As some of you might know, occasionally I have been debating the use of large speaker arrays in Hauptwerk systems. In the most extreme cases more than 100 speakers are used, but 10-20 is more common.
There are some situations in which the use of large speaker arrays is perfectly valid. Some people are using Hauptwerk sample sets as sample libraries to build custom organs that can be optimized to their acoustic environments, e.g. church installations. In such cases, a large number of speakers is often a necessity.
However, many people use Hauptwerk to aim for a faithful reproduction of famous pipe organs in their home environment. I consider myself one of them. In this case, I have some serious doubts about large speaker arrays.
I assume most people here have a limited budget for audio of (maybe) 5000 euro's/dollars/pounds at most, probably less. Apart from a good sound card and a subwoofer, should one spend this money on a large number of not too expensive but reasonable speakers, or should one rather invest it in a single pair (or two pairs for surround) of very expensive high-quality monitors?
Buying all the equipment to make A-B comparisons is very expensive, so I think many people jump into the deep. I am attempting to assess the matter based on solid arguments before making significant investments somewhere in the future. Subjective experience might be impossible to translate from one situation to another, so therefore I am looking for objective observations. This is not an easy matter, therefory I remain open for new arguments. I do not feel qualified to give a final verdict.
My main argument against large speaker arrays is based on the observation that speaker positioning has a very, very large influence on frequency and phase response. Keep in mind that phase and especially frequency response should be as linear as possible for faithul reproduction.
Ideally, a stereo pair of near-field monitors should be placed at ear height, at about 1m distance, forming approximately an equal sided triangle with the listening position. Preferably, no objects should be placed in between, and tweeters should be directly pointed at the ears (for now, I exclude omnidirectional speakers from my analysis). If you compromise on speaker positioning, the influence of room acoustics (reflections from your walls and furniture) rapidly becomes dominant, destroying the frequency response of the speaker. Individual frequencies easily can be boosted or muted by 10db or more! The phase behaviour also quickly deteriorates. Often, excellent speakers give very poor results because people pay too little attention to positioning, or compromise because of visual esthetics.
Apart from positioning, room correction software and (if possible) hardware is very strongly recommended to get better frequency response.
When you keep adding more speakers, it will increasingly become harder to place them at or near the optimal positions (and to perform software-based room correction). At some point, every additional speaker pair will behave worse due to increasingly suboptimal positioning. You run the risk of decreasing the realism and quality of reproduction, rather than increasing it!
So why would someone use large speaker arrays? Some people have referred to an article written by Colin Pykett, http://www.pykett.org.uk/EndOfPipeOrgan.htm#Distortion. This article is worth to be carefully read and provides some powerful arguments for using large speaker arrays in electronic organs. I find this article truly enlightening, yet I think it deserves a critical approach.
The author focuses on two main causes for a lack of realism in electronic organs: intermodular distortion and electronic signal mixing. Here is my first criticism: as mentioned, I think suboptimal speaker positioning combined with the influence of room acoustics is actually the most important reason for lack of realism in many situations.
Intermodular distortion is well explained by mr Pykett. If multiple signals with different frequencies are sent through a single speaker, spurious signals with non-harmonic frequencies will be added due to non-linearities of the amplifier and speaker components. These spurious frequencies are sums and differences of (multiples of) the original frequencies. For example, a 4kHz and 5kHz signal might result in unwanted signals at (5-4)=1kHz, (5+4)=9kHz, (2*4-5)=3kHz, (2*5-4)=6kHz), etc.
Especially in organ music, we sent not just two, but many many more frequencies at the same time through the same speaker. So there will be many spurious frequencies all over the spectrum, forming a disturbing background noise that becomes worse as we play larger chords and use larger registrations.
How significant is this intermodular distortion (IMD) problem? According to mr Pykett, the situation is very bad. IMD specs are often not provided, but he thinks 1% distortion is typical for a reasonable speaker, and in appendix 2 he provides a calculation to show that the total power of unwanted IMD signals easily reaches a level of just 22 dB below the level of the original signal, if just two notes of a trumpet rank are played. This would certainly be audible.
What would be the solution? IMD does not occur if separate sources are used for each signal, and is therefore absent in real pipe organs. So what if we can route each sample to a different speaker? We would need many speakers, but the IMD problem would be eliminated. Mr Pykett suggests to use one speaker for each of the notes in a octave, so 12 speakers in total, or preferably 12 speakers per division. Because in Hauptwerk we aim to reproduce stereo samples, we would actually need 12 speaker pairs, or 24 speakers in total for a stereo sample set.
