Consideraciones acerca de medición y ecualización

Aquí otro comentario, esta vez de Sean Olive:

I am the main author of study so I can respond to the questions.

This was a study that compared different room corrections implemented by different products applied to the same loudspeaker system: a B&W N802 and a JBL 1500 Subwoofer. The final curves you see are based on the results after the products measured, processed and applied their corrections to the speakers. We followed the manufacturers' instructions to the letter, and called one of them to get advice and clarification for best results. In some cases, we manually intervened because the product decided to cross over the subwoofer at say 150 Hz in stead of 80 Hz because the algorithm decided the B&W N802 didn't have sufficient bass to be crossed over at 80 Hz (clearly this was not the case). Some of the differences in low frequency roll-offs are based on automated decisions within the algorithm based on the output capability of the subwoofer.

We tried to be as objective as possible, and not intervene too much with how the room correction was implemented. If you read the paper we came conclusion that none of the products were refined to the point that you can rely on them to always make good decisions, and therefore some form of intervention is necessary to get the best results. That means the person doing the room correction needs some knowledge and expertise..not your typical consumer.

Some products did poorly because of the target curve they chose to use, others because they chose to do full-band EQ making the on-axis sound much worse after EQ due to filling in spectral holes in the sound power response. If a speaker has constant directivity then the on-axis/sound power response generally has similar bumps and dips and you can have you cake and eat it too. This is not the case with the B&W, which has a decent on-axis response but a less-than-perfect sound power response along with a bumpy DI. This speaker is particularly problematic for full-band blind room correction EQ as demonstrated in this study.

The lesson learned: if you start out with a decent loudspeaker with good on-axis, smooth sound power and directivity there is nothing you can to do improve it by equalizing above 500 Hz or so (below 500 Hz you have the potential to truly improve the sound at the listening seat by compensating for loudspeaker-room interactions). On the other hand, doesn't have these qualities you can do considerable harm to the speaker's sound by equalizing above 500 Hz as shown in this study. The other lesson learned is that the target curve is an important consideration.

Cheers
Sean Olive,
Senior Research Fellow, Harman X

Floyd Toole:

I don't intend to step on any response that my old friend and research colleague Sean will have to say about this, but there is a fundamental point that you have obviously missed if you have followed this thread from the beginning. There is the direct sound - the first arrival from the loudspeaker which is not - not - influenced by the room above the transition frequency, and then there is the steady-state sound that is made up of the direct sound plus reflections from off-axis sound radiated by the loudspeaker. The direct sound dictates localization - imaging, and is the dominant influence on timbre - sound quality. Reflected sounds are not irrelevant, adding mostly to spatial imaging effects, and contributing secondarily to sound quality. Over most of the frequency range the measurement by an omnidirectional microphone is a steady-state measure: a poor substitute for two ears and a brain - all of which are engaged in the perception of soundstage and imaging. This is a simplification, but it is an essential starting point to understanding why various room EQ systems differ in their results and why, unless one has detailed anechoic data on the loudspeakers, room curves cannot be interpreted.

The claim that Room EQ can turn any loudspeaker in any room into a flawless system is rubbish, but there are people who believe this. This paper, and my book explain it: Toole, F. E. (2015). “The Measurement and Calibration of Sound Reproducing Systems”, J. Audio Eng. Soc., vol. 63, pp.512-541. This is an open-access paper available to non-members at www.aes.org AES E-Library » The Measurement and Calibration of Sound Reproducing Systems

Floyd Toole:

Almost always a cancellation is a "local" phenomenon, occurring only at one or a few locations. Filling such a dip adds a boost to the direct and all other sounds radiated from the loudspeaker, making it a worse loudspeaker and filling the room with excessive sound at the boosted frequency. All such acoustical interference dips, and peaks, are non-minimum phase phenomena and cannot be corrected by EQ. It is the principal reason why room curves are not definitive descriptors of sound quality.

EDIT: I should add that these non-minimum-phase acoustical interference ripples in room curves are almost always not audible problems. They may look like "comb filtering" but because the direct and delayed sounds come from different directions two ears and a brain perceive them as spaciousness, not coloration. Equalizing them flat/smooth is a mistake - it results in a degradation of the direct sound, which is the most important event.

Bueno y aquí una pregunta directa:

Quote:
Originally Posted by Deboi

Are these negative considerations about Room EQ above the Schroeder frequency also valid for manual EQ (measuring the response in the MLP with an omnidirectional microphone)?



The "negative considerations" about broadband EQ apply to all measurements, but the ones you do yourself may be the most trustworthy. And the MLP is where performance matters most.

