markus76 said:
Well yes, but I was doing this well before that paper came out, so I usually don't reference it. I sent a letter to the editor with my comments on Welti's paper. I agreed with most of it, but the completly symmetrical locations studied and sample on I took exception to. I don't recommend the symetrical palacements which "won" the Welti study, because I have found that a more random placement works better. Its like sampling a regular grid with regular points, it in fact reduces the degrees of freedom. I get the same benefit with three subs as Welti did with four. Otherwise we agreed completely that more is better.
pjpoes said:
I most definately agree that the fewer times we A/D and D/A the better (much better since these steps are usually the most problematic) and the active crossover boxes tend to agravate this. I run a digital signal to the digital input of my receiver and it is only D/A once in the whole chain. If I could strip off the signal in the PC and process the LF part to use in my system then I would do that and just send the PC processed LF signal out to the subs and leave the rest alone. But this takes software that is unavailable and would hve to be developed. Real time processing on a PC is a whole other level of complexity that I am not willing to dive into. But I would love to have someone do this for me!
The crossover that I used did not use the same definitions for Q that I do - admittedly for a parametric EQ there is no standard definition. So when I would run a curve and find the poles and zeros, I could only guess at how the processor settings needed to be set, they always required looking at the actual transfer function and "tweaking" it in.
It's mostly a cost thing, if you change location/system you change crossover settings and then...also if you want to try out new crossover points or change the settings, you aren't going to be able to do that easily. Besides, we are talking about full units that are designed to be used with anything in their limits, have PC connections, screens and can be changed real time. They're just very different things imo. None of the active crossovers discussed save for the DEQ is meant to be used specifically by diyers let alone in a hifi situation.
gedlee said:
I most definately agree that the fewer times we A/D and D/A the better (much better since these steps are usually the most problematic) and the active crossover boxes tend to agravate this. I run a digital signal to the digital input of my receiver and it is only D/A once in the whole chain. If I could strip off the signal in the PC and process the LF part to use in my system then I would do that and just send the PC processed LF signal out to the subs and leave the rest alone. But this takes software that is unavailable and would hve to be developed. Real time processing on a PC is a whole other level of complexity that I am not willing to dive into. But I would love to have someone do this for me!
The crossover that I used did not use the same definitions for Q that I do - admittedly for a parametric EQ there is no standard definition. So when I would run a curve and find the poles and zeros, I could only guess at how the processor settings needed to be set, they always required looking at the actual transfer function and "tweaking" it in.
Yeah latency in real time processing is another major problem, especially with movies. It's gotten so bad as to cause lip sync problems in some cases. The latency becoming in the 10's of milliseconds, its a very real issue. Some processors have the ability to slow the video for sync, but I haven't seen one that can do this with HDMI, which is what I would like.
I happen to still enjoy listening to Analogue records, so another reason for my dislike of the a/d/a issue is that it means I'm taking an inherently analogue signal and converting it to digital, only to convert it back to analogue. Even if you argue that the sound of records is wrong, it still doesn't change the fact that, at least subjectively for me, it changes the sound converting it to digital. I've never liked record transfers to digital, I've never liked the sound of my records through a pre/pro with no analogue pass through, and when I had the DEQ2496 in my own system for some time, I didn't like listening to records anymore.
Ok well this has gone on such a tangent, any progress on the speakers. I'm curious how they are going for you Markus? Given the trouble you have apparently had getting everything to fit together, may I recommend that you make judicious use of bondo, woodfill, and whatever sealing compound you choose. Earl may correct me, but I believe these are a sealed design, and slight box leaks can be a pretty big problem. While it won't have a huge impact on the frequency response, it can create audible distortions from the chuffing out of the leaks, and can also effect the sound of the bass and lower midrange notes.
pjpoes said:
This, of course is true, but with the design, there is very little chance of a leak through the box. I am not so sure about the woofer. The surround on these units has always concerned me as you can see right through it. This does not mean that it leaks air, because light is not air. I often paint a little rubber cement (diluted with laquer thinner) on the surround. Sometimes this changes the response sometimes not. On the last driver that I tried this on it did not make a difference, so I guess that one was sealed. But it still bothers me when I can see through it. The enclosure when done is quite solid, almost overkill IMO.
