QUASI PURE SINE
Added Lissajous waveform file (zip/wav) to the download page.
Added Lissajous waveform file (zip/wav) to the download page.
An externally hosted image should be here but it was not working when we last tested it.
The 6k is quite similar to the 1k, though the 3rd harmonic distortion isn't as bad. The 1k file has (relatively) pronounced 3rd, 5th and 7th harmonics (about -133dB, -141dB and -136dB respectively ref the fundamental). They would be cleaner if you added 24 bit dither of course, since you would then remove the correlated quantisation noise, but even undithered they should be about 8 dB lower than they are. You can save similar (though slightly cleaner) files from REW's signal generator.
Anything linked with the Gnu -lm library should give the same answer for the same format. As an experiment we tried Matlab, Python, and Mathematica and all were exactly the same. Lots of programs especially ones written a while ago will use 32 bit rather than 64 bit IEEE floats which could explain many differences.
As Damien said 64bit floats are almost -300dB noise floor and dithered to 24 bits have no harmonics at any frequency.
Last edited:
Attachments
Thank you pointing out about this! This graph's are as 3-D just using same stimulus frequency and using in addition an 3-D Y-offset too!
Here is the graph's without the applied Y-offset! 😀
Hp
Attachments
My Multitone 2k to 90k (Source is 1.228 Vrms)
no dithered (up) and dithered (down)
The floor in no dithered multitone is about
170
the floor in dithered multitone is about
165
the 'dithered' multitone seems a bit cleaner in the contents.
OT @ HpW
HPW-Works have a economic version like SpectraPLUS-RT ?
no dithered (up) and dithered (down)
The floor in no dithered multitone is about
170
the floor in dithered multitone is about
165
the 'dithered' multitone seems a bit cleaner in the contents.
An externally hosted image should be here but it was not working when we last tested it.
OT @ HpW
HPW-Works have a economic version like SpectraPLUS-RT ?
Last edited:
While avoiding the multitone and
considering to stay as close to your scale with a single 48k signal source,
my 48k/4dBu signal source in realtime mode and dithered:
(to go lower than that, in realtime mode, I need to a cryogenic room 😀😀😱)
considering to stay as close to your scale with a single 48k signal source,
my 48k/4dBu signal source in realtime mode and dithered:
(to go lower than that, in realtime mode, I need to a cryogenic room 😀😀😱)
An externally hosted image should be here but it was not working when we last tested it.
🙂
Last edited:
Hi, there is now joke at all... the (digital) signals provided are for a 24 bit signal of a low quality and requires a re-engineering.
Hp
Just all of them, as mentioned: "the (digital) signals provided are for a 24 bit signal of a low quality and requires a re-engineering."
Cheers
Hp
how does this measure?
1 second of 24/44100 1002.273Hz
small (1/100sec, iirc) leadin and leadout
1 second of 24/44100 1002.273Hz
small (1/100sec, iirc) leadin and leadout
For best results one should use "prime frequencies" of the FFT size, and no FFT window, of course. This is also intrinsically self-dithering.
No sample value ever occurs twice in the whole sequence (FFT size) but the sequence itself must repeat perfectly (for the non-windowing to work).
This is whole idea and recipe for the construction.
No sample value ever occurs twice in the whole sequence (FFT size) but the sequence itself must repeat perfectly (for the non-windowing to work).
This is whole idea and recipe for the construction.
- Status
- Not open for further replies.