Predicting room modes

Predicting room modes can't be that hard can it?

Here is an example using REW (Room EQ Wizard), compared to a recent measurement.

Green: predicted

Red: measured

The simulation isn't too bad down to 70 Hz. Below that point it becomes useless. It doesn't account for the bottom end gain which extends the subs down to 15 Hz even when their nearfield plot is rolls off around 25 Hz. The massive 45 Hz mode was not predicted at all and I've found it to be the most audible issue. If it isn't pulled back the subs boom.

Red: measured response

Green: simulated waterfall

The irony of room mode simulation is that you actually need a real measurement of the space to be able to tweak the simulation and get it reasonably close. Otherwise, it's difficult to know where you can trust the model and where you can't. With a bit more time I might be able to figure out what is going on, but if you are using the simulation to show which of different room designs are better, will it really help? At this point it's difficult to say.

Audio room ceilings

Audio Tools

Here is a list of handy tools to help you with audio projects.

Free Software

WinISD

Flareit.exe

Cut list - for working out how to best cut up a sheet to get the most out of it

Why do cheap speakers sound cheap?

My answer might surprise you.

Above: The woofer from my first system

Cheap speakers cut all the corners that can be cut, starting with cheap drivers, an unbraced resonant box and the cheapest crossover possible. But what happens when the box is improved and a decent crossover is used? That's when the surprise comes in - suddenly they start to sound hifi. For a long time I had blamed the drivers, thinking they were so cheap they could never be worth using. In a recent experiment, I used my first ever speakers. They are cheap 5" paper midbass drivers that I measured as a teenager for a high school project, almost two decades ago. The drivers probably cost only a few dollars each.

This is a nearfield measurement of a Hitachi 5" midbass. It was taken on an open baffle full range with the crossover in place. There is a small cone tweeter on the rear that might do something about the 2.5k dip. In this case I only wanted it to run up to 2k. A little EQ and this driver could behave quite well. The dip just below 800 Hz probably isn't too bad - dips are less obnoxious than peaks.



I set this mid up in a 4 way active speaker, with a waveguide loaded compression driver above 2k and a sealed Vifa woofer coming in at 400 Hz. Just one notch filter to take out the dipole peak was needed to extract reasonable performance. My original plan was to modify the driver and measure the result, but it turned out that it was good enough to try without any mods.

The crossover is the problem

Often the biggest weakness in cheap speakers is the crossover. You might have a 3 or 4 way speaker with the bare minimum of crossover components, often just a cap on the tweeter. Cost cutting is the name of the game. Many speakers never have a chance to sound decent because the crossover is so poor.

In older speakers you might also see some strange designs that include dual tweeters and dual mids side by side. It may be a good idea to disconnect or re-arrange these, since mid and tweeter drivers should never be placed side by side. Where more than one tweeter or mid is used, they should only be placed in a vertical line.

What to do with cheap drivers?

Do you have some tired old speakers collecting dust? Want to give them new life? You will need:

• basic measurement tools (mic, mic preamp, camera/mic tripod, free software)
• MiniDSP active crossover
• enough amp channels (low cost low power amps are fine)

 You may be surprised how good they can sound. A few mods to the box (bracing + damping) and a decent crossover can make a huge difference.

Geddes foam experiment

Update: I now have some proper reticulated foam, the right stuff for this test. You can read about it here >

Here is a quick test of Geddes-style foam. Above you can see my "Miniwave" speaker which is a test of using a waveguide loaded compression driver in a small format. It looks even less pretty with some dacron added. Before foam:

Why foam?

The idea behind this is to damp what Earl Geddes calls HOMs (higher order modes). It's a simple concept. In a horn or waveguide there is the intended axial sound as well as waves which reflect off the walls. Designs that use diffraction to achieve constant directivity create even more HOMs. If we insert foam, the HOMs will be attenuated more than the axial wave because they travel a longer path. In this way the foam is selective. It also has an impact on the response. I noticed a fair amount of top end roll off as a result, but it isn't difficult to correct. The end result is shown here:

The shaded area shows a +/- 3 dB window which has been achieved with the response at 1m. In this case I measured slightly off axis to show the kind of angle that I listen to. The green plot is measured in the listening position. As it's a room plot, it's not as smooth and you can see the treble response above the crossover (2k) is shelved down.

I have moved them closer to the wall and used some damping behind them which acts as a bass trap as well as a mdirange absorber.

