ author ] More information about the SC-04-08 mailing list Expand 000159.html 9 KB HTML - Click to view Options Copy Download Link [SC-04-08] WG meeting report New York 2013-10, MLS pre { white-space: pre-wrap; /* css-2.1, curent FF, Opera, Safari */ } [SC-04-08] WG meeting report New York 2013-10, MLS Thomas Lagö thomaswowwe at gmail.com Wed Jan 15 14:58:28 MST 2014 Previous message: [SC-04-08] WG meeting report New York 2013-10, MLS Next message: [SC-04-08] WG meeting report New York 2013-10, MLS Messages sorted by: [ date ] [ thread ] [ subject ] [ author ] Hi, That filter is NOT the question. If one is trying to make a “Noise Generator” without consideration of the "sync issue” MLS is not necessarily the answer. All “end of sequence” etc will play a major role. You CANNOT use an FFT together with an MLS since its length is 2^N-1 and the FFT is 2^N. The “-1” makes ALL THE DIFFERENCE! In the discussions, all these issues are “mixed” and the it becomes “impossible”. How do we straighten it! Best regards, Dr. Thomas L. Lagö QirraSound Technologies Europe AB Chief Scientist & Chairman thomaslagoqirra at gmail.com On 15 Jan 2014, at 22:25, Smithers, Michael <MZS at dolby.com> wrote: > Hi, > > If the MLS sequence is going through a shaping filter to make it more pink, then it begs the question, why start with MLS at all? > > If there are requirements for a spectral shape (say "pink"), a maximum peak-to-RMS and a particular length in samples, one can simply start with the desired magnitude response and random phase, then iterate between time and frequency domains - FFT/IFFT and at each cycle nudging the magnitude frequency response and maximum sample - until the requirements are met (within some tolerances). This results in a cyclic time signal with a fairly low or well controlled peak-to-rms level (and thus good SNR) and with the desired spectrum. > > My apologies if I am missing other constraints with regards the proposal. > > Best regards, > > Michael Smithers > Senior Staff Engineer, Research > Dolby Australia, Pty. Ltd. > Level 3, 35-51 Mitchell St. > McMahons Point, NSW 2060, Australia > T +61-2-9101-7981 F +61-2-9101-7800 M +61-402-349-774 > www.dolby.com | michael.smithers at dolby.com > > > -----Original Message----- > From: sc-04-08-bounces at standards.aes.org [mailto:sc-04-08-bounces at standards.aes.org] On Behalf Of Andreas Silzle > Sent: Wednesday, January 15, 2014 10:10 AM > To: Working group on Measurement and equalization of sound systems in rooms > Subject: Re: [SC-04-08] WG meeting report New York 2013-10, MLS > > Hi, > > Am 15.01.2014 18:28, schrieb John Woodgate: >> In message <52D6C1A8.103 at iis.fraunhofer.de>, dated Wed, 15 Jan 2014, >> Andreas Silzle <andreas.silzle at iis.fraunhofer.de> writes: >> >>> Please find attached a paper about the difficulties of MLS measurements. >> >> Attachments are not permitted. > Sorry, this I have not known. > Here the reference: > Krarup Olesen, S., et al. An improved MLS measurement system for acquiring room impulse responses. in NORSIG 2000, IEEE Nordic Signal Processing Symposium. 2000. Kolmården, Sweden. >> >> The proposal is to derive a predictable pink-noise signal from an MLS, >> not to use an MLS directly. >> >>> The cyclic error is one problem, >> >> What is a 'cyclic error'? > When you repeat the same MLS sequence several times, you get in the measurement result spikes in the time domain above the normal noise floor, when you have the slightest distortion of e.g. the loudspeaker in the processing chain. You can reduce this by using different MLS sequences of the same length, see paper. >> >>> the high sensitivity to create distortion in the reproduction chain >>> by the rectangular form of the signal is the next. >> >> After the signal has been through the 'pinking' filter, it no longer >> has a rectangular form. > Ok, then it is maybe less critical. It should be tested and still compared to the sine sweep method. >> >>> With the sine sweep method you can separate the response from the >>> distortion and select by a window the response alone. >> >> There should not be significant distortion during the measurement. If >> there is, it is being done at too high a level. > There is the question, what SNR do you need for your measurement results. > For room impulse responses for convolution you need something greater than 70dB and this is difficult to achieve with MLS. > To get a frequency response to setup an EQ less is enough. > > Best regards > > Andreas > > -- > Dr.