I definetely rushed things a bit when ordering 400hz crossovers from Marchand. The XO point is too high and the sound loses realism. I´ve tried 180 and narrower angles. None of them improve things over my former set-up. Bear in mind that I have no crosstalk cancellation under 400. So I´ve decided to forego the modification and return to where I was before.
Extended OSD?
I was intreagued by the posting of an abstract for a talk by Takeuchi & Nelson last summer and published in " Acta Acustica united with Acustica, Volume 94, Number 6, November_December 2008". Since I've not read the article all I have to go on is this sentence from the abstract:
"It is shown here that the advantage of the OSD is further enhanced by separating in-phase and out-of-phase components through an alternative system design."
which triggered some thoughts about adding a centre mono speaker under the tweeters (described earlier in the thread) and changing the processing to suit.
So I decided to add an open baffle (I had wood enough for that but not a box) with a B&C 10PS26 covering 100Hz to 1.3 kHz and a BMS4548+ on an RCF H100 horn very well stuffed with wadding covering from 1.3kHz to 10 kHz or so. The crossover is active (close enough to LR 4th order acoustic) and there is some parametric filtering. The target response (average in room) is a gentle slope downward to hf.
It did not take too long to integrate this into the previously described OSD setup. On most recordings the difference is rather subtle, is is certainly too early to declare success (lots of ways this could be done, and I've only tried a very small part of the idea space).
Also I dug out an old pair of KEF Uni-Q speakers and have tried them as rear speakers-so far only with either sum (mono) or difference (L-R to both). On a few recordings this shows promise, but mostly it is just an effect (sometimes quite fun).
Strikes me that as I have 13 channels of amplification (and 16 drive units counting the KEFs as 4), to listen to stereo tracks, perhaps it is time to stop. The great thing with DSP is rapid turn around for measure/modify/measure - otherwise it would take many months just to get the crossovers right.
Now I'll concentrate on getting everything pretty(ier) - some of the parts are quite ugly.
Ken
I was intreagued by the posting of an abstract for a talk by Takeuchi & Nelson last summer and published in " Acta Acustica united with Acustica, Volume 94, Number 6, November_December 2008". Since I've not read the article all I have to go on is this sentence from the abstract:
"It is shown here that the advantage of the OSD is further enhanced by separating in-phase and out-of-phase components through an alternative system design."
which triggered some thoughts about adding a centre mono speaker under the tweeters (described earlier in the thread) and changing the processing to suit.
So I decided to add an open baffle (I had wood enough for that but not a box) with a B&C 10PS26 covering 100Hz to 1.3 kHz and a BMS4548+ on an RCF H100 horn very well stuffed with wadding covering from 1.3kHz to 10 kHz or so. The crossover is active (close enough to LR 4th order acoustic) and there is some parametric filtering. The target response (average in room) is a gentle slope downward to hf.
It did not take too long to integrate this into the previously described OSD setup. On most recordings the difference is rather subtle, is is certainly too early to declare success (lots of ways this could be done, and I've only tried a very small part of the idea space).
Also I dug out an old pair of KEF Uni-Q speakers and have tried them as rear speakers-so far only with either sum (mono) or difference (L-R to both). On a few recordings this shows promise, but mostly it is just an effect (sometimes quite fun).
Strikes me that as I have 13 channels of amplification (and 16 drive units counting the KEFs as 4), to listen to stereo tracks, perhaps it is time to stop. The great thing with DSP is rapid turn around for measure/modify/measure - otherwise it would take many months just to get the crossovers right.
Now I'll concentrate on getting everything pretty(ier) - some of the parts are quite ugly.
Ken
final update (probably)
After extensive listening I find that my implementation of (my guess at) extended OSD does not help very much at least on the best recordings, in the sense that the image is equally convincing with or without the extra mono centre channel. So two drivers and two amplifiers can retire. The original OSD seems just as good - perhaps because I implemented it better - there are, after all, very many parameters to get right (and adding a few more only makes things harder).
On the other hand I find that the rear speakers, when dialed to the right level (which depends slightly on the recording) can at least add some life to some of the poorer recordings, good fun.
Ken
After extensive listening I find that my implementation of (my guess at) extended OSD does not help very much at least on the best recordings, in the sense that the image is equally convincing with or without the extra mono centre channel. So two drivers and two amplifiers can retire. The original OSD seems just as good - perhaps because I implemented it better - there are, after all, very many parameters to get right (and adding a few more only makes things harder).
On the other hand I find that the rear speakers, when dialed to the right level (which depends slightly on the recording) can at least add some life to some of the poorer recordings, good fun.
Ken
this is zany - your tweeters are just sixteen inches apart, and in front of you? And then the upper mids are two and a half feet off center, and then the lower mids are a whopping twelve feet apart?!
Whoah.
And you get "real" left to right cues?
