Pseudo-anechoic (windowed impulse) measurement on Smaart

Frank Koenig

Frank Koenig

Sophomore
Mar 7, 2011
187
0
16
Palo Alto, CA USA
www.dunmovin.com
I've explored Smaart 7 pretty thoroughly and I do not think there is any way to do this. Simply reducing the size of the transform -- that is using a power-of-2 length rather than "MTW" -- does not do it.

I know "horses for courses" and I have another tool, ARTA, which does this and is very nice, but Smaart is just so damn convenient.

So, first, is there a trick I'm missing? And, second, has anyone else desired this feature?

I don't think it would be too hard for the Rational folks to do. The "live" impulse response is already there. We just need the ability to multiply it by a (rectangular) window and transform it back to the frequency domain.

--Frank
 
Re: Pseudo-anechoic (windowed impulse) measurement on Smaart

Frank Koenig said:
I've explored Smaart 7 pretty thoroughly and I do not think there is any way to do this. Simply reducing the size of the transform -- that is using a power-of-2 length rather than "MTW" -- does not do it.

I know "horses for courses" and I have another tool, ARTA, which does this and is very nice, but Smaart is just so damn convenient.

So, first, is there a trick I'm missing? And, second, has anyone else desired this feature?
Click to expand...

There was a great thread a while back in which Nick Hickman went into this very topic: #37

The short story is, Smaart specifically does not exclude late arriving energy like you might think, in order to better correlate with the way our ears perceive that energy. If you need to completely exclude energy use a different tool.
 
Re: Pseudo-anechoic (windowed impulse) measurement on Smaart

Not sure if I understand exactly what you're looking for, but in your measurement config, you have several options for fixed FFT windows for each individual measurement engine. By default I believe that it uses a Hann window. You might want to check with the folk at Rational Acoustics to be sure.
 

Attachments

  • Screen shot 2012-08-01 at 12.50.26 PM.png
    Screen shot 2012-08-01 at 12.50.26 PM.png
    76.1 KB · Views: 2
Re: Pseudo-anechoic (windowed impulse) measurement on Smaart

Bennett Prescott said:
There was a great thread a while back in which Nick Hickman went into this very topic: #37

The short story is, Smaart specifically does not exclude late arriving energy like you might think, in order to better correlate with the way our ears perceive that energy. If you need to completely exclude energy use a different tool.
Click to expand...

Bennett, thank you. I remember that thread, I just didn't remember that it made it into that territory. Good stuff.

As an aside, I've observed that Smaart using MTW correlates surprisingly well with ARTA set to exclude reflections in a low reflection (outdoor, speaker up in the air) environment. Not so in my experiment mixing in an electrically created artificial reflection, where there is a huge difference.

--Frank
 
Re: Pseudo-anechoic (windowed impulse) measurement on Smaart

The ability to do a time-windowed TF (with an arbitrary length time window) is in the Impulse Response mode of Smaart 7. To do a windowed response, you must select the time chunk of the measured impulse response by zooming on the impulse response and viewing a "Frequency view" of the zoom. You can select any size time window using the zooming tools (you are not restricted to power of 2 FFT's). This feature has been in v7 since its release. My head is a bit squirrely right now (slept on a plane last night), but I will try to post a more complete description tomorrow. Note, the time chunk you are zoomed in to is indicated by the vertical lines in the top "nav" plot.
Windowed TF.png
 
Re: Pseudo-anechoic (windowed impulse) measurement on Smaart

Oh yeah - there are basically two different types of time (data) windowing to consider here. 1) windowing the single channel signals (meas & ref) that are used to create the response measurement. Here, the windowing serves basically to determine how early vs late arriving information impacts the measurement, as signal or as noise 2) windowing a response measurement to specifically include/exclude energy based on its arrival time - "what does this measurement look like without that reflection" - aka a simulated anechoic measurement
Wahoo,
-j
 
Re: Pseudo-anechoic (windowed impulse) measurement on Smaart

I got it to work. To summarize: you bring up the "Impulse" mode, not "Live Impulse", zoom on the area if interest, and click on the little downward pointing triangle that produces a drop down menu, from which you select "frequency".

The trouble is that the frequency response display is magnitude only. No phase. Which makes it pretty useless for speaker settings, which is the purpose for which I need a pseudo-anechoic measurement in the first place. I also had a bit of trouble getting a stable, repeatable display, but that may be due to my not getting all of the measurement parameters dialed correctly (transform size, etc.). The updates are pretty slow, so there really isn't any advantage over measure-and-look apps like ARTA, and I found selecting the region of interest a little fussy.

Anyway, for the time being I'll use ARTA for lab work, and Smaart for field work (aligning subs to mains, mainly).

Thanks for the help. Comments welcome.

--Frank
 
Re: Pseudo-anechoic (windowed impulse) measurement on Smaart

Ahh yes, phase. We did plot the phase info for the "Time-Windowed Transfer Function" measurement in previous versions of Smaart (v5 & v6) - and that capability was coded into the v7 TF engines, but the feature has not yet been added to v7's Real-Time mode. However, I would say that, given the time windows used for the MTW process, and its averaging & smoothing algorithms, I haven't needed a pseudo-anechoic measurement to get a stable, accurate phase traces in lab conditions - or even under reasonable room conditions.

Frank, I am curious, when you say lab work, what type of acoustical environment and measurement distance are we talking about? DTR? How loud are the reflections/reverb relative to the direct energy arrival. What frequencies are we talking about? 100 Hz? 1k? 10k? What size time windows are you using? Not trying to hassle you - I guess I am just asking because you are needing a psuedo-anechoic measurement - just trying to get an idea of the measurement conditions that are requiring it.

Also, feel free to contact me offline at Rational Acoustics and I can get you a copy of v6 to see if that has what you are looking for. As I said/implied, this is a feature that will eventually get activated in v7, it is just low on the priority list right now.
wahoo,
-j
 
Re: Pseudo-anechoic (windowed impulse) measurement on Smaart

First off, I'm just an old washed-up non-audio engineer learning my way around these things for the fun of it. I do have a background in DSP and Fourier analysis, although somewhat dated, so you can talk dirty to me.

As for my "lab" situation, for low-to-mid crossover work (300 - 500 Hz) I'm putting home made speaker systems (a bad idea, I know, but I'm learning, my next ones will be store bought) up on ladders and other tower like things outdoors and measuring (on axis, so far) at ~2-3m.

For example, on a 12 ft ladder my first ground bounce is at least 21 ms out, so by windowing the impulse response to eliminate the reflection I can get a frequency resolution of 47 Hz, or ~ 1/7 octave at 470 Hz, enough to useful, I think, and better as I go higher.

As has been pointed out, and as I've observed in my limited experience, you can get pretty decent measurements outdoors even in the presence of modest reflections, so maybe my concern with eliminating all reflections is overblown. Furthermore, and correct me if I'm wrong, the combing that results from distant (late) reflections has little effect on the smoothed magnitude at high frequencies, since the ripples get smoothed out. I think the phase still gets scrambled.

For low frequency work, where the speaker system is much smaller than a wavelength, I use either a ground plane or near field (mic ~5 mm from dust cap) technique, depending on what I'm trying to do. A 60 ft high (10 Hz resolution) tower would be cool, but my friends would get me committed if I tried to build one.

I also use Smaart to get detailed impedance measurements of woofers in cabinets, which has proved very useful in detecting and chasing down small, in-band cabinet resonances.

The last couple of days I've been trying to understand, and prove-out, my measurement tools a little better, so I've been simulating various scenarios using a DSP (Ashley NE 8800) in place of the electroacoustic (speaker-room-mic) piece. Fun and educational -- everyone should try it. Thanks for listening. --Frank
 
Re: Pseudo-anechoic (windowed impulse) measurement on Smaart

I'm not an expert in the precision that you're looking for, but if you only have the one bounce to worry about why not try to absorb or deflect it?
if you're only talking about one speaker and one mic the location of the bounce that makes it to the mic is very small. even something as simple as a piece of plywood in the right place would do wonders.

Jason
 
Re: Pseudo-anechoic (windowed impulse) measurement on Smaart

Jason Lavoie said:
...but if you only have the one bounce to worry about why not try to absorb or deflect it?
Click to expand...

Thank you Jason,

The reason I am concerned with the ground bounce is that the physical arrangement guaranties that there is no earlier reflection (except off the tower, which is small in area and the therefor highly attenuated, I hope). It is the worst case. If I clip the ground bounce, then all later reflections are clipped, too.

I think the difficulty with ad hoc modifications of the acoustic is that you don't know what effect they have since the speaker under test is unknown. You have to start somewhere, and, I suppose, if you're a real speaker company, that place is an anechoic chamber or a tall, skinny, and absorbent tower. Having said that, I acknowledge that a simple ground bounce is pretty easy to recognize and to verify by moving the mic a bit and seeing the effect. But it also blots out a range of frequencies you might be interested in.

I know this is all old hat for folks who routinely do measurement. Thanks for indulging me. I'm posting largely to get some feedback whether I'm anywhere near the right track. And this is the DIY forum, after all.

--Frank
 
Top Bottom