I have almost no practical experience with anything functional (listenable) in this regard, mostly theoretical analysis.
In my analysis I was trying to determine how electronics could be used to augment the poor modal interactions found in small room at LFs. I found that a random series of impulses will decorrelate each source from each other. The pulses need to decay in time. I noticed that this impulse train looks very much like a reverb algorithm.
Now to what extent any existing unit will do what's necessary, I cannot comment.
Experimentation will be required.
I would say that the initial delay should be around 50 ms., decay rate will have to be determined by listening tests.
The one and only time that I did try this with real hardware, I found that yes, it enhanced the perception of bass, even though the bass level did not change and it made the bass clearer. But the reverb was also noticed and this I thought was not so great. I think that we needed a faster decay, but yes the technique did seem to work.
I would make each sub have a different delay and decay so that they are all decorrelated to the maximum extent. This would take several channels of reverb however and may not be practical in a real system. You will just have to try it and find out.
Only the subs should have reverb, the mains should not, and, of course, the mains must not be HP'd as this won't work right.
Good luck and let us know how it works out.
In my analysis I was trying to determine how electronics could be used to augment the poor modal interactions found in small room at LFs. I found that a random series of impulses will decorrelate each source from each other. The pulses need to decay in time. I noticed that this impulse train looks very much like a reverb algorithm.
Now to what extent any existing unit will do what's necessary, I cannot comment.
Experimentation will be required.
I would say that the initial delay should be around 50 ms., decay rate will have to be determined by listening tests.
The one and only time that I did try this with real hardware, I found that yes, it enhanced the perception of bass, even though the bass level did not change and it made the bass clearer. But the reverb was also noticed and this I thought was not so great. I think that we needed a faster decay, but yes the technique did seem to work.
I would make each sub have a different delay and decay so that they are all decorrelated to the maximum extent. This would take several channels of reverb however and may not be practical in a real system. You will just have to try it and find out.
Only the subs should have reverb, the mains should not, and, of course, the mains must not be HP'd as this won't work right.
Good luck and let us know how it works out.
Last edited:
To find the correlation one needs to do a two channel test and then a cross correlation between these two data sets. But I am not sure what that will tell you. I never actually do this, I just find the best set of parameters for a flat response in a given situation. The idea of correlated sources is useful in theory to think about what is happening at LFs in rooms, but in practice it isn't too useful.
The reason why I asked is that in steady state measurements the effect of reverberation cant be seen due to long averaging time, so, wanted to see the impact of reverberation on de-correlation in frequency response of each sub or some other way.
A satirist might say, "I built the perfect speaker (at 5 cm) but the darn room spoiled it (at my chair)".
A builder might honestly say, "the room makes a mess of the lovely FR curves I get up close and that match my sim perfectly." But the room is not the problem, that's simply the complex reality.
Your post is not the analysis. Your post is the real question: what do I measure to get data that relates to my human experience of hearing of reproduced music at my chair, reverberation and all?
B.
I'm sorry but I don't follow your idea.
At low frequencies our hearing has a "long time average" so you are measuring what it is you want.
Reverberation becomes decorrelated to the source, but this happens slowly at about 10+ periods of the sound (fast at HFs, slow at LFs)
At any rate, I have no idea what a measure of the correlation between subs would mean.
Please be sure that you understand what correlation of sound signals is, because you may be confused about it.
Yes, exactly my point - almost.
Now, I suppose that somewhere there are very complicated tables relating physical loudness AND decay to perceived loudness at various frequencies.
Likewise, you can choose the width of your time-window on your measurements. Maybe it will relate to those tables or maybe off a lot.
Tuning to a room is according to your ear since there is no solid way to connect the two realms of perceived loudness and measured. But you can do FR measurements in a stable manner just as you can play the same familiar recordings. And interactively arrive at the sound colour you like.
B.
The sliding window that many measurement systems can do is a good approximation to our hearing down to about 400-500 Hz. Below that it is not as accurate, but better than a fixed window.
That's exactly the kind of link between acoustics and perception that I'd like to know more of. (I think REW offers that preference hidden away in the fine-print.) Thanks.
B.
Windowing or gating - below 500hz - in a normal/small listening room?
I thought it was to make a time based window, that "removes" unwated reflections in the measurement - which is kind of impossible when that wavelengt becomes to long in relation to room dimensions....
