Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[jjazdk]

Hi All,

I am seeing a strange issue (a bug, I think) with the convolver.

For convolver channels 0 to 9 everything works as expected, but channel 10 and above exhibit a strange behaviour.

I have a multichannel soundcard (Focusrite Scarlett 18i20) and all channels work flawlessly when the convolver is not enabled, or if the convolver script only contains configuration for channel 0 to 9.

What happens is a little difficult to explain, but here goes.

If the convolver script contains configuration for channels above 9 (f.x. 10 & 11), input from channel 1 (right) is copied to these channels and convolved?!?
Something similar happens for channels 12 & 13, the input is still from channel 1 (right), but the output level is now higher?!?
I don't get any of this strange "bleed through" from left/center/sub or any other channel.

If I disable the convolver, everything works just fine and I can do whatever routing and filters in the "parametric equalizer" that I feel like without any issues.
One thing worth noticing is that no matter what I do in the "parametric equalizer" I see this issue with the convolver, it doesn't matter if I disable the "parametric equalizer" or if I try to mute each and every channel.

I am very confused...

For now I am using 10 channels, since I have a need for more channels I would like to know what I am doing wrong, or if there is a bug that needs to be fixed.

An example on a script that exhibit this strange behaviour is here:
44100 16 16 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
N:\Apps\rePhase\2014-12-16_Impulses\XLS_03_LP250_24_HP100_48_taps8k_FFT16k_middle_blackman_44k1.wav
0
9.0
9.0
N:\Apps\rePhase\2014-12-16_Impulses\XLS_03_LP250_24_HP100_48_taps8k_FFT16k_middle_blackman_44k1.wav
0
10.0
10.0
N:\Apps\rePhase\2014-12-16_Impulses\XLS_03_LP250_24_HP100_48_taps8k_FFT16k_middle_blackman_44k1.wav
0
11.0
11.0
N:\Apps\rePhase\2014-12-16_Impulses\XLS_03_LP250_24_HP100_48_taps8k_FFT16k_middle_blackman_44k1.wav
0
12.0
12.0
N:\Apps\rePhase\2014-12-16_Impulses\XLS_03_LP250_24_HP100_48_taps8k_FFT16k_middle_blackman_44k1.wav
0
13.0
13.0

[mattkhan]

I don't understand exactly what your setup is. You mention that the input from channel 1 is copied but the example given applies convolution to channels 9-13 only. I suppose we need to know;

- the output format
- what the source channels are
- any other DSP blocks that are applied
- exactly what comes out the other end (your comment implies that channel 1 is copied to all channels above channel 9 but not sure if this is what you mean)

[jjazdk]

Output format is 16 channels:


Source Channels:
Not sure I understand this correctly, I am playing 2-channel music or using the multichannel test sounds available in Tools -> Advanced Tools -> Audio Calibration -> Volume calibration (narrow band pink noise)

No other DSP blocks applied, this happens even with output format and parametric equalizer disabled.

It seems like channel 1 is copied to all channels above 9, yes.. But the level of the copied signal is not the same in all channels, which is strange. Furthermore the convolution from the script is also applied to the channels. As such it appears the convolution is working, but the input channel is not correct.

[jjazdk]

Correction, if I disable the "Output Format" the strange convolving does not happen.

A screenshot of the Analyzer showing the odd behaviour:


Actually it is also worth noting that the convolution is incorrect, the filtered response is not what the assigned filter should do!?! (The filters work correct for all other channels)

This is the convolver configuration:
44100 16 16 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
N:\Apps\rePhase\2014-12-16_Impulses\XLS_03_LP250_24_HP100_48_taps8k_FFT16k_middle_blackman_44k1.wav
0
9.0
9.0
N:\Apps\rePhase\2014-12-16_Impulses\XLS_03_LP250_24_HP100_48_taps8k_FFT16k_middle_blackman_44k1.wav
0
10.0
10.0
N:\Apps\rePhase\2014-12-16_Impulses\XLS_03_LP250_24_HP100_48_taps8k_FFT16k_middle_blackman_44k1.wav
0
11.0
11.0
N:\Apps\rePhase\2014-12-16_Impulses\XLS_03_LP250_24_HP100_48_taps8k_FFT16k_middle_blackman_44k1.wav
0
12.0
12.0
N:\Apps\rePhase\2014-12-16_Impulses\XLS_03_LP250_24_HP100_48_taps8k_FFT16k_middle_blackman_44k1.wav
0
13.0
13.0
N:\Apps\rePhase\2014-12-16_Impulses\XLS_03_LP250_24_HP100_48_taps8k_FFT16k_middle_blackman_44k1.wav
0
14.0
14.0
N:\Apps\rePhase\2014-12-16_Impulses\XLS_03_LP250_24_HP100_48_taps8k_FFT16k_middle_blackman_44k1.wav
0
15.0
15.0

[mattkhan]

I don't understand how that convolution configuration has anything to process. How does an input channel arrive at channel 9? This is what I am missing here, why is your configuration for channel 9+ if the input comes from channel 1 and 2?

