The poor man's RTA

THE MICROPHONE

The only critical item is the microphone. Unlike other tests such as Ethan Winer's stepped sine wave project or Bob Damiano's excellent RT-60 tester (both free and both highly recommended), an RTA needs to have a flat, wide-response microphone.

If you go out and buy a hardware RTA, it will come with a calibrated microphone. But since the objective of this exercise is a no-cost RTA, the next best thing is a quality condenser microphone, preferably one capable of an omnidirectional pattern.

Omni mode is preferred for two reasons: first, all microphones exhibit a flatter frequency response in omni mode; second, you want to capture the effects of all reflections in the room, from all angles.

IMPORTANT: Make sure any pad or high-pass filter on the microphone is switched off. If you're going through a preamp, also make sure any equalization and compression has been bypassed.

You'll want to take your measurements from the same location in the room that your head normally occupies while mixing. Measuring at different points in the room will yield different results, so you want to specifically measure what you normally hear while mixing and tracking.

To summarize the microphone considerations:
- Use the best condenser mic you have access to
- Use omni mode
- Switch off the pad
- Switch off the HP filter
- Bypass EQ and compression on the mic pre
- Position the mic at the mix position, at ear level

THE SONAR PROJECT

Open a new SONAR project and add two mono audio tracks. You won't need anything else, not even a master bus. The most important settings are noted by a (!).

Track #1:
- mono interleave
- input: None
- output to main speakers
- import your test noise wave into this track

Track #2:
- mono interleave
- input is your microphone
- output: None (!)
- insert Voxengo SPAN in the effects bin
- enable input monitoring (!)

VOXENGO SPAN

The following SPAN settings are chosen to best emulate a hardware RTA. Starting with the top row

of buttons:
- Block: 2048
- Speed: 8
- Slope: 0.0db (!)
- Channel: Avg
- Monitor: RealT, Instant, 1/3
Of these, all are default settings except for the slope and the 1/3 octave display mode

The bottom buttons:
- MeterMode: Pure
- Low/High Freq: leave at defaults, 20Hz - 20KHz
- dB Scale: adjust to suit your preference
- Peak Hold: Off

ANALYSIS

So now you have test results. What do they mean?

First, don't freak out if it's not flat. No room is flat, even those that are equipped with extensive acoustical treatments. Deviations less than 5 or 6db are on a par with some of the best control rooms in existence. Only worry about deviations over 10db, as those are the ones that'll sabotage your EQ. They're also the ones you're mostly likely to be able to do something about.

Start by identifying the largest deviations and note their center frequencies. For low frequencies, divide a frequency into 1130 to obtain the approximate wavelength. Once you know the wavelength, determine which room dimension most closely matches it. That'll tell you where adding bass traps and absorbers will have the greatest impact.

Take a look at the screenshot from my own test:



It shows that my response is reasonably flat above about 300Hz. This is typical for most small rooms - problems usually start at 300Hz on down.

There is a peak around 150-200Hz. I know from previous tests using Ethan Winer's stepped-sine project that I have a nasty resonance at 179Hz that corresponds to the wavelength of the width of the room. It's a nuisance, too, as I have to constantly fight the urge to EQ out the resultant mud I hear in that band. But if I did adjust for it using equalization, then the mix would sound thin on other playback systems. Frustrating, but typical of most small rooms. More bass traps are the only cure.

I also show a rise at around 60Hz. It's probably not resonance in this case, because the
wavelength at 60Hz is about 19 feet, and the longest dimension of my room is only 13 feet. The maximum response is below 40Hz, too low to be caused by room modes. The reading most likely indicates that I simply have my sub turned up too loud.

EDIT:
The snapshot above was taken early this morning. People were still asleep in the house, so I did the test at low volume and the mic was off to one side, in front of the subwoofer (which is off to one side).

A subsequent re-test at 85dbSPL and the mic properly positioned did not show the same low-end bump. The range below 60Hz was fairly flat.

The moral: subwoofers really are somewhat directional, despite the common wisdom that they aren't. If you put your mic right in front of it, it'll measure louder than what you actually hear at the mix position.

Great instructions. I really appreciate the time you took to write this out. I also think your interpretation of your results is very helpful.

I'm assuming you aim the mic basically between the monitors?

I thought that pink noise was the preferred noise for audible testing

Either white or pink noise may be used. It's all about how you interpret the display.

Pink noise is just filtered white noise. The idea is to reflect the way our ears respond. But compare a pink curve to Fletcher/Munsen and you'll immediately see that pink noise is a very crude approximation of reality. For testing, all it really does is tilt the graph to one side. It does not convey any additional information, it simply displays the same information differently.

That's not to say you shouldn't use pink noise - if the result is a graph that's easier for you to comprehend, then that's the way to go. Personally, I prefer a straight line.

