1 Topic by kasaudio (edited by kasaudio 2013-01-03 13:56:40)

  • Registered: 2007-06-16
  • Posts: 214

Topic: RME test signal(s)

Hello to everyone and happy new year 2013!

While developing audio stuff and comparing different things I decided to compare the outputs from my sine generator and test tones from RME's site.

I focused on the 0_16.wav (the first one). It is a pure sine of 1 kHz, sampled at 44.1 kHz, 16 bit, dithered.

I checked the frequency spectra using my simple StFFT analyzer with FFT size of 8192 and 75% overlapping.

I got the following:

[img align=0_16.wav]http://s4.mojalbum.com/17376454_17428820_20378600/audio/20378600.jpg[/img]

Same spectra but zoomed around the spike:

[img align=0_16.wav (zoomed)]http://s4.mojalbum.com/17376454_17428820_20378599/audio/20378599.jpg[/img]

Don't get confused with the slider on the right side, it was just for test purpose and is disabled...

Then I created my own sine of exact 1000 Hz, sampled at 44100 Hz, dithered and noise shaped to 16 bit.

I got the following:

[img align=1000.00 Hz]http://s4.mojalbum.com/17376454_17428820_20378615/audio/20378615.jpg[/img]

Zommed around the spike:

[img align=1000.00 Hz (zoomed)]http://s4.mojalbum.com/17376454_17428820_20378614/audio/20378614.jpg[/img]

As you can see, there is a huge difference in the spectra of literally the same signal.

Then I opened both files with Wave Lab, manually aligned them out of phase and mixed them together. As a result I got a heavy beat tone, with slowly varying amplitude from -inf to 0 dBFS. I concluded, that the files don't have the same frequency.

Then I created my own sine again, but with frequency of 1001.29395Hz which is a frequency of 187th FFT bin (zero-based counting). 44100Hz sampling and noise shaped dithering to 16 bit gave the following result:

[img align=1001.29395 Hz]http://s4.mojalbum.com/17376454_17428820_20378645/audio/20378645.jpg[/img]

That looks pretty the same as the first picture and zooming around the spike proves that.
Mixing this signal out of phase with 0_16.wav in Wave Lab results in a very low (<-90dBFS) constant sine tone and noise. I guess this is due to different dithering methods, but it is quite fair to say that the frequency of both signals is the same.

My final conclusion is, that the test signal which is claimed to be pure, low distortion  1kHz sine has a slightly higher frequency than claimed.

Any comment from experienced guys?

Best regards, kasaudio

2 Reply by MC

  • MC
  • MC
  • Administrator
  • Offline
  • From: Germany/Thailand
  • Registered: 2007-04-06
  • Posts: 25,599

Re: RME test signal(s)

The frequency has been generated digitally with the lowest possible THD, and that indeed requires a small frequency deviation which depends on the used sample rate (the higher the more accurate/near 1,000 Hz).

On the other hand I am not aware that it is said that this frequency is exactly 1000 Hz with zero deviation. Which is impossible anyway as you will get a different frequency dependent on your playback system's quartz deviation.

Regards
Matthias Carstens
RME

MC's Website

3 Reply by kasaudio (edited by kasaudio 2013-01-03 21:33:15)

  • Registered: 2007-06-16
  • Posts: 214

Re: RME test signal(s)

Tkank you MC for your reply. Thank you for proving the frequency in the test files is not 1.00 kHz. It is never explicitly stated that it is exactly 1000.0 Hz indeed. It is only stated that it is extremely pure sine with 0.001% THD + N.

I also focused on the 0_16_96.wav test file. 1 kHz, SR=96kHz, 16bit, mono, dithered. The result was almost the same as before, but the deviation from 1000.0 Hz is larger - 3.90625Hz. Actual frequency seems to be 996.09375 Hz

Please don't get me wrong, I don't want to blame or complain against anything, this is just interesting for me.

For generating pure sine I used two methods, first with standard Math.Sin function (.net framework 2.0) where measured THD+N were around 0.00023% and 32 bit floating point 8192 FFT inversion with 0.00030% THD+N. The resulting audio samples from both generators were 32 bit floating points.

According to that (and the teory I know) I would also say that deviating 1000.00 Hz frequency for some small amount in aspect to the sample rate would not bring any significant THD+N loss or gain. I guess this is true for the method I use for generating sine signal. If I create 0 dB FS 1000.00 Hz sine in 44100 Hz SR, 8192 FFT frequency analyzer displays smeared peak that dances around nearest bin to 1k in time. But it is still perfect sine with the same THD+N figures as stated above. THD+N figures are true for 32bit float samples, truncating to 16bits adds distortion and noise, but this is part of different story.