Not being sarcastic here, this is just wild stuff tho. Very intriguing.
Yes, I think it works ...
... recording dependent, of course. Complex multi-miked or multi-tracked recordings often fall apart (others work better than they should). On some recordings the image is spectacularly precise and consistent (particularly the Mike Skeet ones with a dummy head, even those where electronic sounds are mixed in - though I've also got dummy head recordings that don't work well at all).
A few multi-tracked recordings fall apart at high frequency - too many of the "high" percussion instruments are near the centre line (while the same track on headphones has some percussion off to the side). I guess they have messed up the phase info in the recording.
On good recordings there is no sense of the sound coming from any of the speakers, though sometimes the brain tries to convince you that 12" basses are producing cymbals.
The zone where it works at its best is quite restricted (left-right, not front-back) but the sound is still reasonably balanced off-axis, and still appears to come from the correct end of the room. One nice thing is when the image stays fixed when rotating your head (i.e. it does not rotate with the head) - that is a good test for correct listening position.
Is it perfect? No, of course not - there are relatively few perfectly-matched recordings, and many might sound better in stereo. The rear speakers help most with recordings that are very messed up. I suspect one of the reasons it works well is that room reflections (and I've avoided many bad ones) do not cover the whole band (highs reflect from completely different places compared to mids, etc.).
Does it work for everyone? No idea. It takes a lot of care to set up. Probably about half a dozen people have really listed to my system (i.e. been willing to sit in the hot spot and interested enough to listen), and comments have ranged from neutral to very positive. Like many who share our (community's) affliction, I don't really care - it is for me (and in theory there should be adjustment for head size, though I don't think that is very critical).
Why did I write about it? Apart from showing off a little, I was very impressed by the Takeuchi and Nelson work, and was not aware of a similar amateur attempt to get it working, so thought it might be of interest. It was just shear luck that at the time I read the remark by "dwk123" my speakers were active and modular, so I could separate the horns from the mids and both from the basses and start to experiment. Buying two more tweeters, mounting them and building an extra two channels into my amplifiers was cheap and easy. I was already using the computer and the DSP crossover.
All these comments refer to the final version, after I'd understood the OSD paper and implemented it correctly.
I still come to DIY Audio out of general interest, but it was only really when I made my comment in the other thread that I realised any incentive to make changes has gone (which surprises me, as I've been tinkering for about 27 years).
Ken
Oh, and it definitely does not solve the "significant other" problem to have 8 speakers plus subwoofers in the room.
Oh, and I never did get a good picture posted - I'll see what I can do. I've got a couple of stitched panoramas of the room, but a bit large for posting here without some work.
... recording dependent, of course. Complex multi-miked or multi-tracked recordings often fall apart (others work better than they should). On some recordings the image is spectacularly precise and consistent (particularly the Mike Skeet ones with a dummy head, even those where electronic sounds are mixed in - though I've also got dummy head recordings that don't work well at all).
A few multi-tracked recordings fall apart at high frequency - too many of the "high" percussion instruments are near the centre line (while the same track on headphones has some percussion off to the side). I guess they have messed up the phase info in the recording.
On good recordings there is no sense of the sound coming from any of the speakers, though sometimes the brain tries to convince you that 12" basses are producing cymbals.
The zone where it works at its best is quite restricted (left-right, not front-back) but the sound is still reasonably balanced off-axis, and still appears to come from the correct end of the room. One nice thing is when the image stays fixed when rotating your head (i.e. it does not rotate with the head) - that is a good test for correct listening position.
Is it perfect? No, of course not - there are relatively few perfectly-matched recordings, and many might sound better in stereo. The rear speakers help most with recordings that are very messed up. I suspect one of the reasons it works well is that room reflections (and I've avoided many bad ones) do not cover the whole band (highs reflect from completely different places compared to mids, etc.).
Does it work for everyone? No idea. It takes a lot of care to set up. Probably about half a dozen people have really listed to my system (i.e. been willing to sit in the hot spot and interested enough to listen), and comments have ranged from neutral to very positive. Like many who share our (community's) affliction, I don't really care - it is for me (and in theory there should be adjustment for head size, though I don't think that is very critical).
Why did I write about it? Apart from showing off a little, I was very impressed by the Takeuchi and Nelson work, and was not aware of a similar amateur attempt to get it working, so thought it might be of interest. It was just shear luck that at the time I read the remark by "dwk123" my speakers were active and modular, so I could separate the horns from the mids and both from the basses and start to experiment. Buying two more tweeters, mounting them and building an extra two channels into my amplifiers was cheap and easy. I was already using the computer and the DSP crossover.
All these comments refer to the final version, after I'd understood the OSD paper and implemented it correctly.
I still come to DIY Audio out of general interest, but it was only really when I made my comment in the other thread that I realised any incentive to make changes has gone (which surprises me, as I've been tinkering for about 27 years).
