This has become a juicy thread. The comb filter effects at the high frequencies, when at all off axis, have a problem similar to the comb filter effects in a typical living room down at the Schroeder frequency area, but for different reasons. Both apparently have comb filter effects (nulls) that don't get filled in at the listener position, and both benefit from physically displaced/delayed added emissions, whether from room reflections or added transducers.
The comb filter effects at the Schroeder frequency region (80 - 250HZ) have too few secondary room reflections to fill in the acoustic comb filter nulls, so you get boominess, and forever think tone controls only make things worse (which they might, but not because of the tone control).
According to another diyaudio thread on multiple woofers, I think it was Earl Geddes that showed with graphs that adding multiple woofers around the room makes a substantial improvement in bass flatness at the listener position. Each woofer creates comb filter effects, but the nulls are at different frequencies, so they tend to fill each other in, resulting in a smoother bass response at the listener position.
At the high frequency end, where the wavelengths are so short that being off axis horizontally just a foot can make comb filter effects noticeable, any attempt at phase shuffling or adding delays in the recording/mixing process won't change the fact that the two tweeters at the playback end will still have this issue. Sidewall reflections are apparently helpful to reduce this off axis comb filter effect at the high frequencies, but will muddle imaging in lower frequencies (1kHZ - 6kHZ ? ) where embedded imaging should dominate.
So apparently a good way to improve the phantom center would be to pad side walls in a way that is still highly reflective for frequencies above 6kHZ, to create additional reflections at the highest frequencies, much as one might do with multiple woofers handling below 150HZ or so, but also be absorptive at least in the range of 1kHZ - 6kHZ so stereo effect imaging is minimally damaged. And/or add rear firing tweeters that add into the mix another comb filter mechanism above 6 kHZ (assuming the speakers are near a front wall so there will actually be added reflections from the rear tweeter reaching the listener with time delay), so there is something to help fill in the off axis comb filter nulls above 6kHZ. Due to the size of the wavelengths at 6kHZ the improvement may be small, but maybe still worth it.
The comb filter effects at the Schroeder frequency region (80 - 250HZ) have too few secondary room reflections to fill in the acoustic comb filter nulls, so you get boominess, and forever think tone controls only make things worse (which they might, but not because of the tone control).
According to another diyaudio thread on multiple woofers, I think it was Earl Geddes that showed with graphs that adding multiple woofers around the room makes a substantial improvement in bass flatness at the listener position. Each woofer creates comb filter effects, but the nulls are at different frequencies, so they tend to fill each other in, resulting in a smoother bass response at the listener position.
At the high frequency end, where the wavelengths are so short that being off axis horizontally just a foot can make comb filter effects noticeable, any attempt at phase shuffling or adding delays in the recording/mixing process won't change the fact that the two tweeters at the playback end will still have this issue. Sidewall reflections are apparently helpful to reduce this off axis comb filter effect at the high frequencies, but will muddle imaging in lower frequencies (1kHZ - 6kHZ ? ) where embedded imaging should dominate.
So apparently a good way to improve the phantom center would be to pad side walls in a way that is still highly reflective for frequencies above 6kHZ, to create additional reflections at the highest frequencies, much as one might do with multiple woofers handling below 150HZ or so, but also be absorptive at least in the range of 1kHZ - 6kHZ so stereo effect imaging is minimally damaged. And/or add rear firing tweeters that add into the mix another comb filter mechanism above 6 kHZ (assuming the speakers are near a front wall so there will actually be added reflections from the rear tweeter reaching the listener with time delay), so there is something to help fill in the off axis comb filter nulls above 6kHZ. Due to the size of the wavelengths at 6kHZ the improvement may be small, but maybe still worth it.
I am not really reflection free, but have them down by at least -20 dB in the ETC plot. I used absorption, 3 (pretty huge) panels total, placed strategically and hidden except for one panel that doubles as a poster. That was the limit for my spouse. The rest is processing.
My system doesn't function without EQ. It's a key design feature (lol). It's all in that huge thread of mine. I guess you could say it's LEDE as I was definitely inspired by that concept, but in an artificial way here and there but also including a Haas kicker (artificial as well). Learned a lot though, and I trust my measurements to guide me.
