Sine Wave test

[Music Manic]

Hi guys

I've just conducted a test with sine waves at different frequencies and found something interesting with the vst oscilloscope

At High frequencies, around 10kHz and above there seems to be extreme amplitude modulation with the frequency in question. There also seems to be errors in the smoothness of the sine wave too. Not sure that's a graphical thing.

The Low Mids are very stable

The Upper mIds, the modualation is slight.

The lows look clipped but I guesss that's graphics

Could anyone expand or tell me which measurement tools are true and how to read them properly?


Thanks

[valis]

You sure you're not seeing the effects of dither?

[Music Manic]

You sure you're not seeing the effects of dither?

Well it’s an internal test so I don’t think dither applies.

[Music Manic]

Would a power supply create this phenomenon?

[valis]

It's difficult to say without examples and your process. You can attach files to posts...

[garyb]

does the music sound bad?

[fra77x2]

Have you recorded the waveform and observe it in a wave editor?

[David]

Gost - Analyzer?

[valis]

Examples please. We can analyze on our end and report our feedback.

[Spindrift]

Can only guess since I don't know neither what you use to generate or analyze the sine wave.

But generally, unless the oscilloscope applies a sinc function, high frequencies will become jagged since unlike the output from the DAC which has sinc applied, it will just draw a straight line between samples.

For exampe it looks like oscilloSCOPE doesn't apply sinc, so high frequencies will not look right at all.
However oscilloscope in bitwig grid does, and will show the waveform as it will be coming out from the DAC, not as numerically represented in the data.

[Music Manic]

It's difficult to say without examples and your process. You can attach files to posts...

Here's some pics that show you what I mean:

I tried it at 96kHz and it explains it's a samplerate problem.

[Music Manic]

does the music sound bad?

Not bad Gary but I've been testing it for an hour or so and at 96kHz the Highs and Lows are much cleaner and crispier. Pulsar is stunning in its quality still. It's a shame 3 PCI's cant run 96kHz smoothly for me but my system is old. I will upgrade to my MSI and see if it performs better in the future.

My guess is that the "brittle" sound comes from those highs around 10kHz when pushed hard, which people talk of in the digital domain.

[Music Manic]

Can only guess since I don't know neither what you use to generate or analyze the sine wave.

But generally, unless the oscilloscope applies a sinc function, high frequencies will become jagged since unlike the output from the DAC which has sinc applied, it will just draw a straight line between samples.

For exampe it looks like oscilloSCOPE doesn't apply sinc, so high frequencies will not look right at all.
However oscilloscope in bitwig grid does, and will show the waveform as it will be coming out from the DAC, not as numerically represented in the data.

I've used the Pulsar Scope control room to generate the sine wave internally Spindrift and also a vst one. They both produce the same results. I haven't tried D/A conversion or a D/D Loopback.

Could you expand more on how the software gives us our information and if graphics cards play any part in it too? Any pointers or links would be appreciated in learning how things work.

Thanks

[Music Manic]

Gost - Analyzer?

Yes I've tried analysing it with GOST and it shows sine wave amplitude modulation clearly at high frequencies unless sidebands are creating that. I'm not sure, maybe you experts could say.

[Music Manic]

Have you recorded the waveform and observe it in a wave editor?

No I'll try it.

[Spindrift]

Can only guess since I don't know neither what you use to generate or analyze the sine wave.

But generally, unless the oscilloscope applies a sinc function, high frequencies will become jagged since unlike the output from the DAC which has sinc applied, it will just draw a straight line between samples.

For exampe it looks like oscilloSCOPE doesn't apply sinc, so high frequencies will not look right at all.
However oscilloscope in bitwig grid does, and will show the waveform as it will be coming out from the DAC, not as numerically represented in the data.

I've used the Pulsar Scope control room to generate the sine wave internally Spindrift and also a vst one. They both produce the same results. I haven't tried D/A conversion or a D/D Loopback.

Could you expand more on how the software gives us our information and if graphics cards play any part in it too? Any pointers or links would be appreciated in learning how things work.

Thanks

Yes, looking at you picture this is what is happening.

At the very foundation of digital audio is the Nyquist theorem. It states that "a sinuisoidal function in time or distance can be regenerated with no loss of information as long as it is sampled at a frequency greater than or equal to twice per cycle".

If you just glance over it, it might seems intuitive and somewhat obvious. A waveform at 20kHz the frequency between the valleys and peaks in the signal will be 40kHz...makes sense.
However when you think about it, it seems wrong. First of all it would be a triangle wave not a sine, as as you know a triangle contains harmonics, and there should be "no loss of information", so we should get a perfect reproduction, ie a perfect sine. And if it is not exactly half the frequency the waveform will be very jagged, like you see in your pictures.

But the important part is that sound is a "sinuisoidal function". When the signal goes thru the DAC a sinc function will be applied, interpolating between the values in a sinuisoidal fashion.

Some oscilloscopes and editors will display a visual representation including the application of a sinc function, but most wont and instead draw straight lines between each sample.

Here is a video visualizing the concept:
https://www.youtube.com/watch?v=-jCwIsT0X8M

[Music Manic]

Can only guess since I don't know neither what you use to generate or analyze the sine wave.

But generally, unless the oscilloscope applies a sinc function, high frequencies will become jagged since unlike the output from the DAC which has sinc applied, it will just draw a straight line between samples.

For exampe it looks like oscilloSCOPE doesn't apply sinc, so high frequencies will not look right at all.
However oscilloscope in bitwig grid does, and will show the waveform as it will be coming out from the DAC, not as numerically represented in the data.

I've used the Pulsar Scope control room to generate the sine wave internally Spindrift and also a vst one. They both produce the same results. I haven't tried D/A conversion or a D/D Loopback.

Could you expand more on how the software gives us our information and if graphics cards play any part in it too? Any pointers or links would be appreciated in learning how things work.

Thanks

Yes, looking at you picture this is what is happening.

At the very foundation of digital audio is the Nyquist theorem. It states that "a sinuisoidal function in time or distance can be regenerated with no loss of information as long as it is sampled at a frequency greater than or equal to twice per cycle".

If you just glance over it, it might seems intuitive and somewhat obvious. A waveform at 20kHz the frequency between the valleys and peaks in the signal will be 40kHz...makes sense.
However when you think about it, it seems wrong. First of all it would be a triangle wave not a sine, as as you know a triangle contains harmonics, and there should be "no loss of information", so we should get a perfect reproduction, ie a perfect sine. And if it is not exactly half the frequency the waveform will be very jagged, like you see in your pictures.

But the important part is that sound is a "sinuisoidal function". When the signal goes thru the DAC a sinc function will be applied, interpolating between the values in a sinuisoidal fashion.

Some oscilloscopes and editors will display a visual representation including the application of a sinc function, but most wont and instead draw straight lines between each sample.

Here is a video visualizing the concept:
https://www.youtube.com/watch?v=-jCwIsT0X8M

I’ve read up all week on what you’ve pointed out and that indeed is what’s happening. Is there an software oscillator that gives a DFT plot?

Thanks for information btw.

[Music Manic]

Have you recorded the waveform and observe it in a wave editor?

I’ve just tried re-recording the sine wave at 10kHz through the analogue I/O and it was a perfect sine wave.
It is purely a graphical mis-representation.

You learn something new every day.