Rational Acoustics



Arthur Skudra
July 8th, 2008, 02:58 PM
In light of a recent thread (http://www.rationalacoustics.com/forums/showthread.php?t=22) we kinda went off on a slight tangent on the use of directional microphones in measurement and the limitations in doing so, spurned some thought on my part on how much of the "room" we want to include/exclude in our indoor measurements.

A method of varying the influence of the room in your measurement is to adjust the size of your measurement window. However there is a trade-off: when you decrease the size of the window (increasing time resolution), you decrease resolution in the lower frequencies and vice versa. Pat Brown teaches this as the "Uncertainty Principle" as you increase resolution in time, you decrease resolution in frequency, likewise as you increase resolution in frequency, you decrease resolution in time.

I love the FPPO mode in Smaart, but too many use this feature without thinking of the implications in doing so. At low frequencies the window is huge, at high frequencies the window is small. Big implications here.

If you have been around the industry long enough, I'm sure you've heard the philosophy "you can't eq the room, only the loudspeaker system" and thus techniques using FFT windows and TDS became very popular, seeking as "anechoic" a measurement as possible. Fast forward to last year and Syn-Aud-Con's EQ07 seminar in Salt Lake City, and some of the research Dr. Leishman presented that demonstrated a different point of view:
"A room is an integral part of the audio transmission system. While its frequency response is typically a complicated function of both frequency and space, it has many global (or local) properties that can be beneficially equalized through appropriate procedures." (Slide 27) Wow!

Prior to my Smaart days, I optimized systems using a combination of a TEF analyzer and Ivie IE-30, kinda gave me the best of both worlds of time blind and time specific measurements. Now I use Smaart and a Ivie IE-35 as my tools of choice, though quite frankly Smaart alone covers all the bases well. I jump back and forth between FPPO mode, small windowed measurements, and the spectrograph to get not only a picture of what the speaker system is doing itself, but also to evaluate the influence of the room in my measurements, and carefully make decisions on what should be done with the system equalization.

Well enough of my babbling, I would love to hear other's opinions/thoughts on the use of time windows, and the influence of the "room" in their measurements.

Harry Brill Jr.
July 9th, 2008, 01:04 PM
This is the whole point of the decimated transfer function (FPPO). With a well designed FPPO, which Smaart has, we can really start to trust the data on the screen assuming other factors are taken care of, for example mic position. With FPPO a flat TF is an accurate system. I'm sure as the years go by we'll be able to have increased resolution (maybe 48PPO or even 96), and as we understand the hearing mechanism better, perhaps we'll find target curve implemented, and perhaps even SPL based curves that follow FM.

I'm curious if people prefer to haystack the system or use the console EQ to get good phat bottom end. If I'm on a console with decent EQ, I set up the system as linear (flat TF) as possible. I've found radical EQ on the vocal mics completely unnecessary. Mid range instruments come through just fine. Clarity is improved over all. When it comes to the bass, keys, kick drum, tracks, CDs, etc I EQ them to taste.

I don't drive subs from an AUX, but prefer a different output such as the MONO on some consoles, or a subgroup. Aux sends can be changed on a per channel basis. I like the ability to keep unneeded mud out of the subs.

Jamie
July 11th, 2008, 01:55 PM
You are certainly the the "coffee talk" host - "FPPO, friend or foe? . . . discuss." :D

Ok, so I guess I'll comment here on two topics you touched on. First, FFT TC and TimeWindow length and its impact on resolution and noise floor. As you have stated, the longer the time window you use (either chosen via FFT size/TC or through Time-windowing you IR measurement) the higher the frequency resolution of your resulting data in frequency domain. FR = 1/TC. This is one of those mathematical limitations that holds true for all measurement systems - similar to the need for your sample rate to be at least twice that of the highest frequency you wish to sample (Nyquist). This FR=1/TC restriction means that, the more reverberant information we wish to exclude from our data (the tighter we wish to time window our measurement) the poorer the low frequency resolution will be in our measurement. (Not to mention that the sub and LF energy from typical professional sound systems will tend to lag significantly behind HF due to system phase response - and will therefore cause even the direct LF energy of the system to fall outside of any relatively short windowed measurements) Zoinks.

However, besides the impact of TC on FR, what I would like to point out is that increasing TC size also has a significant impact on the noise floor of our measurements - this is particularly evident in IR measurements. As a quick experiment, do an impulse response measurement of a speaker in a room (view ETC). Start with a 4k FFT, then repeat with a 16k, 64k, and 256k. In each progressive measurement of increasing FFT (TC) size, you should see the noise floor drop with each increase.

NOTE - There is much more that I wish to post on this, but I have run out of time:(, so I will pick this up in a subsequent post.

until then . . .
-j