Loading ...
Sorry, an error occurred while loading the content.

Good sound

Expand Messages
  • Robert
    Re Roman Z;s remarks: I suppose I am somewhat worried about the phrase good sound . This seems to me a bit of a moving target when one gets down to details of
    Message 1 of 1 , Oct 29, 2010
    • 0 Attachment
      Re Roman Z;s remarks:
      I suppose I am somewhat worried about the phrase "good sound".
      This seems to me a bit of a moving target when one gets
      down to details of what happens.
      In a sense, we know what bad sound is:bad sound is a big bass boom followed by a big hole in the lower mids, a projection around 2k,
      and a rising top end. It is in short what all too many speakers
      exhibit, including a great many expensive ones.
      Looking at Sphiles in room measurements gives one a somewhat
      jaundiced view to say the least of what is actually going on
      in audio.

      But once these obvious problems are solved in general terms,
      then it is somewhat ambiguous what ought to happen. This is the point I have been trying to make a lot recently(and I am afraid
      repeatedly , to the point of some tedium).

      Now the thing to note, to my mind, about bad sound as I have
      described it is that it is both easily measured in the lower frequencies and very easily heard, too. It is not too much of a trick
      to figure out what gives. Really the sub 300 area can be done with a SPL meter and some warble tones, pretty well. Certainly it is the case that a great many of the speakers measured in Sphile in room are making errors of this type, that would be easy to detect that way.
      Rives did not suggest this for nothing!

      And if you measure the speakers close up, the same it true pretty much for higher up. The big errors are easy to determine.

      What is confusing is that when you get rid of the big errors,
      exactly what to do begins to be ambiguous as noted. I do not
      see much objection to getting the speaker anechoic flat by DSP
      as a base line and then fine tuning the effects of early reflections and so on. But I would be reluctant to see this done by fiat, the fine tuning part.

      I know this is frustrating from the engineering viewpoint. But the situation is very complicated and to model it correctly would
      involve an enormous number of measurements at different points
      and a lot of processing of the resulting data.

      Moreover, it would involve eschewing correcting things with high resolution in the frequency domain that were spatially unstable.

      Let me describe an experiment. Take an automated system with high res in the frequency domain that shows you measurents. Insert in the measurement set up also a 1/3 octave EQ device. Adjust the 1/3 octave EQ so that the resulting measurement curve is as close to the target curve as possible. Then do the correction on the resulting system --so that no broadband corrections are actually happening.

      One will still hear some change when the DSP correction is applied.
      But will it sound better in any reasonable sense?

      It might if it does speaker correction alone(like the Essex unit).
      In fact it probably will. But if it does one point correction
      at the listening position--ummm, I think whether it will be better
      is highly debatable. Of course if you KNOW which one is which,
      you will probably convince yourself. But if you do not....

      Anyway, people should try this. In principle the high res
      correction is more accurate--at one point. But one point is all it is.
      And you do not hear sound according to a one point measurement.
      Moreover, you do tend to hear sound in 1/3 octave bands(this is the critical bandwidth of the ear). The one exception I think is that peaks which happen over areas of space--which almost always arise in the higher frequencies from anechoic errors of the speakers--
      benefit from being smoothed out specifically with high resolution.
      But this is Essex style correction of the speaker, not room correction. And it is better done using anechoic measurements.

      Look at those Soundstage NRC measurements. The on axis errors mostly persist out to 15 or even 30 degrees. Not much is totally unstable.
      And it is these persistent errors that one hears.

      In short, anechoic measurements over a spatially distributed measurement window(albeit not a super wide one) is what one ought to correct first in the higher frequencies and then one ought to fine tune by ear.

      Bass--that works ok most ways one does it, but even that is unstable
      to the point that it makes little sense to correct it with high res.
      Look at those measurements 6 inches apart in my digital room correction survey! (end of my digital correction survey).

      Good sound--I suppose we all want it. But it is an elusive idea.

    Your message has been successfully submitted and would be delivered to recipients shortly.