[Joe Bloggs]

A few years ago you may have heard a bit about that crazy guy Joe Bloggs who claimed to be able to hear frequency response curves of earphones by ear and tune anything to sound good, including that one time he took some 10RMB buds from the Huaqiangbei mall and posed them as advanced IEM prototypes at the nearby Chinese headphone HiFi show of the time and was offered $1000 for them...

Well time has flied and that Joe Bloggs has now worked for FiiO and now HiByMusic, authored the much-loved (?) MageSound 8-ball music mastering board on the HiBy players, and continues to hide behind his elusive computer workstation when listening to music, doing god-knows-what.

Well here's "god-knows-what"!

1. Continued investment into general spatialization HRTFs

Ever since Viper4Android came out with 4-way HRTF support, I've been experimenting with measuring and creating generalized HRTFs for "public" consumption (as public as the V4A community can be). These days @jaakopasanen's excellent Impulcifier project makes it easy for more people to roll their own, but for anybody who can't, you're welcome to ask me for one to try!



2. Surround upmixing
Given that once you're set up, HRTFs of any number of speakers can be had literally with a turn of the head, and given the difficulty of front localization even with a full stereo pair HRTF, I think there will at some point be a demand for surround upmixing of stereo music for headphone listening. So here it is: (VST plugin)

(the interface is obviously as janky as the tech is advanced XD )

The main takeaway here being that you can tweak the upmixing ratio to have the soundstage be anywhere from all in front of you to 360 degrees around you, for any given piece of music...

3. Dynamic Range Expansion

And finally, an attempt to fight a war that may be behind us these days, the dynamic range wars...

(the interface is obviously as janky as the tech is advanced XD )

The main takeaway here being you can tweak the Effect slider to adjust the dynamic range of any music you're listening to up and down... (within reason)

And finally, yes, that's about how much I care about bitperfect audio streams on a PC

Resources for download:
the combined dynamic range expansion / surround upmixing VST plugin
https://www.dropbox.com/s/8awse518v0gw2ti/RealTrident8_DRE_custOvL_octSmooth.dll?dl=0

(DR effect slider on lower left, upmixing soundstage ratio slider on upper right)

Sample HRTF
https://www.dropbox.com/s/jgaw1mib1pexhi8/20180807-7.7-retimed-jDSP order-crop.wav?dl=0
https://www.dropbox.com/s/k4sovfwo77eif8b/20180807-7.7-retimed-surround-jDSP order_fronthack-crop.conf?dl=0
(configuration file for MCFX convolver http://www.matthiaskronlachner.com/?p=1910)

 

[rlw6534]

I really need to learn how to use Audacity.

 

[morgin]

Sample HRTF
https://www.dropbox.com/s/oibm3f6y2rabbii/20180807-7.7-retimed-jDSP order.wav?dl=0

This sample .wav gives me the voice in the right and music in the left when playing 7.1 movies. The surround is nice just be better if it was further away but its not as close as the atmos or other hrir's But the clarity and the extra sounds I'm hearing it gives is amazing

you're welcome to ask me for one to try!

I have my impulcifer hrir but would love to ry what you can do

 

[Joe Bloggs]

This sample .wav gives me the voice in the right and music in the left when playing 7.1 movies. The surround is nice just be better if it was further away but its not as close as the atmos or other hrir's But the clarity and the extra sounds I'm hearing it gives is amazing



I have my impulcifer hrir but would love to ry what you can do

You would need to reorder the wavs as follows:
(I guess you're using equalizerAPO)
Copy: 1=1+0.6*3+0.6*4 2=2 3=0 4=5 5=6 6=7 7=8 8=1 9=2+0.6*3+0.6*4 10=0 11=5 12=6 13=7 14=8
Channel: L R 3 4 5 6 7 8 9 10 11 12 13 14
Convolution: 20180807-7.7-retimed-jDSP order-crop.wav
Copy: L=L+R+4+5+6+7 R=8+9+11+12+13+14

 

[morgin]

You would need to reorder the wavs as follows:
(I guess you're using equalizerAPO)
Copy: 1=1+0.6*3+0.6*4 2=2 3=0 4=5 5=6 6=7 7=8 8=1 9=2+0.6*3+0.6*4 10=0 11=5 12=6 13=7 14=8
Channel: L R 3 4 5 6 7 8 9 10 11 12 13 14
Convolution: 20180807-7.7-retimed-jDSP order-crop.wav
Copy: L=L+R+4+5+6+7 R=8+9+11+12+13+14

I've never reordered or edited a wav file can you explain how its done? Yes I'm using EQapo and hesuvi

 

[Joe Bloggs]

I've never reordered or edited a wav file can you explain how its done? Yes I'm using EQapo and hesuvi

You can simply copy the lines I wrote into your eqAPO config file

 

[musicreo]

Using the VST in EQ-APO results in a crackling sound for me. Playing with the bufLen setting (bufferlength?) resulted in a crash of the Eq-APO configuration editor. Actually I had to delete the VST to restart the editor.

