Further to your questions... I believe the BBC dip is to ameliorate the difference in hearing sensitivity between direct and reverberant sounds that becomes noticeable when both are replayed over the same loudspeakers. It's shuffler-like properties might then also be expected?
Good answers. 🙂 I figured not many would really read the thread Pano linked to earlier.
This should give many readers something to ponder about.
This should give many readers something to ponder about.
The caveat is that stereo also relies on the sound arriving from each speaker being identical in amplitude and overall delay, and that only occurs when the listener is equidistant from each speaker. Therefore stereo speakers fail to place virtual sources correctly when the listener is no longer seated equidistant from each speaker. HRTF can compensate for the direction of the speakers relative to the listener but it cannot compensate for the difference in distance. Therefore stereo reproduction still 'works' for a narrow or wide placement of speakers or if you turn your head on the spot, but begins to fail if one speaker is moved significantly closer or further away from the listener.
When stereo speakers play independently an off-centre listener can determine the direction of the left and right speakers correctly (due to HRTF), but when the speakers play a coherent/mono sounds together, the virtual centre source is no longer perfectly centred between the speakers as the virtual centre location has become influenced by the imbalance in amplitude and time of arrival at the new listening position. When you become very close to one speaker, the virtual centre becomes basically at that speaker because the listener is now heavily influenced by the much higher amplitude and earlier time of arrival of sounds coming from it (ref post #54) - it has become perceived as a mono, single speaker from the listening position. Getting rid of stereo crosstalk may significantly reduce the amount of suppression of the delayed/attenuated channel by our auditory system.
Centre speaker solves that issue in the case of needing to produce perfectly centred sounds from all positions. Exactly same problem as above still exists if you now want to produce a sound in a virtual location which is between the centre speaker and one of the stereo speakers. To accurately produce surround sound from every listening position you need an infinite number of speakers/channels.
Last edited:
The consideration of non-ideally placed listeners is somewhat off-topic in this thread but raises some interesting points.
The caveat you describe is not a particular caveat of stereo, but of any multi-speaker replay arrangement that employs a sensible number of speakers and cannot then be considered as an accurate (wideband) means of wavefront construction. As you state, such wavefront reconstruction requires an enormous number of loudspeakers. The alternative transaural approach is not intended for non-ideally positioned listeners.
However, even for someone sat in an ideal listening seat, stereo images can still be drawn to one or other loudspeaker at an increasing rate as frequency increases and as the panning angle from centre increases. For non-ideally placed listeners the effects are undoubtedly worse but can also be improved by higher order decoding such by as implementing a shuffler or similar. Again (and not unsurprisingly) Gerzon's 1990s application of his localisation theory was adapted to multispeaker stereo for just such non-ideal listeners to good effect.
And where higher order decoding is employed, so localisation becomes more accurate for ideally placed listeners with increasing order and in some cases for non-ideally placed listeners too - until such time as we arrive at the impractical solution of wavefront construction that requires lots of loudspeakers. Compromise is inevitable. But "when stereo speakers play independently" is simply when each is acting instead as a singular sound source and "getting rid of stereo crosstalk" for less-than-ideally placed listeners will not only reduce localisation errors, it will reduce localisation altogether. It is not a good solution therefore.
But a centre loudspeaker can improve these matters significantly - for non-ideally placed listeners and for images panned anywhere across the usable stage width. The accuracy in localisation increases with the number of loudspeakers subject to the caveat that their inputs are decoded appropriately: With more loudspeakers also comes the increasing possibility of making matters worse as the 'logical' extension of stereo to quadrophonics demonstrated.
I think there also exists in this thread some confusion between a centre loudspeaker used as a discrete channel and a centre loudspeaker fed from a matrix encoder such as stereo?
The caveat you describe is not a particular caveat of stereo, but of any multi-speaker replay arrangement that employs a sensible number of speakers and cannot then be considered as an accurate (wideband) means of wavefront construction. As you state, such wavefront reconstruction requires an enormous number of loudspeakers. The alternative transaural approach is not intended for non-ideally positioned listeners.
However, even for someone sat in an ideal listening seat, stereo images can still be drawn to one or other loudspeaker at an increasing rate as frequency increases and as the panning angle from centre increases. For non-ideally placed listeners the effects are undoubtedly worse but can also be improved by higher order decoding such by as implementing a shuffler or similar. Again (and not unsurprisingly) Gerzon's 1990s application of his localisation theory was adapted to multispeaker stereo for just such non-ideal listeners to good effect.
