FSAF (Fast subband adaptive filtering) measurement

miniDSP and AV Nirvana
FSAF said:
I did not know that about MOTU, thank you for sharing this very valuable insight!
Click to expand...
If you are talking about the TRS vs XLR for 48V phantom power, I've used 3 different audio interfaces, and found this to be true of all of them. Focusrite 2i2, Steinberg UR22mkii, and Motu M4, 48V on XLR pins only, so use TRS for non-mic inputs and you can never make a mistake :)

I misread your comment, I read "true current driving amp" as "dividing amp" and was confused. I understand what you meant now.

I may experiment with a smaller Rsense value and an active instrumentation amp circuit, but that may take some time. Future plans..
 
FSAF said:
after I measured a few dozens of loudspeakers, I got completely confused. I am not saying that nobody can extract valuable data from it - only that I was unsuccessful.
Click to expand...

Can you please expand on this? Was it related to room / environmental noise?
Or listening fatigue?

Did this include drivers like your Audax or Focal?

@dcibel made recordings of two drviers
Dayton DA115 (alu cone, basic motor) and Wavecor WF120 (paper cone, optimised motor) and when I listened to the residual with just +20dB gain, using pair of headphones with passive noise isolation…
Well the differences were like night and day.

I will never use the DA115 in the future!
 
Confused = can't make much sense = can't explain. If I could explain, I would not be confused:-)
Yes, including Focal PS130
 
The previous version of it, PS130F, without Evo. This is how it behaves (H3 on 80 dBSPL@1m) with 0 | 6 | 12 | 24 Ohm series resistor. Nice but very expensive.

3rd.png
 
dcibel said:
It's taken a long time to get to the root cause of my previous measurement inconsistencies plaguing this thread. It wasn't anything to do with my test jig, but wiggling and jiggling the wires in the process of troubleshooting certainly "helped". Today I determined the cause - a cold solder joint on the audio input on my test amp. It wasn't being easily identified in ESS measurements, but FSAF was clearly highlighting the issue. Intermittent problems can be awful to determine. Today is a good day, as I can move on from this headache.
Click to expand...
This examplifies why i am interested in fsaf.
 
Thanks to DCIBEL i am busy getting FSAF to work. I am still digesting all the previous posts in this thread.

The first objective is to become proficient in the functionality, and to get a stable and reproducable result.

My REW FSAF rig is a w11 laptop via ASIO connected to a RME Babyface Pro FS.

REW is voltage calibrated.

With loop back from both outputs to inputs i tested FSAF with White Noise, no filtering and different sampling frequencies:
48kHz:
FSAF-RBFP-48k-NF-WN-DIST.png


96kHz:
FSAF-RBFP-96k-NF-WN-DIST.png


192kHz:
FSAF-RBFP-192k-HP-WN-DIST.png


If i apply a HP filter, the low frequency hump reduces, 96kHz example:
FSAF-RBFP-96k-HP-WN-DIST.png


Why is this hump present? And why is the 48K version showing some ripple and more noise (i think)?
 
Here a screencopy of the FSAF setting:
FSAF-RBFP-96k-HP-WN-Settings.png
 
I tested the filtering option in FSAF panel:
48kHz:
1+2 FSAF-RBFP-48k-HP3464Hz-WN-DIST.png


96kHz:
1+2 FSAF-RBFP-96k-HP3464Hz-WN-DIST.png


192kHz:
1+2 FSAF-RBFP-192k-HP3464Hz-WN-DIST.png


The 48kHz appears to have the lowest noise-floor, but the TD+N shows a bump with the shape of the fundamental. The 96 and 192kHz do not show that hump. How to find out the cause?
A sweep with 48kHz does not show any anomality:
1+2 SWEEP-RBFP-48k-NF-WN-DIST.png
 
If 96Khz/192 KHz works well, but not 48Khz, I wonder if the drivers are playing up.

Had you tried troubleshooting by testing 44.1Khz sampling and multiples of this?

The HP filter should reduce IMD and thus get a better response in a DAC/ADC conversion.

One way to tell is to use a Optical Out to Optical In- does your audio interface support that? If you don't have that, to eliminate the audio interface altogether is to download a Virtual Audio Cable and install ASIO4ALL to simulate how well REW is running on your system (or how well your system is running REW)
 
Last edited:
I can't speak to your problems @JanRSmit, there's lots of possible causes for what you are encountering. Sine sweep or RTA measurements may help in the analysis.

I can provide you a point of reference however, from my Motu M4 interface. Here's a FSAF loopback with white noise, input gain is set at a fairly typical 10 o'clock position, and output level adjusted for about -10dBFS at the input (using "check levels"). Headroom was 1dB during white noise FSAF measurement.
1740702772930.png


Here's a loopback measurement with my test amplifier, using a 4 ohm resistive load, and 0.1 ohm current sensor. My test amplifier is not a super Hi-Fi piece of equipment, but it's decent - Audiotrak AT-300. I adjusted the output level to get about 2.8V across the resistive load, and increased the input gain to keep around the same input level.
1740703351463.png
 
On my 2i2 Gen 3, white noise at -25dBFS RMS, balanced out -> balanced in :
5.png

FSAF shall suppress the residual (aka THD+N) all the way down to the noise level (except for very low frequencies, where the IR tail exceeds the allowed limit, and close to Fs/2 where lots of aliasing occurs):
4.png


REW shall be identical. If you see THD+N rising above noise for audio interface loopback, something is not quite right.