Pano's current room looks like this:
Following the link you posted above I was expecting to see his previous room, Altec A5 setup in a pretty dedicated room, but I've seen that one before too. His new cave has some nice qualities though! And yes, it's a private one deserving the name "man cave".
http://www.diyaudio.com/forums/multi-way/269864-caveman-speakers-troglodytes-6.html
The above linked page has some of my plots there as well.
That would be pretty hard to do in Pano's room!
It is something on my mind (has been for quite a while) though.
My system doesn't function without EQ. It's a key design feature (lol). It's all in that huge thread of mine. I guess you could say it's LEDE as I was definitely inspired by that concept, but in an artificial way here and there but also including a Haas kicker (artificial as well). Learned a lot though, and I trust my measurements to guide me.
Pano's current room looks like this:
Following the link you posted above I was expecting to see his previous room, Altec A5 setup in a pretty dedicated room, but I've seen that one before too. His new cave has some nice qualities though! And yes, it's a private one deserving the name "man cave".
http://www.diyaudio.com/forums/multi-way/269864-caveman-speakers-troglodytes-6.html
The above linked page has some of my plots there as well.
That would be pretty hard to do in Pano's room!
It is something on my mind (has been for quite a while) though.
Last edited:
Earl method of adding multiple subs may very well help for comb filtering, but id personally like to see the ETC. his LF must be absorbed and without bass traps, it seems that adding subwoofer will help the FR response but the ETC will show a very different story. I really wonder how Earl method can help to reduce decay time below the Schroeder area. In any room, you must have as low decay time below the Schroeder area as the rest of the frequency range.
it seem that while smoothing the FR, earl method do not address decay time, his method may even adds more problem. I know that acoustician find ETC much more informative about the room response then FR.
I agree that you must retain some HF reflections in the room, but they need to be diffused or absorb enough at the LP.
absorption should be used only for early reflection points and bass traps (while also fronting the bass traps with kraft paper or any similar membrane to brings back the highs in the room). I have done the mistake to over damp my room resulting in a dead room which I didnt like at all.
''I would still personally try to at least avoid early strong reflections and try to create some kind of diffuse return to the listening position in order to keep some life in the room instead of ending up with a “dead” sounding space (unless this is what is requested naturally)"
Last edited:
I'm not sure what ETC stands for, but I'll guess time domain response. Although multiple signals with random individual time delays can fix up the frequency domain response, it would be at the expense of time domain response, which apparently is the least of the evils. Any room is going to destroy the time domain response anyway. Apparently that can be a good thing, except for the 1kHZ - 6kHZ region where we might want the embedded imaging info to dominate.
" “Decay time” might be a valid term sometimes but in general, all we are really interested in is the shape of the ETC (and perhaps the slope of the linear fit (after sparse reflections) of the Schroeder integral, in order to evaluate the “decay times” at higher frequency bands."
it is ime absolutely detrimental
you do want reflection back at the LP, but you want them diffused (-10 to -30 db).
the more you let early reflection untreated, the more you will have a uneven ETC across all the range. how that can be seen as a "good thing" is very illogical to me.
it is ime absolutely detrimental
you do want reflection back at the LP, but you want them diffused (-10 to -30 db).
Last edited:
youknowyou No, that is not my room nor anyone I have ever visited. Don't know where that photo came from. See wesayso's photo above
I have a lava tube cave that is very reflective, but highly chaotic. No parallel surfaces, almost no smooth surfaces. The sound is dry, not reverberating much at all. It's surprising, considering that it's a cave.
I have a lava tube cave that is very reflective, but highly chaotic. No parallel surfaces, almost no smooth surfaces. The sound is dry, not reverberating much at all. It's surprising, considering that it's a cave.
The signals in an electrical cable are transmitted at close to the speed of light. Between a 10 foot and 20 foot wire, there would be a delay of roughly 10 nanoseconds. This would cause destructive interference (half wavelength cancellation) starting at 50mHz. Differences in cable length are simply a non-issue as far as timing / phase at audio frequencies.
