Article: Audiolense Digital Loudspeaker and Room Correction Software Walkthrough

[omid]

Hi @mitchco,

Thank you for your article and ebook. They have been very helpful. I've redone my 2.1 with digital XO for the sub and tweeter using audiolense. I'm very happy with the results.

Unfortunately, I've only managed to get measurements using windows mme or directsound (I do click 'using separate playing and record streams'). Bernt said ideally ASIO4all, or Wasapi should be used. Unfortunately, I always get an error message saying 'perhaps too many channels', or PaInvalidChannelCount, unexpected host error etc... I've tried AL 5.5, &6.0 both in 32 bit and 64 bit. I've installed Jriver, activated WDM, etc but having no luck.

I must be missing something, or maybe I'm worrying about nothing and should stick with what works?

Do you have any suggestions?

Omid

[omid]

For measuring, with audiolense 5.5 (32 bit), my laptop runs on Windows 10, 64 bit. My input is a Calibrated USB mic (UMM6). The output is a Xonar U5 external sound card, set to 5.1, 48kHz, 24 bit (both windows and the sound card are set to the same sample rate). The sound card's front output goes to the analog input of a benchmark DAC2 --> AHB2 --> Tannoy Stirling GR woofer. The R+L rear outputs go to the analog input of a single velodyne DD12 subwoofer (all equalizers, crossover defeated). The Center+Subwoofer outputs go to the analog input of a marantz NR1609--> Tannoy Stirling GR tweeter. 

The channels in AL are set to 4,5 for L+R subwoofer, 0,1 for woofers, 2,3 for tweeters, which works correctly when the driver is winmme, or directsound. Just can't get it to work with JRiver, Wasapi or ASIO4A.

For playback, I use Roon Rock and its DSP. USB output to Benchmark DAC -->woofers. Internal roon rock soundcard -->subwoofer. Xonar soundcard-->marantz--> tweeter. I then use Adobe audition to measure by how many milliseconds  the signal is delayed for each output and correct for that. I play sweep test tones through Roon, and measure with REW. I'm getting a decent step response and house curve matching frequency response. After DSP correction, the volume does drop by 10dB or so, which is annoying (I use partial correction, only correcting up to 9000Hz, so as to avoid a larger drop). 

Please let me know if I left out any info you need.

[omid]

Thank you. That helps. 

I moved my subwoofer yesterday, and tried to repeat the measurements. The first set of sweeps worked. When I tried another sweep, the sound stuttered (buffer under run?). After that, I started getting error messages PaDeviceNotFound or something like that (Puzzling since I hadn’t touched the driver or any settings). Tried numerous pc restarts and relaunchings of  AL, same result. 

I’ll try safe mode today, maybe there is a driver conflict. I was just curious if you knew of any tricks to sort out my problems. 

Wrt volume, Roon does let you amplify. It also has a clipping warning. Although the signal is diminished by 10dB or so, the signal clips if I amplify it (i guess some frequencies are boosted by more than 10dB).  

Benchmark DACs and amplifiers don’t have a lot of gain, so 10dB matters. 

I’ll eventually get rid of the xonar soundcard and marantz for the tweeter and get another DAC and amplifier (or maybe I’ll finish building my Nelson pass F5 v3). You’re lucky your Hilo can simultaneously process bass, mid and treble streams at the same time. I technically need 3 DACs. 

 

Cheers. 

[omid]

I’m having a puzzling problem, I’m hoping someone with DSP knowledge (hint, @mitchco) may help me solve. 

 

I made a convolution filter through AL making sure ‘all in one’ is selected (so R and L channel corrections are correlated). My mic was equidistant within 1/8 inch when I recorded and generated the correction.

 

I then implemented the convolution in Roon and recorded the output from the sound card directly into my laptop’s line in. The sound card handles the twitters (ie 1800Hz and above).  I recorded the signal using adobe audition. R and L channel signals start and stop at the same time without any DSP, whereas filter corrected sounds have 1 ms delay on the R channel (34 cm difference). The delay is not frequency dependent, ie not likely to be a phase correction related delay.  