PS> I agree, we do need 96+ for SineSweep tests. Yet, FSAF was designed to operate on music, which is predominantly 44.1. You can operate FSAF on Fs>48kHz but... I have not tested FSAF nearly as thoroughly on those conditions. Sh*t may happen for "not tested = does not work".
 
So the question is whether the 48kHz is the right outcome, and thus the 96 and 192 are wrong.
Or the RMEdevice and its asio driver is not good, or REW has issues.
If it is limited to 44.1 kHz it is of limited value.
I can test focusrite 2i2 2nd gen and 18i20 3d gen , stay tuned.
 
Here some measurments this morning:
44k1
1+2 FSAF-RBFP-44k1-NF-WN-DIST.png

1+2 FSAF-RBFP-44k1-NF-WN-Sample.png

48kHz:
1+2 FSAF-RBFP-48kHz-NF-WN-DIST.png

1+2 FSAF-RBFP-48kHz-NF-WN-Sample.png


88k2:
1+2 FSAF-RBFP-88k2-NF-WN-DIST.png

1+2 FSAF-RBFP-88k2-NF-WN-Sample.png

Now rush off to a meeting.
 
Try increasing IR length to 500ms, I can reproduce similarly bad results with IR length too short. I think perhaps 300ms provides the best balance between processing time and ability to capture accurate results at low freq.
1740753584075.png
 
Last edited:
When i started playing with fsaf the ir length was more, but gave errors. Will try again , will also check the asio buffer.
 
If you've not allocated enough RAM for REW at time of install, it will produce an out of memory error if IR length is too long. For default install of 1GB allocation, it will be fine for 500ms IR length.
 
I increased the IR duration in steps of 100msec, and from 800msec the TD+N stayed the same, being without humps, more or less the avarage of the noise-floor:
96k
1+2 FSAF-RBFP-96k-Filter-WN-IR800-DIST.png


48k
1+2 FSAF-RBFP-48k-Filter-WN-IR800-DIST.png

96k no filter:
1+2 FSAF-RBFP-96k-NoFilter-WN-IR800-DIST.png

The above rsults were with ASIO buffer is 2048, below a measurement with ASIO buffer set to 1024:
1+2 FSAF-RBFP-96k-NoFilter-WN-IR800-ASIO1024-DIST.png


So thusfar the liength of the IR should be in my case 800mSec or more. The ASIO buffer was not the cause.

Here with some music:
1+2 FSAF-RBFP-96k-FilterYes-Music-IR800-ASIO1024-Sample.png

1+2 FSAF-RBFP-96k-FilterYes-Music-IR800-ASIO1024-DIST.png


I belive my setup is stable and with reproducable results.
 
That's better. I'd advise to increase the digital signal level to see where it breaks.

On my 2i2, it starts to break when the amplitudes hit above -4dBFS, regardless of the hardware analog attenuator:
This is the noise floor for max=-7dBFS:
6-85.png

and this is for max of -2dBFS:
6-90.png

with everything else being exactly the same.
 
Thanks FSAF, itcwas your post #244 that made me test with increasing the ir duration. Will do your suggestion as well, to know also that limit.
The real first test wil be the possible distortion of cables.
 
Oops, big mistake, post #288, thanks for pointing.
 
What I am finding with this testing, is that drivers obviously have very different suspension systems, so free air testing through bass range and around Fs is not exactly an equal comparison between drivers. Even though the midrange SPL may be equal, they can have very different cone excursions at low frequency which throws off a comparison. The driver either need to be properly loaded in a cabinet for similar FR (for same size of driver), or some EQ applied to ensure equal driver excursion through the bass range, or cut the bass range with LR-48 filter at 100Hz to ensure the result is low excursion, not ideal.

As it is, clear differences between motors can be easily detected, but when drivers are very similar in performance, chances are the one that exhibits lower excursion played free air may appear to be the better driver.

This is doable, but requires some extra steps of measuring TSP for each driver, and model in infinite baffle, and apply EQ for equal excursion. VituixCAD enclosure tool, save driver excursion as overlay, and then use the filter section to design a filter to align the driver excursion. I should probably also add some EQ for the difference in driver inductance too...
 
Last edited:
You must log in or register to reply here.
miniDSP and AV Nirvana

Popular tags

20th century fox 4k blu-ray 4k uhd 4k ultrahd action adventure animated animation bass blu-ray calibration comedy comics denon dirac dirac live disney dolby atmos drama fantasy hdmi 2.1 home theater horror kaleidescape klipsch lionsgate marantz movies onkyo paramount pioneer rew romance sci-fi scream factory shout factory sony stormaudio subwoofer svs terror thriller uhd ultrahd ultrahd 4k universal value electronics warner warner brothers well go usa
Back
Top