A few comments on room treatments / early reflections / time behaviour below the transition frequency:
Treating early reflections often messes up the ETC by absorbing more high frequencies than mid frequencies. Good reason to avoid treating the sidewalls with say, that egg-crate foam. It dampens the high frequencies but nothing else and effectively messes up the tonal balance of the early reflections. There is also substantial support for the notion that early reflections are benign and/or helpful when they are spectrally correct, and need not be diffuse.
Below the Shroeder (transition) frequency the room effect becomes modal, not reflective. It also becomes minimum phase - that is, when you correct the frequency response you also correct the time response. Fix one, fix the other.
Hard proof that equalization kills room modes - Acoustic Frontiers
Treating early reflections often messes up the ETC by absorbing more high frequencies than mid frequencies. Good reason to avoid treating the sidewalls with say, that egg-crate foam. It dampens the high frequencies but nothing else and effectively messes up the tonal balance of the early reflections. There is also substantial support for the notion that early reflections are benign and/or helpful when they are spectrally correct, and need not be diffuse.
Below the Shroeder (transition) frequency the room effect becomes modal, not reflective. It also becomes minimum phase - that is, when you correct the frequency response you also correct the time response. Fix one, fix the other.
Hard proof that equalization kills room modes - Acoustic Frontiers
The best way to fix the phantom center is to have an actual center channel
You can't, unless you have a discrete center channel source (ie. a 5.1 mix) that you are downmixing to stereo.
Messing around in Mid/Side is an inherently lossy or destructive process, as is messing around in unlinked L/R.
Imagine you have 4 specific sounds mixed into stereo:
-Left sound (L)
-Right sound (R)
-Center sound (mid - equal in both L and R and same polarity)
-Out-of-phase sound (side - equal in both channels but polarity inverted in one)
The mid and side sounds can be extracted and operated upon independently by mid-side processing, but the discrete L and R sounds cannot. Similarly, in unlinked "dual mono" L and R sounds can be extracted and operated upon independently, but the mid and side channels cannot.
Pick your poison... or just accept that you have run into a hard limitation of stereo reproduction and pursue multichannel instead.
You can't, unless you have a discrete center channel source (ie. a 5.1 mix) that you are downmixing to stereo.
Messing around in Mid/Side is an inherently lossy or destructive process, as is messing around in unlinked L/R.
Imagine you have 4 specific sounds mixed into stereo:
-Left sound (L)
-Right sound (R)
-Center sound (mid - equal in both L and R and same polarity)
-Out-of-phase sound (side - equal in both channels but polarity inverted in one)
The mid and side sounds can be extracted and operated upon independently by mid-side processing, but the discrete L and R sounds cannot. Similarly, in unlinked "dual mono" L and R sounds can be extracted and operated upon independently, but the mid and side channels cannot.
Pick your poison... or just accept that you have run into a hard limitation of stereo reproduction and pursue multichannel instead.
Hehe, so it seems I'm doing everything in the right order. Broad band acoustic panels from GIK, two corner soffit traps for the bass, and then a miniDSP for the subwoofer.
I didn't realize that the "room correction" EQ I was adding would also positively affect the phase and group delay though, and I thought it was just kind of masking the room nodes. It's good to know it can actually prevent the ringing altogether. It would correlate with how good everything sounds when I finished though.
I like it!
Erik
I didn't realize that the "room correction" EQ I was adding would also positively affect the phase and group delay though, and I thought it was just kind of masking the room nodes. It's good to know it can actually prevent the ringing altogether. It would correlate with how good everything sounds when I finished though.
I like it!
Erik
Last edited:
I do think Mike has a point. There's no real clean cut applying mid/side EQ on a stereo signal. I've shown it on my thread somewhere. Boost the side info compared to the mid and you'll get an inversed copy in the opposite channel. It will be way down in level depending on the amount of boost.
If that's what Mike means by lossy he would be right. If a mix was made in 5.1 from the start you wouldn't have the same losses as you would have a pure center, left and right channel and no shared content. Except anything panned between those pure channels would suffer the same faith, say something between the hard left and center channel. You'd have similar losses once you decide to EQ one of those channels separately as the shared content between both channels would be altered as well. No free lunch anywhere except when you are the one making the mix.
