How relevant is measuring noise to calibrate speech output from a computer

[Laron]

Hello,

I want to measure/calibrate speech output coming from computer files. These are played back through home studio foster near field speakers.

I am looking for an efficient way to calibrate and measure the output,

1. How relevant is measuring pink noise to calibrate speech output from a computer

1. What would a set up look like?

2. How should I measure Spl, dba, dbc, fast slow? Why?

I'd love to hear from you!

 

[Blumlein 88]

You'll need to supply more information about what you are trying to accomplish. Calibration to what for what?

Are you asking about how loud to playback files for good intelligibility?

Are you asking how noisy the files can be and still be understood well by listeners?

Is it special situation like being a transcriptionist listening to files and preparing transcripts from them?

 

[Wayne A. Pflughaupt]

I want to measure/calibrate speech output coming from computer files. These are played back through home studio foster near field speakers.

I am looking for an efficient way to calibrate and measure the output,

1. What would a set up look like?

You can use the measurement software Room EW Wizard (REW), available for download at AV NIRVANA or Home Theater Shack, along with a calibrated USB mic from Cross Spectrum. The software is free, the mic is about $100.

Using REW’s RTA (real-time analysis) feature you can get a frequency response graph reading in real time as the speech is being played (which is why it’s called “real-time analysis”).

2. How should I measure Spl, dba, dbc, fast slow? Why?

Those are all related to measuring SPL (i.e. how loud a signal is). That’s not the same thing as frequency spectrum analysis, which is what your #1 is referring to.

3. Is it a good idea to test with pink noise instead of my speech audio files? How would I go about it?

Pink noise is used in conjunction with the RTA to determine the frequency response of you speakers. Has nothing to do with determining the anything about speech or voices.

Hope this helps.

Regards,
Wayne A. Pflughaupt

 

[DonH56]

Not enough info as Blumlein88 said. What do you mean by "measure/calibrate speech output"? If you want to level-match the files, the best way would be to use Audacity or some program that will perform real-time or off-line analysis of the files so you can match levels.

Pink noise is equal energy per octave and is normally used to calibrate a full-range audio system. I do not see how it applies to the problem of level-matching speech files, if that is what you want to do.

A-weighting is most appropriate for speech but not sure it is relevant to what you want to do. Actually, not sure what you do want to do.

 

[Laron]

Thank you Wayne! This is all good info and I will check out the REW.

And, if I just needed to calibrate playback volume in SPL dBA or dbC -- do think pink noise would be good? You see, calibrating the speech output has a wide dynamic range - so I am thinking it would be more reliable and consistent to use Pink noise (spa in dab or dbc). I heard dbC Fast is the right way for speech by dbA seems to model the ear curve better so I am leaning towards dbA. what are your thoughts?

 

[RayDunzl]

Voice fundamentals are in the area around 100-300Hz.

"This is me, speaking into my measurement microphone"

 

[Blumlein 88]

Wayne's advice is good.

You also might use a sound level meter app or RTA meter app on your phone. Cell Phone microphones usually roll off below 200 hz, but so do most voices and small speakers. Play pink noise over the speakers and measure with the phone app.

 

[Laron]

Thank you both for your responses! I will try to be more clear.

I need to calibrate the speaker playback volume for playing back speech files to be ~63db.

For this, I am using a SPL meter - I currently start playing the speech files on a computer -- and get a wide dynamic range when measuring. It is almost impossible to get a good consistent reading. So, I am toying with the idea to use pink noise for setting the volume levels.

Specifically, I think setting playback levels would be more reliably done by using a -14dB pink noise source (from the same playback machine) - and then calibrating the playback speaker to a level setting of ~63db.

So, my questions is,
-- is this method any good for setting up playback levels for speech - using pink noise vs the original speech files as a source.
-- should the SPL meter be set to dbA or dbC, fast or slow (assuming I would use pink noise) to calibrate speaker playback volume for speech files.

Thanks so much!

 

[Blumlein 88]

Speech is rather dynamic, more so than music most of the time. It typically will have a 50 db dynamic range with basic conversation. So when you say you need playback volume of 63 db, is that -63 db on the volume control, is it 63 db SPL (sound power level like with a sound level meter) or what exactly. Your purpose still seems a bit unclear to me. Are the speech files done at a consistent level themselves?

 

[RayDunzl]

Playing full range pink noise here (with a slight tilt in the "room curve":

63.1dBA SPL = 66.3 dBC = 69.1dBZ (unweighted)



Measured at 10 feet from the speakers.

 

[Blumlein 88]

For instance you could look at the files in Audacity, a free sound editor. Under one of your choices is Contrast. This will give an RMS average level for any mono track you select. I am not sure this is suited for your purpose however as I'm unclear what your trying to get done.

Here is an example. Two tracks, from two albums, and the RMS average level of each. Yet if you played this back with your sound level meter even set to A and slow you'll get lots of fluctuation in it.

 

[Blumlein 88]

What might be useful to you is to playback with Replay gain. That attempts to adjust gain of various sound tracks to be same/similar in perceived level. Then you set your volume to what is comfortable on one track and it should be fine with any others of speech. Replay Gain is built into many computer sound playback softwares like Foobar or Audacious or Clementine and others. You can turn it on or off.

 

[Wayne A. Pflughaupt]

Thank you both for your responses! I will try to be more clear.

I need to calibrate the speaker playback volume for playing back speech files to be ~63db.

For this, I am using a SPL meter - I currently start playing the speech files on a computer -- and get a wide dynamic range when measuring. It is almost impossible to get a good consistent reading. So, I am toying with the idea to use pink noise for setting the volume levels.

It’s common to use some compression or limiting in this situation, to keep signal peaks below the desired threshold.

So, my questions is,
-- is this method any good for setting up playback levels for speech - using pink noise vs the original speech files as a source.

In my opinion no, because pink noise has extremely limited dynamic range.

-- should the SPL meter be set to dbA or dbC, fast or slow (assuming I would use pink noise) to calibrate speaker playback volume for speech files.

I’d go with C-weighting (because its frequency response characteristics better mirror the voice – see Ray’s graph in Post #6) and slow, which will make it easier to get an actual read on the peaks.

Regards,
Wayne A. Pflughaupt