And where higher order decoding is employed, so localisation becomes more accurate for ideally placed listeners with increasing order and in some cases for non-ideally placed listeners too - until such time as we arrive at the impractical solution of wavefront construction that requires lots of loudspeakers. Compromise is inevitable. But "when stereo speakers play independently" is simply when each is acting instead as a singular sound source and "getting rid of stereo crosstalk" for less-than-ideally placed listeners will not only reduce localisation errors, it will reduce localisation altogether. It is not a good solution therefore.
But a centre loudspeaker can improve these matters significantly - for non-ideally placed listeners and for images panned anywhere across the usable stage width. The accuracy in localisation increases with the number of loudspeakers subject to the caveat that their inputs are decoded appropriately: With more loudspeakers also comes the increasing possibility of making matters worse as the 'logical' extension of stereo to quadrophonics demonstrated.
I think there also exists in this thread some confusion between a centre loudspeaker used as a discrete channel and a centre loudspeaker fed from a matrix encoder such as stereo?
Last edited:
Hi Pano, I'm curious about your L-C-R stereo speaker setting. I would like to try it, but I have no knowledge about multi channel speaker configurations. The center channel is the same speaker as the side ones? I also wonder how you send the stereo signals to those 3 speakers.
I'm very interested in the 'best' way to generate the optimum three channel ourput from a standard 2CH digital input. The two I have heard positive things about are Trinaural and Trifield.
Trinaural is Bongiono's analog in and out processor that still appears to be in production:
Trinaural Processor — SST
The second, Trifield is an implementation based on Michael Gerzon's work. It has been in Meridians products for over 20 years. I have heard it and it seems to work quite well. There is an add-on component available for Foobar 2000:
foobar2000: Components Repository - Trifield LR→LRC Decoder DSP
I have installed this and processed a song into a 3ch wav file that I can't presently play (the next issue)
Presumably using a player to send one of these files to an AV receiver or prepro should work. I will probably work in this direction. Foobar2000 and JRiver should work, and SPDIF is likely the least problematic protocol.
This is pretty much OT off to the side. If there's interest it could be split off on its own.
I'm eager to learn from those who have done this.
Skip
Trinaural is Bongiono's analog in and out processor that still appears to be in production:
Trinaural Processor — SST
The second, Trifield is an implementation based on Michael Gerzon's work. It has been in Meridians products for over 20 years. I have heard it and it seems to work quite well. There is an add-on component available for Foobar 2000:
foobar2000: Components Repository - Trifield LR→LRC Decoder DSP
I have installed this and processed a song into a 3ch wav file that I can't presently play (the next issue)
Presumably using a player to send one of these files to an AV receiver or prepro should work. I will probably work in this direction. Foobar2000 and JRiver should work, and SPDIF is likely the least problematic protocol.
This is pretty much OT off to the side. If there's interest it could be split off on its own.
I'm eager to learn from those who have done this.
Skip
I have used passive matrix and active. Active is a little easier and you can use whatever works. I've used a DCX-2496 crossover, a 4 channel sound card and a Yamaha HT receiver which all work about the same. Have also used a Soundcraft mixing console when doing L-C-R P.A.
It is best for all speakers to match. But course I didn't always do that. 😉 I have had just a horn in the center with a level just enough to fill in. Or sometimes a small P.A. cluster. Those worked OK, too.
Additional center speaker for stereo is certainly a frontier and it sounds like a lot of experiments can be done! 🙂
True that. There is the "Correct" way to do it and then all the other ways. Many of those others have a lot of merit and you might be happy with them.
I've had excellent results with very different speakers. What was the same though was I used the same reference curve while designing them all. The tweeters shared a very smooth and extended response past 20 kHz. One set used a Peerless ring radiator, another Mundorf AMTs.
I think if I had tried this with your average boutique speaker, with their ragged upper end signatures this would not have worked.
I didn't even use EQ except for the mid-bass in the center channel.
If you have a surround set-up try turning off the L + R listen with just center and surrounds.
Rob 🙂
Rob 🙂
Agreed. Fortunately, I have first hand knowledge in a number of experiments in the home and in the theater. Personal experience and experimentation is a great way to learn and settle an argument. 🙂
Ha ha. Similar things have happened to me. When I finally built and excellent system and an excellent room - I several times realized "Hey, this recording is in mono!" Not always obvious before. 🙂
Me too just goes to show just how important the Center channel is to typical surround mixes.
Rob 🙂
I use a DOLBY STEREO CP200 in my THX cinema when it comes to Laserdisc Star Wars originals in Dolby 424 matrix wow processor ever made.
Attachments
When I worked at UCI cinema late '89 the CP55 was to me back then wow the centre channel was great and better with good A-type or some SR prints that often came in. Dialogue panning is great to listen too as it is positioned or steers though the matrix 424.
I got myself Dolby Stereo CP55 for fond sounding memories. 🙂
I got myself Dolby Stereo CP55 for fond sounding memories. 🙂