I can live with the side effect (no pun intended) of mid/side processing, as I've played around with it for quite a bit to get a grip on the theoretical drawbacks. Here's an example:
Compared here are a left tone, a center tone and a right tone with the sides boosted. It's clear to see the inverted signal appearing in the opposite channel.
The clean version without any boost looks like this:
So, in theory and practice you do get a penalty for playing with mid/side processing. But perception is another matter. That inversed copy enhances the sound heard on the other side. This kind of mid/side processing was mentioned by Blumlein to increase the perceived width in stereo. Often referred to as a Blumlein shuffle.
http://www.audiosignal.co.uk/Resources/Stereo_shuffling_A4.pdf
Once you're aware of all of that it's up to the end user to use it or not. Personally I do not have a problem with it as it does exactly what I want from it. Make the side content more coherent(*) and solve my tonal differences between the center and the sides by using a variation of the "S" curve. Two birds...
(*)= that one is explained in the S-curve document linked earlier in this thread. The part about a difference in width perception depending on frequency.
If that's what Mike means by lossy he would be right. If a mix was made in 5.1 from the start you wouldn't have the same losses as you would have a pure center, left and right channel and no shared content. Except anything panned between those pure channels would suffer the same faith, say something between the hard left and center channel. You'd have similar losses once you decide to EQ one of those channels separately as the shared content between both channels would be altered as well. No free lunch anywhere except when you are the one making the mix.
I can live with the side effect (no pun intended) of mid/side processing, as I've played around with it for quite a bit to get a grip on the theoretical drawbacks. Here's an example:
Compared here are a left tone, a center tone and a right tone with the sides boosted. It's clear to see the inverted signal appearing in the opposite channel.
The clean version without any boost looks like this:
So, in theory and practice you do get a penalty for playing with mid/side processing. But perception is another matter. That inversed copy enhances the sound heard on the other side. This kind of mid/side processing was mentioned by Blumlein to increase the perceived width in stereo. Often referred to as a Blumlein shuffle.
http://www.audiosignal.co.uk/Resources/Stereo_shuffling_A4.pdf
Once you're aware of all of that it's up to the end user to use it or not
. Personally I do not have a problem with it as it does exactly what I want from it. Make the side content more coherent(*) and solve my tonal differences between the center and the sides by using a variation of the "S" curve. Two birds...(*)= that one is explained in the S-curve document linked earlier in this thread. The part about a difference in width perception depending on frequency.
Last edited:
Thanks for clarifying with those graphics, wesayso.
That's exactly what I meant by "lossy" - MS processing cannot yield a "pure improvement" since any adjustment involves crossfeed, either normal polarity (add mid) or inverted polarity (add side). Because of this, it will never be a best practice for the stereo record/reproduction system - more of an aesthetic tool, or in terms of correction: a last resort!
Independently EQ'ing the mid or side channel can do strange things to the stereo image too. Stereo is *capable* of reproducing a smooth panorama through:
not-center (L/R inverted polarity) - hard left (L only) - center (L/R same polarity) - hard right (R only) - not-center (L/R inverted polarity)
By independently EQ'ing / processing either mid/side, you "pull apart" the image which creates a hole in between center and L/R. At extreme settings you will end up image that has only two areas - center, and not-center.
As for 5.1 mixes, I wouldn't worry too much about the areas between center and L or center and R. Center is usually quite independent from the L/R mix. But, a good point nonetheless. Depends on how the mix was produced.
Of course it's cool that this technique is yielding a subjective improvement for you, and I've also used it myself to compensate for when my speakers are placed too close together. (but mostly in mixing/mastering) I'd like to read up on that Blumlein and S-curve stuff, but first I'm also going to run some tests along the lines of what you've been doing. Easier for me to understand than theorizing!
That's exactly what I meant by "lossy" - MS processing cannot yield a "pure improvement" since any adjustment involves crossfeed, either normal polarity (add mid) or inverted polarity (add side). Because of this, it will never be a best practice for the stereo record/reproduction system - more of an aesthetic tool, or in terms of correction: a last resort!
