Arta and phase measurment

[DJData]

Hello , i am building a 2 way speaker with some friends and we are using Arta to measure what we do with filters and so on.

We are now at the point were we need to make sure the speakers are in absolute fase. Or to get them there how much we have to move the top speaker for or backward to get it there.

The problem is we can not seem to get a good measurement out of Arta,and from what i read this has to do with a feedback loop that is necessary to get an accurate measurement, i wich the feedback is then the reference to see the delay timewise form the generated signal compared to the recorded signal.

We use a computer with Arta and a emu 0404 external dac to wich the mic is connected.

We can't figure out how to get the feedback loop into Arta / how to connected everything so Arta gets the feedback loop.

I hope my question is clear, its a complicated matter .

We hope somebody hear has done this with Arta already and can explain what we are doing wrong / how we can get the correct measurements .

Thank you, data.

 

[DJData]

I think its not allowed here to put up a link to another forum that has most of my question covert , if it is allowed ,let me know please then ill put up the link here.

 

[CharlieLaub]

Sure, you need to use this method to know the relative phase of the drivers:
measure impulse response of driver
gate the impulse to remove reflections
use the single channel frequency response option in ARTA, when converting the impulse data to the frequency response
Save the frequency response as minimum phase (amplitude and min phase responses)

Repeat the same procedure for the other driver(s) in the system

Also repeat the measurement procedure with pairs of drivers (tweeter+mid, mid+woofer, etc) operating at the same time with the same input signal.

Use a program to import e.g. tweeter data, midrange data, tweeter+midrange data. This program should be able to assign a delay, or use spatial coordinates of drivers, to determine the offset (front to back) of the acoustic center of each driver given one driver (usually the tweeter or whichever driver has its acoustic center farthest forward) as the "reference" and others relative to that. Once you know this information for all drivers in the system you will be able to model the phase response perfectly. Then you will know if your crossover system will work properly.

Programs that can do this include my Active Crossover Designer (ACD), and Jeff Bagby's Passive Crossover Designer (PCD). ACD includes a tutorial for an example 2-way system that you can practice on. You can download the ACD files from the Software page on my web site (link in my signature). You also need to understand how to produce a frequency response from a gated impulse response measurement in ARTA. I wrote a tutorial on this that you can find here:
http://audio.claub.net/tutorials/FR measurement using ARTA.pdf

 

[DJData]

Ty for the reply Charlie we will be looking into it .

I think with sure, you ment that i can post the link so ill do it here:

It is form the diy mobile division .

Measure time delay (T/A) with aRTA and RoomEQ - Car Audio | DiyMobileAudio.com | Car Stereo Forum

 

[kimmosto]

Semi-dual channel measurement is the easiest and fastest way, but possible Mix potentiometer or other sound card setting could corrupt signals. Semi-dual is explained in ARTA's manual, but generic help can't cover settings of soundcards.

Here is one instruction containing some details for capturing timing and phase differences from multi-way with ARTA: Measurement Preparations.pdf. Unfortunately there are also some other stories which could confuse.

Single channel measurement is not recommended with common measurement mic and soundcard having two analog input and output channels. Single channel needs minimum phase calculation and other processing with some software which is unnecessary and could cause errors.

 

[DJData]

Ty Kimmosto

 

[CharlieLaub]

Single channel measurement is not recommended with common measurement mic and soundcard having two analog input and output channels. Single channel needs minimum phase calculation and other processing with some software which is unnecessary and could cause errors.

I'm not sure what you mean "could cause errors" and why a "common measurement mic and soundcard having two analog input and output channels" has something to do with it (error in measurement). Can you explain in detail please?

 

[kimmosto]

I'm not sure what you mean "could cause errors" and why a "common measurement mic and soundcard having two analog input and output channels" has something to do with it (error in measurement).

Minimum phase calculation is partly based on assumptions or guesses by user or slope detection logic of the software which makes the calculation. Frequency response below and above measured range is unknown, but needed for calculation. I'm sure you know this very well.
Transducer under test is usually close to minimum phase device, but could be something else. We'll never know for sure if amplitude response is converted to minimum phase by default.

Mic and soundcard didn't have anything to do with the error. It was just my recommendation: user should not use single channel mode for far field measurements of multi-way design if semi-dual measurement is available. Single channel mode causes extra work, especially when designing with full 360 deg polar responses in both planes.

 

[CharlieLaub]

Minimum phase calculation is partly based on assumptions or guesses by user or slope detection logic of the software which makes the calculation. Frequency response below and above measured range is unknown, but needed for calculation. I'm sure you know this very well.
Transducer under test is usually close to minimum phase device, but could be something else. We'll never know for sure if amplitude response is converted to minimum phase by default.

Bingo. You are absolutely correct about that. Mostly this will effect the low and high frequency extremes within a couple of octaves of the valid frequency limits of the measurement. The middle range is still pretty accurate with min phase.

I wrote the code/calculations in the FRD Response Blender that do all the phase extraction. With this approach you estimate low frequency response (e.g. 50-500Hz or so) using a box model, and then assign slopes to the tails of the response at high and low frequency before recalculating the min phase response. All of these "pieces" of the frequency response over the entire DC to infinity range are either spliced or blended together to create a seamless FR (see image below) from which the phase (minphase) is extracted. The result can be very accurate especially at low frequencies.

But in general I agree that the min phase will be somewhat to very inaccurate away from the middle of the frequency response when taken directly from the measurement package (e.g. ARTA).

