Basic acoustic measurement primer v2.1

[DanDan]

Room Analysis Software Primer V2.1
Revised August 2010

Room Analysis Software of decent quality has become very affordable. This has led to many attempting to use it to evaluate their studio or other listening room. The immediate results can be confusing and disappointing. To get the best from any tool, some understanding of the principles and some operating skills are needed. More importantly, let's back up a bit, I recommend that the basic room treatments be done before any attempt at measurement. Corner Bass Traps, RFZ, including Cloud. When these basics are in place let's then measure and tweak positions to get the absolute best from the space. This article is intended as a simple primer on how to do the measurements, and how to view them. As of Feb 2010 it has been substantially revised to accommodate new wisdoms, the results of practical experiments and discussions on the fora. Please remember it is a Primer, intended to help you get going quickly and easily. I have simplified a few issues in order to keep it tight.

What to expect
There is no silver bullet here unfortunately. Expect to see horrible looking graphs. Software can excel at comparative tasks, such as finding the best speaker and listener positions. However it takes an experienced brain to evaluate how a room sounds or errs, and what to do about it. Let’s not expect a health check and a treatment prescription from these programs. They are tools not room doctors.

The Graphs
I will deal with only a few of the easiest understood graphs. For the rest, RTFM.
Frequency response is the most common and is easily understood. The Waterfall is a perhaps more powerful one. Most of our rooms are dominated by strong modes. Tuned by the dimensions of the room, these modal frequencies ring on longer than their neighbours. Conversely there are modal nulls which lead to missing information. These modes are powerful and they wreak havoc. They cause individual Bass notes to stick out or to vanish. This makes mix decisions regarding the bottom end, where most of the energy in music resides, very unreliable. The Waterfall plot displays this behaviour so graphically that it hardly needs explanation. It is like a series of Frequency response snapshots taken at short intervals. Like a cartoon, these snaps portray action over time.


Decay graphs such as EDT, T20,30 are to be regarded carefully. Even in Labs, decay measurements under 125Hz or so are quite uncertain. LF modes dominate in small rooms, so the Decay here is not at all random and diffuse like Reverb in a large space.
The concept of RT60 cannot even exist in small rooms.
REW has a Topt feature which tries to address these difficulties with some success.
FM3 has also improved greatly at Decay and now includes third octaves.
The BBC used third octave decays as an indicator of room quality. No third octave decay length was permitted to differ from it’s neighbours by more than 10%. Note that this test will show up problematic modes. Any room passing this test would be a very fine one, IMHO.

Given time and experimentation, one learns how to interpret these different views of the room sound. Graph changes caused by say a change of speaker or listener position, are however very easily evaluated as an improvement or not.

The Microphone
You may well have a useable mic in your closet already. A multi-pattern condenser switched to omni will work fine. The frequency response bumps of say an 87 or 414 are tiny compared to the 30dB anomalies we find in real rooms. I recommend jumping in. Get a demo of the software and go for it with whatever mic you have at hand. Be aware that directional mics may have poor LF response compared to an OMNI. They can also underestimate some modes due to position and direction. With that caveat, most mics are OK for comparative tests. In time, if you develop a liking for measurement, you might want a more suitable mic or software. Measurement mics are omnidirectional Small Diaphragm Condensers. The best of these are optimised to have a very accurate and stable frequency response, unfortunately this is often at the expense of very poor noise figures. I recommend only the ones which are also quiet enough for recording. DPA and EarthWorks spring to mind here.


Sound Level Meter
SLMs, even cheap ones, are very useful for many purposes. They are well worth having around. There are many recent cheap units way better than the venerable Broadcasting Shed one. Many SLM's can output their Mic signal via the AC OUT socket. This is fine with decent units. Recent experience has shown it to be unreliable with cheap ones. Some have very skewed frequency response. Even worse, the AC OUT can have electronic noise or DC on it.
Not recommended.

Calibration
It is not necessary to calibrate your rig to standard SPL's.
It is however, wise to optimise the S/N ratio.
Adjust your speaker volume to drive the room very well. You may wish to block your ears with headphones or earplugs. Watch out for overload lights on powered monitors, particularly at HF.
This puts our source signal well above background noise, computer fans, etc. Longer sweeps or multiple sweeps averaged can also be used to increase S/N ratio.
With a decent sweep level established, now adjust your mic pre gain using trial measurements and/or the meters on your interface. Get a decent level, graph peaks should end up within say
-6dBFS to OdBFS.
This is really just normal digital recording level setting.

