ARTA

Hi Ivo,

Hopefully you can help me with a problem I'm having with two channel measurement mode in ARTA. There seems to be some sort of bug causing inconsistent results.

Scenario: I am measuring a ribbon tweeter which cannot safely be measured without at least a 12dB/oct high pass filter. My idea is to use Dual Channel mode to cancel out the response effect of the 4Khz high pass filter (at least for a couple of octaves) so I can get a valid amplitude and phase response down to 1Khz to use in XSim modelling.

However this does not seem to work reliably due to incorrect impulse measurement and I cannot find out why. Here is a single channel measurement of the microphone on the left channel showing the tweeter acoustic response with high pass filter:

attachment.php


Here is a single channel measurement of the electrical high pass filter on the right input channel:

attachment.php


Here is a dual channel measurement taken with the Microphone in left and the direct electrical response of the high pass filter in the right channel:

attachment.php


This is exactly what I would hope to see, with valid data down to about 1Khz, and I would like to export this as FRD to use in XSim.

However if I try to repeat the exact same measurement without changing any settings I now get this:

attachment.php


As you can see, the measurement is useless. When I compare the PIR of the good and bad measurement the difference seems to be DC bias in the impulse response. Here is the bad impulse belonging to the bad frequency response above:

attachment.php


As you can see it has a negative DC offset. (But the offset changes at random with each measurement)

Here is the good impulse - the only difference seems to be the DC offset:

attachment.php


Is there anything that can be done to work around this ? At the moment I have my measurement environment set up but I am stuck as I cannot get a reliable reproducible response for the tweeters.

Another problem I am having is if I try to use swept sine in dual channel mode it crashes the whole program, but I'll leave that to another post.
 

Attachments

  • tweeter with highpass single channel.png
    tweeter with highpass single channel.png
    59.9 KB · Views: 821
  • high pass filter.png
    high pass filter.png
    60.4 KB · Views: 811
  • tweeter 2 channel.png
    tweeter 2 channel.png
    63.5 KB · Views: 797
  • 2 channel tweeter bad.png
    2 channel tweeter bad.png
    71.5 KB · Views: 815
  • bad impulse.png
    bad impulse.png
    30.2 KB · Views: 809
  • good impulse.png
    good impulse.png
    30.2 KB · Views: 801
Off Axis Measurements - Ivo, help please

Hi Ivo,
I was looking at the automated off axis measurement piece of ARTA and it seems to be somewhat less user friendly than it could be. I mentioned this to Kimmo, author of Vituix software and also the ARTA recorder, as his program (ARTA recorder) seems more intuitive and more directly applicable to a manual turntable for measurements.

My respectful request is that you might perhaps make this portion of the program more user friendly for those who use a manual and not an automatic off axis turntable for measurements.

Kimmo also made some suggestions:
Details which need fixing in ARTA 1.9.0 in my opinion:
1) Settings in both windows should be saved for the next recording, exporting and sessions (to registry). Exceptions are settings in Export spatial.. which are picked from opened measurement: Start position, Gate length and Delay for phase correction.
2) Font size of Next angle: should be at least five times bigger that user could see the value at least few meters (Arta Recorder has 48pt).
3) "_deg" text should not be required in the filename because these features are serving also other applications and needs which do not require (or accept) "_deg" text. Value could be just a series of numbers 1,2,3,... separated with space from root filename. User can add "_deg" in the end of root filename if he's using for example Directivity pattern function in ARTA (while "_deg" requirement is not removed also from there).

Thanks so much for your consideration,
Jay
 
Hopefully you can help me with a problem I'm having with two channel measurement mode in ARTA. There seems to be some sort of bug causing inconsistent results.

Scenario: I am measuring a ribbon tweeter which cannot safely be measured without at least a 12dB/oct high pass filter. My idea is to use Dual Channel mode to cancel out the response effect of the 4Khz high pass filter (at least for a couple of octaves) so I can get a valid amplitude and phase response down to 1Khz to use in XSim modelling.
That does not work, because at low frequencies there will be no signal in both microphone and loopback input, except for noise. So you have almost zero signal to noise ratio at low frequencies and ARTA still tries to calculate the impulse response from that information.
 
Hi Simon,

"As you can see, the measurement is useless. When I compare the PIR of the good and bad measurement the difference seems to be DC bias in the impulse response. Here is the bad impulse belonging to the bad frequency response above:"

I'm not sure if my observed problem in my post #395 is the same as yours, but Ivo's response may provide an answer for you.

Regards

Peter
 
That does not work, because at low frequencies there will be no signal in both microphone and loopback input, except for noise. So you have almost zero signal to noise ratio at low frequencies and ARTA still tries to calculate the impulse response from that information.
All speakers have near zero output at very low frequencies, so if this was the case then the dual channel measurement mode would be useless for any acoustic measurements...

I tried changing the high pass filter to a shelving function which shelved down at 12dB/oct and leveled out at about -40dB to increase the low frequencies available for analysis in the loopback signal but I still see the same problem.