The first reason for the "negativity" is well described in my book. It shows that steady-state room curves are highly predictable from anechoic data - it is the off-axis radiation - early reflections - that dominate the curve shape. If there are irregularities in the room curve caused by an otherwise well designed loudspeaker having amplitude response problems, EQ can help. This is a rare event, as most competent designers aim for flattish, smooth on-axis response. If, as is more common, an irregularity is caused by frequency dependent directivity problems, EQ cannot fix it - you need a better loudspeaker, or an acoustically dead room.

The second reason involves discrete reflections that acoustically interfere at the measuring location, generating ripples. If these ripples are smoothed by EQ you have changed the on-axis - direct sound - performance of the loudspeaker which is the most important starting point for good sound. Two ears and a brain react very differently to acoustical interference than an omni mic. Humans distinguish between direct and reflected sounds on the basis of direction of arrival and timing. The mic and analyzer do not.

Iniciado por Deboi

Floyd Toole:

Almost always a cancellation is a "local" phenomenon, occurring only at one or a few locations. Filling such a dip adds a boost to the direct and all other sounds radiated from the loudspeaker, making it a worse loudspeaker and filling the room with excessive sound at the boosted frequency. All such acoustical interference dips, and peaks, are non-minimum phase phenomena and cannot be corrected by EQ. It is the principal reason why room curves are not definitive descriptors of sound quality.

EDIT: I should add that these non-minimum-phase acoustical interference ripples in room curves are almost always not audible problems. They may look like "comb filtering" but because the direct and delayed sounds come from different directions two ears and a brain perceive them as spaciousness, not coloration. Equalizing them flat/smooth is a mistake - it results in a degradation of the direct sound, which is the most important event.

Lo que no dice es que si muestras un RTA de 1/3 o 1/6 octava y desprecias los valles y picos bruscos en frecuencias colindantes al ecualizar, en realidad bien poco vas a corregir que sea de fase no mínima (menos aún, si la sala está tratada = casi todo lo que te llegue en la zona media y alta a 1/3 o 1/6 será prácticamente la respuesta anecoica de la caja)


Un saludete

Iniciado por Deboi

Bueno y aquí una pregunta directa:

Quote:
Originally Posted by Deboi

Are these negative considerations about Room EQ above the Schroeder frequency also valid for manual EQ (measuring the response in the MLP with an omnidirectional microphone)?



The "negative considerations" about broadband EQ apply to all measurements, but the ones you do yourself may be the most trustworthy. And the MLP is where performance matters most.

The first reason for the "negativity" is well described in my book. It shows that steady-state room curves are highly predictable from anechoic data - it is the off-axis radiation - early reflections - that dominate the curve shape. If there are irregularities in the room curve caused by an otherwise well designed loudspeaker having amplitude response problems, EQ can help. This is a rare event, as most competent designers aim for flattish, smooth on-axis response If, as is more common, an irregularity is caused by frequency dependent directivity problems, EQ cannot fix it - you need a better loudspeaker, or an acoustically dead room. .

The second reason involves discrete reflections that acoustically interfere at the measuring location, generating ripples. If these ripples are smoothed by EQ you have changed the on-axis - direct sound - performance of the loudspeaker which is the most important starting point for good sound. Two ears and a brain react very differently to acoustical interference than an omni mic. Humans distinguish between direct and reflected sounds on the basis of direction of arrival and timing. The mic and analyzer do not.

Aquí ya saca un poco la "doble vara de medir" que en ocasiones reluce con ese:

"Si hay irregularidades en la curva de la sala causadas por un altavoz bien diseñado que tiene problemas de respuesta de amplitud, EQ puede ayudar. Este es un evento raro, ya que la mayoría de los diseñadores competentes buscan una respuesta plana y suave en el eje. Si, como es más común, una irregularidad es causada por problemas de directividad dependientes de la frecuencia, el ecualizador no puede solucionarlo: necesita un mejor altavoz o un acústico habitacion muerta . "




... porque es él el primero que sabe perfectamente que:

1- Las cajas no son planas +/-0dB... sino más bien "las buenas" suelen oscilar casi todas dentro de un típico +/-2 o +/-3dB (es decir, hay desviaciones entre máximos y mínimos que todavía oscilan entre 4dB y 6dB en la respuesta anecoica). De 4dB a 6dB de diferente intensidad es audiblemente un abismo (cualquiera que haya utilizado un EQ o haya modificado la forma de curva variando el filtro en un HUM lo sabe perfectamente... él el primero)


2- Por otro lado, los problemas causados por la directividad de la caja según su propio criterio (que por cierto, ya hemos visto en este hilo padecen el "95%" de las cajas comerciales) dice "se necesita un mejor altavoz o una sala acústica"... luego de nuevo está diciendo que para el 95% de las cajas del mercado hay que acondicionar la sala si quieres obtener alta calidad.