My impression has been that the sound has improved as a result of the foam, although it's hard to isolate as I did some other changes. The speakers have a different sound stage due to increased distance. The damping has an obvious effect, particularly in the bass. The top end is very revealing and detailed. I'm starting to get the impression that dome tweeters as a general rule are veiled in comparison. One surprise has been that simply swapping the tweeter from a dome to a waveguide loaded compression driver has created a much more dynamic speaker. The mid driver isn't efficient and the subs are the same and yet the sound has become effortless.

Simulating waveguides in Hornresp

An experiment with simulating waveguides in Hornresp. Unfortunately the datasheets don't provide all the parameters needed, so take this with a pinch of salt. Some guesses were used.

Here is a 15" oblate spheroid:

 

 Now here is an overlay of SPL, beamwidth and the directivity plot. The scale on the x axis for the coloured directivity plot is close but not an exact match to the others. 

 

You may notice that the beamwidth is quite jagged and that it narrows around 3 - 4k. The SPL plot shows that the sim isn't quite right. This is due to the incorrect parameters that I had to guess. However, I expect that the directivity would be dominated by the waveguide itself in this model rather than the driver, as long as 
the throat area is correct.

Now for comparison here are the plots of the oblate spheroid that we measured of a similar size:

This version is "normalised" so that the axial response is dead flat. What you see is more of an indication of it's dispersion relative to flat on axis.

The simulation doesn't look much like the measurements. The sims show narrow dispersion around  3 - 4k and elsewhere around 80 degrees. The measurements show narrow dispersion around 2 - 3k and considerable change throughout the range. The normalised plot shows that if EQ were used to make it flat on axis, the dispersion would vary so this makes for a difficult situation.

Waveguide shootout

Jump to the next Waveguide Shootout >

In one day we compared a bunch of compression drivers and waveguides. We first started measuring outdoors. You can see Andi's canon ... oops I mean diy mic pointing at the flashy 12" waveguide. Our measuring was followed by some listening comparisons with a few torture tests indoors. The results were quite interesting. Did the ones that measure the best sound better? Did we all agree?

All four attendees were StereoNET members - Gainphile the host (Andi), Antripodean, Fury and myself.You can read about the even on StereoNET >

Measurement Setup

The speakers were elevated on a rotating turntable with 11 degree increments marked. The mic was also elevated and a Laptop setup with Arta, which proved quite powerful and revealing.

 

We used Andi's DIY mic which is based on the well known Panasonic electret. Andi shares more detail about his measurement system on his blog >

Our aim was to investigate various driver and waveguide combinations, see how they measured and how they sound. The crossovers were all done by Andi with MiniDSP and a level of phase correction was applied on the spot.

Understanding the plots

The directivity plots shown will be unfamiliar to many, but they are not difficult to understand. They are essentially a frequency response plot which measures from on axis right out to 90 degrees. This plot features an open baffle mid and a dome tweeter.

As you move across the x axis, you see how the directivity changes. The X axis shows frequency as with any frequency response plot, but the y axis shows the angle.

Region A (200 - 1kHz) shows where the mid has constant directivity due to the influence of the open baffle. This is very good performance and can't be achieved in any other practical design in this bandwidth. In this area it has 100 degrees dispersion, which means the response is 6 db down compared to the axial response. Region B shows a problem - the directivity is no longer contstant but narrows until it reaches 1.6k, then expands again. In region C it has constant directivity over a small range, but this time the dispersion is quite wide at 150 degrees. In D it becomes narrower again and in the top octave (E) it narrows from 60 to 40 degrees.

Plots such as these show problems in the crossover and problems with drivers integrating. A directivity plot of a perfect speaker would have constant directivity into the low midrange region. Across this region the directivity should not narrow at any point or reveal crossover points.


S15 Econowave

 

Compression driver: Selenium D220Ti (Titanium)

Waveguide: JBL clone

Woofer: Eminence Beta 12

The plot displays very good performance where directivity is controlled well over a wide bandwidth, with a little narrowing at very high frequencies. The achilles heel of this speaker appears to be the compression driver, which becomes quite harsh on dynamic material. I won't comment too much on the woofer because it was used in all of them.

18 Sound XT1086 + Faital Pro compression driver 

Compression driver: Faital pro

Waveguide: 18 Sound XT1086

Woofer: Eminence Beta 12

This waveguide was quite different. It appeared to have a diffraction slot at the throat,  then  it transitions to an elliptical mouth.  The driver has a novel orange phase plug. 

