Horn vs. Waveguide

Status
Not open for further replies.
Jump to Latest
#864
Carl_Huff said:
Earl,

Your post #847 was thoughtful and very well written. Kudos to you. I for one appreciate your hard work.

Simpl Dug said ...
"How can you be a Phd in acoustics and be a two channel guy?"

I can do you one better. I am a two channel guy even tho my day job is to design and build surround gear for a company whose name you would certainly recognize. What's more, I expect that most of the engineers in the business would say the same about themselves. To be more specific, surround sound is mostly about splashing noise in support of what you are seeing with your eyes. Surround Sound is about cinema, whether it be in a commercial setting or your home. Most music is purposely recorded as a two channel experience. That is especially the case for recordings that are more than 15 years old. Audiophiles use words like 'imaging' and'soundstage' that are most meanful when listening to 2 channels. Think about it, what we call imaging and soundstage is really about reconstructing in our brains what we would be seeing with our eyes if we were watching the performance live. And when watching that performance, unless you are watching live unamplified musicians perform, chances are what you are hearing is a mostly monophonic mix delivered over multiple loudspeaker stacks (ie: No imaging) and is an entirely different listening experience with what you have at home.

I would go on to say that most emerging products such as the new 10 channel surround is mostly about marketing (ie: exchanging your $$ for their spin and paying the wages of guys like me) and mostly serves to fill in the holes made by the crappy left and right mains of your system. That is why the research that Earl has been doing is so important. Horns and now waveguides do a great job of recreating the edge and transients of a live performance. For me, Earls work is about recreating the soundfield of that live experience in a way that traditional horns cannot.
Click to expand...

Yes I understand Carl, but I am already on my second hand on designers/national sales managers who I have demonstrated their failure to realize the potential of their own products. I demonstrated this on the fly at trade shows. Surround Processors setup completely wrong by the designer of the product. Yeah they were quite shocked especially when visitors to their room felt compelled to suddenly voice their opinion on the sound which they had no idea was in basic Prologic II!

But I refuse to discuss surround on the internet any longer because it is pointless waste of time, I am outnumbered and I do not speak the technical language necessary to make a strong case in print.

For me the Meridian processing does a good enough job for me even if the industry only makes two channel recordings my clicking back and forth between two channel and multi-channel there has never been an argument with any of my guests which one sounds better and more like a "Live" experience. Infact most guests don't even know they're listening in surround. So keen is the audiophile ear.

My typical system is meridian and ATC active speakers, I point this out so you have a reference to the type of equipment I am accustomed to to achieve this level of sound. Although Cambridge Audio and DIY speakers can have a similar effect.

If you think Soundstage and imaging are better with unadjustable two channel than three speakers, well we have a serious disagreement. And the advantages of this configuration certainly should be evident in someones paper, somewhere.

I have already learned my lesson, if I cannot demonstrate it then there is no worthwhile discussion to be had. I appreciate your input even though you haven't told me anythig I haven't been told many many many times before. When the doubter gets a demo...the perspective changes.
 
#865
gedlee said:



Nice plots, but they miss a critical factor - timing. In the Geddes setup the nearer speaker gets softer as the timing gets shorter, this helps to keep a central image over a much larger area. In the "classic CD" setup the opposite occurs causing an almost immediate collapse of the image to the closer speaker. I've tried to draw this effect, but its not easy.
Click to expand...

Do you mean the decay in level due to the distance from the waveguide?

That's factored in; it was actually the most difficult part. For instance, when the speakers are positioned eight feet away, and the listeners are seated on a couch that's eight feet wide, the left listener is a full meter closer to the left speaker than to the right.

So you need to factor in the decay in level due to distance to get a proper idea of how closely the levels will be.

If anyone wants the originals in xara or pdf, that would be a lot easier to read; I can only post images that are 800 pixels wide.
 
#867
Simpl Dug said:


~~~~~~~~~~~~~~~Snip~~~~~~~~~

If you think Soundstage and imaging are better with unadjustable two channel than three speakers, well we have a serious disagreement. And the advantages of this configuration certainly should be evident in someones paper, somewhere.

~~~~~~~~~~~~~~~~~Snip~~~~~~~~~
Click to expand...