Correction
As one becomes familiar with the Software, one may want to avail of some little sophisticated extras. It is possible to input a microphone correction curve to adjust a known mic response to flat.
e.g. If your mic comes with a printed frequency response curve showing -2dB at 100Hz, you can apply an automatic +2dB correction in the software. Similarly there is a loop back correction facility to automatically correct for sound card or interface anomalies including delay. FuzzMeasure has a Create Time Domain Average facility. It is essential to use the loop back correction for this to function properly.


Acoustic Analysis Software
I will only mention the free or cheap packages that I know. There are many out there, from Freeware to extremely expensive Lab standard. However this little group have democratised the process by being both competent and affordable.
ETF and it’s current variations was one of the earliest programs. The Acoustisoft website has great walk-through tutorials on actual uses.
FuzzMeasure Pro is a similar product for the Mac. This one is very easy to use, but is fully competent. Another strength is it’s very sophisticated averaging of measurements. e.g. Take measurements at several different listening positions, then average them to get a broader sense of the room’s response.
REW now works with Mac and PC. It is very comprehensive, which results in a learning curve. It costs a nice round figure, 0. The manual deserves special mention, superb writing.
ARTA is PC only. It appears to be very comprehensive and reasonably priced. The manual is pretty much a text book, it explains the measurement techniques in hand to the very deepest levels.

Other Tools and Helpers
I use a collection of must haves together to achieve a holistic evaluation of the room.

A CD of reference Tracks. I bring mine with me everywhere.

The SoundCheck CDs by Alan Parsons.

LabMeter, a Mac Frequency meter from rustykat.com

A Room Mode calculator such as this one http://www.bobgolds.com/Mode/RoomModes.htm
There are many RM calcs, with different strengths. Test drive a few.

The free MiniRator and Test tones at RealTraps - Home
I use sine waves to drive the modes crazy. SignalSuite is nice and smooth. Try the Signal Generator plug-in in your DAW, sometimes the frequency jumps make this difficult. REW has a unique Sweep follows Graph feature on it's Signal Generator. This has a wonderful smooth action.
I use an ancient steam driven analogue one with a real sweep dial.
Oneself- the body can generate tones of varied pitch and duration, and most importantly can easily alter the source (and listening) location. This human moving tone generator uniquely reacts to the results of its own output. No loudspeaker has this ability to move about, provoking and reacting to room response. Short pitched tonal bursts such as grunts, barks, or vowels, can very strongly stimulate the booms honks and rings. Because of the variable stimulus location, we can identify exactly where they live. Labmeter can show the frequency. Handclaps are excellent for finding Flutter echo, or single slap echoes, again by moving into the actual paths and hot spots.
You might want to be alone when doing all of this!
A sound absorbent panel, temporarily held in place by a friend or mic stand, can categorically prove the paths of these flutters and such. This will coincidentally show exactly where to put treatment. Lastly, a caveat; some software graphs allow clicking on, say a peak, to identify it’s frequency. The resulting decimal point readout suggests great accuracy. I have found them to be way out, they are not meant to be accurate. For example P-z shows a 10 percent error at 100Hz. I wouldn’t use these readouts while mixing or to create a Room Eq. for instance.

How to do it
I am assuming a rectangular room, with speakers at the narrow front wall. Identify the zones at 3/8 (38%) of room length from the front wall and ditto from the back wall. These zones mathematically should have the best balance of room modal activity. Another rule of thumb suggests there is little Low Bass at the room centre. These are useful, often but not always correct, guidelines. However Measurement always trumps Theory. Use these guides only as suggested starting points. Prove by measuring. Be ready for surprises. Using masking tape, label the floor at all significant listening spots, e.g. Engineers seat, Producers seat, Couch. Use descriptive names and numbers for your chosen spots. I use names like L38FC (Left Speaker Front Centre), L38BL (Left Speaker Back Left) and so on. Establish your own system and stick to it. 8 or so spots seems appropriate in a small room. Don’t be afraid to change your spots in response to the graphs. Mount the Microphone or SLM on a stand or tripod. Seated ear height is good. If you use a Mixing Desk and like to prowl around, then include standing ear height. Mix and match heights if you like, but do remember to use fully descriptive labels. I tie a thread with a small weight to the microphone or tripod. The dangling weight over the marked floor spot guarantees repeatable location and height.

