gedlee said:
Fair enough on the dictum vs. recommendation point. I thought I detected a change in your position over time on that, after finding that your high sub wasn't working in your room or something like that.
In my previous home and in the previous iteration of this system, I did use a high sub as well as a second "broadband" sub. In my previous home, it was easy to hide the high sub. In this one, it was ugly. Eventually I got tired of looking at it, and ended up replacing both of my "broadband" subs with something much prettier.
gedlee said:
I would think it an assumption that anyone serious about getting the bass (or any other part of the spectrum) right in a room relies heavily on good measurements taken in situ.
Great discussion, and more of it than I expected to see when I arrived home!
Hello Wayne! Long time, no see. As you can tell, I've abandoned the huge basshorns for home and heading for the multi-sub approach. The bass horns are pretty spectacular, but better suited for other applications. I am sold on the idea of more uniform response rather than spectacular dynamics at this point. Multiple subs do a fine job still and don't sacrifice integration in the process.
A lot of good thought coming out of this topic, with some really great minds involved. Looking forward to learning more with regards to individual unit symmetry/non symmetry and room placement in both the horizontal and vertical plane.
Jerrod
Hello Wayne! Long time, no see. As you can tell, I've abandoned the huge basshorns for home and heading for the multi-sub approach. The bass horns are pretty spectacular, but better suited for other applications. I am sold on the idea of more uniform response rather than spectacular dynamics at this point. Multiple subs do a fine job still and don't sacrifice integration in the process.
A lot of good thought coming out of this topic, with some really great minds involved. Looking forward to learning more with regards to individual unit symmetry/non symmetry and room placement in both the horizontal and vertical plane.
Jerrod
Good to see you Jerrod, yes, it has been a while.
I tend to hang out mostly on AudioRoundTable.com and the Pi Speakers forum mostly. I come here from time to time too.
Not much new to report on my end. Still the same basic models. They pretty much do everything right.
Four Pi mains with three Pi subs in my main living room.
Two Pi towers in my bedroom.
Seven Pi cornerhorns in my office.
A warehouse full of 12Pi hornsubs, some out for for rentals.
Plans are still free, kits damn near.
I tend to hang out mostly on AudioRoundTable.com and the Pi Speakers forum mostly. I come here from time to time too.
Not much new to report on my end. Still the same basic models. They pretty much do everything right.
Four Pi mains with three Pi subs in my main living room.
Two Pi towers in my bedroom.
Seven Pi cornerhorns in my office.
A warehouse full of 12Pi hornsubs, some out for for rentals.
Plans are still free, kits damn near.
Yes, when you've got something that works, keep workin' it! Still got lots of really good parts here. Two Pi towers in the garage. Many itterations of midhorns, enough JBL parts to do several "pro Pi" speakers, big custom Fitz horns, really big dipole's with 5 15" drivers a side, and a bunch of other really fun stuff. After many horn variations, been doing high efficiency open baffle stuff for a few years. Been helping with some crossover work, testing, tuning, etc. for the Hawthorne Audio people. Been working with lots of the same old great parts, just different implementations(yeah, a few new parts too!....) Will have to drop in at ART and pull up a seat!
Back on topic
Back on topic
I think the fundamental approach between Geddes' and Welti's original papers were different:
1. Welti's recommended setups in the Harman paper used the passive approach of node placement and/or active approach of mode cancellation. The main problem I had with this paper were the use of only a single real-world room for an investigation, which for the record, did have a door, and which generally supported the modelled results (I always point out that the single corner subwoofer did surprisingly well in this investigation).
2. Geddes' original approach (four subwoofers randomly placed, including one off the ground) in his white paper was a bit of a blunt force approach relying on random differences in phase and distance to excite modes in different phases. The problems I had with this paper was the location of the listener in the center of the room and the focus on a single listener, as I recall.
Obviously, both have modified their approaches since then. As noted, Sound Field Management relies on signal processing, which wasn't part of Welti's Harman paper but his subsequent one. Geddes' approach now relies on spatial averaging that wasn't discussed in his original white paper.
I imagine a future where the room shape and dimensions are entered into a computer, a subwoofer is placed into one of the room corners, someone walks around the room with a wireless microphone to record frequency response below 80 Hz, subwoofers are then placed into the primary horizontal nodes, the computer suggests volume and phase settings depending on the subwoofer characteristics, and signal processing takes care of the rest.
1. Welti's recommended setups in the Harman paper used the passive approach of node placement and/or active approach of mode cancellation. The main problem I had with this paper were the use of only a single real-world room for an investigation, which for the record, did have a door, and which generally supported the modelled results (I always point out that the single corner subwoofer did surprisingly well in this investigation).
2. Geddes' original approach (four subwoofers randomly placed, including one off the ground) in his white paper was a bit of a blunt force approach relying on random differences in phase and distance to excite modes in different phases. The problems I had with this paper was the location of the listener in the center of the room and the focus on a single listener, as I recall.
Obviously, both have modified their approaches since then. As noted, Sound Field Management relies on signal processing, which wasn't part of Welti's Harman paper but his subsequent one. Geddes' approach now relies on spatial averaging that wasn't discussed in his original white paper.