Marcus did have some boards that could have fit better and I'm cutting him new ones next chance I get, but unfortunately that won't be for another week or so. I have to share my work space between sawdust generation and paint booth. Right now its a paint booth and I can't generate any dust until the paint is complete. The paints that I use don't dry very fast - they are not two parts they are water based and its humid here. So I have to move at the rate that the drying allows.
Dr. Geddes, I could make those enclosures for you. We have a constant stack of 3/4" furniture grade MDF at the shope, 5 routers on the table, 4 free routers, a Delta table saw with a Bisslermier fence, and we are a Whitesides router bit dealer...we ave a 1" rounder over router bit at the shop for 1/2" collet routers. Not to mention we get MDF cheap- we have connections with a lumber yard that sells to all of the furniture companies up here in Hickory, NC. Right now we are paying $10 a sheet for a sheet of MDF.
With one template made, it would only take a minute to do the face and less than that to do the other sides.
And I would like to talk to you about those car waveguide molds.
With one template made, it would only take a minute to do the face and less than that to do the other sides.
And I would like to talk to you about those car waveguide molds.
" I get the same benefit with three subs as Welti did with four. Otherwise we agreed completely that more is better."
This strikes me as a bit out of context; Welti's conclusion was
"Since putting even 50 subwoofers in a room is not practical,
this approach is not practical."
The take-away for me was that the most practical solution is two subs on opposite mid-walls, which give almost as good a result as four.
"I most definately agree that the fewer times we A/D and D/A the better (much better since these steps are usually the most problematic)"
I'm surprised to hear this (not disagreeing, not my area) but I thought A/D and D/A were mature with very high levels of conversion fidelity.
This strikes me as a bit out of context; Welti's conclusion was
"Since putting even 50 subwoofers in a room is not practical,
this approach is not practical."
The take-away for me was that the most practical solution is two subs on opposite mid-walls, which give almost as good a result as four.
"I most definately agree that the fewer times we A/D and D/A the better (much better since these steps are usually the most problematic)"
I'm surprised to hear this (not disagreeing, not my area) but I thought A/D and D/A were mature with very high levels of conversion fidelity.
winslow said:
Getting someone to make them is not a problem, its the extra cost that is. There is no way that I could absorb this extra cost and it would have to be passed on to the customer. I asked if people would rather pay more for a better fit and I'll have to say that only Markus said "yes".
The DE500 is not a direct replacement for the de250. I do send an enclosure drawing and crossover schematic with the baffle only option.
noah katz said:
I don't see where there is any conflict in my statement and Welti's. But four is better than two as is three better than two. And five is better than four. It's a 1/N situation with ever diminishing returns. But symmetric on opposing walls is not what I would recommend. If you only use two, put one in the far corner and the other near the midpoint of the opposite wall or anywhere else you want, but one should be in a corner and the other one not.
The digital chain is virtually perfect and ANY error that occurs is going to occur at the D/A. Maybe its not audible, but it is the weak point.
This is a somewhat separate issue, but I was talking with one of Tektronix engineers about measuring distortion and signal noise using a device of theres, and his response to me was that right now the resolution is limited by the quality of A/D converters. He said that especially the "for audio" A/D converters are substandard and the primary limiting factor. He told me that none of the units they tested could meet their claimed bit rate and that most of them had noise floors much higher than the D/A converter. I remember a similar conversation with the guys from Smith and Larson (Woofer Tester), and so I've had the impression that the A/D converters are the bigger limiting factor. I also read an article put out by Phillips on the problems with decoding 16 bit data with 24 bit decoders, or maybe it was the 44.1khz with 96khz decoders, I don't remember what the issue was, but they claimed the main reason for up-sampling and over-sampling was not supposed improvements in sound, but simply fixing problems created by using the wrong chips in the first place. This is definitely out of my area so its possible that was marketing more than science.