 

[Joe Bloggs]

Using the VST in EQ-APO results in a crackling sound for me. Playing with the bufLen setting (bufferlength?) resulted in a crash of the Eq-APO configuration editor. Actually I had to delete the VST to restart the editor.

what's your processing sample rate?

I also had this editing issue with various other commercial plugins (the crashing after editing). It seems to result from the configuration string becoming too long. My hunch is that unchecking "apply automatically" would help. Also, you can delete the "ChunkData "blahblahblah"" thing at the end of the line manually in a text editor to fix.

As for the stuttering itself, the bufLen shouldn't be an issue. It should either be the totLatency needing to be increased or the processing sample rate needing to be set at 44100 or 48000. Or just my plugin not being optimized enough to run on your computer yet

 

[musicreo]

The processing sample rate is 48kHz. I could run the VST with VSTHost but I usually prefer to use EQ-APO.

Why does your impulse response file have 28 channels?

The mcfx vst convoler also gives me a stuttering sound when I use it in EQ-APO or in foobar2000 (the 32bit version) with my 16ch impulcifer file. In your *conf file the length is set to 100000. What is this setting?

 

[Joe Bloggs]

The processing sample rate is 48kHz. I could run the VST with VSTHost but I usually prefer to use EQ-APO.

Why does your impulse response file have 28 channels?

The mcfx vst convoler also gives me a stuttering sound when I use it in EQ-APO or in foobar2000 (the 32bit version) with my 16ch impulcifer file. In your *conf file the length is set to 100000. What is this setting?

>The processing sample rate is 48kHz. I could run the VST with VSTHost but I usually prefer to use EQ-APO.
Did you try connecting your own HRIRs to the VST's output?
>Why does your impulse response file have 28 channels?
There were 7 eye-level channels (2 channels for each virtual speaker) and 7 height channels (again 2 channels for each virtual speaker).
1st 14 channels were as follows
In-Out
L-L (where in Out L stands for left ear, R stands for right ear)
R-L
C-L
RL-L
RR-L
SL-L
SR-L
L-R
R-R
C-R
RL-R
RR-R
SL-R
SR-R

the next 14 channels would have been the same thing but at height. However, I personally found (even though the IRs were recorded on my own ears then balanced for neutrality) that taking 0.6*L-L and 0.6*R-L as replacement for C-L and C-R (or simply doing phantom centre) gave better results than the actual measured centre speaker, so that's what my conf file was set for.

>The mcfx vst convoler also gives me a stuttering sound when I use it in EQ-APO or in foobar2000 (the 32bit version) with my 16ch impulcifer file. In your *conf file the length is set to 100000. What is this setting?

That's the maximum length of an IR to process. The actual length convolved is this number or the actual length of the IR, whichever is lower.

I was also having trouble processing the IR using mcfx on eqAPO. You can use this config snippet with eqAPO for up to 7.1 input (channel assignments of L R C LFE RL RR SL SR from ch1 to ch8)

Copy: 1=1+0.6*3+0.6*4 2=2 3=0 4=5 5=6 6=7 7=8 8=1 9=2+0.6*3+0.6*4 10=0 11=5 12=6 13=7 14=8
Channel: L R 3 4 5 6 7 8 9 10 11 12 13 14
Convolution: 20180807-7.7-retimed-jDSP order-crop.wav

 

[musicreo]

Did you try connecting your own HRIRs to the VST's output?

Yes I did but the VSTHost gives me only 2channel output so there is no benefit for up mixing. The VST perfomance in the VSTHost is also far from perfect. Still sometimes stuttering.

the next 14 channels would have been the same thing but at height. However, I personally found (even though the IRs were recorded on my own ears then balanced for neutrality) that taking 0.6*L-L and 0.6*R-L as replacement for C-L and C-R (or simply doing phantom centre) gave better results than the actual measured centre speaker, so that's what my conf file was set for.

How did you do the measurement? What microphones have been used? In my personal favourite measurement the center speaker is the best part.

I was also having trouble processing the IR using mcfx on eqAPO.

For foobar2000 I use ConvolverVST but a newer alternative would have been nice.