Controversy
There are conflicting views on the following, which gives rise to confusion.
This will all clear up shortly as my world domination plan takes effect.

I point the mic directly at the tweeter. This will give the best result in the vast majority of situations. You may see mics pointed at the ceiling, particularly in the USA. The issues behind this are quite complex so I have dealt with them here http://www.gearslutz.com/board/7257133-post1.html
Simply put, if you have a normal Free Field measurement or recording mic, point it at the tweeter. If you are using the internal mic of a USA ANSI standard Sound Level Meter, or any other Diffuse Field Mic, point it at the ceiling.
For Frequency Response and ETC drive one speaker only. (Or 1.1 with the Sub)
For Waterfalls and Decay measurements drive both or all speakers. This is more likely to get all the modes going. A Y lead may be needed, I use a Mono switch on my controller. You could drive the modes even better by placing the speakers in the floor tricorners.
Feel free to use the graph tools. I like to use smoothing to view trends, e.g. the slope of a curve, the biggest anomalies in a Waterfall. Without smoothing HF on graphs can be an unhelpful mess. Later on, when searching forensically for individual anomalies, use little or no smoothing, in combination with zooming.
On Waterfalls, limit the Frequency Range to 300 Hz to see the modes clearly.


Tips
A human body close to the mic causes strong midrange errors. Up to 6dB around 400Hz. So stay at least a metre away from the mic. Similarly keep the mic at least 1 metre from surfaces.
http://www.gearslutz.com/board/7467237-post1.html

If you see a suspected problem, e.g. a dip or peak perhaps caused by a reflection from your desk or floor. Try temporarily covering the reflecting surface with an absorbent or diffusing panel. If the problem vanishes, your suspicions were correct.

Some of the default view settings can be a bit odd so I recommend over-riding them.
Adjust the Window setting to the same length as your Sweep and Graph.
For Waterfalls set the duration to say 1000mS for an untreated room, 500mS for treated. Tweak to fill the screen nicely. The lowest visible level threshold interacts with this. Set it to say -50dB. Again tweak to fill the screen nicely while ensuring that the end tails of the Modes submerge nicely down into the noise floor.
Be aware that many mix rooms are quite noisy due to fans and such. 45dBA is common, so use that lower limit control as you would a noise gate.




How to view it
Optimistically! Take a look at the Frequency Response, third octave smoothing.
Ideally this should be a nice even slope from say +3dB around 100Hz to perhaps –3dB around 10K or so.
Next restrict the viewing frequency range to say below 300Hz. Lighten the
smoothing or turn it off entirely. Observe the Peaks and Nulls.
Remember 40/80Hz is around the low E on electric Bass and electric Guitar.
Kick Drum and Bass action is hot from 40-120.
Now let’s expand the spectrum again, say up to 18K.
Look for signs of comb filtering. This will have a repeating series of dips and spikes,
thus the comb. Experiment, use the view changing tools of Smoothing and Zooming,
to see right into areas of interest.

Don’t panic
The curves almost always look awful. Frequency response graphs show scary peaks and dips. Waterfall plots will often show one very low extremely long decay accompanied by a confusing array of peaks and dips up through the spectrum. For some perspective let’s take a real world example.

This room was from hell. It has mostly concrete surfaces, it is asymmetrical, and worst of all it has alcoves. It is a tuned indoor swimming pool.
The Green Before curve shows the room with elementary treatment; the Red After curve is with much more considered treatment, of considerable quantity and quality. We did the lot; four corners, alcove corners, Ceiling Cloud, RFZ. Sadly, the After curve has a very similar shape to the Before. Disappointing and very common. Let’s look closer focussing on the musical crucial 100 Hz zone. Here we find a broad 6dB improvement. Now, consider if you were to eq a full mix with such a broad 6dB boost….. This is a big and welcome change. This room is used every day professionally, with great success. Perhaps the curves generated by software are not great at fully describing a sonic experience. A survey of Engineers gave this a 9. The graphs however deliver very clear warnings. E.g. Be careful of Bass decisions in the 100Hz zone. Try to find workarounds. Find the listening spot spots in the room where the graph is flattest. Perhaps refer to top quality headphones to hear the mix without any room or speaker.