The 1% distortion figure was mr Pyketts estimation for a "reasonable" speaker. What if we use one pair of expensive, excellent speakers instead of an array of 24 "reasonable" speakers? On this forum, the Mackie HR824 (price around EUR 900 for a single unit) has been recommended for Hauptwerk. Specifications: http://mackie.com/sites/default/files/P ... MK2_SS.pdf. These specs give a figure of <0.035% for distortion, including (but not exclusively!) IMD. That's almost 30x better than Mr Pyketts 1%. According to the formula used by mr Pykett, the power of the distorted signal is a function of the square of the distortion percentage. So the Mackies will have almost 1000x (30 squared) lower IMD power (or much less!) than mr Pyketts "reasonable" speakers! This decreases the IMD level by a further 30 dB. In total, the IMD would be at least 50dB lower than the original signal for small chords. I don't think that would be very audible. For large chords, the situation could be somewhat worse, but still quite acceptable.
My preliminary conclusion is that it could be well worth to invest in expensive but excellent speakers to solve the IMD problem, instead of building a speaker array. Moreover, other aspects of speaker quality also would be improved.
What about my own set-up? I have M-Audio's DSM3 as front speakers, and BX8a as rears. I did some analyses using a measurement mic and the free software REW. I was able to detect several IMD frequencies. The individual IMD peaks were 50-60 dB lower than the original signal. I am therefore not worried about IMD in my audio set-up. I did however discover quite a number of other spurious frequencies (no harmonics or sum or difference frequencies). Although I am not sure, I think those result from wind noise in the individual organ pipes/samples. If I am right, undesirable IMD would be far less significant than desirable sampled wind noise in my set-up.
This leaves me to consider mr Pyketts second argument concerning the problems of electronic signal mixing. I don't think these problems are significant for us because per-pipe stereo sampling is used for Hauptwerk samples, but maybe I should try to analyze this issue in more detail later.
Meanwhile, I hope people find this analysis useful. Again, this long story is not meant as some final verdict, but hopefully it will be the start of a discussion which will generate deeper insights.
There are some situations in which the use of large speaker arrays is perfectly valid. Some people are using Hauptwerk sample sets as sample libraries to build custom organs that can be optimized to their acoustic environments, e.g. church installations. In such cases, a large number of speakers is often a necessity.
However, many people use Hauptwerk to aim for a faithful reproduction of famous pipe organs in their home environment. I consider myself one of them. In this case, I have some serious doubts about large speaker arrays.
I assume most people here have a limited budget for audio of (maybe) 5000 euro's/dollars/pounds at most, probably less. Apart from a good sound card and a subwoofer, should one spend this money on a large number of not too expensive but reasonable speakers, or should one rather invest it in a single pair (or two pairs for surround) of very expensive high-quality monitors?
Buying all the equipment to make A-B comparisons is very expensive, so I think many people jump into the deep. I am attempting to assess the matter based on solid arguments before making significant investments somewhere in the future. Subjective experience might be impossible to translate from one situation to another, so therefore I am looking for objective observations. This is not an easy matter, therefory I remain open for new arguments. I do not feel qualified to give a final verdict.
My main argument against large speaker arrays is based on the observation that speaker positioning has a very, very large influence on frequency and phase response. Keep in mind that phase and especially frequency response should be as linear as possible for faithul reproduction.
Ideally, a stereo pair of near-field monitors should be placed at ear height, at about 1m distance, forming approximately an equal sided triangle with the listening position. Preferably, no objects should be placed in between, and tweeters should be directly pointed at the ears (for now, I exclude omnidirectional speakers from my analysis). If you compromise on speaker positioning, the influence of room acoustics (reflections from your walls and furniture) rapidly becomes dominant, destroying the frequency response of the speaker. Individual frequencies easily can be boosted or muted by 10db or more! The phase behaviour also quickly deteriorates. Often, excellent speakers give very poor results because people pay too little attention to positioning, or compromise because of visual esthetics.
Apart from positioning, room correction software and (if possible) hardware is very strongly recommended to get better frequency response.