Ken
Oh, and it definitely does not solve the "significant other" problem to have 8 speakers plus subwoofers in the room.
Oh, and I never did get a good picture posted - I'll see what I can do. I've got a couple of stitched panoramas of the room, but a bit large for posting here without some work.
picture
here is a picture of an early setup: the first one with the extra tweeters. Now the basses, seen here under the mids, are at +/-90 degrees against the walls, and the mids are on tall stands, farther apart than shown (the "amp rack" is now behind the right mid). The upper mids are also on stands (not old speakers as here) and the tweeter holder has been improved (they are now height-matched with the upper mids and mids). The listening-half of the room has more damping (sofas, bean-bags etc.), and the listening position = camera position, is far from the rear wall.
The main tapped horn subwoofer is off to the left and cannot be seen, the one in the left corner fills in some response holes. The basses are now at ear height, and overlap with the subs a bit, to even out modes. Everything is precisely time aligned.
I can't get a good picture of the final setup in 300kB, nor will I post one elsewhere.
Ken
here is a picture of an early setup: the first one with the extra tweeters. Now the basses, seen here under the mids, are at +/-90 degrees against the walls, and the mids are on tall stands, farther apart than shown (the "amp rack" is now behind the right mid). The upper mids are also on stands (not old speakers as here) and the tweeter holder has been improved (they are now height-matched with the upper mids and mids). The listening-half of the room has more damping (sofas, bean-bags etc.), and the listening position = camera position, is far from the rear wall.
The main tapped horn subwoofer is off to the left and cannot be seen, the one in the left corner fills in some response holes. The basses are now at ear height, and overlap with the subs a bit, to even out modes. Everything is precisely time aligned.
I can't get a good picture of the final setup in 300kB, nor will I post one elsewhere.
Ken
Attachments
Minor update: 5 band OSD
While working on the cosmetics of the tweeter mount, I thought I'd try a 5 band OSD setup (I had an extra two channels of amp around, so only two cheapish vifa tweeters were needed to complete the system). It turns out that, as might have been expected, the difference from 4 band (details above) to 5 band is subtle.
I also re-optimised for 6m listening distance, and the angles etc. given here are for that new distance. (Mids and basses are closer, but delayed to give correct arrival times to within a couple of cm, and their amplitudes are reduced appropriately.) The longer distance gives a bigger sweet spot. The speakers are mostly close to walls (but with some care taken about reflections), so most of the room is free for other purposes.
Anyway, in case someone wants to try this, the angles etc are as follows:
Upper tweeter: 7.9kHz up, angle = 5.5 degrees so dist = 6m, spacing = 0.58m. (Vifa BC25SC55-04)
Lower tweeter: 3.2 to 7.9kHz, angle = 11 degrees so dist = 6m, spacing = 1.15m. (SEAS H1499)
Upper mid: 1.1kHz to 3.2kHz, angle = 27 degrees so dist = 5.96m, spacing = 2.78m.
(2" dome)
Mid: 400Hz to 1.1kHz, angle = 66 degrees so dist =3.8m, spacing = 3.97m.
(B&C 8PE21)
Bass: up to 400Hz, angle roughly 165 degrees so dist ~ 1m, spacing = 4m.
(Prec. Dev. 12SB30)
"dist" means distance along centre line to line joining the two units in one band, "spacing" means the length of that line.
Each "band" covers about 1.5 octaves, crossovers are 24dB/octave LR (acoustic, near enough).
The spurious accuracy on the frequencies comes from calculating them as 10^x where the log values are read from (e.g.) figure 18 of Takeuchi and Nelson. I optimised for up to 17kHz (beyond my hearing), and the angles and bands were chosen to give "n" values between about 0.6 and 1.4 (roughly, with some constraints from the room towards the lower/mid).
As before the Hilbert filter is accurate to around 150Hz, between 50 and 150 Hz is stereo, below 50 Hz a pair of subwoofers play mono.
The 4 tweeters are mounted on a baffle-board which is 135cm long by 12.5 cm high, supported on aluminium columns at its ends, the other speakers are on stands (or on the floor in the case of the subs).
The extra cost of 5 band over 4 is small, and so is the benefit - on some tracks high percussion does tend to blend into the overall image slightly better.
The cosmetics are now OK too, so perhaps I'm done.
While working on the cosmetics of the tweeter mount, I thought I'd try a 5 band OSD setup (I had an extra two channels of amp around, so only two cheapish vifa tweeters were needed to complete the system). It turns out that, as might have been expected, the difference from 4 band (details above) to 5 band is subtle.
