Member
Joined 2003
Member
Joined 2003
Just for a bit of additional information, I can provide some comparison with SoundEasy and REW included as well. I took a little bookshelf speaker, and measured in ARTA with sine sweep, 4 measurements overlaid, hanning window at 4.3ms.
ARTA Sine Sweep, not very good consistency / repeatability at low frequency:
Same sine sweep measurement in SoundEasy, same hanning window at 4.3ms provides very good consistency / repeatability:
Same result in REW, sine sweep presents no issue with consistency / repeatability, in fact the measurements are perfectly overlaid that you can't see the other three, you'll have to take my word that 4 traces are shown.
Based on this, I don't see the problem as anything inherent to a dual channel sine sweep measurement, or the Motu M4 hardware.
ARTA Sine Sweep, not very good consistency / repeatability at low frequency:
Same sine sweep measurement in SoundEasy, same hanning window at 4.3ms provides very good consistency / repeatability:
Same result in REW, sine sweep presents no issue with consistency / repeatability, in fact the measurements are perfectly overlaid that you can't see the other three, you'll have to take my word that 4 traces are shown.
Based on this, I don't see the problem as anything inherent to a dual channel sine sweep measurement, or the Motu M4 hardware.
Hi Dcibel,
thank you for your measurements.
I have 3 interfaces. Two works OK (RME and Roland Rubix) while third (Komplete Audio 2) behave just like your Motu M2.
I will surely spend some time to find what happens.
Ivo
thank you for your measurements.
I have 3 interfaces. Two works OK (RME and Roland Rubix) while third (Komplete Audio 2) behave just like your Motu M2.
I will surely spend some time to find what happens.
Ivo
Member
Joined 2003
Thanks for taking the time to look at this, Ivo. Myself and others with similar issues appreciate it!
Thanks for the correction DcibeL. My version of SE was v21. Sine Sweep had been removed and I had asked for its reintroduction. I suspect others had as well since it was returned to the program.
Member
Joined 2003
I don't recall which version it was re-introduced, but it's in v25 for sure. I don't really use SE anymore to be honest, the combination of ARTA and VituixCAD covers my needs much better.
Member
Joined 2003
I can comment on the REW results in case that helps. REW does not perform what I would consider to be a dual-channel FFT measurement, the differences are more than semantic. It is carrying out two single-channel log sweep deconvolutions on the input channels and uses the windowed IRs resulting from those measurements to correct the primary channel. That means harmonic distortions have largely been excluded from both results before the correction happens, which may be a factor.
I use scarlett 2i2 sec gen, and it is horrible, like samp.es already shown. But it also depends on the sequentie length. Still testing possible setups.
Yesterday and today i did the same tests, to see how much it varies.
In short the sine sweep has what appears as a DC offset effect, rising with ~6dB/oct towards lower frequencies, also with occasional a serious outlier ?!
Periodic Pink noise is the best option. Periodic white noise is worse. Lower sampling frequencies and/or lower FFT periods does not significantly change this behaviour.
Sweep: (IMP-Setting-10 ; IMP-Setting-10-IMP-View ; IMP-Setting-10-SeveralTimes-FR),
Periodic Pink Noise:
(IMP-Setting-11 ; IMP-Setting-11-IMP-View ; IMP-Setting-11-SeveralTimes-FR) although at very low frequencies (<20Hz) again a rising curve of ~ 6 dB/oct:
At lower sample frequencies and FTT length, in essence the same:
(several measurements litterally the same output!)
Periodic White noise is worse compared to periodic pink noise:
In short the sine sweep has what appears as a DC offset effect, rising with ~6dB/oct towards lower frequencies, also with occasional a serious outlier ?!
Periodic Pink noise is the best option. Periodic white noise is worse. Lower sampling frequencies and/or lower FFT periods does not significantly change this behaviour.
Sweep: (IMP-Setting-10 ; IMP-Setting-10-IMP-View ; IMP-Setting-10-SeveralTimes-FR),
Periodic Pink Noise:
(IMP-Setting-11 ; IMP-Setting-11-IMP-View ; IMP-Setting-11-SeveralTimes-FR) although at very low frequencies (<20Hz) again a rising curve of ~ 6 dB/oct:
At lower sample frequencies and FTT length, in essence the same:
(several measurements litterally the same output!)