The measurement shows that it's performance has a few glitches and isn't working quite as well as we might hope. At 3k it becomes quite narrow and very narrow at very high frequencies. However, in our listening session it sounded audibly better than the S15. It had a smoother sound and we suspected the drivers were the issue here. 

Miniwave 3 way - OSWG6 + B&C DE250 + Vifa C17

Compression driver: B&C DE250

Waveguide: DIY oblate spheroid 6"

Midrange: Vifa C17 (see below)

Woofer: Eminence Beta 12

Vifa C17 modified

 

It was a surprise to the group that it could work well down to 2k, but this was what I had expected. The performance is generally quite good for a small waveguide, but there is a problem around 10k. It should be noted that the plots shown have been normalised. In other words, they show what the response would look like if the axial response was ruler flat. Here is the plot that has not been normalised:

 

This plot shows the response as quite good, with two issues. The first is a narrowing around 4.7k. It was necessary to use some EQ around here to get it flat on axis. The other issue is an axial hole around 9k caused by diffraction at the edge of the waveguide. This is an issue with most axi symmetric waveguides.

The solution is to toe them in to avoid the axial hole. Then only the narrow dispersion around 4.7k remains.

In the subjective comparisons this setup was preferred by all, although by this stage Antripodean was not able to stay. The sound was smoother and had less bite.

OSWG12 + BMS

Compression driver: BMS

Waveguide: Fibreglass oblate spheroid 12" 

While we did listen briefly in mono, it was difficult to form an opinion. 

Normalised plot:

You can see in both plots that it doesn't maintain constant directivity, but around 1.3 - 4.5k it becomes narrow.

Dayton 10" + Selenium CD

This combination has the best measured performance:

 

While it does narrow a little towards the top, the transitions are all smooth. Overall very impressive, but we didn't subjectively evaluate.

In our next event, we started using different drivers on the same waveguide, and one result in particular measured even better than these.

Jump to the next Waveguide Shootout >

Off the shelf passive crossovers

Should you buy an off the shelf passive crossover? This question often gets asked on audio forums. My answer to this is an emphatic "NO!" It's never a good idea to buy one of these units, unless you want to learn by experience the lesson you can learn here for free.

Why not?

A crossover need to be a custom design. A generic crossover fails in two ways. Firstly, it will fail to do what it attempts to do 99% of the time. Secondly, it omits many necessary aspects which require the box and parts to be chosen first. The result is that your DIY project will fall short of a commercial speaker at the same price.Generally a DIY speaker if well designed will match much more expensive speakers.

My two points here will make more sense with further explanation. Let's consider the two main aspects of passive crossover networks.

1. Dividing networks

The main function of a passive crossover is to divide the full range signal into separate bandwidths for each driver. This is the only function provided by typical generic crossovers and they do even this aspect poorly. One reason is they don't work is that the driver acoustic response adds to the equation. They assume that if you add filters at 3kHz to a tweeter and woofer that the crossover will happen at this point. Here is an example:

The crossover in this case does occur at 3kHz. However, the filters necessary to get this acoustic crossover were not placed at 3 kHz. The filters are applied closer to 2 kHz. If filters were simply applied at 3k, it would not work.

It's necessary to choose the filters to suit the drivers, and the actual filters will usually need to be applied at different frequencies for each driver.


2. Impedance and contour networks

The above example uses active filters, which are simpler to implement. With passive components, the performance is also influenced by impedance which varies with frequency. Often impedance correction is performed to flatten the impedance curve so that the drivers will response to the dividing networks in a more predictable way.

A good crossover design will often need contour networks as well, and this aspect is especially driver-specific. Often the response of a driver needs flattening. It may be a notch filter to tame a peak. Sometimes with crossoverless full range drivers, contour networks are still needed.

A decent crossover will also feature an allowance for speaker placement and baffle width. This is called bafflestep compensation.

One size doesn't fit all

I hope that by now it's clear to you why an off the shelf crossover isn't going to do the job. There are just too many things a crossover needs to do that can't be done in a generic unit. At this point you have two good choices. One is to learn to design your own - be ready for a learning curve. The other is to let someone else do the design and get to work building a speaker that has already been sorted. There are plenty of choices available.

A third option is to design your own but make it a little easier by going active.

McGurk effect - can you trust your ears?

