Sweep or Pink Noise for setting BFD EQ

Discussion in 'Home Theater Projects' started by Darryl_St, Mar 6, 2003.

  1. Darryl_St

    Darryl_St Extra

    Joined:
    Aug 15, 2002
    Messages:
    21
    Likes Received:
    0
    Trophy Points:
    0
    I'm setting up a BFD through the midi port with the BFD software ( very cool!)and RTA software ( w/ mic corrections)and my RS meter.

    I can play pink noise or a sine sweep and adjust my filters in real time to see my changes in the RTA software, I also play Pink and sine sweep Vobs in DD 5.1 and DTS.

    Question: I get different responses in the RTA when I use Pink or Sine waves, but which is closer to how movies and music will be affected by my filters? Slow sweeps show higher bass below 40Hz, longer sweeps show even more.

    Is there an agreed upon standard to use for the sweep time?

    Thanks

    Darryl
     
  2. AndyDC

    AndyDC Agent

    Joined:
    May 30, 2003
    Messages:
    38
    Likes Received:
    0
    Trophy Points:
    0
    I had the same question. Did you ever get any answers somewhere to these questions? I suppose the slower the sweep, the more it is like using a sequence of sine wave test tones with the RS meter, as many people seem to do. Did you find that the pink noise trace is closer to the faster sweeps that it is to the slower sweeps?

    Thanks,

    Andy
     
  3. Wayne A. Pflughaupt

    Moderator

    Joined:
    Aug 5, 1999
    Messages:
    6,207
    Likes Received:
    56
    Trophy Points:
    9,110
    Location:
    Katy, TX
    Real Name:
    Wayne
    I’ve used both and I prefer sine wave signals. Pink noise is random, so the readings fluctuate a lot. This makes precise readings and EQ settings much more difficult.

    Regards,
    Wayne A. Pflughaupt
     
  4. How about 100 shot averages of pink noise?
     
  5. AndyDC

    AndyDC Agent

    Joined:
    May 30, 2003
    Messages:
    38
    Likes Received:
    0
    Trophy Points:
    0
    Exactly. I've done pink noise, using linear averages of snapshots taken over say 20 seconds (I don't remember how many snapshots per second, but the average seems to stabilize after about that amount of time).

    I've also done 1/6 octave sine wave tones and measured with the SPL meter.

    I've not done the sweeps (maybe tonight!).

    Of course, there is always MLS analysis for a fourth method.

    I got somewhat different results from the first two methods, though I wasn't systematic enough to be able to compare exactly.

    My impression from other threads is that this may come down to how room modes are excited. Long (e.g. 5 second) sine waves will excite the modes completely. Pink noise jumps around a lot and so presumably doesn't do so as much? Sweeps would be somewhere in between, depending on speed.

    Even if this is so, what I don't know is which is the best in that it will produce the most pleasingly flat response. (Let's not get into house curves etc.). For example, the sine wave method may produce stronger spikes here and there. If you use pink noise to calibrate, you won't fully smooth them out. Is that a problem? I don't know.

    Of course, given enough time, I could experiment with all this, but I'm 41 years old so I don't know if I have enough time left to complete the work on my own . . .

    It seems that the Harmon people with the Rabos system are proposing the use of roughly 1/12 octave sine wave tones to calibrate. But this may just be because they don't want to assume everyone has a Windows PC and sound card etc. so could run RTA software.
     
  6. RichardHOS

    RichardHOS Second Unit

    Joined:
    Mar 11, 2003
    Messages:
    454
    Likes Received:
    0
    Trophy Points:
    0
    I don't think pink noise would be the best method, for the reason stated above: it may not completely energize a particular room mode(s).

    Sine tones or long, slow sweeps should fully energize each mode (provided the tones are spaced closely enough, or the sine sweep is slow enough). Since music may contain a sustained note at a problem frequency, I don't want a resonance problem that only shows up during these sustained notes.

    I'm not sure how MLS methods would figure into this.
     
  7. TimForman

    TimForman Supporting Actor

    Joined:
    Dec 4, 2002
    Messages:
    847
    Likes Received:
    0
    Trophy Points:
    0
     
  8. AndyDC

    AndyDC Agent

    Joined:
    May 30, 2003
    Messages:
    38
    Likes Received:
    0
    Trophy Points:
    0
    [Thisis a repeat of my post on a similar thread over at AVS forum. I don't see how to post the graph here. It's over there, though.]

    AVS post (with picture)

    It mostly talks about whether you need to adjust the RS line out, as some there were doubting. Also compares different techniques of measuring frequency response. Preliminary!]

    I have some preliminary results on different trace techniques and whether to adjust the RS line out readings. The bottom line is that (1) it seems that the line-out of the RS mike requires adjustment factors, as does the meter reading (as Tom V. says) but (2) the traces vary substantially depending on technique (sweep, pink noise, sine wave tones). The line out readings are much closer to the sine wave meter readings when the adjustment factors are applied to both than when they are only applied to the sine wave tones. If you have to choose between applying the usual factors and applying none at all, apply the adjustments.

    I compared 5 traces: (a) 1/6 octave sine wave tones measured for several seconds each using the RS meter, applying the usual adjustment factors; unadjusted (b) and adjusted (c) frequency response traces from Spectraplus based on a ~1 minute spectra plus sine wave sweep input through the RS microphone line-out to the sound card; unadjusted (d) and adjusted (e) readings similar to (b) and (c) but using Spectraplus generated pink noise, averaged until the trace settled down (about 20 seconds is enough). Sub location was the same for all runs, cut-off at 80hz, and the outputs were calibrated so that they all were about 65 Db at 63 Hz.