[jjazdk]

I don't understand how that convolution configuration has anything to process. How does an input channel arrive at channel 9? This is what I am missing here, why is your configuration for channel 9+ if the input comes from channel 1 and 2?

See, that is what I do not understand either, why does the signal from the right channel arrive at the higher number channels when the convolution is active?! (and there is no parametric equalizer that routes the signal to the higher number channels)

If I disable the convolution, all I have is silence (apart from the actual signal on channel 0 & 1).

This configuration was only to show that the issue exists, even when no other channels are in the script. My usual configuration script is of course very different (using the first 12 channels, apart from channel 6), though it does not work as expected.

[mattkhan]

I will try to reproduce on my system.

Just to check, you mean that if your convolution configuration includes channel 1 then channel 1 still appears on the other channels as well as wherever it is meant to go?

[mwillems]

If I disable the convolver, everything works just fine and I can do whatever routing and filters in the "parametric equalizer" that I feel like without any issues.
One thing worth noticing is that no matter what I do in the "parametric equalizer" I see this issue with the convolver, it doesn't matter if I disable the "parametric equalizer" or if I try to mute each and every channel.

Do you mean to say that if you mute all sixteen channels before the convolution module you still get sound after the convolution?  If so something is very wrong somewhere.  If not, what happens when you manually route sound to the higher channels with PEQ and no convolution?  Does it work perfectly?  Your post suggests it does, but double checking would be handy.

Bracketing that issue out, some more information about your device's output routing would be good; have you checked the focusrite mixer to see what's happening there?  What physical outputs are you using on your device (it only has 8 analog outputs)?  Are you using an ASIO output?  What's your channel offset?

Two of the devs have the same interface as you, so if you've found a bug they should be able to reproduce it, but I'm still unclear on what exactly is happening here.

For your reference, the analyzer is not a good representation of what's actually happening (it's pretty crude)

[mattkhan]

This looks like a bug to me. I will double check my config to be sure but I quickly reproduced as follows;

- configured my RME FF800 to have 16 channels
- setup a convolver config as per the example given
- no other DSP in the chain
- confirmed totalmix routing has no mixing going on
- played a stereo track
- output is on channels 1,2,11,12,13

[jjazdk]

I will try to reproduce on my system.

Just to check, you mean that if your convolution configuration includes channel 1 then channel 1 still appears on the other channels as well as wherever it is meant to go?

Yes, exactly.

Example is here, view from Analyzer and the convolver script including filters for channel 0 & 1. I am playing back a 2-channel music file, and output format is set to 16 channels.



44100 16 16 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
N:\Apps\rePhase\2014-12-16_Impulses\mid-tweet_03_HP250_24_taps8k_FFT16k_middle_blackman_44k1.wav
0
0.0
0.0
N:\Apps\rePhase\2014-12-16_Impulses\mid-tweet_03_HP250_24_taps8k_FFT16k_middle_blackman_44k1.wav
0
1.0
1.0
N:\Apps\rePhase\2014-12-16_Impulses\XLS_03_LP250_24_HP100_48_taps8k_FFT16k_middle_blackman_44k1.wav
0
9.0
9.0
N:\Apps\rePhase\2014-12-16_Impulses\XLS_03_LP250_24_HP100_48_taps8k_FFT16k_middle_blackman_44k1.wav
0
10.0
10.0
N:\Apps\rePhase\2014-12-16_Impulses\XLS_03_LP250_24_HP100_48_taps8k_FFT16k_middle_blackman_44k1.wav
0
11.0
11.0
N:\Apps\rePhase\2014-12-16_Impulses\XLS_03_LP250_24_HP100_48_taps8k_FFT16k_middle_blackman_44k1.wav
0
12.0
12.0
N:\Apps\rePhase\2014-12-16_Impulses\XLS_03_LP250_24_HP100_48_taps8k_FFT16k_middle_blackman_44k1.wav
0
13.0
13.0
N:\Apps\rePhase\2014-12-16_Impulses\XLS_03_LP250_24_HP100_48_taps8k_FFT16k_middle_blackman_44k1.wav
0
14.0
14.0
N:\Apps\rePhase\2014-12-16_Impulses\XLS_03_LP250_24_HP100_48_taps8k_FFT16k_middle_blackman_44k1.wav
0
15.0
15.0

[mattkhan]

I'm quite sure this is a bug

I tried a few other variations, note that input channels in the numbering below are 0 indexed and output are 1 indexed (so 1:1 means input 1 to output 1). I played a stereo track and then switched convolver config so that

- convolver applied to 1:1 and 2:2 = output seen on 1 & 2
- convolver applied to 3:3 and 4:4 = output seen on 1 & 2
- convolver applied to 8:8 and 9:9 = output seen on 1 & 2
- convolver applied to 8:8 and 10:10 = output seen on 1 & 2 & 11
- convolver applied to 8:8 and 10:10 and 15:15 = output seen on 1 & 2 & 11 & 16

[jjazdk]

Do you mean to say that if you mute all sixteen channels before the convolution module you still get sound after the convolution?

Yes, if I mute all sixteen channel with PEQ, I still get sound after the convolution.

