Optimal crossover point for my measurements

2234rew

Member
Thread Starter
Joined
Nov 24, 2020
Messages
180
This is imo a very fine result. I would have easily lived with this kind of performance.
Thanks for your help in getting to this point.

There's no urgency from my end to solve this low end pre-ringing because I'm loving SQ.

I'll park this aside and focus on the HDMI drama
 

juicehifi

Audiolense
Staff member
Joined
Feb 5, 2018
Messages
232
Quite observant of you, jjazdk, :T

I usually run a «development license» which basically exposes a lot of stuff that has gone out and some stuff that has never gone in .... and some stuff that was only for development purposes.

I can assure you that every parameter where I believe a custom setting could improve the sound quality is exposed to the user.
 

jjazdk

Member
Joined
Aug 17, 2018
Messages
62
Thanks :-)

I have not given up on Audiolense, and am still toying with the measurements. Downloaded 6.17 yesterday, and succeded in making consistent measurements through JRiver (Direct Sound Out to JRiver, Direct Sound In from my OmniMic USB microphone).
My idea was to experiment with really short TTD (1344 taps = 15ms delay@44k1, to be within the critical lip sync window). But.. The corrections goes totally bogus when I make them with less than 2048 taps :-/




Quite observant of you, jjazdk, :T

I usually run a «development license» which basically exposes a lot of stuff that has gone out and some stuff that has never gone in .... and some stuff that was only for development purposes.

I can assure you that every parameter where I believe a custom setting could improve the sound quality is exposed to the user.
 

hulkss

Active Member
Joined
Feb 12, 2020
Messages
181
Thanks :-)

My idea was to experiment with really short TTD (1344 taps = 15ms delay@44k1, to be within the critical lip sync window). But.. The corrections goes totally bogus when I make them with less than 2048 taps :-/
Try the Audiolense Convolver. I use 65000 (approx) taps and have only 7 ms latency in the convolver filters. Frequency correction only, minimum phase target, and min delay crossovers if you use them.
 

jjazdk

Member
Joined
Aug 17, 2018
Messages
62
Try the Audiolense Convolver. I use 65000 (approx) taps and have only 7 ms latency in the convolver filters. Frequency correction only, minimum phase target, and min delay crossovers if you use them.
I did try that, and it did not at all work in my system. Mid left+right bass drivers (100-250Hz) ended up not summing at all, while the rest of the frequency spectrum summed nicely.

Also, 1344 taps is enough to account for a lot of phase issues.
 

juicehifi

Audiolense
Staff member
Joined
Feb 5, 2018
Messages
232
TTD doesn't always work well on very short filters. 1344 taps is very scarce to clean up the time domain and then fix the frequency response.
 

jjazdk

Member
Joined
Aug 17, 2018
Messages
62
TTD doesn't always work well on very short filters. 1344 taps is very scarce to clean up the time domain and then fix the frequency response.
It is, I agree.

But, the correction with 1344 taps really screws up the response in a very mystical way. The frequency response from my nicely corrected (with MSO) subwoofers goes totally haywire. I can make a few plots if it is of interest?
 

juicehifi

Audiolense
Staff member
Joined
Feb 5, 2018
Messages
232
You may find something that works if you apply short TTD window. It has to be well inside the total taps available.
 

jjazdk

Member
Joined
Aug 17, 2018
Messages
62
Ahhh, I did not think of that. I will try that and see if the simulation comes up with some sensible results. Thanks.
 

jjazdk

Member
Joined
Aug 17, 2018
Messages
62
You may find something that works if you apply short TTD window. It has to be well inside the total taps available.
Hmm.. Can you elaborate on this. How do I apply a TTD window that is short enough for a 1344 tap filter at 48000kHz (14ms filter length) ?

Quite often Audiolense comes out with filters that make no sense at all, as in correction filters that allows the subwoofer signal through, but attenuates the front speaker 70dB.

