It is quite a relief having you saying that things worked at a good level with my project!
The problem is that as I mentioned before, all the lumber work was done with the assistance of my father who currently has moved about 300 km away till next winter, so there is literally no way to do any changes or new constructions at this point. In addition, there is another issue. The house belongs to my parents and it was already a bit of difficult to convince them to allow me to make holes in the walls and ceiling. Thus I'm afraid it will be quite unlikely they will let me go further with this, so I only hope that any further modification will not demand making any more holes
OK. Understood! So, maybe when your dad comes back for the winter, he could help you modify the cloud, and just us the same holes to hang it in the same place again? Just an idea...
The problem is that I thought that it is secondary compared to the FR analysis. I mean I thought it wouldn't make much sense as long as the FR was still a mess. But after reading your reply I realized that this is not really the case. Thanks a lot for clarifying!
Right! Tis is confusing for many people, because "frequency response" is just logical and easy to understand, and you see it all over the place, with graphs or specs for pretty much every type of studio gear, from speakers to mics to consoles to outboard gear: Everywhere you look, it's all about frequency response. And that's fine for purely electronic stuff, but for acoustics, the "time domain" response is just as important, and personally I think it is even more important. It's not intuitive for many people.
But I have a technical question. What is the point of negative time values in the Spectrogram? I suspect it has something to do with phase, otherwise it makes no sense to me, since there is no point in having negative values. But maybe I am missing something here...
It's more a mathematical oddity than a "real" thing, in the sense of something you could hear BEFORE the impulse, but it does have real world implications. If you look at the impulse response graph, for example, and zoom out a lot, you'll see that over on the left (negative time: before the impulse), there's a series of small "copies" of the impulse itself. That's a result of the way the mat works, and those actually are related to distortion in the system... but of course, they did note really happen before the impulse! That would be impossible. It's just an artifact of the way the math works. But REW does use that information in calculating the distortion graphs. Each of the "miniature compressed copies" of the impulse is actually one of the distortion harmonics, but working backwards from right to left. So the first one you see (closest to the impulse itself) is the second distortion harmonic, the next one to the left is the third harmonic, then the fourth, and so on. Another way of thinking of it is that the main impulse response shows the rooms linear response, and the others show it's non-linear response. The main impulse that starts at time zero is the response of the room without any distortion, and the distortion harmonics appear before that, out to the left "before" time zero... but only because that's the way the math works out.
It's complicated!
Also, how can you get the waterfall analysis to show you both before and after measurements simultaneously
This is a bit convoluted, but very useful! First, make sure that you have actually generated the waterfall plots for both of the data sets that you want to see. If you have not yet generated a graph for one of them, you won't be able to overlay it. Then, select one of the waterfalls that you want to use, and click on the "controls" button, on the far right hand side of the window, just above the end of the graph. It looks like a gear wheel. That opens up a sub-window with some controls you can use to change the way the graph appears. Down in the bottom left corner of that, there's three controls for handling overlays. The first time you see it, the top one will just say "overlay transparent", the second one tells you the percentage transparency, and the bottom one is where you can select the OTHER waterfall that you want to overlay here. IF you click on the down arrow next to that, you get a list of all of the data sets in the current REW session. The ones that have NOT been generated yet are shown in gray, and you cannot select those. The ones that HAVE been generated are shown in white, and you can select any one of those to overlay. Then you can adjust the other two controls above that one, to change the way it is displayed. When you don't want to see the overlay any more, then just go back to the bottom control, click on the down arrow again, and select "No Overlay".
I thought -10db would be fine, but -20 is quite challenging! Is it really so much important to get it that low?
There's some debate among acousticians about what is "acceptable" in a studio, but what the science actually shows is that it needs to be -20 dB down for the first 20 ms in order for it to not cause a problem, psycho-acoustically. -10 dB means that it is half as loud as the direct sound (subjectively), but one tenth of the intensity (objectively). -20 dB means that it is one quarter as loud as the direct sound (subjectively), and one one-hundredth of the intensity (objectively). That's a big difference in intensity, even though it doesn't seem that big subjectively. But your ear works on the absolute intensity if the sound waves BEFORE it gives you its subjective impression of how loud things are, so really its the intensity that matters here. So a reflection at -10 dB is actually ten times stronger than one at -20 dB. The stronger the reflection is, the more problematic the issue becomes.
There's also the issue of timing: Psycho-acoustic studies show that if the reflection arrives at least 30ms after the direct sound, your ear can probably already determine that it is a distinct, separate sound. Certainly, after 50ms the two sounds are distinguishable. going the other way, at 20ms it might still be possible to discern the direct and reflected sound as being separate and distinct, but for the majority of people, they hear both together as one single sound that is sort of "blurred" in time, comes form a different direction, and has a different frequency balance. As times get shorter, the problem gets bigger.
There's also the issue that the relationship between these two parameters is not fixed: A louder reflection will always cause a bigger problem, even for longer delays, while a softer reflection might not cause any problem at all, even if the delay is very short.
Add to that the issue of "masking", where a loud reflection can actually prevent you from hearing what comes right after it, under some circumstances, as well as "forward masking" where a very loud reflection can also suppress your perception of sounds that came just BEFORE the reflection did, which is a bit unexpected. It all has to do with the very complicated way that our ears and brains work. And some of this stuff is used in things like "MP3 compression": it removes sounds from the audio stream if w would not be able to hear them due to "masking", as well as other similar psycho-acoustic artifacts. If our ears did not work like this, then MP3 compression would not work either.
So, that's a long way of saying: it's complicated!
Here's a diagram that might help make it clearer. It shows how the various levels and delays affect our ability to determine all of this:
- Reflection Audibility.jpg (30.69 KiB) Viewed 47487 times
- Reflection Audibility.jpg (30.69 KiB) Viewed 47487 times
You can see where the "20-20" criteria comes from on that graph! If the reflection is -20 dB down,for delays up to about 20ms it is not audible. But if the level is -10dB, it registers for ALL time delays, and it fools your brain into hearing the sound as though it were more spacious than it really is. In your case, your reflection is very much in that area of "spaciousness", meaning that it is messing with your ability to determine direction and frequency, and giving you a false sense that the music is more "surrounding" than it really is.
Related to the above, there are several "philosophies" or "concepts" for building control rooms, with names like "NER", "RFZ", "CID", etc. Each of those precisely defines the acoustic response characteristics that the room must achieve. One of those is the ideal "20-20" criteria, which means "20ms delay, 20 dB down". (And that's also where my screen name comes from! "SOUNDMAN2020" has nothing at all to do with the year 2020.... I've been using it that name more than a decade now! It's all about this "20-20" criteria that a room must meet in order to be good: no reflections greater than -20 dB in the first 20ms). Now, that's the ideal situation: to not have any reflections at all for the first 20ms, but to do that properly means that you need a large room. Sound travels roughly one foot for every millisecond, so you need to have at least ten feet between your head and the closest reflective surface behind you, plus there's often a 3 foot deep bass trap behind that reflective surface, so you need at least 13 feet distance just behind your head, plus a corresponding distance in front of you... thus, a room that fully meets the 20-20 criteria needs to be abut 20 feet long. That's a bit more than 6m long. Most home studios are not big enough to do that, so we have to make do with shorter times. 15ms is reasonable, and down to about 12ms is still acceptable. Now, as I mentioned before, as the delay time gets shorter, the level needs to be lower in order to not produce serious psycho-acoustic artifacts. So, while a level of -10dB might be fine for a large room with a long 25ms delay, you do need more attenuation for a small room.
Unfortunately, there's no simple rule that says you need xx db for every zz ms... it would be nice if there were, but there isn't: people are different, rooms are different, perception is different. There are only guidelines. All I can suggest is measuring the distance from your head to the closest reflective surface behind you, and if that is less than 10 feet (3m), then make sure that there are no reflections at all for the time that it takes for sound to get back to your ears (ie, less than -20dB), then after that time, keep the level under -15 dB, and allow it to decay at the correct natural rate for your room, which is easy to calculate. If you happen to have a room that is large enough to allow more than 20ms for the first reflections, then you can relax that levels a little, and allow something slightly higher, but certainly not as high as -10 dB. That's too strong, even for a large room.
Wow! I didn't know REW could present IR graph with just sweeping tone measurement analysis. I thought it demanded a momentary pulse tone or a "clap", "shot" etc... I remember doing such measurements back in the University days along with fourier transforms and moving from time to frequency domains,
Exactly! And that exact sam Fourier transform is reversible! You did it one way, but it is also possible to go the other way: take the frequency response, and use the reverse Fourier transforms to get back to the impulse response. That's what REW does. Time and frequency are flip sides of the same coin, and it is simple to convert between them. Which is why REW can do a log frequency sweep, then convert that to the actual IR (Impulse response), which it then uses to calculate many of the other graphs.
The problem with trying to do measure a true impulse response, is that it is impossible!
In theory, the impulse needs to be infinitely short, and therefore it has to be infinitely loud... which, of course, is impossible. All real-world impulses, such as a balloon bursting, or a gunshot, or wood planks slapped together, are non-zero (they do last for a certain time, even if it is very brief), and thus they are not true impulses. Also, they are not loud enough! Even though they are very loud, because the last a certain non-zero time, they are not loud enough. They would need to be something like 190 dB SPL (in other words, almost a shock wave) to create a true impulse. Once again, it is impossible to actually generate such a sound in real life. And even if you could, such extreme over-pressures would probably damage the room...
So the only realistic way to derive the impulse response for a room, is by using some type of sound at a lower level, for a longer time, and covering all frequencies. There are various types of sound that can be used, such as white noise, pink noise, MLS, and swept sin waves. Each has its advantages and disadvantages. REW uses the swept sine technique, which is a good compromise between the limitations of the other methods.
Please forgive me if I am mistaken but I have an objection: According to my calculation the further distance travelled (assuming c=343m/s and 3.08ms time travelled) should be 1.056m , not 1.1m .
True... assuming that sound really is traveling at 343 m/s in your room! Did you take into account the air temperature?
Also, the measurement can't be extremely accurate, because you are not measuring over a continuous unbroken time period, but rather a series of samples, separated by time (which depends on your audio interface), plus the reflection is not a sharp impulse but rater spread over time, and in fact your reflection is not just one single reflection, but two at nearly the same time! And the timing changes between measurements, and as the treatment changes. REW tries to estimate a correction for the timing differences, and in your case it says that the actual impulse is delayed by -0.053ms, which is 18mm. If you tell it to adjust for that presumed delay, and zoom in for higher level of detail, this is what you get:
That's the actual reflection, and you can see that there are two, very close together, at different levels, with the adjusted time peak at 3.15ms. Assuming 343 m/s, that would be 1080mm, but the second peak is at 3.29ms, which is 1139mm. Average gives you 1109ms, but the first peak is larger magnitude, so the average deserves to be "weighted" a bit. So 1100ms is a "best guess" that should find the spot that is causing BOTH of those reflections. They will be very close to each other (just a few mm apart), so it really doesn't matter if the string only gets to one of those points, or the other one, or somewhere in between: once we find that spot and treat it, we will be treating BOTH of those issues, since the treatment will be much, much larger than the possible error in locating the point.
I had an idea of how things can improve by proper sub placement, but your words are much more enlightening. For instance, I didn't know that even 10cm can make such a great difference, mainly because we are talking large wave lengths here, significantly bigger than this distance (10cm).
That surprises a lot of people, but it is very real. Here's a real-world example:
That's from a room that I'm in the process of precision-tuning right now for a client in the USA, and shows exactly what I'm talking about. The frequency range here is just the low end: 12 Hz to 500 Hz. In this case I am using two identical subs to create a plane wave bass array (which is a bit complex to explain right now), and that graphs shows a series of tests where we were sliding one of the subs in steps of just 2" (10cm) without moving the other one at all. The major differences are clearly visible. In this particular test I was looking for the spot that best eliminates the phase-related issue at 80 Hz, while also reducing the modal issue at 37 Hz. You can see that I did find a good spot for doing both of those...
This often comes as an eye-opener for many people, since you don't really expect to see any difference from moving a sound source by 2 inches (10cm) when it is generating waves 14 FEET long (430cm), but there's the proof! The reason this works is because of the interaction between the speaker and the room: if you did the same thing out in the open, with nothing at all around the speaker for 100 feet in any direction, then the changes would be miniscule. But because the speaker is inside a room, and is also not alone: there are other things in the room generating the same waves, but at different places (the main speakers, and the other sub in this case), all of those interact to create patterns of interference. So it's just a matter of moving the sub around to get the most beneficial pattern.
The room in question is in the final stages of tuning right now, and the owner is a forum member.... who I'm trying to persuade to create a thread about the design and construction of his place, as he has quite a story to tell! He says that maybe he'll do that one day, but for now he doesn't want to be identified... but I can still use data like this to demonstrate the principles of acoustics.
But there are two limiting factors that make the process difficult. One is the lack of space in the room, where as you can see clearly things are quite squeezed.
Fair enough. .... but there must be at least a FEW places you could try! Here's the data from another client, in the UK, whose room I am also doing precision tuning, and he has a similar situation: there's furniture and acoustic treatment in the way, so he could only get his sub to 6 specific positions in the room... as you can see, "position A" is by far the best. Before, he had it at "position D", so you can see why he said there was a lack of bass at the mix position. I fixed that very simply, just by moving the sub off to the side, to "A", and now the bass is roaring! There's also enough treatment in that room now, that I'll probably be using digital tuning to slightly reduce the peaks at 48 Hz, 76 Hz, and 100 Hz. , to smooth out the low end even more. However, there's also phase to take to account: I need to be sure that there are no phase issues at any of those frequencies, because if there are then it is not possible to apply EQ there...
But anyway, the point is that even if you can only get your speaker to half a dozen different spots in your room, try them all anyway.... as soon as you can buy longer cables!
Ok, didn't know that treatment was not meant to correct FR, that's why I felt dissapointed in the first place. Thanks for pointing out.
Some treatment can change frequency response yes, but mostly there isn't a lot of effect. Acosutically, everything effects everything! Here's a graph from an excellent book on acoustics, that shows how the modal response for a room shapes the entire frequency response for the room:
- TICHY--Modal-contribution-to-room-response.jpg (32.36 KiB) Viewed 47487 times
- TICHY--Modal-contribution-to-room-response.jpg (32.36 KiB) Viewed 47487 times
Here's another graph from the same book, showing what happens to that when you install bass trapping, to dampen the modal issues:
As you can see, even though the modes are being damped in the time domain, there's not much happening to the frequency response! The curve just smooths out a bit, with the peaks and dips getting a little lower in intensity, but the overall shape of the frequency response doesn't change much at all, just from damping (bass trapping). I'm not sure if you saw the article I wrote on this a while back, where I say that it is impossible to get rid of the modes in a room: this is why. You can damp them, and you can reduce the intensity, but you can't get rid of them, and they still shape the overall response of the room, even if they are damped.
I think you misunderstood me. By "removing the character of ther speakers" I was speaking literally about the speaker it self, excluding the room interference. This is why I mentioned the speaker brands in my comment, meaning that every brand exhibits its own character, and that includes its own FR as well. NS-10s for instance, they exhibit quite a unique FR, yet they are still considered as an irreplaceable tool by many engineers. So why correct it with DSP?
I did understand what you meant, yes, but I guess I didn't explain the point I was trying to make: if you use EQ at all in the signal chain, it will affect the room response! Even if the purpose was to correct a the deficient response of a not-so-great speaker (such as the NS-10, for example...
), then that will ALSO have an affect on the room response, if the room is not treated. For example, let's say that you attempt to correct the notoriously "missing bass" of the NS-10 (which falls off the edge of a cliff below about 100 Hz) by adding just a little 6 dB boost at 70 Hz, to try to get at least some low end out of it.... Now, if your room happened to have a modal resonance at 70 Hz, you would be triggering that more strongly (even though you did not intend to do that! You only wanted to extend the low end of the NS-10...) So now you would have a frequency response peak there, sure, but you'd also have excessive modal ringing. On the other hand, if your room had a modal null at 70 Hz, then your EQ boost would do nothing at all! 70 Hz would not get louder, because you'd be pouring energy into a deep hole that you cannot ever fill.... However, the energy doesn't just disappear: a modal null is a phase problem, and a position problem, so the energy you pour into that "hole" would appear at another point in the room, that is NOT in the modal null at the mix position, and it would appear with a phase shift.... So you would make matters worse, not better. Even though you were just trying to "fix" a lousy speaker by eq'ing it, if the room is not treated properly then that is doomed to failure, because it will always have unwanted consequences for the room itself, even though it was intended just for the speaker.
Gosh!!! This is from another world!!! You definitely nailed it there, Stuart!!! Well done!!
Thanks! That's an example of what can be done with very careful precision tuning in a well-treated room. It took a LOT of time to get that right, and Rod and I worked late into many, many nights, tweaking here and tweaking there. But once again, it would not have been possible to get those results just from the digital tuning alone: the room first had to be treated to the point that made it possible. You would not believe the truck loads of treatment that went into the ceiling of that room, to damp the low end....
Ok, so next I will track down that nasty first reflaction to see where it comes from and post again
Great!
- Stuart -
) 
. Now I know that probably it isn't going to make much of a difference no matter which one I choose of the two since little differences in the absorption coefficients don't influence acoustic efficiency that much (ie. the bobgolds coefficient chart we all get advise from for years now whenever we need to clearly states in the 3d line of the 2nd sentence: 
.