ORIGINAL: bitflipper

I thought that pink noise was the preferred noise for audible testing

Either white or pink noise may be used. It's all about how you interpret the display.

Pink noise is just filtered white noise. The idea is to reflect the way our ears respond. But compare a pink curve to Fletcher/Munsen and you'll immediately see that pink noise is a very crude approximation of reality. For testing, all it really does is tilt the graph to one side. It does not convey any additional information, it simply displays the same information differently.

That's not to say you shouldn't use pink noise - if the result is a graph that's easier for you to comprehend, then that's the way to go. Personally, I prefer a straight line.

Ah, I think I see what's going on: we must use our Frequency Analyzers differently. When I load up mine, pink noise is a straight line.

When I load up mine, pink noise is a straight line.

That's because the analyzer automatically compensates, reversing the 6db/octave rolloff of pink noise, so that the normally-tilted graph is straightened out. Theoretically, it should give very close to the same results as white noise without the compensation (setting SPAN's "slope" setting to zero turns off the compensation).

Nice one BF!!!!

Guess what I'll be doing this weekend.

Thanks Dave for that. One day I'll get around to using your creative RTA.

(Right now my near-monitors are in violation of proper placement (being too near my work-desk-surface) ... and my room is not fully-ideal for producing.) ... Your RTA may one day help me apprehend better translations from my monitors/room.

Right now my near-monitors are in violation of proper placement (being too near my work-desk-surface) ... and my room is not fully-ideal for producing.) ... Your RTA may one day help me apprehend better translations from my monitors/room.

I just recently rearranged my room just so I could pull the monitors further out from the wall.

Man, that was a major chore! This is also my day-job work office, and it's packed. Two desks, computer gear, keyboards, and loads of 703 panels. I had to untangle cables that had been secretly copulating like worms under my desk for years. It basically meant rotating everything in the room 90 degrees and then figuring out where all the cables used to go.

But it was definitely worth it. There was an immediate improvement. Though I have a brand new anomaly in the new position that wasn't there before, overall the low end is now smoother, more predictable and much less muddy.

I also put the speakers on stands. Previously, they'd been on a shelf, which I'm sure was an audio source itself. I was also in constant dread that the weight of the monitors would bring the shelf down with disastrous results. Now, the speakers are finally at the proper height, distance and spacing, and three feet from any wall front or sides, on heavy (homemade) stands that won't fall over.

Plus, I got rid of a big pile of dust bunnies from under the desk. All in all, it was two days well spent.

Man I gotta get those bass traps going asap!!! This look baaaaad lol... I also need to move speakers from wall more like bitflipper just got done doing.

And from the highs, it looks like I need some good diffusion and high frequency absortion as well...

I used the MCA SP-1 for this test as my rode mic was probably more colored since it has a tube in it.

Actually, the Rode might work better. The MCA is a cardiod pattern, and the test works best with an omnidirectional pattern. You might check its frequency response chart (if it came with one), to see if it's got a hyped high end or anything.

I am generally distrustful of most mics in the 3-7kHz range, especially ones marketed primarily as vocal mics, since they all tend to have a bump somewhere in that range. I have a Rode NT-1A and it has an annoying bump at 3Khz. It's fine for male vocals, but that's about all it's good for. I used a KSM-44 for my test, which is extraordinarily flat in omni mode.

Fortunately, you don't need to worry near as much about the upper mids and highs. It's below 500Hz, that's where you have to get aggressive.

Well that's good. It means you can trust that the microphone had little effect on the measurement.

I guess that shouldn't be too surprising. Typically, most of the little hills and valleys in your mic's frequency response are going to be under 3db, while room anomalies are going to be much, much larger. Same for speaker response.

Maybe you could go as far as to say that the speakers and microphone that you use for this test are almost irrelevant, since their imperfections are going to be trivial next to your room's imperfections.

Well I went ahead and did the test today. The instructions were excellent as I believed earlier. You might put in a link for the Voxengo Span VST. A simple google search finds that in a hurry.

I ended up with a roughly 10 db bump around 150 hz. That isn't surprising as I have virtually no room treatment.

I used a Audio Technica ATM 31a which is a Cardioid Mic since I didn't have a Omni mic available. I tried pointing it in different directions and the results were surprisingly consistent.

Thanks again for writing this up!

Actually, the minor details you are saying there are nothing compared to some real room treatment that will transform the sound making frequencies way more true to your ear.

A perfect example of this is that 2 of us tried different mic situations and the response was pretty much the same, the curve looked the same in Span.

So, my point is, room treatment isnt really "detail", it is basic necessity, and I am going to make bass traps VERY soon and also add some diffusion and foam treatment as well