Anyway, I would argue that digital is becoming a mature science, but that, be it cost cutting measures, or the fact that we are still learning, it seems like new digital products sound better with every new generation. I never thought much of those jitter reduction devices, yet I see companies like TI, AKM, etc all now putting a great deal of money into jitter reduction for their chips. I think people have realized that outboard jitter reduction devices didn't really work, but that jitter is a very real issue. I read one of TI's white papers on one of their new chips and recall them saying that none of their 24 bit chips could reach there theoretical limits in actual applications because of limitations caused by jitter. This would make sense as I've never seen a converter capable of true 24 bit resolution.
Anyway, I would argue that digital is becoming a mature science, but that, be it cost cutting measures, or the fact that we are still learning, it seems like new digital products sound better with every new generation. I never thought much of those jitter reduction devices, yet I see companies like TI, AKM, etc all now putting a great deal of money into jitter reduction for their chips. I think people have realized that outboard jitter reduction devices didn't really work, but that jitter is a very real issue. I read one of TI's white papers on one of their new chips and recall them saying that none of their 24 bit chips could reach there theoretical limits in actual applications because of limitations caused by jitter. This would make sense as I've never seen a converter capable of true 24 bit resolution.
"I don't see where there is any conflict in my statement and Welti's."
You're right, and I failed to sufficiently articulate my point.
Which is that the cases of 2 and 4 that I mentioned do not get their benefits from the 1/n random factor, but from their being placed in specific locations which result in the elimination of odd-order modes between the front/back and side walls.
I guess in a room where even-order modes are worse offenders that may not necessarily be the optimum solution.
You're right, and I failed to sufficiently articulate my point.
Which is that the cases of 2 and 4 that I mentioned do not get their benefits from the 1/n random factor, but from their being placed in specific locations which result in the elimination of odd-order modes between the front/back and side walls.
I guess in a room where even-order modes are worse offenders that may not necessarily be the optimum solution.
noah katz said:
Mathematically, because the two symmetrical locations are exact mirrors of each other they are highly correlated. The 1/n assumption holds for uncorrelated sources. Thus Welti's 1/4 partially correlated sources are about the same as my 1/3 more uncorrelated sources.
Now, no two sources in a room can be completely uncorrealetd at these frequencies so this raises the question: Could one uncorrelate the sources and achieve an even better response with, say, electronics. The answer is "Yes". But decorrelation filters are not trivial to impliment. However, a PC could certainly do it.
winslow said:
$10 per sheet?!?! Is this possibly for only 1/4" mdf? We're paying nearly $25 a sheet wholesale now for 3/4" MDF. If you would be able to PM me or email me(john at aespeakers.com) with a contact for the lumber yard I would love to have it. If it really is $10 per sheet I'd seriously consider buying a full truckload and bringing it up here.
John
"Mathematically, because the two symmetrical locations are exact mirrors of each other they are highly correlated. The 1/n assumption holds for uncorrelated sources. Thus Welti's 1/4 partially correlated sources are about the same as my 1/3 more uncorrelated sources."
It sounds like I may have still not communicated my point.
The benefits of the one or two pairs on opposite mid-walls has nothing to do with numerous uncorrelated sources and everything to do with the unique characteristics of those locations.
For those who haven't read the paper, the odd-order modes are eliminated thusly: In the opposite wall location each speaker generates a mode that is everywhere in opposite phase from the other's, so they cancel; and because the mid-wall location is a null (or is that node?) point for the odd-order modes between the side walls, they're not generated in the first place.
It sounds like I may have still not communicated my point.
The benefits of the one or two pairs on opposite mid-walls has nothing to do with numerous uncorrelated sources and everything to do with the unique characteristics of those locations.
For those who haven't read the paper, the odd-order modes are eliminated thusly: In the opposite wall location each speaker generates a mode that is everywhere in opposite phase from the other's, so they cancel; and because the mid-wall location is a null (or is that node?) point for the odd-order modes between the side walls, they're not generated in the first place.
markus76 said:
Hi Markus,
I'm curious, you are going to use 4 subs? do you have any info on the quality of the subs you are going to use? they seem like a very good deal, about 100euro per sub (excl. amps and XO), I might consider 2 of them if I decide to build the Nathan10 or the 12inch version. Wonder if they can keep up with the main speakers though, when do they run out of breath? (at which SPL)?
Cheers,
Jacobus
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.