ARTA

To Ente: Feature Request

Good morning. I had asked for this feature some time back but realized that I was quite unclear. It is a feature in LIMP in which I am hoping to be able to use the F3 button to evaluate which Impedance Model is best suited for a particular measurement. I was hoping to be able to add multiple overlays with different colors using each of the models to compare to the measurements. I hope that this makes sense. Thanks, Jay
 
What use is a "true" minimum phase response when measuring with a 2 channel system?

For most design purpose, just like absolute SPL, minimum phase is not required at all, only relative phase between drivers, so simply lock the FFT window start and off you go. If your measurements have a decent SNR you can avoid HBT process altogether. SoundEasy unfortunately forces this process for all measurements, so it becomes a "min phase" that you then add delay back in to keep the measured phase result which I feel is a tedious and unnecessary process.

When I asked the question of the purpose of min phase for 2 channel design to Bohdan, the answer was simply so that you can keep a driver measurement library with a common phase result, and drop into your designs with the necessary delays. The way I look at it, is that similar function can be achieved by keeping accurate record of each impulse response for distance of mic to baffle surface. Reference to baffle surface is usually a common plane for all drivers so results can be achieved that way by saving only impulse response files for records and processing as needed when needed. Some reference point is needed regardless to determine the delay requirements when combining multiple drivers with min-phase response in a project, so why not keep the impulse measurement and record of mic distance instead?

Anyway, on to the process...

The Inverse HBT process described here appears to simply compare the phase result of the HBT to the measured phase and provide an "error" result. The intent would be to adjust the delay of the measured result until the error is as low as can be, which I believe is what the "automated" process document is trying to describe. It is simply an iterative process. The result of min phase derivation is that the phase of the HBT result is always min phase, and accuracy is only as accurate as the HBT slopes chosen.

As an aside, Bohdan's previous paper on minimum phase implies that the start of the FFT for min phase is basically 10 samples back from the peak of the impulse, within the error of +/-1 sample. This was a bit confusing when looking at the reference impulse in SoundEeasy, as 10 samples back from the reference impulse does not locate the cursor at t=0ms. t=0 when looking at the reference impulse turned out to be only 8 samples from the peak of the impulse, and Bohdan did not reply when questioned about this. One would think that t=0 would result in the min phase location for the reference impulse, but it does not in SE. I digress.

Moving on to ARTA, there's a few things to note. First is that it has no HBT feature at all, however there is a "min phase" button under the frequency response "view" menu, as well as some other phase features. You will find a description of this "min phase" option in the help file under section 6.1.3.

"A simple definition of minimum phase is: A system phase characteristics for which the equivalent system with the same magnitude characteristics and a minimal phase changes can be realized (over all frequencies. The difference between the phase and the minimum phase characteristics is usually called excess phase.

Mathematically, the minimal phase can be estimated from the magnitude of the frequency response using the Hilbert transform. ARTA, as well as other similar programs, use the DFT to calculate the Hilbert transform. It introduces periodicity in the estimation of the minimal phase and gives the result that is close to the true minimal phase only at frequencies below fs/4."


Another thing to consider is that when applying a microphone calibration, only the frequency response is compensated for in ARTA, the phase remains as measured which will create inconsistencies when comparing measurements made using different microphones even if the frequency response measured is exactly the same. I did question the developer on this, and the response was simply that there is not agreement in the industry of how to apply phase compensation for microphones, so ARTA just doesn't do it at all, and the recommendation is to use a microphone with good linearity to begin with.

A friend showed me his method for determining min phase using ARTA that I will share with you. First simply process the frequency response as you normally would, then go to view menu and select "excess phase". Now, under the edit menu, go to "Delay for phase estimation" and adjust the delay to achieve a "min phase" result, basically where phase has no slope through the "pistonic" driver region where the Frequency response is flat. Un-check the "excess phase" option and you should now be viewing a minimum phase result, as you've applied a delay to remove the excess phase from your measurement. This process would be somewhat similar to the process described by Bohdan IMO, but without any HBT processing, simply the measured phase is being shifted by a delay value to remove the excess delay, measured response and phase is maintained across the entire spectrum which isn't the case when HBT is applied.
 
Here is excess phase shown with the delay adjusted to remove the excess.
 

Attachments

  • Screenshot_20210725_114433.png
    Screenshot_20210725_114433.png
    29.1 KB · Views: 280
Last edited:
Hi DcibeL,
you have many right points.
In real measurements, in environment with reflections it is impossible to get/calculate minimum phase.
Why we need phase that is close to minimum phase?
We need it to see maximally unwarped phase to better see some phase parts that points to possible resonance. And your approach is good.
The more important question is: Why we need correct phase. The answer is: to have correct crossover response summation, or for estimation of system poles and zeros.
The simplest approach is to measure multi-driver response from the same distance of microphone, but there will always be some error due to off-axis and diffraction response.
A more complicated way is to interpolate 3D response, but there always will be some measurement and interpolation error.
Using the minimum phase in crossover design is not recommended.
Ivo
 
Greetings. I have been using STEPS to generate frequency response graphs for my reel to reel recorders. Even with the coarse 1/6 octave sine waves, the results are excellent. The problem is that STEPS crashes after running about two graphs. One system is a recent Win 10 desktop with a Focusrite Scarlett 2i2 and another is a Win 7 laptop with an E-MU 0202. Both eventually crash with the blue screen.

