Audiolense Correction of a Multi-channel Tri-Amplified Loudspeaker System

[Brad]

I just took the new Audiolense 4.4 beta for a test drive and built a new filter set with 64 bit floating point, 48 kHz, 65536 tap filters. This filter set has 19 filter paths and performs all of the system bass management including the new 10 dB amplification of LFE. The filters control and correct a multi-channel home theater audio system. The system has the typical seven speaker layout with two additional subwoofers so I guess you would call it 7.3. The front three channels are tri-amplified and the three subwoofers are on separate channels, making a total of 10 loudspeakers with 16 measured and corrected audio channels.

 

After setting up the system, I used the all in one measurement with a microphone (Earthworks M30) aimed at the ceiling. This mic has very good omnidirectional response. I took three measurements without touching the mic and used them as a multiseat measurement to average out any anomalies.

 

To develop the correction procedure, I used a frequency only correction and progressively shortened the time window (cycles) from the defaults watching the filtered measurement and the simulated result (frequency response) looking for the best filter performance. I then switched to True Time Domain correction with a very short time window (0.1 cycles) and then progressively increased the "cycles" until the impulse response was corrected. I found that a minimum phase instead of linear phase target worked best for me.

 

You can see in the attachments the mess Audiolense had to clean up. Keep in mind the measurements are of direct connected loudspeaker drivers with no correction or filters of any kind. I removed all passive crossovers and bypassed the amps in the stacked Epik subs as they had filters built into them.

 

The target response I used is included in the final response plots. The pile 'o subs I have is pretty much flat to 10 Hz. I rolled off the target response below 16 Hz. As a test, I left the target flat with prevent bass boost selected and the response simulation was flat down to 8 Hz.

 

The configuration and filter files are "plugged in" to the digital signal processing convolver of JRiver Media Center for audio playback. A 16 channel, 24 bit, 48 kHz audio stream is output from an HTPC over usb 2.0 to a 16 channel Lynx Aurora DAC.

 

Some comments and questions for Bernt,

 

I was able to generate multiseat filters using three measurements with the new memory setting at 40%. This can only be done once at the beginning of a session. I get memory errors if I try to generate filters back to back. I must quit and restart the program between iterations when using multiseat correction. Other than that, everything worked great!

I found this in the help file "Since Audiolense 3.0, a psychoacoustic interpretation of the frequency response is being used for the frequency correction." Obviously this can have a big effect on setting targets that sound good. Does it use a method similar to that at the bottom of page 5 here? FFPO Technique

[Bernt Ronningsbakk]

Great work on the new filters, Brad. I really appreciate that you share your approach and experiences.

 

I don’t think I am able to do more with the memory problem at the moment. But I’ll stress my math lib supplier and see if he has more tricks to offer or if he has learned something new since the last time we spoke of this problem. The memory usage problem goes deep. To tell the truth I don’t know where the problem sits: In the math lib I’m using, in intel’s math instruction set, or perhaps in Microsoft’s x86 environment on 64 bit OS – or perhaps there’s a bad synergy between all of those. I hope to get out a 64 bit version before Christmas and hopefully the problem will go away – at least it will increase the memory usage limit substantial.

 

To your other question about FFPO. As far as I can see, FFPO is frequency dependent window. Long time window at long wavelengths and short at short wavelengths. That makes a lot of sense physically and psychoacoustically – and musically too. Frequency dependent windowing has been with Audiolense since the first launch. But things are more complicated than that, and Audiolense 3.1 was a big step forward as far as psychoacoustics are concerned. I wrote an article a while ago that also covers this topic. You can find here: http://www.hifizine.com/2010/12/sound-correction-in-the-frequency-and-time-domain/

 

Kind regards,

 

Bernt

--
Audiolense User Forum.
http://groups.google.com/group/audiolense?hl=en?hl=en
To post to this group, send email to audio...@googlegroups.com
To unsubscribe, send email to audiolense+...@googlegroups.com

[Mitch Global]

Hey Brad, I hope you don’t mind me saying that your system is the ultimate in high efficiency.  The dynamic impact must put a smile on. Thanks for sharing.

You mentioned a minimum phase target worked best for you.  I would be interested to hear more about that if you have more to share.

With respect to your correction procedure, I understand your method, but do you have a few subjective words about the sonic differences while incrementing cycles?   Plus, what factors determined 25 ms @ 10 Hz the right setting for your system?

Cheers,

Mitch

--

[Brad]

Hi Mitch,

 

Yes, my audio system is dynamic and can reproduce a live concert quite well. I am mainly listening to Blu-Ray concert recordings these days. The recording engineers typicall use less compression on the 24 bit lossless audio tracks with DTS Master Audio, Dolby True HD, or multichannel LPCM as compared to typical CD's.

 

I've attached the impulse response from the same filter settings with the same target switched to linear phase. There is more ringing present. From what I've read, minimum phase response sort of naturally results from frequency domain correction of audio systems. The 25ms correction window is all it took to clean up the impulse response longer looked worse on the computer screen.

 

When shortening the frequency window, if you go too short you will loose filter resolution. Just keep an eye on the filtered measurement for too much smoothing. I did not stop and listen to the filter iterations, but I can say the changes in the simulated results were significant and the end result sounds pretty good. I think the big hitters to Audiolense success are:

 

1. Get your system and room as good as you can before using Audiolense.

2. Measure with a microphone that has a good impulse response.

3. Use microphone placement without early reflections from the seats messing things up.

4. Design a good target response for your system and personal preferences.

5. Use digital XO if at all possible.

6. Experiment with Audiolense - this can go on without end it seems.

[Mikkel]

Hi Brad,

 

Great post you wrote. I will give it a go this week and see what can be achieved in my own system (which is pale compared to yours - I'd love such a system AND room!).

 

Have you made an A-B listening between linear and minimum phase? I have tried but have no way to leave out the placebo effect of thinking there is a difference :-)

 

Btw and out of curiosity: looking at the single picture of your room, have you treated it with any sort of passive materials or...?

 

 

Best regards,

Mikkel

[Brad]

Hi Mikkel,

I have tried both linear and minimum phase targets but not in a way to compare them (too many other changes). The current beta release of Audiolense is the first fully working version for me so I will experiment more now.

I have 7 bass traps in the room. I use 2x4 foot MiniTraps by RealTraps
I could use some more. I'll post a photo in a few days.