Dug,

I'm not a party to this thread, but concerning your quoted statement above, I believe that Paul Klipsch of Klipschorn fame wrote several articles championing the use and advantages of three channels over two, at the very beginning of the stereo era.

I'm sure you can Google, or perhaps go to the Klipsch Forum for the information.

Best Regards,
TerryO
 
#868
For the record, I am a strong believer in three forward channels - again IF done properly, but that is a big "IF". Most center channels are derived by summing the left and right, which is incorrect, and I would never use that technique. The correct way to do it would be very complex and I don't know anyone who does it (but that is certainly NOT my specialty). But the surround speakers, just delaying the two channel signals and adding reverb does not appeal to me. I prefer the natural reverb in my room to anything synthesized. So I am not at all oppsed to the idea of multichannel, I just have not see or heard one that is acceptable. There are some very nice multichannel DVD's of live music that are mixed very well across the three front channels and have a natural room reverb in the rear, but these are five channel recordings, mixed that way from discrete signal sources, not derived from a two channel source.

Way off topic here, and I am personally very interested in the topic. Maybe someone should start a thread somewhere to discuss it.
 
#869
noah katz said:
"Do you mean the decay in level due to the distance from the waveguide?"

I think he means the precedence effect, where the sooner-arriving of two equal level sounds is perceived as louder.
Click to expand...

Ugh, hadn't considered that. In the car waveguides I want to build, I'd tweaked the coverage angle so that the left and right are equal in level for both the driver and passenger.

http://www.diymobileaudio.com/forum/diy-mobile-audio/60146-creating-perfect-soundstage.html

But based on what you've suggested, the speaker closest to the listener should be a bit quieter.

And *how* much quieter?

1db? 3db?

Assuming a perfect waveguide and a distance from the listener to the speaker of 8ft, the closer speaker is 3db quieter.
 
#871
gedlee said:
For the record, I am a strong believer in three forward channels - again IF done properly, but that is a big "IF". Most center channels are derived by summing the left and right, which is incorrect, and I would never use that technique. The correct way to do it would be very complex and I don't know anyone who does it (but that is certainly NOT my specialty). But the surround speakers, just delaying the two channel signals and adding reverb does not appeal to me. I prefer the natural reverb in my room to anything synthesized. So I am not at all oppsed to the idea of multichannel, I just have not see or heard one that is acceptable. There are some very nice multichannel DVD's of live music that are mixed very well across the three front channels and have a natural room reverb in the rear, but these are five channel recordings, mixed that way from discrete signal sources, not derived from a two channel source.

Way off topic here, and I am personally very interested in the topic. Maybe someone should start a thread somewhere to discuss it.
Click to expand...

Earl,
I don't know about multichannel recordings, but using variations of the Hafler system, which eliminates out of phase signals in a stereo recording, generally will allow for a center channel that's not too bad. I'm certainly no expert and there are probably people that can come up with better ideas, but this does work.

Best Regards,
TerryO
 
#872
JoshK said:
As an FYI, here is one algebraic processor for 3 channels.
Trinaural processor
Click to expand...

Hey Josh

Close, but no cigar. he clearly understands the problem, but unless I am wrong, it can't be done analog, or I don't see how.

What one must do is to find the correlated signals in the left and right signals - this is the center channel, and then subtract these signals from the left and right. This makes the three signals uncorrelated and there isn't an interference problem. But finding the correlated signal between the left and right is not trivial. Its not hard in DSP, but I don't know of an analog process that will do that.

The writup itself admits that what he does, which he really doesn't define, leaves some coherent signals in all of the channels. This is precisely what the correct processing would not do.
 
#873
Patrick Bateman said:
But based on what you've suggested, the speaker closest to the listener should be a bit quieter.
Click to expand...

Thats correct. When the closest speaker gets louder, by even a dB, then the two effects - loudness and precidence - make the image jump to the nearer speaker. This is the classic effect that I found so strong in the Wilson Sound Puppies that I heard. They had a good (almost Summa like) image when the head was dead center, but it collapsed almost immediately when the head was moved left or right. It truely was a sweet spot - with the emphasis on the word "SPOT", as in "point".