Octave Reverb Time Measurements varied a lot with position, some were zero. The modes were so clearly so overpowering that these Decay views were of little use.

The Waterfall plot told another story. On a cursory glance the After again looked very similar to Before in shape, just generally shorter. Not very impressive.
However, with focus we find the good story. Before, the Waterfall showed a 1.3 Second long boom at 37Hz. After, it was reduced to 0.7 S. Go SuperChunks. Sonically, this changed a Kick drum from a chest massage to a nice ‘subby’ thump. Note there is almost no sign of this huge anomaly in the averaged Frequency Response graph. Waterfalls are IMHO the most useful of all the graphs.


Conclusions
Software alone cannot evaluate a listening space. It does however bestow great certainty when making comparative choices. When treating a room there is simple clear advice on the usual websites as to where to put treatment and why. They all agree on the basics; Broadband or Bass Traps in the corners, a Ceiling Cloud, Side Reflection points. This is not voodoo and it doesn’t change from room to room. Treat the Room first, then use Software to make comparative choices such as best speaker position, best seating positions, best speaker eq settings, etc.

Dan FitzGerald AMIOA MAES

Irish Acoustics
Sound Sound - Homepage

[jhbrandt]

Excellent, Dan.
Cheers!

[avare]

+1! Great! Even better than before! Thanks!

!,
Andre

[Glenn Kuras]

Outstanding Dan!!!

[Weasel9992]

Fabulous Dan!

Frank

[jd_]

Thank you!

I am just entering the devilish world of room acoustics and this will help me a lot.

[Ethan Winer]

Originally Posted by DanDan

Room Analysis Software Primer V2
Substantially Revised Feb 2010

Sticky! Sticky! Sticky! Sticky!

[Jeffrey Hedback]

Bump...Sticky!

[DanDan]

Thank you gentlemen for the kind words. The work was all in V1. However one learns, so this one is hopefully a lot more useful. You will of course recognise the substantial revisions. They are direct results of our discussions here, which kinda says a lot. Hopefully this is now more of a consensus with the aim of minimising confusion amongst the target audience. There is no 'right' way to do this, but we are gradually establishing one.
Best Regards, DD

[jhbrandt]

BUMP!

I just wanted to get the back to the top.

[Heavyfunker]

You rock dude! great work and thank you for this thread!

[rosevision]

+1

[tonio]

DD,
Nice primer.

Question on driving both speakers for waterfall / decay measurements. How do you do that in Fuzzmeasure? You can only set up 1 channel at a time.
Do you just take 2 readings (L & R) and average them?

How to set up for ETC ? A full sweep or a short chirp, @ what frequency?

Thanks!

[DanDan]

Hi Tonio, thanks. Chris has good reasons why he didn't include two speaker drive in FM. You can live without it just fine. Two speakers should excite the modes in the room a bit more evenly. One could be in a null, but two is less likely. More even mode stimulation should give better Waterfalls.
Two drive is not exactly the same thing as two individual responses averaged. Two speakers driven will fight somewhat, just like in real life with music!
I harbour the opinion that Room Measurements should ideally be done with two speakers running BUT with a Dummy Head Binaural Microphone recording in Stereo. FM can do this. This should give results more relevant to how and what we hear. I currently simulate this by moving the mic about half a head distance to the Left and Right of my measure spot, and repointing it directly at each tweeter. Ears.

Finally to answer your question :-) I use the Mono switch on my speaker controller to drive both speakers. If you don't have that facility, make a Y cable. For ETC drive one speaker. Full range sweep. For all measurements it can be worth trying longer sweeps, say 10 Seconds or/and synchronous averaging. Better S/N ratio, important for EDT.

DD

[tonio]

Thanks DD! Suppose I missed the mono switch thumbsup Makes sense now. Does it matter on the smoothing algo in Fuzzmeasure? I've read its best to use 1/12 octave in FM? If using 1/6 octave (as in your example) it really smooths it out, but if switch to 1/12 octive the peak/nulls get deeper.

Also, (in FM v3) Chris has the impulse response on the bottom of the GUI. Is it recommended to move the start point to the initial amplitude reading? Moving the starting amplitude changes the frequency response (somewhat).
Its like a science in just reading/setting it up.

