SINAD Measurements

[Rja4000]

The used tones are not harmonics from each other so have no phase relation.

Aren't they per 1/3 octave?
If they are, they are pretty much harmonics...

I guess @amirm is doing something similar than for the 7 tones tests, as he explains here:
https://www.audiosciencereview.com/...-of-hifiberry-dac-pro.8555/page-2#post-217754

 

[solderdude]

https://www.audiosciencereview.com/...-of-schiit-audio-dacs.5487/page-3#post-121956

 

[Rja4000]

https://www.audiosciencereview.com/...-of-schiit-audio-dacs.5487/page-3#post-121956

Thanks
Any idea where the phase definition comes from?
There should be a formula of some sort ?

I know the formula to get the frequencies, function of sample rate and FFT window size, but I don't know for the phase.

 

[NTK]

Thanks
Any idea where the phase definition comes from?
There should be a formula of some sort ?

One of the papers: https://web.stanford.edu/~boyd/papers/pdf/multitone_low_crest.pdf

 

[Rja4000]

One of the papers: https://web.stanford.edu/~boyd/papers/pdf/multitone_low_crest.pdf

Thanks
Will need to take a breath, and dig into that...

 

[Blumlein 88]

Meant to ask before if there is a copy of Amir's multitone file that can be downloaded. You could create your own, but copying is so much easier.

 

[DonH56]

I'd be interested in the algorithm to generate the tones.

 

[solderdude]

Would be fun to test headphones with a multitone as well.
A WAV file would be great.

 

[Rja4000]

Meant to ask before if there is a copy of Amir's multitone file that can be downloaded. You could create your own, but copying is so much easier.

True
But you'd need a different one for each sampling frequency/FFT window size
Although I guess he's always using the same (which is not the one published here)

 

[amirm]

Meant to ask before if there is a copy of Amir's multitone file that can be downloaded.

Get it while it is hot: https://www.dropbox.com/s/j0b8twx8pu7ga1o/APx555 Multitone 32 192 khz 24 bit.wav?dl=0

It may not be there forever if someone is reading this months from now.

Above is at 192 kHz sampling by the way.

 

[Rja4000]

Get it while it is hot: https://www.dropbox.com/s/j0b8twx8pu7ga1o/APx555 Multitone 32 192 khz 24 bit.wav?dl=0

It may not be there forever if someone is reading this months from now.

Above is at 192 kHz sampling by the way.

Would you have the algorithm to compute the phases?

 

[amirm]

Would you have the algorithm to compute the phases?

No. These come from Audio Precision.

 

[Rja4000]

I guess I can find more here:
https://www.ap.com/download/technote-18-multitone-waveform-creation-utility-for-use-with-fasttest/

 

[Rja4000]

I'd be interested in the algorithm to generate the tones.

You mean to select the frequencies?

 

[DonH56]

You mean to select the frequencies?

Mainly for the phase. I use the IEEE Standard for frequencies, but do not recall it specifying phase. Having the signals "line up" to maximize the amplitude (or crest factor) was routine when I was designing and testing (RF) data converters; I had not considered if a particular phase relationship could be established to reduce the peaks for a given set of tones and record length. Rather the opposite; we intentionally designed test signals to hit all codes possible for a given frequency and record length.

 

[scott wurcer]

Would you have the algorithm to compute the phases?

Here is one source http://class.ece.iastate.edu/djchen/ee509/2018/CHAPTER 8 Yuming MultitoneMinPAPR.docx

and another https://www.researchgate.net/public...synthesize_multisines_with_arbitrary_spectrum

The .wav files are very easy to generate in Octave, Python, or other signal processing environment.

 

[Rja4000]

The frequency is easy.
Using Excel formula, if A1 is Sampling rate, A2 is FFT Window size and A3 is your target frequency, the formula is
=ROUND(A3/A1*A2;0)*A1/A2
As an example:
if Sampling rate = 192000, FFT Window size= 65536 and target frequency is 1000Hz,
the selected frequency is ROUND(1000/192000*65536;0)*192000/65536 = 999.023438 Hz

The Phase, I don't know (yet), and I'm interested too...

 

[KSTR]

I'd be interested in the algorithm to generate the tones.

Find attached a quick&dirty very-much-work-in-progress C code that I use for generating a specific multitone sequence. It's literally a set of cosine oscillators.
The bin population was "stolen" (reverse-engineered from one of REW's "NID" multitone sequences with all the IMD components being (multiples of) 7 bins apart (so that they can be discerned from the noise).

Generates a 64bit floating-point ultra-high precision output array (pure to way below -200dB), which can be converted to whatever WAV format with the SoX utility (if Sox says 'clipping' then reduce the scale factor).
Compiles with MinGW, a mimimalist compiler for Windows (no GUI). Edit the file to your needs (like setting specific phases and amplitudes, atm the phase is random -- specific phase tailoring to get defined crest factors could be inserted here -- , FFT-size to be used is 16k, only a single block is created, scale factor is unity with all phases set to 0, etc), I didn't bother to write an options interface or setup file read-in.
The combination of sample-rate applied to the output and the block size determines the component frequencies and the frequency headroom for the upper products.
Don't populate bins at and above FFTsize/2 ;-)
gcc -march=native -Os -s -Wall GenMTone_7.c -o GenMTone_7.exe

 

[scott wurcer]

The frequency is easy.
Using Excel formula, if A1 is Sampling rate, A2 is FFT Window size and A3 is your target frequency, the formula is
=ROUND(A3/A1*A2;0)*A1/A2
As an example:
if Sampling rate = 192000, FFT Window size= 65536 and target frequency is 1000Hz,
the selected frequency is ROUND(1000/192000*65536;0)*192000/65536 = 999.023438 Hz

The Phase, I don't know (yet), and I'm interested too...

If you simply want 30 tones spaced at 1/3 octaves (with 1000Hz included as closely as possible) this works.

sample_rate = 192000.
fft_len = 65536.

base = sample_rate/fft_len

for n in range(-16,14):
freq = round((1000*pow(2.0,n/3.0)/base))*base

Which gives...

23.4375
32.2265625
38.0859375
49.8046875
61.5234375
79.1015625
99.609375
125.9765625
158.203125
199.21875
249.0234375
316.40625
395.5078125
500.9765625
629.8828125
793.9453125
999.0234375
1259.765625
1587.890625
2000.9765625
2519.53125
3175.78125
3999.0234375
5039.0625
6348.6328125
8000.9765625
10078.125
12700.1953125
15999.0234375
20159.1796875

 

[Blumlein 88]

Get it while it is hot: https://www.dropbox.com/s/j0b8twx8pu7ga1o/APx555 Multitone 32 192 khz 24 bit.wav?dl=0

It may not be there forever if someone is reading this months from now.

Above is at 192 kHz sampling by the way.

Thank you for this. It will come in handy. Should be fun trying this with Pkane's Deltawave.