 

I imagine the twitter signal should arrive at one’s ears at exactly the same time for a centered voice. Am I missing something, or is Audiolense tripping?

[omid]

No worries. I've been through so many iterations that I can't recall what the issue was at the time. I have figured it out though. 

I'm just going to summarize a few of my findings so far in case it helps anyone:

Audiolense has some bugs/driver incompatibilities that are frustrating (and frankly should not be there for the price of the software). Bernt is thankful very helpful.

ASIO4ALL comes with a optional extra software that I hadn't downloaded. The extra software is needed if you have a USB microphone, make sure to download it. Set the buffer to 2048 bits.

'Check speakers' is good for channel mapping but it has bugs. Play with winmme directsound etc until it works for mapping the channels. Once you've done so, you don't need to check again. You can switch the driver to ASIO4ALL after that. 

Playing sweeps of more than 10 seconds does weird things, and results in alarming knocking sounds. I think it's due to clock drift. Once that happens, reset the delays for your speakers to 0, or you'll keep on getting weird knocking and distortions during your sweeps.

The low and high frequency of the target curve needs to carefully follow your speaker's measured curve or your step response looks weird. Also the more dB correction is needed to match your target curve, the lower your response curve will be. The DSP correction can easily attenuate your signal by 10 to 20 dB. Make sure your pre-amp/amp have enough dB gain (and/or your speakers are sensitive enough) to achieve a decent maximum loudness upon playback.

Although the option to' detect and correct polarity' was checked off, looking at my sub's response, I realized it was inverted. Once I switched the polarity, my step response improved a lot.

 

 

 

[omid]

Thank you @mitchco. 

The longer sweep was to get a better signal/noise ratio (the measurement goes from good to excellent). I realize the extra resolution is unnecessary.

I attached the measured step with neg polarity on the sub. Not sure why it makes a difference (I figured the DSP should easily correct for that, but it didn't). The simulated step response comes out funny (attached).

I'll attach the corrected step response and corresponding step response, and REW. I use 6 dB correction. 12 dB and 18 dB give me similar curves (slightly higher peak on the step response).

I do add some gain back in Roon (just short of clipping).

I'm happy with how it sounds. I've ordered a Motu 8a and a new amp for the tweeters.

[omid]

8 hours ago, asdf1000 said:

Hi again @mitchco

 

I watched your fantastic recent YouTube video and you mention the best DRC software solutions don't boost any frequencies.

 

But in your screenshots in this article I see the max correction boost is set to +6dB.

 

Would you recommend setting this max boost to 0dB ?

 

 

Hi @asdf1000 

Can you post a link to the video? I'd love to watch it. 

Thanks.

[omid]

Thanks for the link @asdf1000, and thanks for the nice explanations @mitchco !

[omid]

Hi Mitch (@mitchco),

 

I ‘ve been re-reading your book and wondered where you sit at this point on near field corrections and beam forming. Both concepts make a lot of sense to me. Just wondering with time, experience and switching to Audiolense whether you feel they are worthwhile. 

I recently did do a Nearfield measurement and correction in Audiolense. I loaded that in Audiolense convolver, then I did a new measurement at a distance . I then loaded nearfield and farfield cascading filters in Roon, changing the CFG file as needed, but was disappointed to see that Roon was not properly handling it (or maybe I’m doing something wrong). Before I spend too much time on this I thought I’d ask you if it’s a waste of time. 

The second part of my question is about multi seat measurements. I usually measure in a few locations right, left, front and back of my normal sitting position. In order to take advantage of beam forming would it be smarter to do these measurements front to back exactly at the midline using the accurate microphone alignment tool?

 

Thank you 

 

Omid

[omid]

Thank you for the thorough answer. 

 

I’m using Audiolense for my digital XO. Based on your answer I will not complicate things and keep my filters as they are, ie based on the listening position sweep rather than breaking it in  two parts for Nearfield and farfield. The multi seat filter will include positions to the right and left not just in the middle line.

 

Thanks to you, I don’t have to worry about cascading filters, and losing 7 to 10 dB with each filter. 
 

Cheers.