 

[kimmosto]

But in general I agree that the min phase will be somewhat to very inaccurate away from the middle of the frequency response when taken directly from the measurement package (e.g. ARTA).

One problem with "single channel measurement + minimum phase extraction + acoustic center detection" -method is that measurement programs usually export responses with normal/measured phase - not the minimum. If polar response measurement set contains e.g. 108 measurements (3-way, hor plane only, steps 10 deg), it will take a while to export all one by one as minimum phase. Radically faster option would be exporting all with normal/measured phase and using batch converter. VituixCAD Calculator can convert all responses to minimum phase at once. After that user has MP responses, but he can't be sure how much acoustic centers have drifted in the off-axis data compared to reality, and how accurate minimum phase result is at the top octave because off-axis responses have directivity effects which make slope detection and estimation almost impossible.

Most of the work, obvious errors and remaining uncertainty will be avoided by using dual or semi-dual channel measurement and known (constant if possible) measurement distance from mic to baffle. That saves timing in measurement to all directions and with non-minimum phase systems.

 

[CharlieLaub]

One problem with "single channel measurement + minimum phase extraction + acoustic center detection" -method is that measurement programs usually export responses with normal/measured phase - not the minimum. If polar response measurement set contains e.g. 108 measurements (3-way, hor plane only, steps 10 deg), it will take a while to export all one by one as minimum phase. Radically faster option would be exporting all with normal/measured phase and using batch converter. VituixCAD Calculator can convert all responses to minimum phase at once. After that user has MP responses, but he can't be sure how much acoustic centers have drifted in the off-axis data compared to reality, and how accurate minimum phase result is at the top octave because off-axis responses have directivity effects which make slope detection and estimation almost impossible.

Most of the work, obvious errors and remaining uncertainty will be avoided by using dual or semi-dual channel measurement and known (constant if possible) measurement distance from mic to baffle. That saves timing in measurement to all directions and with non-minimum phase systems.

Right, but that still does not help you if your measurement is only good to 100Hz and you need to extend that lower. Once you try to splice on other data (e.g. a nearfield measurement) you must recalculate the phase. That was the original motivation of my work, which was then generalized and extended.

Measured phase is just fine, however, time=0 is arbitrary, so I don't see a need for 2-channel measurements. As long as you keep the mic in the same exact location for all measurements taken at a particular angle to the "on-axis" position you have all the data you need. This is the case whether you do the extra processing and phase extraction (e.g. FRD Blender) or you just go with the measured phase.

 

[kimmosto]

...I don't see a need for 2-channel measurements. As long as you keep the mic in the same exact location for all measurements taken at a particular angle to the "on-axis" position you have all the data you need.

Unfortunately not. For example ARTA normalizes single channel measurements so that highest IR peak starts at sample #300. Timing can drift depending on shape of IR curve. Off-axis measurements could have slightly different timing than axial because slope of IR is not so steep. Highest peak can also be from reflection with off-axis measurements to rear sector, especially with horns and other directive speakers. Measurement data can be totally useless mess.

REW normalizes single channel measurements same way for the display. Highest peak will be at 0 ms line -> timing depends on shape of frequency response. Fortunately REW can export also single channel responses with common time reference; moment when the signal is sent I suppose. That data could contain some latency variation, but I've never tested how much or little latency variations exists. (That alone does not make REW more valid tool for speaker engineering because few important features are still missing).

So, hardwired semi-dual is the first certain way to catch phase and timing with stable reference. That can also eliminate linear distortion of soundcard with ARTA (sweep signal only) assuming that channels are quite identical.

 

[CharlieLaub]

Unfortunately not. For example ARTA normalizes single channel measurements so that highest IR peak starts at sample #300. Timing can drift depending on shape of IR curve. Off-axis measurements could have slightly different timing than axial because slope of IR is not so steep. Highest peak can also be from reflection with off-axis measurements to rear sector, especially with horns and other directive speakers. Measurement data can be totally useless mess.

REW normalizes single channel measurements same way for the display. Highest peak will be at 0 ms line -> timing depends on shape of frequency response. Fortunately REW can export also single channel responses with common time reference; moment when the signal is sent I suppose. That data could contain some latency variation, but I've never tested how much or little latency variations exists. (That alone does not make REW more valid tool for speaker engineering because few important features are still missing).

So, hardwired semi-dual is the first certain way to catch phase and timing with stable reference. That can also eliminate linear distortion of soundcard with ARTA (sweep signal only) assuming that channels are quite identical.

Wow, I did not know that (!) about ARTA. Do you happen to know if that is documented in the ARTA manuals somewhere?

I probably did not realize that there was any time shift added because I almost exclusively make multiple measurements (single driver plus pairs of drivers) on each axis and then determine the offsets (pathlength differences) from acoustic center to mic for each. Yes, it does take a lot longer to do that!

While we are on the subject, what do you use as the second channel input? Do you split the input to the amp and route the same signal to the input used for the "second channel" in the 2-ch measurement?

 

[kimmosto]

Do you happen to know if that is documented in the ARTA manuals somewhere?

I tried to find but no luck this time.

While we are on the subject, what do you use as the second channel input?

I just loop line output 2 to mic/line input 2 and switch input 2 for line level signal. That is semi-dual, which is very safe compared to full dual with ICEpower etc. power amps which could explode with flashes

CLIO 11 is my main gear at the moment, but earlier it was ARTA (and some others).