Could this be a problem with the demo version? I am hesitant to pay for a license if all I can run is 2 FR plots at a time. ARTA seems to work fine on both setups with only rare crashes on the laptop. Does anyone have any suggestions?
 
Considering moving from REW to ARTA (with a DIY spinorama table) but have a few questions:

1. Do I benefit from an ARTA box? First, I have DATS v2 for impedance measurements and ZMA files already. Is ARTA any more accurate than DATS v3, which upgrade I'm considering? Second, is it really necessary to do full dual-channel measurements including the voltage probe to take the amp itself into account as long as the amp is of good quality, or is it sufficient to do just semi-dual channel measurements like I'm used to? This especially since, and I quote from the relevant application note: "The quality of measurement with this setup can be better than with ARTA Measuring Box, especially when measuring impedance as there is no need for voltage dividers, and the dynamic range is larger (usually the divider reduces input level 20 dB or more)". If still considered useful, I'd get the ready-made ARTA 4.1 module but obviously not if considered superfluous in my case. If it matters, my interface is the Focusrite Clarett 2Pre USB (or EMU 0404 USB if I could just find it…).

2. If I choose to do full dual-channel measurements, I understand that the negative amplifier output needs to have true ground (zero potential between it and input ground) which rules out almost all modern digital amps and bridged Class AB etc. But I see the Thomann PM40C being recommended which LM3886 chip is a "Class AB-A (conjugate) amplifier that has a fully symmetrical structure (push-pull), meaning that the sine waves produced, will produce a +, - output". Not sure what that means exactly, if it's a deal killer, but I'd like a confirmation that it has true ground on the negative output (single chip, not two in parallel bridged) before going down this route.

3. Also for the LM3886 experts, most LM3886 modules take bipolar +/- VAC on their input. I'd like for the amp to be portable (12V battery) however, so I can make measurements outdoors, and consider getting one of these modules that convert 12VDC to +/- 18-26VDC (not VAC) and run one of these modules off of it. Would appreciate if someone could confirm that this (or any other good-value LM3886 module) could be run off this bipolar DC supply into what I assume is an onboard rectifier expecting an AC supply? The seller says this is the case but I've learnt to not trust all Ali sellers when it comes to EE.:p

Thanks.
 
2. If I choose to do full dual-channel measurements, I understand that the negative amplifier output needs to have true ground (zero potential between it and input ground) which rules out almost all modern digital amps and bridged Class AB etc. But I see the Thomann PM40C being recommended which LM3886 chip is a "Class AB-A (conjugate) amplifier that has a fully symmetrical structure (push-pull), meaning that the sine waves produced, will produce a +, - output". Not sure what that means exactly, if it's a deal killer, but I'd like a confirmation that it has true ground on the negative output (single chip, not two in parallel bridged) before going down this route.
My solution to the problem of measuring the output of an amplifier or passive crossover is to use a microphone transformer to provide galvanic isolation and voltage step down.

The one I use has several taps but I use the 1200 ohm impedance tap on the input side and 30 ohm output impedance.

The impedance ratio gives just the right amount of voltage step down from a ~100w amplifier to a line level input without needing to use a diode protection network, nor is a resistive divider needed.

Of course the transformer itself has its own frequency response limits - the one I use measures nearly ruler flat from 5Hz up to about 30Khz so is more than good enough for me.

I use it for the 2nd (reference) channel on dual channel microphone measurements, and also use it to measure the output of passive crossovers.

Due to the full isolation you can connect to any part of any crossover - it doesn't matter if its a series crossover, whether it has an all pass filter (that doesn't share a common input and output "earth") etc.

In my experience even if you use a simple crossover where the negative is passed through to the driver directly and the negative terminal of the amplifier is common to signal ground, directly connecting the line input of a sound card (via a resistive divider) to the amplifier / crossover causes earth loops that cause noise pickup and inaccuracies.

This is particularly apparent if you try to measure a passive high pass or low pass filter - instead of seeing the expected continuous rolloff the response will often rise again due to the ground loop!

So I find a step down transformer very useful for measuring low impedance high level signals with full isolation.
 
I use ARTA, but not the ARTA box. If you want an amp for measurements, go for a single ended non-switching output with common gnd and symm output swing - i.w. half bridge topologoy, not full bridge.
To measure T/S parameter I discovered there is no need for a distinct amp. The line out of my soundcard is sufficient. So all I did was building the simple speaker box with a 100 ohms precision resistor and am done for T/S setup.

I considered measuring full-bridge class-d amps using isolating xformer. After measuring THD of the transformers at low levels I gave up this idea and build a symmetric (i.e. 2-channel) resistor attenuator that I link between power output and symmetric line input. Thislast tool is a must-have imho.
 
Last edited:
To measure T/S parameter I discovered there is no need for a distinct amp. The line out of my soundcard is sufficient. So all I did was building the simple speaker box with a 100 ohms precision resistor and am done for T/S setup.

Unfortunately this is not true. (as I said before)
I can show you quite some comparisons I did, also with references from manufactures. (who show the same behavior)

There is a need for highers voltage in some cases.
Otherwise your T/S can be off by 30-40%

But just in general it's handy to have a bit more juice for all kinds of troubleshooting purposes.
Even something like a LM1875 etc is already adequate.
Silly to be frugal about it.

One can even get a working 2nd hand stereo amp for less than $20-30 these days.

*btw,
Bridged amps can work as well, but that needs a little more circuitry to get it going.
 
Last edited: