Mix Room Acoustic Treatment - Let the fun (again) begin.. Update #12

octatonic

grappagreen said:

octatonic said:

grappagreen said:

octatonic said:

Question- why 1/48th smoothing?

With 1/48 you can actually see what the room is doing where it matters and is most troublesome i.e. the sub/low and, to a certain extent, mid range. It's good if you want to identify issues/challenges and 'attempt' to deal with specifically with them. In reality however the human auditory system is very clever and as such stuff that looks terrible objectively isn't an issue subjectively because of the way our brains process the data. I'll talk a little about the psycho angle in the results/analysis to follow in a sec! Understanding a little about this is actually pretty important as nobody should waste time and effort on tackling problems that don't matter.

Folks sometimes lose sight of the fact that a swept sine and a microphone is nothing like music and your ears/brain! It's easy to go down rabbit holes if we're not a bit careful..

Yes, I know why we use smoothing.
Why 1/48, rather than another value such as 1/24 or 1/6?
1/48 is pretty detailed and is the treatment you are doing going to be so targeted to specific frequencies with that much granularity, or more broadband?

I know you know James 1/48 is very detailed. I will be using a combination of velocity and pressure based specific treatment (and some diffusion/scattering) based on the challenges I find. The posts on decay time and freq response over the next couple of days will bring it all together and make more sense..

OK no worries- watching this with interest.

roberty

Looks like a lovely place to work with all the green outside the window

But mostly I am impressed with your decorative moose head!

grappagreen

roberty said:

Looks like a lovely place to work with all the green outside the window

But mostly I am impressed with your decorative moose head!

Thanks.

Rudolph has to stay under orders from the family..

grappagreen

Update #6b

Found a little more time today so let's have a discussion about decay time and take a look..

Decay time is important.. actually it's way more important than folks realise. There are actually two aspects (remember the two room comment before?). The first is reverb. OK well it's not actually . You see reverb doesn't exist in small rooms. They are just too small. What we do have is a decay in sound energy over time and, as per the EBU standard, we want to control it. We're looking for two things; a decay of broadly 200ms average (calculated from the volume of my room) and a frequency response variation within limits across the spectrum (again with some tolerance up and down per 1/3rd octave band). The second is resonance. I mentioned before that rooms 'ring' at certain frequencies (along with potentially other things as we will see later) and we want to ensure these are also under control so that they don't mask what we hear.

Let's take a look at some data and discuss

The following shows decay time (t30) from before treatment and after.



Let's talk about 200hz upward first. Notice the T30 time has dropped and between 200 and 800hz we're roughly hovering around the 175ms point. That's good but I'd like to see it a bit flatter. Notice from 1.5k upwards the line seems to be steadily rising. This may seem a little strange as I'm using velocity based absorption and generally this kills high frequency first as it's more effective in this range. We most certainly don't want a room with high-mid/high energy depleted as this is what causes rooms to sound 'dead' rather than 'tight'. If you've ever been into a room with all the high end sucked out you'll know how horrible it sounds and feels - sorta like someone has attached a vacuum cleaner to your head! The reason I'm not seeing this is my mineral wool is actually wrapped in very, very thin plastic film. I'm doing this to keep fibers at bay and a positive consequence is that the high-end is retained a bit. I can modify further treatment to bring this to where I want it to be. Much easier to take it away than add it. 

Looking below 200hz, the data is pretty useless. Remember we spoke about the Schroeder frequency? Well, below 200hz ish REW is showing the impact of resonances ringing on. It's fun to look at the above in conjunction with the Waterflow chart that shows decay over time. Let's take a look..

Here's before the treatment



As you can see we have loads of ringing between 30 and 250hz as expected. Let's take a look after the treatment;



This is significantly better (albeit the final one has its lower limit set to the C weighted noise floor measurement in my room i.e. looks a bit better visually and is more in line with what is actually audible). Once I had treated the room (not shown in the above) I had three frequencies still ringing significantly up in the low-mid range at about 175, 210 and 326hz. Little strange I thought. The 175 and 210 immediately shouted 'windows'. I used some spare Rockwool packs in the window recesses behind the speakers and they reduced massively. Need to think about what I'm going to do about this.. The 326hz was more fun to find. I used the signal generator in REW set to 326hz and walked around the room starting and stopping the signal. The decay was obvious and sounded like a bloody bell. Looking at my lamp shade, it was made of thin metal and shaped like, well, a bell! Damping this removed the resonance. I'll be replacing it shortly with something more acoustically friendly

Below 200hz we can start to see modal ringing. Note also that this ringing correlates with what we see in the low end of the decay on the T30 graph shown prior i.e. we see higher decay times on the decay graph where we see tentacles coming forward in the waterfall.