If I repeat the measurement process many times, about 1 in every 20 tries I get a perfect result, and the others I get various random DC offsets in the impulse response which cause various response anomalies.

It seems that in dual channel mode if the loopback signal deviates from a flat response by more than about 20-30dB at low frequencies a stable impulse response without a DC offset cannot be generated, even though the low frequency signal is still about 60dB above the noise floor at that point.

I still think this is a bug, as there is still sufficient signal available at low frequencies. If the loop-back signal is near a flat response there is no issue.
 
Last edited:
Hi Simon,

"As you can see, the measurement is useless. When I compare the PIR of the good and bad measurement the difference seems to be DC bias in the impulse response. Here is the bad impulse belonging to the bad frequency response above:"

I'm not sure if my observed problem in my post #395 is the same as yours, but Ivo's response may provide an answer for you.
Hi Peter,

I had a quick read of your post but I don't think I'm experiencing the same issue for a couple of reasons:

1) I only see a problem with dual channel mode, single channel mode is fine.
2) I'm using Periodic noise as the stimulus not swept sine - in fact trying to use swept sine stimulus in combination with dual channel mode is causing ARTA to crash on me every time - so I can't even test it...
 
Spotted a minor bug in LIMP - when using 1/48th octave and exporting as ZMA, duplicate frequency values are produced at low frequencies with differing impedance and phase, probably as a result of rounding to too few decimal places and then not checking that a duplicate was produced by the rounding process.

This can lead to the ZMA failing to import into some other software such as XSim, which was discussed over here:

http://www.diyaudio.com/forums/software-tools/259865-xsim-free-crossover-designer-8.html#post5233627

It can be worked around by manually deleting one of the "duplicate" lines before import but it would be nice if it can be fixed by increasing the number of decimal places or detecting and discarding lines with duplicate frequencies during export.

Example exported ZMA files attached.

BTW I have quite a number of minor functional bugs that I have noted down and documented for ARTA over a few years but not reported yet, (mainly as I stopped working with speakers and ARTA for a few years, but now I'm back) is this forum thread a good place to report them or should I use email directly to Ivo ? I first have to reproduce each bug to make sure it still applies to the current version of ARTA and has not already been fixed as some date back several years, but I am happy to do that.
 

Attachments

  • ZMA.zip
    20.7 KB · Views: 43
Last edited:
Rich08

I know that it is nice to see the effect of IR gating to FR in the same window.
I'll put your request on ToDo list

Ivo
Hi,Ivo
Why not put a QC measurement package in your software which includes FR,polarity,Rub&Buzz,THD,IMP fast acquisition.It's much more convenient and versatile for production line speaker testing for quality control.
By the way,to make things easy you could combine FR and IMP measurement together and lower the requirement of input & output channel number on soundcard,that is to say only 2 channel acquisition is adequate for most work.
Here is the wiring diagram here,maybe my suggestion above will give you some hints for improvement of your software.Thanks for your advice and consideration.
 

Attachments

  • hardware connection.png
    hardware connection.png
    87.9 KB · Views: 214
THD with ARTA or STEPS

When you measure THD for 1kHz sine signal, both STEPS and ARTA will give the same results but it is important that ARTA uses high order windows and same sampling frequency as STEPS.
For measurement on lower frequencies STEPS automatically adjust acquisition length, while in ARTA the user has to decide which FFT length will be used.
Ivo
 
I have the possibility to lay my hands on a UMIK-1 from MiniDSP:
Acoustic Measurement Tools : UMIK-1

Since a unique calibration file is available, it looks pretty good to me.

However, it is a USB microphone, that means that essentially two soundcards are operating: One to generate the signal, in my case an ESI Juli@ PCI card, and the other (the USB Microphone "Soundcard") to record it.

While I figured out that it is not an issue in telling ARTA to use different devices for input and output, what about other shortcomings, like different delays, maybe different sample rates (UMIK-1 is fixed at 24bit / 48 kHz) or other things I have no clue about?

Does anyone have already tried such a configuration?

Thanks!
 
Does anyone have already tried such a configuration?

Yes. I've tested latency variations with Umik-1 and external sound card (but not with ARTA). Conclusion was that severe timing/phase differences exist between different measurements. Timing may stay quite stable few measurements but then delay might jump ~0.5...>1 ms back and forth. Rebooting or starting another program between measurements was not needed to generate delay jumps, and latency adjustment of sound card did not help.
So, it is not valid gear for advanced speaker measurement needed to simulate multi-way to off-axis, power, DI responses etc. because that requires very stable timing i.e. semi-dual (or dual) channel connection and measurement mode selected in ARTA (or REW).
 
Thank you for the quick and well-founded answer....allowing me to save some money :)
Well, so I have to search for something different as measurement microphone...

Btw, I have a MCE-2000 capsule in my junk box (I believe it is equivalent to Panasonic WM-60) and a suitable preamp/phantom power circuit, but have no idea how to calibrate...
 
Thanks Kimmo.