-Ing. Andreas Silzle Senior Scientist > Fraunhofer IIS Laboratories phone: +49 9131 776-6248 > Am Wolfsmantel 33 fax: +49 9131 776-6099 > D-91058 Erlangen, Germany andreas.silzle at iis.fraunhofer.de > > _______________________________________________ > SC-04-08 mailing list > SC-04-08 at standards.aes.org > <http://standards.aes.org/sc.cfm?ID=91> > _______________________________________________ > SC-04-08 mailing list > SC-04-08 at standards.aes.org > <http://standards.aes.org/sc.cfm?ID=91> > Previous message: [SC-04-08] WG meeting report New York 2013-10, MLS Next message: [SC-04-08] WG meeting report New York 2013-10, MLS Messages sorted by: [ date ] [ thread ] [ subject ] [ author ] More information about the SC-04-08 mailing list Expand 000160.html 11 KB HTML - Click to view Options Copy Download Link [SC-04-08] WG meeting report New York 2013-10, MLS pre { white-space: pre-wrap; /* css-2.1, curent FF, Opera, Safari */ } [SC-04-08] WG meeting report New York 2013-10, MLS Smithers, Michael MZS at dolby.com Thu Jan 16 14:50:48 MST 2014 Previous message: [SC-04-08] WG meeting report New York 2013-10, MLS Next message: [SC-04-08] WG meeting report New York 2013-10, MLS Messages sorted by: [ date ] [ thread ] [ subject ] [ author ] Hi, >> That filter is NOT the question. Agreed. >> If one is trying to make a "Noise Generator" without consideration of the "sync issue" MLS is not necessarily the answer. Agreed. If synchronization is not an issue, for example in the dBC level calibration measurements in cinema, then a random number generator through a pinking filter is reasonable and easy to generate. There is a question of the PDF of the noise generator which I'll discuss later. If synchronization is a requirement, then MLS is problematic for 2^N based analysers. One can resample the 2^N-1 signal to 2^N but some of the special properties of MLS are gone. Passing the MLS signal through a pinking filter also changes the properties of the sequence. If multiple but different MLS sequences are concatenated, then one may as well start with with a much larger N. The iterative signal generating technique I described earlier results in a cyclic noise sequence that holds its spectral magnitude properties regardless of capture sample alignment. It's time sample PDF is ~Gaussian and it's crest factor is controlled. On the question of the PDF of the time signal, I think this is very important to nail down since different PDF's may result in variations in voice coil heating, and thus power compression, measured SPL level and possibly SPL level variation between drivers in a multiway system. I have seen variations in PDF's: for example pink noise from "Sound Forge" is a little narrow in PDF than Smaart V5. How much variation, I don't know. Are there any MIPS/memory constraints with regard to the generation of the standardized signal? BTW I'm roughly familiar with the related work in SMPTE. Best regards, Michael -----Original Message----- From: sc-04-08-bounces at standards.aes.org [mailto:sc-04-08-bounces at standards.aes.org] On Behalf Of Thomas Lagö Sent: Wednesday, January 15, 2014 1:58 PM To: Working group on Measurement and equalization of sound systems in rooms Subject: Re: [SC-04-08] WG meeting report New York 2013-10, MLS Hi, That filter is NOT the question. If one is trying to make a "Noise Generator" without consideration of the "sync issue" MLS is not necessarily the answer. All "end of sequence" etc will play a major role. You CANNOT use an FFT together with an MLS since its length is 2^N-1 and the FFT is 2^N. The "-1" makes ALL THE DIFFERENCE! In the discussions, all these issues are "mixed" and the it becomes "impossible". How do we straighten it! Best regards, Dr. Thomas L. Lagö QirraSound Technologies Europe AB Chief Scientist & Chairman thomaslagoqirra at gmail.com On 15 Jan 2014, at 22:25, Smithers, Michael <MZS at dolby.com> wrote: > Hi, > > If the MLS sequence is going through a shaping filter to make it more pink, then it begs the question, why start with MLS at all? > > If there are requirements for a spectral shape (say "pink"), a maximum peak-to-RMS and a particular length in samples, one can simply start with the desired magnitude response and random phase, then iterate between time and frequency domains - FFT/IFFT and at each cycle nudging the magnitude frequency response and maximum sample - until the requirements are met (within some tolerances). This results in a cyclic time signal with a fairly low or well controlled peak-to-rms level (and thus good SNR) and with the desired spectrum. > > My apologies if I am missing other constraints with regards the proposal. > > Best regards, > > Michael Smithers > Senior Staff Engineer, Research > Dolby Australia, Pty. Ltd. > Level 3, 35-51 Mitchell St. > McMahons Point, NSW 2060, Australia > T +61-2-9101-7981 F +61-2-9101-7800 M +61-402-349-774 > www.dolby.com | michael.smithers at dolby.com > > > -----Original Message----- > From: sc-04-08-bounces at standards.aes.org > [mailto:sc-04-08-bounces at standards.aes.org] On Behalf Of Andreas > Silzle > Sent: Wednesday, January 15, 2014 10:10 AM > To: Working group on Measurement and equalization of sound systems in > rooms > Subject: Re: [SC-04-08] WG meeting report New York 2013-10, MLS > > Hi, > > Am 15.01.2014 18:28, schrieb John Woodgate: >> In message <52D6C1A8.103 at iis.fraunhofer.de>, dated Wed, 15 Jan 2014, >> Andreas Silzle <andreas.silzle at iis.fraunhofer.de> writes: >> >>> Please find attached a paper about the difficulties of MLS measurements. >> >> Attachments are not permitted. > Sorry, this I have not known. > Here the reference: > Krarup Olesen, S., et al. An improved MLS measurement system for acquiring room impulse responses. in NORSIG 2000, IEEE Nordic Signal Processing Symposium. 2000. Kolmården, Sweden. >> >> The proposal is to derive a predictable pink-noise signal from an >> MLS, not to use an MLS directly. >> >>> The cyclic error is one problem, >> >> What is a 'cyclic error'? > When you repeat the same MLS sequence several times, you get in the measurement result spikes in the time domain above the normal noise floor, when you have the slightest distortion of e.g. the loudspeaker in the processing chain. You can reduce this by using different MLS sequences of the same length, see paper. >> >>> the high sensitivity to create distortion in the reproduction chain >>> by the rectangular form of the signal is the next. >> >> After the signal has been through the 'pinking' filter, it no longer >> has a rectangular form. > Ok, then it is maybe less critical. It should be tested and still compared to the sine sweep method. >> >>> With the sine sweep method you can separate the response from the >>> distortion and select by a window the response alone. >> >> There should not be significant distortion during the measurement. If >> there is, it is being done at too high a level. > There is the question, what SNR do you need for your measurement results. > For room impulse responses for convolution you need something greater than 70dB and this is difficult to achieve with MLS. > To get a frequency response to setup an EQ less is enough. > > Best regards > > Andreas > > -- > Dr.-Ing. Andreas Silzle Senior Scientist > Fraunhofer IIS Laboratories phone: +49 9131 776-6248 > Am Wolfsmantel 33 fax: +49 9131 776-6099 > D-91058 Erlangen, Germany andreas.silzle at iis.fraunhofer.de > > _______________________________________________ > SC-04-08 mailing list > SC-04-08 at standards.aes.org > <<A HREF="http://sta... Expand This file was truncated for preview. Please download to view the full file. 0 Comments Public All Members My Connections Only Me PublicAll MembersMy ConnectionsOnly Me Public All Members My Connections Only Me