Let's talk a little psychoacoustics as it's important to know what matters and what doesn't in this region. There's a fascinating doctoral paper written by M Stephenson entitled 'Assessing the audio quality of low frequency reproduction in critical listening spaces'. As the focus is on psychoacoustics the paper focuses on the temporal impact of perceived quality of sound reproduction in the low end. The long and short is that the lower we go, the longer the resonances have to be to be detected as impacting on our perception of quality. In effect, dropping a resonance time below a certain point (at a certain frequency) makes no difference as we cannot perceive it! This is good as treating these things is tough due to the size of the wavelengths. In fact temporal response to ringing is more critical than frequency response at sub frequencies..

In summary then I want to consider the following for improvement from a decay perspective;

• Flatten out decay time from 200hz - 1k. 
• Bring down decay times above 1k
• Reduce model resonances between 50 & 100hz
• Reduce window resonances

It's worth also noting that I need to start being a little careful with adding more absorption. A 175ms average decay time in the mids is already very tight and I don't want to suck the life out of the room.

If you've made it this far well done!

Si

grappagreen

Update #6c

Let's have a look at the last area of interest. Frequency response. Now, what to say about this before we look at the data..

The first thing to point out is that FR is a bit like the DNA of a room. It's literally built into it. This is especially true at low and sub frequencies where certain phenomena dominate what we experience. That doesn't mean to say we can't influence it, we can. It just means that we have to manage expectations and understand that changing some things is literally beyond our reach if our intention is not to either design/build a room from scratch or treat a room to an extent beyond what most folks would consider sane. There is a reason why pro mix/mastering rooms can cost hundreds of thousands of pounds to build. It is possible to take a smaller/existing room and turn it into a extremely high quality environment but it's beyond the scope of my little project because I just cannot do what needs to be done for practical and marital reasons! 

So let's take a look at what we have;

1/48 smoothing



As per previous discussion we have a room of sort of two halves. Below 200hz 'ish' the FR is influenced by standing waves and something called SBIR. Above this frequency, modes/SBIR still influence, however as we move up the spectrum the influence of comb filtering is more prevalent.

The FR we are seeing is pretty good actually. It is pretty balanced across the spectrum and the variance is relatively tight given we are at very high resolution at around +/- 7.5db for the vast part of the spectrum. There are some areas that would be good to improve upon. It's worth bearing in mind that my speakers have a +/- 3db variance from 60hz upwards..

So, what can we do to improve things? If we want to improve then we need to try and understand what's causing some of the lumpiness and that's where things get very difficult. The problem is that the ultimate FR graph is the composite of layer upon layer upon layer of acoustic energy interactions and influences. This limits our ability to a) deconstruct problems and b) implement solutions. Somewhat counter-intuitively as we add treatment we often see things start to look worse in some places than the empty room. The reason for this is that as we remove some of the energy through absorption, other problems hitherto hidden become apparent as they poke themselves out of the melee.

I'll follow up with some more discussion on the above, what we can discern and what we may do about it in the next post. We will also need to empirically test/validate.

I'd finally like to tackle the comment James @octatonic made on 'Why 1/48 smoothing?' It's a very valid question. Why not 1/24 or 12 or 6 or 3 or psy or no smoothing... I suppose my answer to this would be 'What is it we are trying to achieve?' I'm assuming that the source of the question is actually 'Why are you looking in such detail at something you can't understand/change?' Let's go the other way and I'll explain my thinking. I consider psy valid as it provides an illustration of what we actually perceive when we hear. I consider 3 valid as it is a de-facto standard for FR and as such it is useful as a comparison with other data we may see. I consider 1/48 valid as it's the highest resolution that we can actually use in a meaningful way to try and assess what's going on (along with other REW data). I would consider 24/12/6 useful if we were trying to assess against a standard that specified these. I'm targeting EBU 3276 and it actually assesses room response differently i.e. 1/3 octave band filtered pink noise. I'll ultimately assess against this.

