Projecting voice with minimal twang

[kickingtone]

I know many people swear by twang. For some, singing without twang is like beer without alcohol. I am not too keen on twang, myself, though.

 

I am trying to project my voice without using much twang.

 

I've tidied up my last attempt from an earlier thread.

 

https://soundcloud.com/kickingtone/our-last-summer-001le

 

Once I am happy, I will practice for real.

 

As a matter of interest, does it still sound dark in tone?

 

(I don't mind dark, so long as it doesn't impair projection of the voice too much).

 

Could it be explained by the frequency analysis of my voice?.

 

 

All that sub 300Hz stuff is quite strange. Which resonator is doing that?

[Masklin]

What audio are you using to generate that spectrum? The entire song?

 

If yes, I think that might explain the low frequency peaks - I'm quite sure the spectrum is not supposed to be used on an entire song, but shorter passages (a few seconds at most).

 

If you record a single vowel, for example, the spectrum will give you a lot of info. A spectrum of an entire song seems to me a big jumble.

 

If your spectrum really is taken from a shorter passage of your naked voice, the low frequencies could be artefacts from a low quality microphone - do have several microphones to test this?

 

What is the fundamental note you're singing in that spectrum, anyway?

 

:]

 

EDIT:

 

I took two spectra of two different ways of singing the 'Ah' vowel on an A2 fundamental (attached). Note the typical overtone peaks, starting with the octave, then a fifth on top of that, another octave, a major third here and there, bla bla.

 

I noticed though that some of the overtone peaks were oddballs like G or B, or even F and D#. Heh. I think that means my vowel isn't properly formed, or something. Anyway the effect of twang is somewhat apparent. I drew it in red. MS Paint! \o/

 

[kickingtone]

Hi Masklin

 

I am using the entire song. The spectrum is generated using a bucket algorithm which analyzes the sound in short windows and accumulates the result. So you get a fingerprint of your formants. Each time you sing a note that crosses a formant, it adds to the amplitude at the frequency of that formant.

 

(You could also use the spectrum to analyze the partials of a single note).

 

The mic is a cheapo £40 one, for podcasts -- not good quality. But I have analyzed other vocals using it, and the low frequency 'formants' are not there! I am talking 25dB difference + !

[Masklin]

Does the song include other instruments like bass guitar, guitar, drums? Or is it an acapella track?

 

You could sample different parts of the song to track down the source of the low frequencies!

[kickingtone]

Great demo of twang vs no twang, Masklin! My last post crossed with your edit. There is a clear increase in the first and second singer's formants in your demo.

 

My track is a cappella. I did put up a mismatched spectrum, but it doesn't seem to matter. They all have the same low frequency peaks. I tried 4 second samples throughout the track, and each sample had the low frequency peaks.

 

Here is the a cappella track for the spectrum I posted.

 

http://soundcloud.com/kickingtone/wh...and-done-001le

 

and here is the spectrum for the a cappella track in the OP..

 

 

Not much difference.

[Masklin]

Aha. Turn off all effects before you generate the spectrum, maybe? The echoes are low frequency signals.

[kickingtone]

The spectrum in my last post is before any effects.

 

I also tried recording silence. I get a peak at around 100 Hz. If I take off that noise the spectrum would look more like this (red bit is the removed noise)

 

 

...which looks less strange. Peak at around A3 and another at around B4.

[Masklin]

If ambient noise contains the peaks also, then it must be your computer or equipment or some other part of your immediate environment.

 

Or your microphone has gone crazy maybe?

[kickingtone]

The ambient noise contains one relatively tiny peak -56 dB. I've exaggerated the red area. It should be just a slither, as it is a log amplitude scale.

 

It can't be the equipment, because I have tried recording some other people's vocals and not got the peaks. And I get them consistently when recording my own vocals. I think they must be real. I am singing quite low.

[Masklin]

You mean you have recorded other people's vocals with the same equipment and ambient noise as now, and not obtained the low peaks? Are you sure nothing has changed but the singing voice?

 

Try changing the spectrum size to 4096, it will give you a clearer picture, I believe.

 

If you sing low I don't understand why you're surprised about the peak - it probably resembles all the fundamental tones in the vocal melody.

 

 

[davie]

I agree with Masklin, if you sing low enough, you'll likely get more of the low frequencies. But if its anything below 150hz then there must be something else happening. Were the other singers and you singing at the same distance? Just remember that singing close to the mic can cause a bass boost, especially if its a microphone with a cardioid pickup pattern.

[kickingtone]

You mean you have recorded other people's vocals with the same equipment and ambient noise as now' date=' and not obtained the low peaks? Are you sure nothing has changed but the singing voice?[/quote']

 

I play recorded vocals and record the recording through my mic. I don't see the low frequency peaks.

 

But if I re-record my own recorded vocals, the peaks are still there.

 

Try changing the spectrum size to 4096' date=' it will give you a clearer picture, I believe.[/quote']

 

Good call. I've pasted in the noise floor. (No surprise that the mains frequency 50Hz and mains power frequency 100Hz dominate the noise.) Peaks still look real and extend down to zero! (confused)

 

 

If you sing low I don't understand why you're surprised about the peak - it probably resembles all the fundamental tones in the vocal melody.