 

[Joe Bloggs]

Yes I did but the VSTHost gives me only 2channel output so there is no benefit for up mixing. The VST perfomance in the VSTHost is also far from perfect. Still sometimes stuttering.


How did you do the measurement? What microphones have been used? In my personal favourite measurement the center speaker is the best part.


For foobar2000 I use ConvolverVST but a newer alternative would have been nice.

Before I worked on the current version, there was one that was heavily optimized by my friend James based on roughly the same algorithm as the current one, but for upmixing only: https://www.dropbox.com/sh/px56zdw6eb95vga/AAB1QNacGpIZW3fZhd0nBe9Va?dl=0

And there's a stereo dynamics plugin that worked independently of the upmixing: https://www.dropbox.com/s/gwr5389vyayl5rq/DREPlugin_eff.dll?dl=0

Their combined CPU load should be lower than the one on this OP

You can also try tweaking CPU priority for audiodg.exe which is the DSP process of eqAPO

Oh, and for VSTHost, 1. you won't be limited to 2 channels if you use it together with ASIO4ALL, 2. it doesn't really matter since your output IS two channels (left and right earcups). Internally you can upmix internally and render them with mcfx convolver no problem (and mcfx convolver works wonderfully in VSTHost).

>Measurement
Mic'd a log sine sweep and deconvolved. Lots of post processing that basically made the BRIRs tonally neutral when playing the average recording through the upmixer, making them sort of agnostic in regards to choice of headphones. Not really affecting relative positioning, which is always worse the closer to front centre it is for me lol. Mics were / are these: https://soundprofessionals.com/product/MS-TFB-2/ I've snipped off the earhooks and it didn't make much difference either.

>foobar convolver
The standard foobar convolver https://www.foobar2000.org/components/view/foo_convolve can be used, offering an experience similar to eqAPO when used with the Matrix Mixer https://www.dropbox.com/s/i4ukyzw109kgo22/foo_dsp_mm.dll?dl=0

 

[reter]

I'm using Hesuvi but i found that i don't like the stereo upmix settings they made, can i use your vst plugin in EQ-APO without using the Dynamic Range Expansion? because when i enable the plugin, the default settings that comes with your plugin makes my audio very weird

maybe i am missing something?

 

[Joe Bloggs]

I'm using Hesuvi but i found that i don't like the stereo upmix settings they made, can i use your vst plugin in EQ-APO without using the Dynamic Range Expansion? because when i enable the plugin, the default settings that comes with your plugin makes my audio very weird

maybe i am missing something?

You might be missing the part where you have to select a number of channels for processing equal to the total number of output channels by the plugin, BEFORE the plugin.

e.g. if upmixing to No. Of channels = 5 and stereo sub = off then the channels you have to have for the plugin's processing are L, R, C, LFE, RL and RR.

(and yes, if nothing else you will need to modify the following to suit your situation:
1. no. of channels to the no. of main speakers you have
2. "stereo sub" checked only if you don't have a separate LFE (bass will be directed through front L R)
3. adjust the "angle"s to suit the configurtion of your speakers, where -1 is leftmost and +1 is rightmost. e.g. in a 5 main channels setup angles 1...5 may be as follows
-1, -0.4, 0, 0.4, 1 . Angles beyond your number of main channels may be ignored.)

 

[reter]

You might be missing the part where you have to select a number of channels for processing equal to the total number of output channels by the plugin, BEFORE the plugin.

e.g. if upmixing to No. Of channels = 5 and stereo sub = off then the channels you have to have for the plugin's processing are L, R, C, LFE, RL and RR.

(and yes, if nothing else you will need to modify the following to suit your situation:
1. no. of channels to the no. of main speakers you have
2. "stereo sub" checked only if you don't have a separate LFE (bass will be directed through front L R)
3. adjust the "angle"s to suit the configurtion of your speakers, where -1 is leftmost and +1 is rightmost. e.g. in a 5 main channels setup angles 1...5 may be as follows
-1, -0.4, 0, 0.4, 1 . Angles beyond your number of main channels may be ignored.)

Do i need to convert my hesuvi.wav file to make it work directly in eq apo?

Is this the way it should be set in eq apo? i found that the weird audio was the ratio setting in your plugin, after setting that to 0 i don't have the weird sound anymore, can you explain me what that setting does?

Also, with your .wav (and also mine) seems like the angle7 and angle8 does nothing, why there are only 6 channels working?

sorry i'm inexperienced with eq apo, i only used hesuvi with impulcifer, thx in advance!