Si

roberty

grappagreen said:

roberty said:

Looks like a lovely place to work with all the green outside the window

But mostly I am impressed with your decorative moose head!

Thanks.

Rudolph has to stay under orders from the family..

Glad you took that the right way. The last time I complimented a man on his moose head I got a black eye

Winny_Pooh

Great looking room! 

My recollection on reading about the AE22s is that they have great time domain response in the lows (like a modern ns10, not ported).

Do they work in the mid field position where you have them at present? How are they brought 1-2 feet closer? 

At the current distance it'd be tempting to put in something bigger like a set of KH310s or something even bigger. 

grappagreen

Thanks Winton,

The AE22's don't have the flattest frequency response but they are mega fast and have great distortion specs given that they are sealed box. All the positives of NS10's but none of the downsides! I have used them for years and love them. They were a bargain when they came out but you can grab a used pair for about £400 now which is mad when you consider their level of performance.

I have them at the limit distance wise; the specs give max distance of 2.5m which I'm at. I won't move them forward due the impact on the 50hz region from an FR perspective.

Once the room is finished a pair of 310's will be appearing as I want sealed cab, 3 way and they go a bit lower..

Si

Winny_Pooh

 

robinbowes

I'm sure this will be useful for me in the future as I'm looking to buy a new place soon and hope to build myself an office/workshop/den/music room, which will of course need treating acoustically.

The current place we're looking at has a large (18m x 7.5m) outbuilding that should give me plenty of scope to convert:



Yes, that is asbestos, but it's already been removed. If we buy, I'll negotiate the spec of the build-out with the developers, making sure I consult my Home Recording Studio book!

R.

grappagreen

robinbowes said:

I'm sure this will be useful for me in the future as I'm looking to buy a new place soon and hope to build myself an office/workshop/den/music room, which will of course need treating acoustically.

The current place we're looking at has a large (18m x 7.5m) outbuilding that should give me plenty of scope to convert:



Yes, that is asbestos, but it's already been removed. If we buy, I'll negotiate the spec of the build-out with the developers, making sure I consult my Home Recording Studio book!

R.

That looks like a fun project!

Be good to see what you do with the space you finally end up with! 

grappagreen

Update #7

Thought I'd take a moment to take a breath and recap a little.

We've effectively got to a point now where we've significantly improved early reflections, decay time, resonances and as a consequence of removing energy (to achieve the above) have improved the frequency response. Note I said as a consequence here. The unfortunate truth is that whilst it's relatively easy to improve the first 3 it's much, much harder to improve the 4th especially in the low/sub region. Physics is just against us and, as I mentioned before, low/sub energy is like a rampaging elephant and the tools we have are just nowhere near effective enough to stop it; we can slow it a little but beyond that. We will eventually use some tools beyond acoustic treatment to add the final gloss to the FR but we have to understand what is impacting this as some things can be improved using other tools and some things can't. We'll use some further analysis of the data to identify these and treat accordingly.

In terms of outcomes from analysis of the data the following areas are where we want to focus attention for phase 2 in order of priority;

• Reduce reflections around 12.5 and 17ms
• Reduce modal resonances between 50 & 100hz
• Reduce window resonances
• Bring down decay times above 1k to circa 175-200ms
• Flatten decay time from 200hz - 1k. 
• Smooth overall frequency response
• Match speaker response in both FR and time domain

We'll start to work through each of these over the next few weeks and see where we end up.

Finally I thought it might be worth talking about how the room actually sounds subjectively now as after all that's the reason I'm doing it. In short, the change is so profound it's almost a laughable! It's like I'm listening in a completely different room (which I am acoustically speaking!). Imaging is sharp and the the sound literally jumps out of the speakers. 3D sound field presentation is amazing with left/right/front/back/up/down easily portrayed on decent mixes. ER/verb/delay treatments are now exposed in all their glory as indeed is everything else including technical issues, horrible synth patch choices etc My speakers are very fast and the low end is now super-tight and well balanced.