Focusrite Scarlett sure is a well engineered device.

I'll have to think about whether to use my phantom mic amp in combination with Juli@ (a good soundcard btw) or to go the "easy" way and buy an "all inclusive" soundcard. ARTA works well with it, however I made no acoustic, only electronics measurements up to now. The next project will be the refurbishing of a two way speaker, and with your (VituixCAD) and ARTA's help, I'm confident it will end up well.
 
Yes. I've tested latency variations with Umik-1 and external sound card (but not with ARTA). Conclusion was that severe timing/phase differences exist between different measurements. Timing may stay quite stable few measurements but then delay might jump ~0.5...>1 ms back and forth. Rebooting or starting another program between measurements was not needed to generate delay jumps, and latency adjustment of sound card did not help.
So, it is not valid gear for advanced speaker measurement needed to simulate multi-way to off-axis, power, DI responses etc. because that requires very stable timing i.e. semi-dual (or dual) channel connection and measurement mode selected in ARTA (or REW).
In another thread I argued that dual channel measurement was required with any sound card to be absolutely certain that there are no timing/phase errors between multiple measurements, such as when measuring multiple drivers to import into a sim such as Vituixcad, but I was ridiculed for it.. :rolleyes:

What you describe is of course worst when the microphone and sound output are completely different sound devices - in that case there is no true time synchronization between them, and not only can there be a varying sample offset from one measurement to another due to software/driver issues, there will also be sample rate drift between them as they won't have a single locked sample rate clock that is usually the case of a single sound card in full duplex mode.

But these kind of unpredictable timing jumps between output and input can also happen due to software/driver glitches and issues even on a single full duplex sound card - I've had this happen to me before. So while you might "get away with" single channel measurement on a good quality full duplex sound device, you cannot rely on it without a lot of cross checking of measurement consistency.

For this reason I now always do dual channel measurements wherever possible, and particularly if I'm interested in capturing the true phase. Then I can know with confidence that my phase measurements are correct and consistent without constantly performing cross checks between measurements.

Full dual channel mode also helps remove some other sources of error such as imperfections in the frequency response of the test amplifier and speaker cable resistance, if the 2nd channel is taken from the driver terminals or crossover input as applicable for the measurement.

You don't need to pay a lot to get a decent full duplex USB sound card these days either. I only have a Behringer UMC204HD which from memory cost me about £65, and it does the job very well.
 
Last edited:
In another thread I argued that dual channel measurement was required with any sound card to be absolutely certain that there are no timing/phase errors between multiple measurements, such as when measuring multiple drivers to import into a sim such as Vituixcad, but I was ridiculed for it.. :rolleyes:

One of the main issues in practice is that common measurement softwares such as ARTA and REW normalize timing by moving IR peak to 0 ms point if semi-dual or dual channel mode is not selected. Software cannot do much more because input data for cross-correlation function is not available in that location (2nd analog input) they expect.

Normally I don't recommend full dual connection due to possible grounding problems and severe damages if power amplifier is not single-ended: common negative potential/rail with line input and speaker output. User just need to know features of power amplifier typically used while measurement. Compensating of everything; speaker cables etc. might be overshoot because equipment IRL have some linear distortion after speaker is designed.

Single channel gear such as simple USB mics are very common today. They have spread from room acoustics measurements to speaker design which is unfortunate because more and more new users end up to bad, inaccurate/random and very slow design methods. This is not my fight, so I simply recommend to ignore messages and authors supporting or advertising such gear from now on.
 
Normally I don't recommend full dual connection due to possible grounding problems and severe damages if power amplifier is not single-ended: common negative potential/rail with line input and speaker output.
Even when using a single ended amplifier direct connection of earths still gives issues with ground loops where current in the negative cable to the speaker/crossover will influence the measurement. This can cause weird artefacts like a tweeter high pass filter rolling off at low frequencies only so far and then appearing to increase again. :D

Also the crossover output in systems like mine with a lattice all pass filter in the crossover cannot be measured this way.

For these reasons I use a small high quality audio step down transformer to sample speaker level signals to connect to the sound card input in line level mode.

The step down ratio is such that the peak voltage from a 100w amplifier (my amplifier is only 50w anyway) is less than a normal maximum line level signal so no further attenuator or protection diodes are needed.

The transformer is ruler flat from about 10Hz to 40Khz and provides full galvanic isolation, as well as being able to drive the balanced input in the USB sound card correctly. (Which has slightly higher distortion if driven by an unbalanced signal due to the servo input) Distortion is also vanishingly low compared to speaker distortion.

This gives me full freedom to measure anywhere on a crossover including on the output of a lattice all pass filter and completely eliminates any ground loops or other spurious effects.

I use it both for the 2nd channel input when performing acoustic measurements, as well as being the primary input when measuring the electrical output of a crossover. (I actually have two identical transformers so can use dual channel mode here too if I want to)

I definitely recommend this over direct connection using resistors and protection diodes, provided a transformer of suitable quality and step down ratio is chosen.