T

[DanDan]

tonio,
No single smoothing choice will suit all purposes. Thus the choices.
I regard smoothing as similar to zoom tools in a DAW or Sketchup. Zoom out for perspective, to see an overall picture, zoom in to ferret out some little anomaly. At LF it is necessary to use little or no smoothing in order to fully see the peaks and dips. At HF, the same setting results in an almost unviewable mess with hundreds of dangerous looking black lines. Not realistic. Use whatever smoothing delivers the most informative view. Ditto the frequency extents. There is little point in viewing modal activity over say 300Hz on a Waterfall.

A perspective comes to mind, how does this picture relate to the sound, the tonality, how does that sound look, or how does that picture sound.
The ear has a finite resolution. Like pixels in a picture. Nothing smaller than a pixel can be distinguished. Subject to ongoing review, but historically and with good reason, this has been regarded as one third of an octave. I use 1/3 view to get a sense of the tonality of the room. Flat would be quite harsh. A nice downward slope towards HF works well in a small dead non diffuse room. Without smoothing these trends are not easily visible.
Also, without smoothing we can see all sorts of spectacular anomalies, e.g. vicious looking comb filter dips. These often have extremely narrow bandwidth, rendering them relatively insignificant. In that case I regard a smoothed view as closer to audible reality.
But obviously if I were chasing the source of the combing I would turn off the smoothing and use graphical zooming to completely over emphasise the anomaly I am working on. A magnifying glass.
DD

[DanDan]

tonio, I kept this deliberately separate. I recommend that you address FM questions directly to Chris. He is very interested and responsive. John of REW is similarly extremely helpful.
In FM you can Create a Minimum Phase Copy of your Measurement.
As far as I can make out this is always a good idea. It moves the IR and ETC to zero. It also improves the clarity of Waterfalls. Perhaps other benefits.
When doing quick comparative tests I just shoot and view, move along.
When digging in, I always use that MP function.
Perhaps it should be the default behaviour.
DD

[tonio]

Dan,
Thanks for all the insight! I am sure myself and fellow GS' can use all the information to reach audio nirvana.

You are an asset to this community

T

[Rod Gervais]

Dan,

First time I ever saw this - a fantastic job........

Rod

[DanDan]

Thanks Rod. And Heavy and Rose.
DD Blushing but chuffed.

[miner]

I just started treating my room ... lots of jigsaw figures on my screen ...

This post is great, it's a ray of hope

Many thanks DanDan

[Jens Eklund]

Top effort DanDan!

I would like to include the mention of perforated panels as a very effective solution (if applied correctly) for the trickiest bass problems (but also the midrange when you don’t want to absorb the highs).

Sincerely Jens Eklund

[Brian Cares]

Hey guys,

I just posted a thread about this:

http://www.bbc.co.uk/rd/pubs/archive...icpractice.pdf


But i thought it also deserved a link in this sticky.
The document contains the bundled knowledge of the BBC and includes detailed plans and every information you need to build and treat a studio to perfection.

Pure gold

[dave999z]

first, very helpful stuff. thanks a lot for doing this!

second, my only mics are a studio projects C1, an SM57, and an SM58. i presume none of these would make a good measurement mic. are there any affordable omni small diaphragm condensers worth getting (i.e., that would serve their initial brief purpose for room measurement but then actually be useful for recordings down the road)? perhaps in the price range of an avantone ck-1 (approx. $150)? if anyone has thoughts on a good option, that'd be great!

[DanDan]

The Avantone specs look very good. Remarkable value. Take note of the HF boost in omni when viewing measurements. OR, write a correction file from the published frequency response.
Assuming this is made in China (or maybe Burma ) I personally avoid buying such products as much as possible. If everyone continues to support this, i.e. direct copying with no R and D, there will be no new inventions or genuine products. How much is that concept worth?

DD

[kidcarbon]

Thank you for that piece Dan.

I am having trouble using REW.

I have an old Terratec EWS88mt sound card that produced this wonderful looking flat response during calibration.


I then set up a CAD e70 (omni) > focusrite ISA One > EWS88mt > Room EQ Wizard. This chain yielded these graphs from one sweep, at one listening position.


I am having a really hard time interpreting this stuff. I can see the +8db peak at 50hz as well as the -8db null at 60hz. Everything past 300hz seems like nonsense.