As soon as I go from 1344 to 1536 taps, this strange behaviour dissapears.

Also, even if I closely match the targetcurve to the measured subwoofer response, it still tends to make a really strange filter for the subwoofer channel. Quite odd, as I am trying to make it easy for the filter matching algorithm, hmm..

41636
 

jjazdk

Member
Joined
Aug 17, 2018
Messages
62
Missing Front Left Speaker?!?

But here the subwoofer filter matches the target response very well.

41637
 

juicehifi

Audiolense
Staff member
Joined
Feb 5, 2018
Messages
232
When you combine crossover and TTD with such a short filter you need some luck to get a good result. It is basically a matter of trial and error.
 

jjazdk

Member
Joined
Aug 17, 2018
Messages
62
When you combine crossover and TTD with such a short filter you need some luck to get a good result. It is basically a matter of trial and error.
Fair enough. But, look at my last screenshot, the correction filter for the Front Left is missing and so is the Front Left output. Only the subwoofer part of the simulated response is there?!?
 

juicehifi

Audiolense
Staff member
Joined
Feb 5, 2018
Messages
232
It's down there somewhere. These things can happen when you run out of taps.
 

2234rew

Member
Thread Starter
Joined
Nov 24, 2020
Messages
180
Try the Audiolense Convolver. I use 65000 (approx) taps and have only 7 ms latency in the convolver filters. Frequency correction only, minimum phase target, and min delay crossovers if you use them.
Hi @hulkss

Does the same "Frequency correction only, minimum phase target, and min delay crossovers" result in same 7ms latency when you use X-MCFX Convolver on macOS ?

If not, what's the lowest latency you got with X-MCFX Convolver and which AL settings used ?
 

hulkss

Active Member
Joined
Feb 12, 2020
Messages
181
On my M1 Mac I have less overall latency. I can operate with smaller buffers. The convolver itself has the same latency. I am using Jriver into Element (convolver host) into MOTU pro audio interface. The audio is routed from Jriver to Element with Loopback software from Rogue Amoeba.
 

jjazdk

Member
Joined
Aug 17, 2018
Messages
62
On my M1 Mac I have less overall latency. I can operate with smaller buffers. The convolver itself has the same latency. I am using Jriver into Element (convolver host) into MOTU pro audio interface. The audio is routed from Jriver to Element with Loopback software from Rogue Amoeba.
Is there a benefit in using the Element (convolver host), as I thought the optimum solution would be to use the Audiolense convolver?
 

hulkss

Active Member
Joined
Feb 12, 2020
Messages
181
Is there a benefit in using the Element (convolver host), as I thought the optimum solution would be to use the Audiolense convolver?
No benefit to Element unless you want to load other processing plugins. The AL convolver is not supported on Mac OS.
 

jjazdk

Member
Joined
Aug 17, 2018
Messages
62
No benefit to Element unless you want to load other processing plugins. The AL convolver is not supported on Mac OS.
I would like to try a different convolver than JRiver's build-in, so perhaps Element is worth a try (I haven't yet found a VST convolver that worked in JRiver).
 

hulkss

Active Member
Joined
Feb 12, 2020
Messages
181
Perhaps you can now try the Hang Loose Convolver from Accourate Sound, which has VST3 functionality. https://accuratesound.ca/ JCR
The Hang Loose Convolver looks OK, however, from the documentation it appears to be stereo only in and out. No active crossovers or multi-channel home theater possible.
The X-MCFX convolver can support up to 128 channels in and out.
 

jjazdk

Member
Joined
Aug 17, 2018
Messages
62
Thanks.. I will look at the X-MCFX as I need 12 channels of convolution.
 

2234rew

Member
Thread Starter
Joined
Nov 24, 2020
Messages
180
Thanks.. I will look at the X-MCFX as I need 12 channels of convolution.
I have a thread where I discovered X-MCFX and Element and shared by findings:

 
Top Bottom