Independently EQ'ing the mid or side channel can do strange things to the stereo image too. Stereo is *capable* of reproducing a smooth panorama through:
not-center (L/R inverted polarity) - hard left (L only) - center (L/R same polarity) - hard right (R only) - not-center (L/R inverted polarity)
By independently EQ'ing / processing either mid/side, you "pull apart" the image which creates a hole in between center and L/R. At extreme settings you will end up image that has only two areas - center, and not-center.
As for 5.1 mixes, I wouldn't worry too much about the areas between center and L or center and R. Center is usually quite independent from the L/R mix. But, a good point nonetheless. Depends on how the mix was produced.
Of course it's cool that this technique is yielding a subjective improvement for you, and I've also used it myself to compensate for when my speakers are placed too close together. (but mostly in mixing/mastering) I'd like to read up on that Blumlein and S-curve stuff, but first I'm also going to run some tests along the lines of what you've been doing. Easier for me to understand than theorizing!
Speaking of side mixes, has anyone ever recorded any variation of an acoustic band using boundary mics on side walls for below about 900HZ, and a stereo pair "head mic" arrangement in the middle for above about 900HZ? I have a suspicion that this could work well with the shortcomings of a typical playback set-up... (which will have inter-aural crosstalk).
Interesting idea Bob.. I suppose that is to widen the low frequencies? What do you mean by head mic? Surely not a binaural head, but an ORTF/NOS pair? (Spaced and angled) I'm reading the Gerzon paper now to learn more about the goals of the shuffling circuit...
I have been messing around with the Voxengo MSED plugin and some test tones.
The test tone (top track) pans from:
- R(-) to L(+) through side
- L(+) to R(+) through mid
- R(+) to L(-) through side
- L(-) to R(-) through mid
The bottom track is processed through the MS plugin which boosts the side channel by 12dB and attenuates the mid channel by 12dB. A dramatic example.
By looking at the change in slope, you can see how the imaging has become "compressed" or rather, pulled out towards the sides. There are no longer discrete L and R sound sources, but fairly homogeneous variations on the out-of-phase Side channel.
(I'm sure there is a better way to present this information than the dual mono waveforms ... perhaps I will have to get clever with a graphing program.)
I have been messing around with the Voxengo MSED plugin and some test tones.
The test tone (top track) pans from:
- R(-) to L(+) through side
- L(+) to R(+) through mid
- R(+) to L(-) through side
- L(-) to R(-) through mid
The bottom track is processed through the MS plugin which boosts the side channel by 12dB and attenuates the mid channel by 12dB. A dramatic example.
By looking at the change in slope, you can see how the imaging has become "compressed" or rather, pulled out towards the sides. There are no longer discrete L and R sound sources, but fairly homogeneous variations on the out-of-phase Side channel.
(I'm sure there is a better way to present this information than the dual mono waveforms ... perhaps I will have to get clever with a graphing program.)
Attachments
Does anyone know what these guys are doing? They were on Home Theater Geeks about a month ago. One of them is a PhD in acoustics.
immersavtechnology
immersavtechnology
For the 5.1 mixes, in the case I described, the left and center will contain an equal amount to present a queue of something in between those two channels. Eq-ing either one separately, left or center will mess up that balance.
If the content is not the same anymore in the left and center, that panned sound between left and right will exhibit the same hole as a mid/side processed stereo file gone wrong.
The biggest difference between my mid and side EQ is less than 3 dB, I'm not worried
. I'm not widening the stage, and not narrowing it either. I just went for a bit of balance between bass and high frequencies, which makes wide spread backing vocals more full bodied while making the phantom center less dark.
Last edited:
Thanks for this link. I've been wanting this technology to evolve for a long time. They may be doing what I'm going to do with my head mic above 1kHZ and side wall boundary mics for below 1kHZ, but with just a head mic and some fancy digital processing to create the right timing cues below 1kHZ which get lost due to the second occurrence of inter-aural crosstalk at the playback end. Their knowledge and resourses appear to be second to none, so I'm excited to see how this will evolve.