I understand it would probably be best, just throw up as many bass traps as money can buy, and then things will be 90% ok (maybe). I have been following the teachings of Winer, Kuras, Brandt, Andre, and now Yourself, but i just can't seem to wrap my head around what to do with this info.

Bottom line is i would like any suggestions on a remedial route to sweep/impulse interpretation. I would like to look at these graphs and say "hell yeah this room is awful" and then do my best to fix it with appropriate treatment. After treatment, i would like to look at this data and say 'ahhh, that's what that fiberglass is doing."

Before i start exploring all the 8+ possible listening positions for peaks and nulls i would like to truly understand what i am looking at in one position.

Thank you,
-Nick

P.S. The Master Handbook by Everest is light years over my head.

[kidcarbon]

im unsure how to get those fancy images that most people have posted. here are the links to the three images on my previous post.

EWS88mt Calibration

Sweap Measurment

Waterfall Graph

Thanks again,

-Nick

[DanDan]

kid, you have illustrated my main point perfectly.
i.e. These Softwares do not answer broad questions like

How bad (or good) is my room?
What treatment do I need and where to place it?

Having said that, I reckon that waterfall is not at all scary. The full range graphs always have too much information. Use third octave smoothing if you want to see the tonal slope or trends. Use full resolution and zooming to examine particular details. If you drive both speakers there is no point in viewing full range. All you will see is the result of a battle fought with sharp comb filters.

The room treatment and setup articles at RealTraps and GIK will get you there.

DD

[kidcarbon]

Hello All,

First off, thank you to Lupo and DanDan for the pointers and references.

After 3 days of reading hundreds of posts and using REW, i think i finally got over the first foothill. (only 10,000 more feet to go)

The room dimensions are as followed: 24' x 20'. Although this a basement room, the walls are covered in raw ceder planks. the floor has one small area rug at the listening position while the rest of the floor is concrete. the ceiling is a standard grid style drop ceiling. No acoustic treatment has been employed.

The mix desk is centered along the 20' front wall. The monitors sit about a 1' off the wall. The listening position is 5' from the wall

The graphs included in this post were all done with one channel (left). The SPL "c weighting" box was un-ticked. and the loop correction for the sound card calibration was completed.

I did not calibrate the CAD e70 mic that i used because i had no .cal file. from what i have read, calibration is an insignificant gesture, when compared to the modal and spectral anomalies of the room.

I will now make a feeble attempt to extrapolate data from these graphs. i did the best i could to massage the "windowing" to meet the requests of so many senior posters.

THE FREQUENCY DOMAIN

The Frequency Response graph and Waterfall graph are telling me that there is a huge peak in the room at 60hz and 160hz. Looking further into the waterfall i can see that the modal ringing at 60hz is lasting for about 1 second. I think this is a significant finding. The numbers tell me to start adding bass traps to the corners. (everywhere???)

The second really obvious assessment, is the null at ~105hz and 120hz. From most of what i am reading, except for Ethan's posts, I can do nothing about this except move the listening position forward or backward, an inch or two, and try again.

THE TIME DOMAIN

The ETC graph is telling me, via the measuring tool, that the first significant reflection had to travel 2.7 additional feet in order to reach the microphone. This means that a reflection point 1.35ft away from the microphone, is causing this reflection.

Well i got out the trusty measuring tape, and the closest surface i could find at 1.35ft was the desk the monitors are perched on. (it was actually 2 feet away)

The 7ft ceiling was three feet away. I am assuming that the ceiling is the cause for the next reflection seen at about 3.75msec. (going by the math i would expect the ceiling to be ~2 feet away if the total extra distance traveled for the second reflection was 4.21ft.)

Questions:

Is there any ideas on on my modal issues? Peaks and Nulls.

Am i approaching the ETC interpretation correctly?

Am i in the ballpark for locating reflection points or is my math and logic wrong?

Thank you,
-Nick

[elan]

whoa a room with a 12 db or more dip is used professionally everyday? Interesting! So my room is better? :D

Would be interesting to see a not averaged to 1/3 frequency response graph, I would like to see the comb filtering :P

And also a L - R measurement to have an idea of the symmetry

I'm looking for pro studio freq response all the time!! Just to have an idea of what is a good result

Thank you