After generation of high frequency tones, e.g. 15 or 17 kHz, you hear a low frequency tone (estimated about 1 kHz).
I use this tones to measure students upper hearing range in a psychophysics course. It is difficult to measure this range when there are artifacts of easily hearable lower frequency sounds, despite they have lower ampliltude. The tones have ramps, so there is no clicking at the beginning and ending of each tone.
Nevertheless people hear tones when the “official” frequency is well above their hearing range. I have reproduced this artifacts with the German audacity 3.0 as well as with the English version 3.1.3. I have tested 2 USB sound cards, both reproduce this bug. Interestingly, the Spectrogram does not show any tones except the high frequency “original” tone.
I did some testing and I confirm this on my W10 PC Realtek soundcard with both 3.1.3 and the latest alpha master build I have for 3.2.0.
This is not a regression, its has been in Audacity from its very early days.
a) It’s not present in in Audacity 1.0 - but that’s only because there was no tone generator in the early versions
b) 1.2.6 (the next earliest version I have) does exhibit this audible low tone - but also has a hiss. And so does 1.3.0
c) by 1.3.2 you just get the extra audible low tone and this is the same right up to 1.3.1 and 1.3.2 alpha
I tested with a 15kHz tone (beyond the range of my aged ears these days) ad when I analyze the tone with Analyze > Plot Spectrum… I can only see the 15kz but no visible presence of a lower tone confirming your result: This makes me suspect that this is some kind of pulsing caused by the different elements of the generated tone
Note that this low tone also afflicts generated chirps - I tested a chirp starting at 10kHz (I can hear that) and ending at 15kHz and the low hum is there throughout.
I will log this on GitHub so that the Muse developers and QA can see this - as the devs do not visit this Forum.
If you zoom in to sample level on my 15kHz tone - you will observe that the tone is actually made up of 3 separate sine waves:
I suspect that this may be causing the low tone pulsing.
What you are seeing there is a kind of visual “aliasing”. If you switch from “stem plot” to “Connect dots” representation of the samples, you may be able to see why it is happening (it’s a similar phenomenon to “Moire patterns”). This “should” just be a visual phenomenon and not audible, but that requires that the DA converter behaves properly which may not be the case with a standard / cheap computer sound card.)
What are your settings in “Preferences > Quality”?
How are you playing the sounds?
Further testing om my W10 laptop (64-bit HP Envy) shows that I only get the phantom tone if I turn the Audacity volume uo with the output slider in the Audacity Mixer toolbar.
If I have the slider more than say 1/3 or 1/2 along then I can hear no “phantom tone”, but I can still hear the expected test tone. When I move the slider further along then the phantom tone reappears.
I get a similar effect with a very low setting in the Mixer toolbar and turning up the gain control on my wired active speakers (that my be the speakers’ circuitry I’m guessing.
a) the onboard Bang & Olufsen onboard speakers in the HP Envy laptop.
I could detect no phantom tone at any setting of the Mixer toolbar
b) Then I used my wired Sennheiser HD-25-1 studio cans
7kHz: No detectable phantom tone at any Mixer toolbar setting - just the 7kHz tone
15kHz: mostly silent (for me anyway) but at high settings, above about 0.75 on the Mixer toolbar and above a very very faint phantom buzz could be just about heard.
c) wired active speakers
15kHz: mostly silent (for me anyway) but at higher settings, above about 0.45-0.50 on the Mixer toolbar and above the phantom tone could be heard.
d) I then tried my wired active (high-end) Wharfedale Active Diamond speakers
with the 15kHz tone a phantom tone can be heard with any setting of the Mixer toolbar output.
Certainly using my good cans or even the onboard laptop speakers would be fine for the original poster’s psychoacoustic experiments. But my active speakers, neither low cost or expensive, would not.
UPDATE: the same is true on my Monterey Macbook Pro, fine with headphones or onboard speakers - bad with wired active loudspeakers.
Thanks for looking into the problem in replicating it. I did not change preferences.
Default Sample Rate = 44100
Default Sample Format = 32-bit float
Real-time Conversion = medium quality
High-quality Conversion = Best
I have now tested:
Best quality (formerly only medium quality)
32, 24 and 16 bit
Output frequencies 32000, 44100, 48000 Hz
and several mixtures of these preferences:
in all cases there was the identical side tone
The bit format (bit-depth) is not relevant to this issue.
The sample rate may be relevant, but there’s a lot of different things that could cause the issue - “resampling” (changing the sample rate) is one of them.
The reason that the Spectrogram does not show any tones except the high frequency “original” tone is because the “phantom” tones are not present in the audio data. They are an artefact created by the playback system. Windows and the sound card are the two most likely culprits in my opinion.
What I would suggest is taking Windows out of the equation. Export the audio as a WAV file, and play it back on a system that does not involve a PC (for example, make an audio CD and play it on a good quality CD player). Avoid converting to a lossy format such as MP3 as that will introduce a whole other raft of issues.
Is there a possibility that you are clipping (distorting by going over 0dB)?
In order to prevent clipping both signals should be -6dB (50%) or less. Distortion does generate new frequencies.
There is a beat-frequency (amplitude modulation) but it doesn’t really exist as a “signal frequency” and I’m pretty sure that isn’t perceived as a pitch or frequency.
The beat doesn’t show-up if you plot the spectrum, and it won’t be there if you bandpass the mixed frequencies and try to keep just the beat frequency.
Similarly, the beat-frequency can’t be filtered-out with a notch filter.
If you sum a 50kHz frequency and a 51KHz frequency (linearly) you shouldn’t be able to hear the 1kHz beat. (Of course you need a high sample rate to do that.)
You are right: I produced a WAV file with audacity. When playing it back with audacity (import on a different PC) I could clearly hear the side tone. The same is true when I use the Sound Button of PowerPoint. However: When playing it with Win10 Grove Music App, the side tone is not present! Thus, audacity has, probably dependent on the system used, a problem playing, but not producing, tones.
Yes, but the problem is not within Audacity, except for one special case.
When Audacity plays audio, the audio data travels a long path before it appears out of the speakers / headphones. In short:
The “one special case”:
If the track sample rate is different from the Project Rate, then Audacity resamples the audio data to match the Project Rate. If “Real-time conversion quality” is set to “Low”, then the resampling is done with priority on speed over quality, and some amount of aliasing (phantom tones) is likely.
Phantom tones could be produced at any later stage in the chain, or even at multiple places.
Ideally, you would ensure that the sample rate remains the same all the way from disk to DAC, but even then it is possible for the analogue components (amplifier / headphones / speakers) to produce phantom tones due to resonance.