In summary it's already sounding great

Si

grappagreen

Update #8

Before we go any further on the project I wanted to talk a little about the biggest challenge I face in my room; the sub 100hz region. My current speakers roll off at 12db/oct from 100hz and as such in the absence of any impacts from my room/boundaries (a theoretical construct!) I should see this reflected in the recorded frequency response. I don't as obviously I'm in a room! So what's the deal here? The deal is that my room is built from hard stuff and pretty much all of the energy is contained within the space. Why is this such a problem?

Imagine my room was reconstructed at exactly the same dimensions but, instead of the brick/plaster, the walls were actually a wooden frame covered in fabric that allowed energy to flow through it. This hypothetical room is standing in a field with nothing around it and hence energy can freely dissipate (I'm conveniently ignoring the floor here). if we were to pump energy into the room and measure the frequency response at my listening spot it would be very close to the signal coming from the speakers themselves. 

The problem with rooms is that the energy creates pressure at the boundaries and, at frequencies that match the physical dimensions of the room (and its multiples), standing waves are created. These establish themselves and literally form waves that vary in amplitude if you move through them at the relevant frequencies. Think of these waves as ripples in amplitude that overlay each other and, as we go higher up in frequency, more and more overlay to create a complex pattern. We can calculate these standing waves (modes) using a tool like amroc online but bear in mind these are theoretical models and as such reality often diverges. To hear this in action simply calculate a mode (low axial modes are the best), play a sine wave and move around the room. You will hear the response vary widely and at the lowest modes the tone you hear will literally disappear in the middle of the room along the relevant axis and sound way louder at the boundaries.

So how do we fix it? What absorbers are we gong to install to remove all this energy and solve the FR problem? None; we can't! The problem here is that there is no passive or active absorber that has the horsepower to flatten a frequency response by any significant margin below 100hz. Sure if we are lucky and we spend a load of time, money and literally fill the room with deep treatment or high Q pressure traps we might shave a db or two off a peak or a null - given I have swings of circa 10db I have literally no chance at these frequencies. Treatment on all surfaces is good as we go higher in the spectrum as the tools become more effective (and we get some additional benefits), but at the bass/sub..forget it. Remember I am talking frequency response here not time domain.

How do 'proper' rooms minimise this problem then? They have special boundaries (within boundaries) that are designed to soak up as much of this energy as possible - remember our acoustically invisible room example in a field above. 

There is actually another problem. Part of our strategy is at some point to use digital room correction to smooth the frequency response. Easy then, let's just EQ out the problem. Without wanting to get too heavy the very low frequency response isn't minimum phase and as such EQ just won't cut it to any significant degree. We'll look at minimum/excess phase at some point soon.

Is it actually a problem? We know that our sensitivity to changes in amplitude and at low frequencies it is very poor but the problem is we are actually quite sensitive to peaks. We tend to hear those peaks more easily and that's problematic.

So what do we do. Option 1 is to live with it. Don't like the sound of that. Option 2 is to use the only practical tool we can actually implement and we'll be doing that at some point. Welcome to the world of multi-sub woofer room mode management. 

My room will broadly have the same issues irrespective of the speakers I use and their respective FR. It's built into the DNA of the room itself. 

Si

robinbowes

grappagreen said:

robinbowes said:

I'm sure this will be useful for me in the future as I'm looking to buy a new place soon and hope to build myself an office/workshop/den/music room, which will of course need treating acoustically.

The current place we're looking at has a large (18m x 7.5m) outbuilding that should give me plenty of scope to convert:



Yes, that is asbestos, but it's already been removed. If we buy, I'll negotiate the spec of the build-out with the developers, making sure I consult my Home Recording Studio book!

R.

That looks like a fun project!

Be good to see what you do with the space you finally end up with! 

Sadly, we have pulled out of this property - we can't afford it  

Back to the drawing board...

R.

grappagreen

robinbowes said:

grappagreen said:

Sadly, we have pulled out of this property - we can't afford it  

Back to the drawing board...

R.

A little frustrating no doubt! Sorry to hear that..

grappagreen

Update #9

So come on Si, what's the plan? 

We know that we're going to deal with the sub 100hz region using subs (and we may also get some help from other things we do) so what are we going to focus on prior to this.

I highlighted that time domain specifically between 50hz and 100hz is something I need to control tightly and if you look at the FR between 100hz and 300hz it could definately benefit from some smoothing out if possible. I face a couple of constraints though; I don't want to suck out the high end further and I also face physical space constraints. The problem is that velocity absorbers fail to address both of the constraints adequately. The problem with the lower frequencies of this range is that the power of even my 400mm deep panels is just limited and they will literally eat my room physically. At 100hz these panels have an absorption coefficient of about 0.6 so effective but not as effective as I would like.

