ARTA

Hi,
there is no difference in how ARTA record noise and swept sine.
The problem can be with a driver or with laptop.

I've not tested Behringer UMC204HD. Does it have control panel to adjust sampling rate?
It has a control panel which will display the currently active sample rate but not set it. In windows it is set to default to 192Khz 24 bit, however it is not the default windows sound device so ARTA is the only device to access it.

It is configured to allow exclusive access to applications.

I found it only crashes with a 262k sequence length, smaller sequence lengths are OK.

After your suggestions I think I have found the issue. The sound card control panel allows adjustment of "USB streaming mode", which seems to be a latency adjustment, as well as "ASIO buffer size". Buffer size was set to Auto. See screenshots.

With USB Streaming mode set to the default "Safe", ASIO Buffer size set to the default "Auto", crashes were occurring.

With ASIO buffer size manually set to 4096, 8192 or 16384 samples it appears to work fine on the swept sine sweep without ARTA crashing.

If I try to choose a value below 4096, it reports "The selected ASIO buffer size is too small for the current USB streaming mode and the current sampling rate". Strangely no setting for buffer size except "Auto" causes ARTA to crash.

Do you have any recommendations for streaming latency and ASIO buffer size in general or does a higher latency and/or higher buffer size mode not cause any issues with ARTA ?
 

Attachments

  • streaming mode.png
    streaming mode.png
    16.3 KB · Views: 604
  • asio buffer.png
    asio buffer.png
    25.6 KB · Views: 523
Simon,
I usually recommend buffer size from 256 to 2048 samples (the larger sampling rate dictates larger buffers), but it seems that your card has larger latency than is usual in ASIO practice.
Hi Ivo,

Thanks. The interface does have 7 different latency settings from "Low Latency" to "Extra Safe" as depicted in the screenshots, the default setting of "Safe" which I am currently using is the second from highest latency setting.

Unfortunately Behringer don't state the latency in ms for each setting instead using an ambiguous descriptive word. From what I've read online the lowest latency setting is approx 1ms while the highest is about 30ms, so it is probably about 25ms at the moment.

I can easily reduce the buffer size to 2048 samples and then reduce the latency setting as well until it no longer complains that the buffer is too small, but I'm not sure a lower latency provides any benefit unless I am using the RTA mode ? I generally only use impulse measurement mode.

Lower latency modes increase the risk of dropouts if the laptop/OS is slow ? Would ARTA provide any indication of an error if a dropout occurred or would there simply be a random glitch in the impulse recording if this happened ?
 
It's good you found the culprit -- but it's strange that the problem depends on the signal type. Sweep crashes, noise does not crash. How would the soundcard / driver worry about signal shapes in such a way?
I can only assume it's something to do with differences in the recording length as the periodic noise and swept sine methods do not playback/record for the same length of time. But it does still seem strange. I don't know what the Auto setting is doing because it does not crash with any of the manual buffer size settings... a buffer that is too small does corrupt the audio dramatically but still does not crash.
 
Hi Simon,
you guess right, swept sine acquisition system uses double length than noise acquisition. That problem has occures only on 256 k signal length, (acquisition length 512 k) and sampling rate 192kHz.
In ASIO "auto mode" ARTA sets ASIO buffers to 1k length. Obviously it is not enough for this card as you've got report "The selected ASIO buffer size is too small for the current USB streaming mode and the current sampling rate" if it was lower than 4k. So you must use manual mode ? Nothing strange here!
Who crashes: ARTA, ASIO, or windows process thread system I don't know.
The problem arose with newer sound cards that use 24-bit transfer even at 192kHz.
I will certainly in new ARTA version add new checking of required buffer size.
Thanks for report.
Ivo
 
newbie in need of advice

I'm a newbie when it comes to measuring. I have a tascam u122 and am currently waiting on a ecm8000 to be delivered. could any of you point me in the direction or provide a block diagram of how i connect my sound card in conjunction with speakers, any set up values I might need.

Thanks in advance,
Nigel
 
Which calibration file do you use guy for capsule Panasonic MW61A?
Do you recommned any other microfone which is better that "home made" pipe with MW61 but relatively not very expensive?

It depends on what you mean by "not very expensive". A $5 capsule may be acceptable for some measurements, but you will have to send it to a calibration lab to determine a calibration curve for it. Even with that you'll not know if/when the thing starts changing (they do!). The next level would be a Behringer ECM8000 or similar, which is nicely built, but still comes without calibration, and they also change with time (mine did!). If you want a reliable microphone with calibration data, it might be better to just bite the bullet and get one. I like the ISEMCON EMX-7150. It's good, but not as expensive as the B&K or Earthworks offerings.
 
Hi Ivo,

There was some discussion in another thread about using mic calibration profiles in ARTA which have only amplitude data:

https://www.diyaudio.com/forums/software-tools/307910-vituixcad-14.html#post5583046

I had not realized in years of using ARTA with such a mic profile, that no correction to the phase response exported as frd is made if the mic profile doesn't include phase data. (Which many do not)

This means the exported frd is no longer minimum phase even if the original measurement was.