When you keep adding more speakers, it will increasingly become harder to place them at or near the optimal positions (and to perform software-based room correction). At some point, every additional speaker pair will behave worse due to increasingly suboptimal positioning. You run the risk of decreasing the realism and quality of reproduction, rather than increasing it!
So why would someone use large speaker arrays? Some people have referred to an article written by Colin Pykett, http://www.pykett.org.uk/EndOfPipeOrgan.htm#Distortion. This article is worth to be carefully read and provides some powerful arguments for using large speaker arrays in electronic organs. I find this article truly enlightening, yet I think it deserves a critical approach.
The author focuses on two main causes for a lack of realism in electronic organs: intermodular distortion and electronic signal mixing. Here is my first criticism: as mentioned, I think suboptimal speaker positioning combined with the influence of room acoustics is actually the most important reason for lack of realism in many situations.
Intermodular distortion is well explained by mr Pykett. If multiple signals with different frequencies are sent through a single speaker, spurious signals with non-harmonic frequencies will be added due to non-linearities of the amplifier and speaker components. These spurious frequencies are sums and differences of (multiples of) the original frequencies. For example, a 4kHz and 5kHz signal might result in unwanted signals at (5-4)=1kHz, (5+4)=9kHz, (2*4-5)=3kHz, (2*5-4)=6kHz), etc.
Especially in organ music, we sent not just two, but many many more frequencies at the same time through the same speaker. So there will be many spurious frequencies all over the spectrum, forming a disturbing background noise that becomes worse as we play larger chords and use larger registrations.
How significant is this intermodular distortion (IMD) problem? According to mr Pykett, the situation is very bad. IMD specs are often not provided, but he thinks 1% distortion is typical for a reasonable speaker, and in appendix 2 he provides a calculation to show that the total power of unwanted IMD signals easily reaches a level of just 22 dB below the level of the original signal, if just two notes of a trumpet rank are played. This would certainly be audible.
What would be the solution? IMD does not occur if separate sources are used for each signal, and is therefore absent in real pipe organs. So what if we can route each sample to a different speaker? We would need many speakers, but the IMD problem would be eliminated. Mr Pykett suggests to use one speaker for each of the notes in a octave, so 12 speakers in total, or preferably 12 speakers per division. Because in Hauptwerk we aim to reproduce stereo samples, we would actually need 12 speaker pairs, or 24 speakers in total for a stereo sample set.
The 1% distortion figure was mr Pyketts estimation for a "reasonable" speaker. What if we use one pair of expensive, excellent speakers instead of an array of 24 "reasonable" speakers? On this forum, the Mackie HR824 (price around EUR 900 for a single unit) has been recommended for Hauptwerk. Specifications: http://mackie.com/sites/default/files/P ... MK2_SS.pdf. These specs give a figure of <0.035% for distortion, including (but not exclusively!) IMD. That's almost 30x better than Mr Pyketts 1%. According to the formula used by mr Pykett, the power of the distorted signal is a function of the square of the distortion percentage. So the Mackies will have almost 1000x (30 squared) lower IMD power (or much less!) than mr Pyketts "reasonable" speakers! This decreases the IMD level by a further 30 dB. In total, the IMD would be at least 50dB lower than the original signal for small chords. I don't think that would be very audible. For large chords, the situation could be somewhat worse, but still quite acceptable.
My preliminary conclusion is that it could be well worth to invest in expensive but excellent speakers to solve the IMD problem, instead of building a speaker array. Moreover, other aspects of speaker quality also would be improved.
What about my own set-up? I have M-Audio's DSM3 as front speakers, and BX8a as rears. I did some analyses using a measurement mic and the free software REW. I was able to detect several IMD frequencies. The individual IMD peaks were 50-60 dB lower than the original signal. I am therefore not worried about IMD in my audio set-up. I did however discover quite a number of other spurious frequencies (no harmonics or sum or difference frequencies). Although I am not sure, I think those result from wind noise in the individual organ pipes/samples. If I am right, undesirable IMD would be far less significant than desirable sampled wind noise in my set-up.
This leaves me to consider mr Pyketts second argument concerning the problems of electronic signal mixing. I don't think these problems are significant for us because per-pipe stereo sampling is used for Hauptwerk samples, but maybe I should try to analyze this issue in more detail later.
Meanwhile, I hope people find this analysis useful. Again, this long story is not meant as some final verdict, but hopefully it will be the start of a discussion which will generate deeper insights.