What we need here is some treatment that is broadband between enough to cover the region 50-300hz, most effective around 100-150hz and has a small physical depth footprint such that we can implement it without taking up too much space. It needs to be simple, effective and something that we can build at home with available materials. It's a big ask but, luckily for us, some clever folks in Germany have a possible answer..

We've talked about two types of treatment so far; velocity and pressure. Most pressure traps are amazingly effective but they have a couple of significant problems; they have a high Q so they can only really tackle very specific problems and they are hard to build at home and be effective. Not that we can't build them, I've had great success with MLV traps but it can be tough and for every success in the past I've had failures and wasted time/money. The odds get even worse with things like Helmholtz devices.

A little while ago some folks came up with a rather novel idea. Why not take a thin metal plate, attach it to a substrate that had specific properties and damp the vibration of the plate to remove energy as it is excited by sound. This solution is somewhat elegant and in some respects is the answer to many a home acoustics prayer. To be able to build something effective from circa 50hz, has lower Q charateristics and that is only 4" deep seems just too good to be true. The data seems to bear it out however.

I'm going to build 4 off these and we'll find out..

Back soon.

Si

grappagreen

Update #10

Thought I'd have some fun today as progress is a little slow on the passive treatment side.

I mentioned before that I basically cannot do anything about the lumpiness in the low end of my setup with any of the treatment options. Basically there's nothing I can put in the room to improve things. There's only one option here.. multiple subs..

So I rigged up a miniDSP 2x4HD with 2 Eve Audio TS107 subs.. both subs are in positioned in my front corners and I managed get a couple of hours setting up a very basic, quick and dirty setup with the sub integration. The sub positioning isn't technically optimal however, I'm not just interested in a flatter FR; I want to ensure my imaging is as good as I can get it. I'm running the XO a bit higher than I would like as my main speakers roll off from 100hz. When I change them I'll run it down lower.

This also allowed me to move my main speakers to a much better near field position without the 50hz range falling into the abyss.

Using subs like this allows us to excite the room modes more evenly and smooth out the low end in a way that no treatment can. We also get the added benefit of unloading the sub/low from the mains and improving the mid-range reproduction a little..

So.. here's an FR from REW that shows my untreated room (left and right individually) along with a combined L+R+subs from today all at 1/48.. you can ignore the high end as this always rolls off when measuring two speakers at the same time



Starting to look (and sound) pretty sweet!

Deadmau5 playing at the mo so that I can enjoy all that low end..I do love a filter sweep or two 

Si

grappagreen

Update #11

Despite a distinct and frustrating lack of progress from a treatment perspective I thought I'd jump ahead a little and at least implement another piece of the final puzzle. I'll have to do this again but it's fun so..

Basic treatment is in place, ER's are covered, decay time/balance is sorted (to a degree) and subs are in place and well integrated. It's as good as it's gonna get with the current treatment but it just isn't good enough! 

What's a chap to do eh?

Just imagine if we could turn the chart in Update 10 into this..

1/48 oct smoothing:



1/3 oct smoothing:



Impressive eh? The above is only half of the story as it represents frequency only. Step response and group delay are pretty much perfect so we have the time domain covered as well.

As you may be able to imagine the difference is pretty stunning; very balanced with amazing imaging and clarity/detail.

What is this magic? Welcome to the world of 'proper' DSP.. and by proper I don't mean Sonarworks marketing bull.. I mean proper room correction.

Will expand soon..

Si

roberty

That looks super flat @grappagreen nice one 

I used REW and a MiniDSP to set up room correction in my listening room and the results were quite spectacular

grappagreen

roberty said:

That looks super flat @grappagreen nice one 

I used REW and a MiniDSP to set up room correction in my listening room and the results were quite spectacular

Thanks..

Did you use Rephase as part of the process to tackle time domain after working in REW? Are you deploying FIR filters or just exporting EQ from REW into MiniDSP?

If not and you want to take another step up in quality I can guide you to something that is well worth the effort! It's not as good as the tool I use for a number of reasons but you can do something similarish for free..

Si