Here is an effect that will surprise you. It's called the McGurk effect. See it for yourself:




Audio implications

So what does this mean for audio? It illustrates one thing we should have admitted all along - that we can't fully trust our ears. Sure, when we hear a sound system we like, our ears aren't lying. However, it pays not to believe everything your ears will tell you. The important thing to realise is that what we hear is processed by a powerful computer. It's very good at processing some things, not so good at others. In this case, when it comes to speech recognition, what is seen can over ride what is heard. The interesting thing is that this effect is robust enough to occur even when we know about it. It isn't like visual illusions that work only until you realise what is going on.


High end audio

Consider high end audio components. Is there a real acoustic benefit to having a 10mm thick brushed aluminium face plate? Or do ultra thick speaker cables sound better? Somehow a 20mm thick cable looks like it should perform better in some way. I recall hearing of one particular company that sold very expensive power cables who used garden hose to make them thicker. I suspect there are two benefits here. One is that the manufacturer can increase their profit because consumers will pay a great deal more for the perceived benefits. The other is that those who buy them will believe they hear a difference. This isn't the McGurk effect, but rather it's the brain displaying it's talent for pattern recognition. Listen to the same piece of music twice and if you have an emotionally compelling reason to hear it different, and you believe there is a genuine difference, you will hear a difference even if it doesn't exist. I call this the "man in the moon" effect. There isn't a man in the moon, our brain just finds a pattern. The same is true with music. There is also a self-generated chemical reward system where the belief that something is better stimulates the chemical reward system of the brain.

So what does all this mean?

If you want to believe, you will be tricked. Sales people want to sell. Other audio enthusiasts want you to back them up on what they feel sounds better. The question is, do you want to be sold? In many cases, the answer is yes. Audiophiles are always looking for an upgrade. There is of course nothing wrong with that, however, this usually means spending more money on the things that matter the least. The most likely reason is that they are the easy upgrades. It takes no skill or special knowledge to buy a cable based on a good review or a recommendation from an audio forum. However, the biggest improvements require you to invest in some audio education, or pay a professional.

Do you want to get some real improvements that are far bigger? One of the main goals of this blog is to help audiophiles get started on the bigger improvements. Acoustic treatments and speakers that interact better with the room is the best place to start.

LFE sub shoot out

Want to know how to get the most bang for your buck in a LFE subwoofer? Here I'll compare different options for a popular driver, the Acoustic Elegance AV15.

Introducing the AV15

 
Acoustic Elegance have a range of woofers and subwoofers with advanced motor designs. The AV series feature high excursion (23mm one way xmax), high power handling and low inductance.

What can we do with this driver?

For an LFE sub, we have a few choices:

• sealed
• vented
• 6th order bandpass
• tapped horn
• front loaded horn

Power amp

We'll power this sub with a popular pro audio amplifier - Behringer Europower EP4000. It's maximum power output when bridged is 2.4kw into a 4 ohm load or 2 x 950w into dual 4 ohm loads with a more conservative rating.

Conventional options

Here are the main conventional options - sealed (blue), vented (red) and 6th order bandpass (magenta).

You can see the sealed box doesn't stand a chance for LFE use. 106 db isn't bad, but it's not impressive when 120 db is available. All of them simulated within the excursion limts, as shown below:

In each case, a rumble filter was used at 20 Hz to keep excursion under control.  The bandpass has slightly better efficiency and hence a little more output, but considering the extra size and difficulty, it's hardly worth getting excited about.

Now for some horns. First a tapped horn:

 

Around 3db extra output is possible compared to the vented box. While this might not be as much as we might hope for, greater extension can be achieved.

If we're prepared to upsize it, then a front loaded horn blows everything else away.

 

The horn is huge.  The mouth is 63 times the cone piston area, or 2.7m high (8 ft) by 1.9m wide (6 ft). This can only work as a built-in horn. The make it a fair match to the other simulations, I've simulated the horn without the benefit of corner loading, hence the ripples. However, in a corner this horn will be flat in response. In reality, this horn could be reduced in size.

Conclusion

If you are looking for a killer LFE sub, you can rule out the sealed box. It's the expensive way to get the job done and it doesn't pay to be put off by those who claim vented boxes are "slow." The vented box wins in terms of reward for effort. It can hit 120 dB with just one driver and if that's not enough, simply build more of them. Four of them will reach 130 dB. In a real room you may get more or less depending on the specific acoustic properties of your room.

The tapped horn in this case doesn't seem to offer a great deal more than the vented box. Along with a bandpass, it's still worth considering but you will need to be prepared for something very big. Chances are it will be bigger than you can live with, so you might want to think about where you can hide it.