I see that in the new PIR import feature in Vituixcad that Kimmosto has added an option to calculate minimum phase for the microphone profile if phase data is not included in the profile - is it possible that this could be added in ARTA as well ?
 
FR compensation

Hi Simon,

The question is when we should use microphone calibration data in measurement results and when we should use it in crossover response simulation?

In the first version of ARTA I did not implement response compensation with mic. calibration data. After lot of requests I implemented it later. In ARTA only magnitude data are used, as manufacturer calibration data usually does not contain phase information. Why they ignore phase information? Possible answer is that there no exists method for correct mic. phase response measurement. The calibration is usually made by comparison with some reference microphone, but phase obtain that way is not true phase, as reference microphone does not have ideal response.
I must say that true mic. response phase cannot be calculated by minimum phase calculation as we do not have true mic. response neither before or after calibration.

Magnitude calibration is welcomed when we measure total loudspeaker response as indicator of possible. tonal balance.

When we are designing crossover:

1) it is desirable to have microphone that has flat response (within 1dB) at least one octave above and below crossover frequency

2) measurement of all driver’s responses should be made with same mic., on the same positions (some correction of distance is needed if we set mic. close to the loudspeaker).

If both of this condition are met than we do not need to use compensated response in crossover design, on the contrary, it is better not to apply response (magnitude + phase) compensation.

The compensation can be used on summed total response – to give us insight in achievable tonal balance.

All this said, I recommend:

1) Find descent microphone that has flat response 100Hz-10kHz.

2) When designing crossover do not use the frequency response compensation in measurement results, rather use FR compensation in crossover total summed response.

3) If you measure total multi-way response use FR compensation.

Ivo
 
More on FR compensation in crossover design

In the last post I insisted that we need to measure driver's responses with same microphone and on same distance from the box. Importance of this I will explain with following reasoning.

Let analyze two-way system:

W1 = fr. response of first driver without fr. compensation
W2 = fr. response of second driver without fr. compensation
Wf1 = fr. response of first driver crossover filter
Wf2 = fr. response of second driver crossover filter
Wmc = frequency response of mic. response compensation

The total response, assuming both responses are measured with same mic. on same position is:

Wtotal = W1 Wf1 Wmc + W2 Wf2 Wmc = Wmc (W1 Wf1 + W2 Wf2)

Here it is obvious that it is the same if we apply compensation on each driver’s response or on total response calculated without compensation.

The second important conclusion is that phase of compensation does not change the magnitude of total response.

One thing more. I must correct myself from the last post. In CAD systems that do crossover response optimization it is better to apply compensation on each driver response as it is easier to define crossover target function as maximally flat in pass-band.

If we use driver's responses measured with different microphones, then preceding reasoning does not hold and system design becomes complicated.
If we use minimum phase calculation it is not quite correct, as assumption of minimum phase systems does not hold in whole driver's frequency range.

Ivo
 
I must say that true mic. response phase cannot be calculated by minimum phase calculation as we do not have true mic. response neither before or after calibration.

That's an interesting statement, but I don't understand the logic behind it. If the (absolute) microphone response is known, why would we not be able to calculate phase by assuming that the microphone is a minimum phase system?
 
Minimum phase problem

There are two problems when we consider minimum phase:

First….

Minimum phase is characteristic of linear systems that can be calculated from the magnitude of frequency response but only for systems in which there are no wave reflection (such systems can be described by finite number of poles and zeros in plane of complex frequency). This mathematical definition practically means that responses of loudspeakers and microphones does not follow minimum phase in frequency range where dimension of object are larger than quarter of wavelength.

Second ….
I have published paper before 30 years where I analyzed various method for minimum phase estimation. The numerical method that can be used for correct minimum phase calculation from magnitude response requires knowledge of response in frequency range from zero to infinity frequency. As we measure in restricted range we must assume response below and above that range, and that assumption is always a problem.

I know that this explanation is hard to follow but here I will not go with mathematical proofs.

Ivo
 
Ok, a typical microphone capsule is about 5 mm in diameter, which corresponds to a frequency of about 70 kHz. And yes, we usually don't know the frequency response at the "extremes" (for example below 50 Hz and above 20 kHz). My Hilbert transform math is a bit rusty, but... how bad is it in practice? Do the poorly constrained response data (<50 Hz, >20 kHz in my example) really "bleed" much into the interesting range (50 Hz to 20 kHz) during minimum phase calculation?

Could you post a copy of your paper?
 
The paper was copyrighted, but I think now it can be published. My copy is very bad and it will take time to type it again.

I solved the numerical calculation of Hilbert integral by piece-wise integration (as I remember) probably similar to what you have done. I have also analyzed pole-zero system identification technique.

Now, let us be practical. If you use quarter inch mic with minimized body size (to reduce reflections) then you will have mic that does not need FR calibration.

If you have mic with slightly worse characteristic, then phase response will still be unimportant for group delay and resonance detection, compared to loudspeaker phase response (for single mic measurements.)

Loudspeaker exhibit more deviation from minimum phase system then microphone, as there are effect of cabinet diffraction and reflections.

Ivo