    The idea is (1) to see which of b/c and d/e match better with (a) and (2) to see how the traces from the different techniques compare, once the adjustment issue is out of the way.

    A graph of the results is attached.

    As you can see, none of the traces are very similar, except for the pink noise and the sweep over most of the range. They do give similar patterns, at least. The adjusted traces (c and e) are much closer to (a) than are the unadjusted (b and d). Clearly, the different methods seem to excite room modes differently, as expected. It seems likely that the adjustment factors should be applied, however. Otherwise, we would have to conclude that the methodologies differ even more than the adjusted results suggest.

    Future research ļ and comments

    1. I would like to try much slower sweeps, as these should, in the limit, approach the results of the sine wave tones and provide a better test of whether to adjust. I couldn¡¦t get Spectraplus to go slower than about 1 minute (and the avia sweep is about as fast and gives very similar results). I remember somewhere a free/share-ware signal generator that could do this, but I¡¦ve lost track of it.

    2. I would like to use 1/12 tones but I don¡¦t have the corresponding adjustment factors. I could interpolate the usual ones, but has someone done that?

    3. I would like to test MLS, though without shelling out the money I don¡¦t think I can save and easily compare the results from ETF.

    4. I will get the BFD (in the mail!). My tentative plan would be to use the pink noise or sweeps (which are cooler to watch on the screen) for sub location and quick checks, then 1/12 octave for checking and equalizing.

    5. Background noise is not the issue I thought it was. I have a background spike at 60hz at 96 db (this is not properly calibrated for level so is not directly comparable to the readings in the graph), general background at 66 db, and a signal of around 103 db including at 60 hz (this is pink noise plus the background). What would the 103db signal read if there were no background spike? From the magic of logs, 102db. In other words, that big looking 96 db spike doesn¡¦t really add much to a 102db signal.
     
  9. AndyDC

    AndyDC Agent

    Joined:
    May 30, 2003
    Messages:
    38
    Likes Received:
    0
    Trophy Points:
    0
    I've updated these results over at

    AVS forum thread on this topic

    The bottom line: (1) RTA of pink noise is not the way to go, at least in that it does not pick up many of the important modes that sine wave tones would pick up. The last picture I posted on the other thread was very revealing to me on this. (2) Even a one minute sweep does the job, as long as you set up the software to resolve frequencies down to something like 1 Hz. It seems to very closely tracks sine wave tones, though I've only tested with 1/6 octave tones, and it easily goes to much finer gradations, which are very revealing at least in my case. So (3) don't just use 1/6 octave tone measurements.
     
  10. DavidES

    DavidES Stunt Coordinator

    Joined:
    Jun 23, 2003
    Messages:
    116
    Likes Received:
    0
    Trophy Points:
    0
    I used Avia's 20-200Hz sweeps for room/sub only combo eq-ing with BFD. Worked great especially with Repeat pattern on. It really exposed the room booms. Used Avia's wideband pink noise to help match levels after eq-ing since the meter uncompensated causes me to run sub 4-6dB below mains.

    I used SpectraPlus with log and linear frequency scaling (each had it's advantages and disadvantages) displaying the range between 20-200Hz, peak hold, flat (unweighted) response, "uniform" smoothing, 48kHz sampling (I have an Audigy 1 soundcard), and the maximum fft samples available enabled/set which all together had a much finer resolution than 1/12 oct. It was 1.4 Hz resolution.
    It didn't even slow down my computer like Spectragram did at those same settings.

    The only gross errors would come from the RS analog spl compensation file on the internet (corrected on the high freqs. with zero degree axis chart values located in RS SPL meter manual) if there are any errors.

    Bass sounded mucho better after eq-ing. [​IMG] I don't think I'll need a house curve even. Hope this helps ya'll.
     
  11. AndyDC

    AndyDC Agent

    Joined:
    May 30, 2003
    Messages:
    38
    Likes Received:
    0
    Trophy Points:
    0
    That does help. Now, after much trial and error, that's how I do it (same settings and sound card too). It hadn't occurred to me to use the pink noise to match levels across speakers between sub and mains, but that makes sense. You could adjust the gain and watch the curve shift in real time.
     
  12. DavidES

    DavidES Stunt Coordinator

    Joined:
    Jun 23, 2003
    Messages:
    116
    Likes Received:
    0
    Trophy Points:
    0
    Andy (and others if you also noticed this), have you noticed an overall downward slope from low to high frequencies when viewing a log-scaled spectrum of Avia's wideband pink noise or is yours reasonably flat? Even my high-end corrected compensation still shows that slope.
    I can't tell if it's system, software, or what. Puzzling. System seems to sound right but I have high frequency hearing loss above 10kHz and tinnitus so I'm not a great judge of high end sound. Which scale are we supposed to use for pink noise spectrum, log or linear?
     
  13. Greg Monfort

    Greg Monfort Supporting Actor

    Joined:
    May 30, 2000
    Messages:
    884
    Likes Received:
    0
    Trophy Points:
    0
    Pink noise is the same energy per octave (AKA the 1/f noise), so it falls 3dB/octave, just like music. For this reason it's what should be used with an RTA to achieve a ~flat in-room response at the listening position. With the exception of organ pipes, there's no musical content with sinewaves that are of sufficient amplitude down low to wreak havoc with a typical sized room's modes. If a sound track has some artificial LF sinewaves for special effects that's excessively energizing room modes, then dealing with them separately with a parametric EQ is best if room treatment, etc., isn't an option.

    GM
     

Share This Page