If you want something extreme, then it's hard to go past the massive built-in horn. If you make it big enough, it will win every time. Just one driver with one amp will probably beat four drivers running on at least twice as much power. Using more drivers can push the output up to 140 dB. At this point it starts to sound ridiculous, but the point isn't winning an SPL competition, but being able to play as loud as you want with a lot of headroom. You might find that you only hit 115 dB peaks at your listening position, which is the THX reference, with 105 dB in the midrange. This won't cause your ears to bleed on a system well designed for reference levels and it won't damage your hearing. A movie played at this level will have a typical level that is perfectly safe. If you play music all the time with 115 dB in the midrange, then you will have problems.

DIY in-room bass measurement

Achieving accurate bass presents a unique challenge. You can buy great midrange, but you can't buy great bass. It must be tuned to the room and this is where most get stuck. Most audio enthusiasts play a game resembling pin the tail on the donkey. When a change is made it is evaluated subjectively with music and while this will give an indication of what has been achieved, one is still doing it blind. Anyone who wants to achieve truly accurate bass will need to use measurments to help in the tuning process. The good news is that taking measurements is not expensive or difficult.

Why measurements are essential

Some are not looking for accuracy, but instead are happy with the sound they can get while making all choices subjectively. For those that are aiming for accurate bass that is true to the original, measurements are necessary. To ensure that you have achieved accuracy, you will need to evaluate with measurements as well as subjective listening. Both evaluations are necessary to validate the result.

Defining accuracy

Accuracy can't be defined without some objective measures, otherwise it becomes far too open to interpretation.

Firstly, a smooth frequency response should be achieved in each significant listening position. In a 2 channel setup this might be just one sweet spot, but in a home cinema the target may include many seats spread around the room. The latter presents a much bigger challenge. While speakers typically aim for +/- 3 dB in an anechoic chamber, in the bass range we can tolerate a bigger range. I'd suggest that around +/- 5 dB is a suitable goal for the bass range. A typical audio setup that has been done by ear will often achieve only a 30 - 40 dB range.

Above: Red: untreated Purple: treated. Image courtesy of Ethan Winer at Real Traps

 Secondly, a good response in the time domain should also be achieved. In typical untreated room, standing waves cause certain frequencies to decay at a much slower rate. EQ can achieve a flat response in a small area, but the slow decay caused by standing waves can't be fixed in this way.

 

 

This example comes from Ethan Winer at Real Traps. In this example, a small room was treated extensively in a non-obtrusive style.You can see it in his Hearing is believing video.

These waterfall plots show how the response changes over time. In the first waterfall plot, the ridges indicate how the modes decay at a slower rate. In the second plot, you can see a much more even decay.

Don't you trust your ears?

A common argument against taking measurements is that one only needs to trust their ears. The problem with this is that while our ears are the final judge, they aren't able to give the kind of specific feedback that is needed to fully optimise bass. Suppose you were to choose a subwoofer placement for a home cinema with 8 seats. To do this by ear, you would need to listen to a number of tracks and then repeat the process for each seat and again repeat this process for each position. Try try 8 subwoofer positions for 8 seats means 64 evaluations. Obviously the process would be far too tedious and the result would be that it gets done poorly.

By contrast, you can measure in seconds and get useful data and you will see the real impact of each placement.

Basic tools

A basic measurement system includes:

• measurement microphone,
• microphone preamp,
• microphone cable,
• tripod and
• measurement software.

Professional vs prosumer choices

Professionals use high end microphones which cost upwards of $1000. For personal use, these microphones provide more accuracy than is needed. Many enthusiasts use the Behringer ECM8000 which has a street price of around $80. The main limitation of these is that their unit to unit consistency is quite poor. Fortunately, they can be purchased individually calibrated from Cross Spectrum Labs. While all mics come with a generic calibration file, individual calibration is needed so to overcome the unit consistency issues. The extra cost is modest and I strongly recommend purchasing a mic that has been individually calibrated.

These microphones require phantom power, which is provided in the low cost mixer shown above. While it is possible to only use a phantom power unit, the preamp allows gain control. The absence of gain control could mean the correct level can't be achieved.

Introducing the Room EQ Wizard

It turns out the only bass measurement software you will ever need is free: REW - Room EQ Wizard. This application is very easy to use and can be downloaded from the Home Theatre Shack where extensive support is available.


Download REW here >