 

Because I haven't seen the peaks when other singers sing low.

 

(I was actually experimenting, and comparing my 2000Hz+ frequencies, when I noticed the difference at the lower end.)

 

Also, according to the linear vocal tract model, it would appear that at least a quarter wavelength of the frequency has to fit in the vocal tract for you to be able to amplify the frequency. The male adult vocal tract is on average 17cm long. This means that any frequency > 68cm won't get amplified. That's the fundamental of anything below B4.

 

When I pointed this out in a debate (arguing that the linear model cannot be right), I was told that we do not indeed amplify low frequencies.

[kickingtone]

I agree with Masklin' date=' if you sing low enough, you'll likely get more of the low frequencies. But if its anything below 150hz then there must be something else happening. Were the other singers and you singing at the same distance? Just remember that singing close to the mic can cause a bass boost, especially if its a microphone with a cardioid pickup pattern.[/quote']

 

I've posted the expanded spectrum in my reply to Masklin, together with the noise floor. The amplitude is significantly above the noise floor below 150Hz, although my first peak (now expanded) is just to the right of 150Hz.

 

The method I used to examine the other vocals is explained in my last post, too.

[Masklin]

Try recording your voice at various gains and distances from the microphone. Perhaps it's like Davie said - extra bass artificially picked up due to your mouth being too close?

[kickingtone]

Try recording your voice at various gains and distances from the microphone. Perhaps it's like Davie said - extra bass artificially picked up due to your mouth being too close?

 

 

I'm using a usb mic and built in mic. Both give same results, regardless of distance from the mic.

 

I amplified a trace of a single note and I got this:

 

 

Same with either mic. And I'm running the laptop on battery to cut out the mains frequency.

 

I can see stuff at about 3Hz and at about 20Hz. It looks structured, but maybe non-linear, so there are probably lots of low frequencies in there.

 

I don't think it's the sound card causing it, otherwise the low frequency stuff will always show up. But I am going to find a steady sound source to compare with.

[Masklin]

From that picture it looks like you're recording with an extremely low sample rate of 4 kHz. 44 or 48 kHz is standard.

 

Also, I don't think there's any point in recording with a 32-bit bitrate. I bet your microphone only supports up to 16-bit.

 

Try playing around with those settings and see if it helps!

[kickingtone]

Changing the format from 32-bit hasn't made any difference.

 

I've experimented with the sample rate and sample size. 44100 Hz rate, with a sample size of 4096 seems to give something that looks a little more familiar. I suppose the rest may be down to the mic I am using and voice tone.

 

I am at a point where I have found a comfortable basic vocal coordination, that has a reliable feel. It's quite raw at the moment, but I intend to start honing it over the next six months.

 

Of course, I don't want it to be inherently too dark sounding, which is what I was concerned about. My ears were telling me no, but the spectrum was initially telling me yes.

[davie]

Have you tried recording while using headphone monitoring?

 

I find it a lot easier to adjust vocal colour and shape the sound while listening on headphones at the same time.

 

Also, I would not be too overly concerned with the spectrum, especially if the audible sound is fine. Usually when mixing vocal tracks you should be placing a high-pass filter around the 100hz range anyway.

[Masklin]

Mjeah if it sounds good to your ears, that's what matters in the end. Davie's tip about real-time monitoring is good!

 

The spectrum is fun and all, but for aesthetic purposes, your ears are all you need.

[kickingtone]

Have you tried recording while using headphone monitoring?

 

Good call. Not yet.

 

I'd like to thoroughly understand and stabilize the natural feedback first -- something I am going to have to do anyway. Then I'd be in a stronger position when using headphone monitoring.

[kickingtone]

The spectrum is fun and all' date=' but for aesthetic purposes, your ears are all you need. [/quote']

 

More or less. Part of it is about educating the ear as well. So all kinds of feedback count, in the end, but I wouldn't give a spectrum the last word, or anything like that. And hey, I am learning a bit about mics, filters, visual monitoring (twang etc.), etc.

 

[Masklin]

For sure, it's interesting and fun. I remember screwing around with it when analyzing different vowels. It still boggles my head a bit that the relative intensities of different overtones is what differentiates vowels!

 

I imagine that a low-pass filter would make many vowels sound much the same, so weird.

[kickingtone]

I seem to remember reading somewhere that the first two harmonics of a tone are enough to distinguish between the major vowel sounds. I think the basic vocal theory models the vocal tract as a linear filter. If I have understood correctly, the soprano does not have a formant matching the first two harmonics of some notes, so all her vowels sound the same for those notes.

 

As for messing around with the frequency analyzer in Audacity, I am discovering so many different things by mistake. For example, I am trying to get out of the habit of judging the pitch of notes at the edge of my range by the amount of effort I am putting in, lol! I am now starting to hear the note, rather than imagine it. The spectrum would show that I'd hit a note outside my range totally by accident (often while exploring some unrelated thing). I'd think back and recall the feel of the note, and, presto! I'd learn to hit the note consistently. You have to be really aware to be able to recall and exploit such 'accidents'.

[Masklin]

Yes I recognize that soprano thing from somewhere. Interesting! IIRC, every voice has this problem, but it happens much sooner for sopranos - this is probably why it's near impossible to understand lyrics in opera.