Spectral analysis software -- why so bad with bass guitar parts?

[krabapple]

I don't know how many here use tools like Audacity or Audition (or full fledged DAWs) to analyze audio spectra, but I end up doing it fairly often for various hobbyistic and musicianly reasons.

I keep noticing that the spectral/pitch analysis tools of these are often bad at identifying the correct octave of a low note being struck on a bass guitar. Like, though I know for certain that the note being played is the low, open E string (that's 41 Hz, aka E1 in octave-based pitch-naming) or the low G (49 Hz, G1), they show as E2 (82 Hz) and G2 (98 Hz) in these views.

Typically the fundamental note is very dim or even invisible in the spectral view (and frequency plots), whereas its upper first harmonic is prominent. This is true even when I increase the resolution of the spectral analyzer.

Obviously for purposes of transcribing the notes of a bass part, this is quite misleading.

Is this just something I have to live with -- that a tone is 'objectively' dominated by the first harmonic , so the analyzer is telling me that is 'the' note , even though my ears (and fingers) tell me it's an octave lower -- or are there smarter analyzers out there?

 

[Redacted]

What are you using to get the bass signal into Audacity (etc) ?

I'd use my sig gen/pocket oscilloscope to inject a low E into Audacity and compare what Audacity thinks the amplitude is compared with say a signal 2 or 3 octaves above, thus checking the calibration of your signal chain again frequency. The scope only being used to be sure your input signal has the same voltage in all cases.

 

[krabapple]

It's a recording of (mostly) a bass. A .wav file. Simply import into the Audacity etc.

Yes, looking at a synthetic or pure tone at those frequencies is a good idea.

 

[Redacted]

Ah, my mistake. Already a digital file, not playing a live bass!

But then you don't know whether the bass was crunched or went through a pedal - there are many ways by which the apparent fundamental could be less strong than the 1st harmonic.

Try making a digital file of sine tones, at E etc as I said for the none-existent bass, and see what Audacity makes of those.

 

[MAB]

It's a recording of (mostly) a bass. A .wav file. Simply import into the Audacity etc.

Do you have an example .wav file you can post?
I use typically use REW and want to compare to Audacity.

 

[krabapple]

Well this is interesting. I used Musescore to write out a 'score' of a electric bass playing the octave notes E1 F1 G1 A1 B1 C1 D1 E2, and saved it as an audio flac file. Musescore does decent, though not SOTA, instrument synthesis. It sounds like a bass guitar.

Audition spectral pitch view set at high resolution shows it as this:

The red bars are the the spectral view and the blue lines are the pitch calls. It gets these calls right -- these are indeed the synthetic bass guitar notes being 'played', all in octave 1. But every note has signficant harmonic energy, and often there's a harmonic with more energy (brighter red) than the fundamental. The seventh (D1) and eighth notes (E2) are strong examples. For those, the fundamental decays very rapidly compared to the harmonics.

I think what's happening in my real world exxample is that there's too much other stuff going on (it's not a solo bass part, even though I have used tricks to get rid of vocals etc). Also, the 'real world' bass is messier -- a human playing, not a machine. But it's still frustrating to hear the bass part go from G1 --> E1 while the software view indicates
G1 --> E2 (bass going the wrong way!) or G2 -->E2 (right way, wrong octave).

 

[MaxwellsEq]

It's not necessarily a "reproduction" issue or a measurement limitation. I think it's a production or a bassist choice.

Most listeners are not in a position to hear the lowest string fundamental. They listen on portable speakers; they listen in noisy environments; in the kitchen etc. Also humans have low sensitivity to sounds in the bottom octave. Live concerts often deliberately roll off the bottom octave.

So bassist and/or producers use effects and EQ which boost bass guitar harmonics and don't boost (or even suppress) lowest octave fundamentals. They do this so that the bass lines can be heard on music systems with nothing below 80Hz and also so the tune is more in our higher sensitivity range.

 

[RandomEar]

It should also be noted that even with the fundamental missing and only the (correct) harmonics being present, our brain will still tell us that the fundamental is there.

I see no indication that those spectrograms/ FFTs are faulty.

 

[krabapple]

Good replies, thanks. I also glean that the length of the note matters for the analyzer to make a good octave call. The synthetic bass notes were all whole notes of a few seconds' duration each, while many of what I'm transcribing are of much shorter duration.

 

[Redacted]

Good replies, thanks. I also glean that the length of the note matters for the analyzer to make a good octave call. The synthetic bass notes were all whole notes of a few seconds' duration each, while many of what I'm transcribing are of much shorter duration.

A good observation - anyone used to a stringed instrument will have noticed that, once plucked, a string will go out of tune and back, and the harmonics increase (as a percentage) as the string amplitude dies away.

 

[krabapple]

Whereas synthetic bass notes say perfectly in tune for their duration.

 

[Redacted]

Whereas as synthetic bass notes say perfectly in tune for their duration.

Indeed - that's how people with keen ears can spot it!
I do not have very keen ears, but can tell with my guitar and a tuner.

 

[ChrisG]

A suggestion: EQ the bass before running it through the note-finder.

Something like +12dB at 40Hz, Q=1.

 

[krabapple]

Yes, I try stuff like that too.

 

[kyuu]

It's not necessarily a "reproduction" issue or a measurement limitation. I think it's a production or a bassist choice.

Most listeners are not in a position to hear the lowest string fundamental. They listen on portable speakers; they listen in noisy environments; in the kitchen etc. Also humans have low sensitivity to sounds in the bottom octave. Live concerts often deliberately roll off the bottom octave.

So bassist and/or producers use effects and EQ which boost bass guitar harmonics and don't boost (or even suppress) lowest octave fundamentals. They do this so that the bass lines can be heard on music systems with nothing below 80Hz and also so the tune is more in our higher sensitivity range.

I'm not sure it's necessarily entirely a deliberate EQ decision (though that sort of manipulation can certainly be taking place in addition). The strings on a bass guitar or what-have-you are not nearly long enough to actually produce those low notes. Even the strings in a grand piano aren't long enough. Instead, mass is added to the string to get the tuning lower, but the side effect of this is that you don't necessarily get a strong fundamental but instead the harmonics are often as strong or stronger than the fundamental.

 

[Cbdb2]

I'm not sure it's necessarily entirely a deliberate EQ decision (though that sort of manipulation can certainly be taking place in addition). The strings on a bass guitar or what-have-you are not nearly long enough to actually produce those low notes. Even the strings in a grand piano aren't long enough. Instead, mass is added to the string to get the tuning lower, but the side effect of this is that you don't necessarily get a strong fundamental but instead the harmonics are often as strong or stronger than the fundamental.

And strings and playing style make a big difference. Theres probably 10db more fundamentals in flat wounds than super brights and the same goes for playing softly with fingers compared to hitting hard with a pic or slapping.
EQ should help, I would use a low pass to reduce everything above 60 hz.

 

[kemmler3D]

The analyzer settings are important, you have to trade off time and frequency resolution, frequency resolution gets worse in the bass because the wavelengths are longer, so you might have to switch around settings to be sure.

As others have mentioned, roll-offs present in the recording can also be problematic.

If you are transcribing I think you just need to be smart about eyeballing it.

 

[krabapple]

I use maximum resolution to analyze bass frequencies e.g. window size >32K in audacity spectrogram settings. And I'm really only examining from ~320 Hz down. Window type (Hann, Hamming, etc) doesn't seem to make much difference.