I imagine a future where the room shape and dimensions are entered into a computer, a subwoofer is placed into one of the room corners, someone walks around the room with a wireless microphone to record frequency response below 80 Hz, subwoofers are then placed into the primary horizontal nodes, the computer suggests volume and phase settings depending on the subwoofer characteristics, and signal processing takes care of the rest.
youngho said:
Sorry, can't edit posts, meant to say that the computer suggests placement (which may be the primary horizontal modes), volume, and phase settings; more spatially averaged measurements are done at the listening position; and signal processing takes care of the rest.
The SFM device from JBL seems to work with already positioned subwoofers. Anyone had experience with it?
youngho said:
SFM is very light on signal processing and apart from the single parametric eq filter applicable the approach seems pretty much identical to Geddes'.
I'm interested in creating a software tool implementing SFM's brute force search for best combination of subwoofer positions with gain and delay settings. What's the deal with patents and stuff? Could I release such a software for free?
Welti has published multiple papers. His prior papers only considered what is referred to as "positional optimization" in his later paper. When people here discuss "Welti" or "Harmon" sub placement they're generally referring to this approach, not SFM. I believe this is the source of your confusion.
SFM has only appeared in his later papers. While there remain some differences, Geddes approach is clearly a pragmatic version of the same method with a human doing the optimization rather than software (naive exhaustive search vs shotgun hillclimbing with good heuristics if you want to be all grad student about it). I think Geddes discussion of his approach predates Welti's later papers, but I'm not sure of that.
In any case, it's not surprising that their approaches are converging, since they're solving the same problem largely under the same constraints.
My subjective experience with multiple subs is a clearly less muddy sound. I was amazed once i heard 4 symmetrically placed subs. Even two, placed at the right position, was a "never go back to the state before" experience. So, my applause goes to people like geddes, who brought this approach to the diy world... i dont want to imagine how my room would have looked with oldschool fibrous absorbers for bass frequencies
My impression from reading Toole's book is that SFM has developed further since Welti's 2006 paper. Toole used the phrase "substantial signal processing"in describing it, hence my impression that it involved signal processing. What was I thinking?
Also, I did not get the impression that SFM searched for best subwoofer positions but rather ideal settings for level, delay, and one band of parametric equalization for subwoofers already positioned. Geddes' method seems to result in the nearest subwoofer being the loudest and the furthest being the softest, at least in terms of gain. If SFM's brute force search tended to provide similar setups, then the two approaches would indeed tend to converge, but the complexity introduced by seemingly endless possibilities for settings for gain, delay, and parametric equalization make me skeptical without proof.
Perhaps I am mistaken. If so, please enlighten me.
Also, I did not get the impression that SFM searched for best subwoofer positions but rather ideal settings for level, delay, and one band of parametric equalization for subwoofers already positioned. Geddes' method seems to result in the nearest subwoofer being the loudest and the furthest being the softest, at least in terms of gain. If SFM's brute force search tended to provide similar setups, then the two approaches would indeed tend to converge, but the complexity introduced by seemingly endless possibilities for settings for gain, delay, and parametric equalization make me skeptical without proof.
Perhaps I am mistaken. If so, please enlighten me.
Hi Earl
I may have missed this, but are the mains run full-range, to further spread the number of sound sources? The downside would be that you miss the opportunity to relieve the mid-woofers of low bass duty, and gain better sound from them . .
Also, I see why limiting higher output is desirable, eg by a bandpass. If a bandpass isn’t used for whatever reason, wouldn’t a reasonable alternative be a crossover?
Lastly, is sealed preferred because of the slower rolloff, ie more bass down very low, at the expense of typically less output c 30 – 50 Hz. Would a mix of sealed and ported spread the output best?
Thanks
I may have missed this, but are the mains run full-range, to further spread the number of sound sources? The downside would be that you miss the opportunity to relieve the mid-woofers of low bass duty, and gain better sound from them . .
Also, I see why limiting higher output is desirable, eg by a bandpass. If a bandpass isn’t used for whatever reason, wouldn’t a reasonable alternative be a crossover?
Lastly, is sealed preferred because of the slower rolloff, ie more bass down very low, at the expense of typically less output c 30 – 50 Hz. Would a mix of sealed and ported spread the output best?
Thanks
youngho said:
An alternative approach involving a complex filter for each subwoofer was developed and compared to the thing we know as SFM.
I'm not sure if the SFM described in the paper allowed to choose from many possible subwoofer locations, but it would be a trivial task to allow such a possibility. SFM is amazingly simple due to brute force search and simple settings apart from the parametric filter. Only the filter option requires any intelligence from the algorithm.
SFM does optimize for position. More specifically, you measure an individual subwoofer at some number of potential positions, and SFM selects a subset of those. However, in the 2006 paper I believe only one room had a large number of choices (8), the others just had 4 and the best found solution for those rooms used all of them. That may be what you're thinking of.
I don't believe Geddes method results in the closest being the loudest. Unless I misunderstand it the typical result is that the first subwoofer, placed in the corner, dominates the response. That said, following his approach you might end up with any number of subwoofers being the loudest depending on the room and what positions you try.
I don't believe Geddes method results in the closest being the loudest. Unless I misunderstand it the typical result is that the first subwoofer, placed in the corner, dominates the response. That said, following his approach you might end up with any number of subwoofers being the loudest depending on the room and what positions you try.
Thanks! I appreciate the information. If I understand correctly, the SFM approach tends to result in symmetric subwoofer setups, at least in symmetric rooms. The one very asymmetric example still had some symmetry of the two subwoofers with respect to the listening position. I imagine that Geddes' approach does not?