We did experiments on this at Ford. You cannot completely overcome the precidence effect, but 1-3 dB (depending on time difference) does a lot to maintain the image. More than three dB doesn't help the image and begins to confus things. There is a chart in Spatial Hearing that has the exact tradeoffs shown.
 
#874
Indeed, he calls it a "first order" solution...which to me means first order approximation.

Ralph Glasgal takes a different approach in his ambiosonic processor and uses two channels that are decorrelated. He cooperated with TactLabs to put out a processor.

http://www.tactlab.com/Products/Ambiophonics/ambiophonics.html

Ralph's site
http://www.ambiophonics.org/

I've heard one implementation of ambio and it was pretty spectacular for live recordings, especially large hall recordings. Ralph invited me to hear his and I need to do that soon.
 
#876
gedlee said:
What one must do is to find the correlated signals in the left and right signals - this is the center channel, and then subtract these signals from the left and right. This makes the three signals uncorrelated and there isn't an interference problem. But finding the correlated signal between the left and right is not trivial. Its not hard in DSP, but I don't know of an analog process that will do that.
Click to expand...

That's how Dolby Prologic II works. Jim Fosgate originally developed it with vacuum tubes and Dolby converted it to DSP when he sold it to them. I'm not sure of the details (Fosgate is naturally vague about them) but I think it involves summing the L+R and subtracting the L-R so that gives you 4 signals to work with (L, R, L+R, L-R) that can be used with normal or inverted phase and can each be adjusted in amplitude. If I thought about it a bit (feeling lazy today), I'm sure I could figure out how to add and subtract those signals to send L-only info to the L, info in both L and R to the C, R-only info to the right.
 
#877
catapult said:


That's how Dolby Prologic II works. Jim Fosgate originally developed it with vacuum tubes and Dolby converted it to DSP when he sold it to them. I'm not sure of the details (Fosgate is naturally vague about them) but I think it involves summing the L+R and subtracting the L-R so that gives you 4 signals to work with (L, R, L+R, L-R) that can be used with normal or inverted phase and can each be adjusted in amplitude. If I thought about it a bit (feeling lazy today), I'm sure I could figure out how to add and subtract those signals to send L-only info to the L, info in both L and R to the C, R-only info to the right.
Click to expand...

You would have to prove to me that it can be done without doing a cross-correlation which does not have an analog process that I know of. No linear process, again that I know, can find the LR cross correlation from the four signals that you describe. I've talked with many people about the Pro-logic scheme and none of tem can prove to me that it is correct. How much error there is I don't know but I don't believe that it is exact.
 
#878
We should split this topic off from this thread into its own (moderators, could you please assist?).

Its a simple vector math problem but I have zero clue how one does this with analog filters.

Going from very rusty memory, the correlation between two vectors is the cos of the angle between them or related to the projection of one vector onto the other. Does anyone know an analog filter convolution that does similar?
 
#879
Here's a Jim Fosgate interview.

http://www.stereophile.com/interviews/1204fosgate/index.html

I'm really out of my element trying to guess how he did it and we should probably just ignore my wild a** guesses above. In the interview he mentions using negative feedback on the steering logic so I'm sure that's a clue but he's spent decades working on this stuff and I've spent a few minutes. 😉

To my ear, DPL2 Music mode does a pretty good job of separating the front channels. I like to set the 'width' control so it's still leaving some center content in the R&L -- just a personal preference.

Edit: anyone really curious about how it works could read the 200 page patent. 😉

Edit again: these seem to be the relevant patents. I searched for inventor = fosgate and assignee = dolby. The oldest one was filed in 2000 which is about right for when DPL2 came out. A little light reading. 😉

http://patft.uspto.gov/netacgi/nph-...1=INNM&co1=AND&TERM2=dolby&FIELD2=ASNM&d=PTXT
 
#880
The patent is interesting reading and, as I had suspected, its all an "approximation". Typically only 3 dB of signal seperation is possible, with the patent this is improved, but in amounts that vary with signal and no data is given for the level of improvement.

With the technique that I describe a center channel could be derived that was perfectly seperated from the left and right, which is the ideal. But the techniques would be difficult to do and could only be done with DSP. Analog is, as I said, limited to the approximate techniques.
 
Status
Not open for further replies.
Top Bottom