Best regards, kasaudio

4 Reply by kasaudio (edited by kasaudio 2013-01-04 12:12:14)

  • Registered: 2007-06-16
  • Posts: 214

Re: RME test signal(s)

Just a short update:

With the file 0_16.wav I get 0.0014 % THD+N or 0.001% THD+N rounded to 3 decimals, so web specification is OK to my measurement.
Using the same measurement approach and creating 16 bit file with my own generator, I get 0.0022% THD+N with dithering and 0.0039% THD+N with dithering and noise shaping. Sine generator on its own (32 bit float) shows 0.00042% THD+N.

As I don't want to reference RME's forum and RME's downloads anywhere else, I decided to post this topic here, although it is a little away from audio interface area.

BR, kasaudio

5 Reply by MC

  • MC
  • MC
  • Administrator
  • Offline
  • From: Germany/Thailand
  • Registered: 2007-04-06
  • Posts: 25,599

Re: RME test signal(s)

Yes, and please also note that these signals were generated not by our own software, but by using the generator within HpW Works:

http://www.hpw-works.com/

We use this software also for measurements, as you most probably know. It is not cheap, but can do measurements that you won't get or find somewhere else. The author (a swiss guy) spends a lot time optimizing it, and also uses RME interfaces ;-)

Regards
Matthias Carstens
RME

MC's Website

  • Registered: 2008-01-01
  • Posts: 3,025

Re: RME test signal(s)

I usually like to use the free version of Realtime Analyzer for signal generation: http://www.ymec.com/manual/era/overview.htm

Of the free offers I compared some time ago it came out producing the "best looking" signals when analyzing them via Digicheck.

  • Registered: 2007-06-16
  • Posts: 214

Re: RME test signal(s)

Thanks for recommendation! Will take a look

BR, kasaudio

8 Reply by kasaudio (edited by kasaudio 2013-01-15 22:40:14)

  • Registered: 2007-06-16
  • Posts: 214

Re: RME test signal(s)

I spent quite a lot of time investigating different software for audio measurements and to my opinion HpW works beats everything in aspect of functionality and accuracy.

It actually helped me a lot by improving the performance of my own software, not only in aspect of precise measurement today's PRO audio equipment distortion.

To make a long story short, with the file 0_16.wav from RME website now I get 0.00132% THD + N, 0.00010% THD and -98.3 dB signal to noise ratio. No weighting. With my own generated test sine I get 0.0022% THD + N. Loop-back test of the analyzer itself now displays an excelent value of less than -240dB THD+N across the entire frequency range. (Before latest optimisation ~ -140dB THD+N...)

I also measured THD+N, THD and SNR of my own Firefaces 800 and I must say, that both units meet advertised figures.

Two years ago I developed a small audio aplication called RATT (RME Asio Test Tool) which was also available to download within this forum.

RATT thread

The new, much finer (numericaly and graphicaly) is planned for this year. It will definitely contain some basic measurement tools.

Best regards, kasaudio

  • From: Germany
  • Registered: 2007-04-10
  • Posts: 9,774

Re: RME test signal(s)

Just a suggestion: The name RATT may lead to confusion with Microsoft's RATTV3, which is also being linked and discussed in this forum (performance measurement)... Maybe you could find another name at least for the new version?


Regards
Daniel Fuchs
RME

Regards
Daniel Fuchs
RME

RME Support's Website

10 Reply by MC

  • MC
  • MC
  • Administrator
  • Offline
  • From: Germany/Thailand
  • Registered: 2007-04-06
  • Posts: 25,599

Re: RME test signal(s)

So far this tool has never been named RATT anywhere. And I agree with Daniel, there are better names possible for sure.

Good to hear you are still working on it. I use the tool from time to time and missed support for our latest devices (especially the ones with lots of channels). Hopefully you can add that as well.

Regards
Matthias Carstens
RME

MC's Website

  • Registered: 2010-07-17
  • Posts: 62

Re: RME test signal(s)

kasaudio wrote:

" RATT (RME Asio Test Tool)"
The new, much finer (numericaly and graphicaly) is planned for this year. It will definitely contain some basic measurement tools.

Do you expect to complete new tool soon?
Thanks,
Linz

SamProX23 Suite-RME UFX+,FF800;X-Touch,21" Acer Touch,ASUS ROG G750JH-DB71-Win8,MSSurfbook-Win10