I also re-optimised for 6m listening distance, and the angles etc. given here are for that new distance. (Mids and basses are closer, but delayed to give correct arrival times to within a couple of cm, and their amplitudes are reduced appropriately.) The longer distance gives a bigger sweet spot. The speakers are mostly close to walls (but with some care taken about reflections), so most of the room is free for other purposes.
Anyway, in case someone wants to try this, the angles etc are as follows:
Upper tweeter: 7.9kHz up, angle = 5.5 degrees so dist = 6m, spacing = 0.58m. (Vifa BC25SC55-04)
Lower tweeter: 3.2 to 7.9kHz, angle = 11 degrees so dist = 6m, spacing = 1.15m. (SEAS H1499)
Upper mid: 1.1kHz to 3.2kHz, angle = 27 degrees so dist = 5.96m, spacing = 2.78m.
(2" dome)
Mid: 400Hz to 1.1kHz, angle = 66 degrees so dist =3.8m, spacing = 3.97m.
(B&C 8PE21)
Bass: up to 400Hz, angle roughly 165 degrees so dist ~ 1m, spacing = 4m.
(Prec. Dev. 12SB30)
"dist" means distance along centre line to line joining the two units in one band, "spacing" means the length of that line.
Each "band" covers about 1.5 octaves, crossovers are 24dB/octave LR (acoustic, near enough).
The spurious accuracy on the frequencies comes from calculating them as 10^x where the log values are read from (e.g.) figure 18 of Takeuchi and Nelson. I optimised for up to 17kHz (beyond my hearing), and the angles and bands were chosen to give "n" values between about 0.6 and 1.4 (roughly, with some constraints from the room towards the lower/mid).
As before the Hilbert filter is accurate to around 150Hz, between 50 and 150 Hz is stereo, below 50 Hz a pair of subwoofers play mono.
The 4 tweeters are mounted on a baffle-board which is 135cm long by 12.5 cm high, supported on aluminium columns at its ends, the other speakers are on stands (or on the floor in the case of the subs).
The extra cost of 5 band over 4 is small, and so is the benefit - on some tracks high percussion does tend to blend into the overall image slightly better.
The cosmetics are now OK too, so perhaps I'm done.
Reverting to 4 band, with some changes
There was probably no benefit of 5 band over the previous 4 band setup. By accident, during switching between 4 and 5 bands, the "outer" pair of tweeters was used for the highest of 4 bands (rather than the inner pair). Contrary to expectation this seemed to work very well, and led to the best setup so far, in which, on many recordings, all of the 8 main drive units are "not there".
The basic details are as before, but with tweeters subtending 12 degrees covering 3.3kHz up, upper mids subtending 36 degrees covering 1-3.3 kHz, lower mids covering 300 Hz to 1 kHz subtending 60 degrees, and the basses and subwoofers as before. (Though I did change the large subwoofer from tapped horn to bandpass to remove some just-audible distortion which sometimes made the subwoofer locatable.)
This setup has been stable now for 6 weeks or so.
It occurred to me that, replacing the lower mids - which are at the normal locations for stereo speakers - with nearly full range speakers (Geddes Harper) would allow easy switching between stereo and OSD (by changing between saved settings on the DCX2496s). This change is currently in progress, awaiting the Harpers.
There was probably no benefit of 5 band over the previous 4 band setup. By accident, during switching between 4 and 5 bands, the "outer" pair of tweeters was used for the highest of 4 bands (rather than the inner pair). Contrary to expectation this seemed to work very well, and led to the best setup so far, in which, on many recordings, all of the 8 main drive units are "not there".
The basic details are as before, but with tweeters subtending 12 degrees covering 3.3kHz up, upper mids subtending 36 degrees covering 1-3.3 kHz, lower mids covering 300 Hz to 1 kHz subtending 60 degrees, and the basses and subwoofers as before. (Though I did change the large subwoofer from tapped horn to bandpass to remove some just-audible distortion which sometimes made the subwoofer locatable.)
This setup has been stable now for 6 weeks or so.
It occurred to me that, replacing the lower mids - which are at the normal locations for stereo speakers - with nearly full range speakers (Geddes Harper) would allow easy switching between stereo and OSD (by changing between saved settings on the DCX2496s). This change is currently in progress, awaiting the Harpers.
Harpers done - stereo option
Short note: Geddes Harpers implemented either as bass-mid in OSD, or by changing the program on a DCX, as full range - normal stereo if the other DCX and stereo convolver in foobar are turned off. So very easy to switch, and not much tonal difference. I need to listen more before deciding the tracks that work best in each setup (for multiple listeners stereo is probably better though).
Ken
Short note: Geddes Harpers implemented either as bass-mid in OSD, or by changing the program on a DCX, as full range - normal stereo if the other DCX and stereo convolver in foobar are turned off. So very easy to switch, and not much tonal difference. I need to listen more before deciding the tracks that work best in each setup (for multiple listeners stereo is probably better though).
Ken
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.