Periodic White noise is worse compared to periodic pink noise:
Member
Joined 2003
Thanks for your verification. I agree that with sine sweep there is occasional outlier, perhaps 1 in 10 measurements appears very good, but repeatability for exact same measurement conditions is not good, which is a problem in itself in addition to the rising noise floor at low frequency. In any case, it appears that Ivo has been able to reproduce this with some of his own measurement equipment and is looking into it. In the meantime, periodic noise is the solution.
For your PN measurement, below 20Hz you are likely seeing rolloff of equipment frequency response, so this behaviour is expected. single channel loopback will confirm.
If i look at the result measured in the left channel you are talking about microvolts. And the bad ones show a positive dc offset, again very minimal yet significant enough.
The rising results below 20Hz are of little concern indeed.
The rising results below 20Hz are of little concern indeed.
I'm sorry, I sort of lost track of this request... I can't tell from the recent activity if you still need me to do this. ?
Member
Joined 2003
February, 2, 2023; ARTA software version 1.9.6 is published.
Note: To download ArtaSetup196.exe you will have to choose left download menu, then twice choose See more and Keep options.
Microsoft just need few days to approve download.
Ivo
- Measurement with swept-sine signal in dual channel mode is improved by better control of sub-audio noise.
- Several small bugs are corrected
- This version no longer works in Windows XP
Note: To download ArtaSetup196.exe you will have to choose left download menu, then twice choose See more and Keep options.
Microsoft just need few days to approve download.
Ivo
Member
Joined 2003
Thank you Ivo!
I have checked the new version, and while there is still some small error in repeatability of sine sweep measurements, the overall noise floor is incredibly improved. In the image below, same noise floor measurement as previous, with no input and -10dBFS reference channel. Solid green is periodic noise, the overlays are sine sweeps, 1 of 4 measurements is a bit out of place but overall the result is a huge improvement!
I have checked the new version, and while there is still some small error in repeatability of sine sweep measurements, the overall noise floor is incredibly improved. In the image below, same noise floor measurement as previous, with no input and -10dBFS reference channel. Solid green is periodic noise, the overlays are sine sweeps, 1 of 4 measurements is a bit out of place but overall the result is a huge improvement!
Focusrite Scarlett 2i2 2nd gen. One flat, one high-passed and eight muted. Very good.
Tested ARTA 1.9.6:
My conclusion:
Significant improvement over Arta 1.9.4 : a level to noise of 90-100dB, less strong outliers if any. The typical "wing" is below 20 Hz of even below 10Hz and less pronounced.
Case PPink-256k:
Result:
Note: Home button (sampling zero is 300):
If cusror at 270 instead of 300, it makes some difference, still prefer the 300 samples home:
Case Sweep 192kHz:
Quite good as well, but occasionally a not so strong outlier, also the Home (300) vs cursor at 270 makes a not so significat difference.
Case Sweep-96kHz:
Basically same picture with 96kHz-256k:
Note that the position of cursor has an effect, both in level(ampliture) and in pattern of the open input noise(in overlay the bottom two entries (red, level, noise) are with cursor at 300 samples.
My conclusion:
Significant improvement over Arta 1.9.4 : a level to noise of 90-100dB, less strong outliers if any. The typical "wing" is below 20 Hz of even below 10Hz and less pronounced.
Case PPink-256k:
Result:
Note: Home button (sampling zero is 300):
If cusror at 270 instead of 300, it makes some difference, still prefer the 300 samples home:
Case Sweep 192kHz:
Quite good as well, but occasionally a not so strong outlier, also the Home (300) vs cursor at 270 makes a not so significat difference.
Case Sweep-96kHz:
Basically same picture with 96kHz-256k:
Note that the position of cursor has an effect, both in level(ampliture) and in pattern of the open input noise(in overlay the bottom two entries (red, level, noise) are with cursor at 300 samples.