In theory, the window should get longer at LFs than a linear sliding window will. In practice, what I do, is to use the sliding window for everything above Fs and then just steady state below that. This too is not exact, but pretty close.
You've said a couple of times now that you don't use a crossover on the subs.
What do you do then? Let them play fullrange? Isn't that going to interfer with the main speakers.
Read it again
As I recall... there is LP on the subs but no HP on the mains. If I remember correctly, this is to make the mains blend well with the subs - and you get two extra sources to even out the low frequencies.
The mains just run as deep as they can - while the subs are crossed individually - different slopes, different EQ and different levels, so that the total response becomes smooth.
Read it again
As I recall... there is LP on the subs but no HP on the mains. If I remember correctly, this is to make the mains blend well with the subs - and you get two extra sources to even out the low frequencies.
The mains just run as deep as they can - while the subs are crossed individually - different slopes, different EQ and different levels, so that the total response becomes smooth.
Yes I get that about the mains and I don't HP them. They are Geddes speakers by the way.
But you say the 'subs are crossed' with an LP. For me that is not what I understand of no crossover on the sub. (It can be a semantic question)
Hmm... maybe I remember wrong
I was just sure that the subs had to be crossed individually, so that you get the best summation - meaning mains runs down, falling off with a natural roll off. Then the subs are crossed maybe on at 110hz on at 90hz and one at 140hz - undtil everything blends nicely - also with different slopes - meaning that they gotta have some kind of LP filters to do so
Let's not use the word "crossed" as it is misleading here.
Mains are full range (except for a single notice at about 30 Hz where the room will boom badly is I let all drivers play, so I cut out this frequency.)
Each sub will inevitably have a LP filter of some slope somewhere, but they will all be different. This means that by about 150 Hz all the subs are gone.
It's pretty simple really. Getting there is anything but. It takes some practice, I believe, to get the settings right. It's probably not a "one shout deal."
Mains are full range (except for a single notice at about 30 Hz where the room will boom badly is I let all drivers play, so I cut out this frequency.)
Each sub will inevitably have a LP filter of some slope somewhere, but they will all be different. This means that by about 150 Hz all the subs are gone.
It's pretty simple really. Getting there is anything but. It takes some practice, I believe, to get the settings right. It's probably not a "one shout deal."
I only have 8 channels of DSP right now. two 3 way mains and 4 subwoofers. So I have 2 subwoofers on the right and two on the left - but places where they fit and asymmetrical - one in a corner, one next to the sofa, one close to a main and so on.
Two subwoofers on the left are driven in parallel by one amp, and given it's own signal from the DSP - same on the right. This is not perfect, but somehow it works - maybe simply because, an even frequency response is way more important than how we achieve it.
I have +/-5dB from 20hz all the way up.... getting closer to +/-1,5dB as we go past the 500hz mark - which could have a lot to do with both my measuring technique and the shorter wavelenght.
In how big an area around the listening position should one expect and even bass response? and within what limit? +/-2dB?
Two subwoofers on the left are driven in parallel by one amp, and given it's own signal from the DSP - same on the right. This is not perfect, but somehow it works - maybe simply because, an even frequency response is way more important than how we achieve it.
I have +/-5dB from 20hz all the way up.... getting closer to +/-1,5dB as we go past the 500hz mark - which could have a lot to do with both my measuring technique and the shorter wavelenght.
In how big an area around the listening position should one expect and even bass response? and within what limit? +/-2dB?
Last edited:
Think of a circle of radius 1/8 wavelength. Within this circle the SPL cannot vary more than a dB or so. Clearly at LFs this means that the entire seating area is pretty stable, but as the frequency goes up this radius becomes so small that spatial variations of significance can occur.
For this reason, I do not do much spatial averaging at LFs as there just isn't much change. At the Fs frequency we might see some variation that should be accounted for but not so much below that. Still a +/-2dB response would be very good, and mostly hard to achieve, but doable.
Ahh... I see. Thank you
When I first started to experiment with multiple subwoofers, I noticed that sometimes corners in even larger rooms, would stop with the typical "buildup" of the lower frequencies . exactly like you point out - it evens out and the big variation that used to be huge - now seem to be smoothed out considerably.