La otra opción es comprar uno de esos poquísimos altavoces que cumplen su criterio de directividad (aunque personalmente critico que por muy homogénea que sea esa directividad de la caja, el reflejo de las paredes en salas normales sin tratar sigue sin serlo = sigue habiendo claramente coloración tanto en los primeros ms (que afectan a la imagen y timbre) + también espacial en reflexiones más retardadas (porque el rebote no mantiene esa homogeneidad que las cajas "buenas" envían hacia las paredes en salas normales)




- En lo que sí estoy de acuerdo es en que el oído es al final el que te dice si una EQ es o no correcta (por eso mejor una EQ manual en la que puedas tocar todo poco a poco y no una que te lo haga todo de golpe según su criterio... a no ser que la automática tenga potencia suficiente como para variar en el ajuste pre-EQ la impulsiva en frecuencia según nos la zona que nos interesa sobre la que ir trabajando el convolver). De las pocas EQ automáticas que conozco tienen esa potencia está el gratuito DRCop (pero no se cómo esta el tema en multicanal). No se qué tal el software de ecualización de probablemente los monitores más utilizados y reconocidos de estudio: Genelec (software que por cierto ecualiza en toda la banda)


Un saludete

Iniciado por Deboi

Some products did poorly because of the target curve they chose to use, others because they chose to do full-band EQ making the on-axis sound much worse after EQ due to filling in spectral holes in the sound power response. If a speaker has constant directivity then the on-axis/sound power response generally has similar bumps and dips and you can have you cake and eat it too. This is not the case with the B&W, which has a decent on-axis response but a less-than-perfect sound power response along with a bumpy DI. This speaker is particularly problematic for full-band blind room correction EQ as demonstrated in this study.

En ese estudio lo que se demostró en su día es que tras ecualizar las B&W 802 por encima de 500hz (incluida la zona del bache comentado) el sonido mejoró en el blind test subjetivo

A parte, insisto en que los estudios de Floyd Toole se suelen centrar en sistemas multicanal (4.x o 5.x la gran mayoría; ya que suele ir en contra de colocar muchos más canales)... y para que cumpla no sólo en una posición fija de escucha, sino que por lo general busca en los 6 asientos/butacas que suele colocar en alrededores del centro de la sala (el criterio de una buena EQ es muy diferente si buscas buen sonido para "un único punto de escucha dulce; el típico sentado centrado en el sistema"... que si buscas prioridad se mantenga la coherencia en 6 asientos de alrededor (un espacio mucho más amplio): en éste último caso la ecualización necesaria para que en toda la banda funcione no puede ser tan agresiva como en el primer caso)



Audio Musings by Sean Olive: The Harman International Reference Listening Room

Con más detalle:




https://docs.google.com/presentation...t#slide=id.i13


Sigo sin ver tanta variedad de acústica probada: dónde están las trampas de graves y los paneles de grosor considerable colocados en las primeras reflexiones... para obtener con todo ello un RT compensado?


Porque sí, también estoy de acuerdo en que para poner paneles con esos espesores (que no actúan con suficiente eficacia hasta al menos frecuencias cercanas a la FC de la sala) mejor no poner nada porque vas a descompensar todo (de hecho, soy el primero en no recomendar ese tipo de tratamiento en los foros, porque lo he tenido y se perfectamente que ambientalmente deja la zona media y alta demasiado muerta respecto la zona media/baja y baja = muy descompensado)


Un saludete

Iniciado por Deboi

Interesante artículo acerca de la ecualización en sala:

http://www.aes.org/tmpFiles/elib/20190101/17839.pdf

Estoy intentado leerlo pero me reporta error 404, no found. Se ve que lo han quitado o cerrado. lo tendrías por otro medio?

Iniciado por kalymocho

Estoy intentado leerlo pero me reporta error 404, no found. Se ve que lo han quitado o cerrado. lo tendrías por otro medio?

Pues no sé ahora mismo exactamente qué paper es, tampoco lo tengo guardado.

Pues yo puedo abrirlo sin problemas.... igual la página web estaba caída o en mantenimiento cuando lo habéis mirado.

Un saludo del Oso