If so something is very wrong somewhere.  If not, what happens when you manually route sound to the higher channels with PEQ and no convolution?  Does it work perfectly?  Your post suggests it does, but double checking would be handy.

Yes, if I route the signals with PEQ (which is my preferred method, I only use convolving for actual filters) and don't use convolving, then everything works perfect.

Bracketing that issue out, some more information about your device's output routing would be good; have you checked the focusrite mixer to see what's happening there?  What physical outputs are you using on your device (it only has 8 analog outputs)?  Are you using an ASIO output?  What's your channel offset?

Two of the devs have the same interface as you, so if you've found a bug they should be able to reproduce it, but I'm still unclear on what exactly is happening here.

For your reference, the analyzer is not a good representation of what's actually happening (it's pretty crude)

Focusrite MixControl is only used to route the outputs, the mixing capabilities are not used at all.

The interface is chosen as "Focusrite USB 2.0 Audio Driver [ASIO]" in JRiver, and channel offset is zero.

The Scarlett 18i20 interface has 10 analog outputs, and I use the SPDIF output to an external DAC to get 12 outputs in total. I have verified the issue by routing the higher number channels to the analog outputs.

[jjazdk]

Your observations go hand in hand with mine.

I guess very few people use convolving of more than 10 channels, since I am the first to report the issue :-)

I'm quite sure this is a bug

I tried a few other variations, note that input channels in the numbering below are 0 indexed and output are 1 indexed (so 1:1 means input 1 to output 1). I played a stereo track and then switched convolver config so that

- convolver applied to 1:1 and 2:2 = output seen on 1 & 2
- convolver applied to 3:3 and 4:4 = output seen on 1 & 2
- convolver applied to 8:8 and 9:9 = output seen on 1 & 2
- convolver applied to 8:8 and 10:10 = output seen on 1 & 2 & 11
- convolver applied to 8:8 and 10:10 and 15:15 = output seen on 1 & 2 & 11 & 16

[mattkhan]

I then tried with 24 output channels

if I added a convolver path for channel 23:23 then no extra output appeared
if I added a path for 16:16 then output did appear on channel 16

I then added paths 1 by 1 and found that the rogue outputs stopped appearing once I hit this path

Code: [Select]

// this one creates a rogue output
C:\Users\Matt\Documents\Acourate\Filters\20150406-Aligned\Cor2S48_LR.wav
0
19.0
19.0
// this one does not create an additional output
C:\Users\Matt\Documents\Acourate\Filters\20150406-Aligned\Cor2S48_LR.wav
0
20.0
20.0

[mwillems]

You're quick on the draw mattkhan, and you saved me the trouble of messing with my device routing  

It sounds like you've reproduced the issue; someone should report this as a bug in the build thread.  Both Matt and Hendrik have an 18i20 (or did a little while ago), so they should be able to reproduce it if you give them a step-by-step (not that it's necessarily interface specific, but it helps that they have devices with the requisite number of output channels)

[jjazdk]

Very interesting. Unfortunately my interface (Scarlett 18i20) supports a maximum of 20 output channels, so I am unable to try that on my own setup.

Which reminds me, it is a little annoying that JRiver insists on either 16 or 24 channels, why not 18, 20 and 22... I for one might buy an extra DAC for the ADAT output on the Scarlett, but will only be able to use 16 channels unless JRiver will allow me to choose 20 in the "Output Format".

I then tried with 24 output channels

if I added a convolver path for channel 23:23 then no extra output appeared
if I added a path for 16:16 then output did appear on channel 16

I then added paths 1 by 1 and found that the rogue outputs stopped appearing once I hit this path

Code: [Select]

// this one creates a rogue output
C:\Users\Matt\Documents\Acourate\Filters\20150406-Aligned\Cor2S48_LR.wav
0
19.0
19.0
// this one does not create an additional output
C:\Users\Matt\Documents\Acourate\Filters\20150406-Aligned\Cor2S48_LR.wav
0
20.0
20.0

[mattkhan]

You're quick on the draw mattkhan, and you saved me the trouble of messing with my device routing  

It sounds like you've reproduced the issue; someone should report this as a bug in the build thread.  Both Matt and Hendrik have an 18i20 (or did a little while ago), so they should be able to reproduce it if you give them a step-by-step (not that it's necessarily interface specific, but it helps that they have devices with the requisite number of output channels)

the convolver is dear to my heart, all bugs must be squashed!

I will go and post something on the build thead

[mattkhan]

see http://yabb.jriver.com/interact/index.php?topic=99073.msg688023#msg688023 (correct me if anything looks funky in what I wrote)

[jjazdk]

see http://yabb.jriver.com/interact/index.php?topic=99073.msg688023#msg688023 (correct me if anything looks funky in what I wrote)

Looks good to me, apart from your first routing table, which goes:
C:\Users\Matt\Documents\Acourate\Filters\20150406-Aligned\Cor2S48_LR.wav
0
0.0
1.0
C:\Users\Matt\Documents\Acourate\Filters\20150406-Aligned\Cor2S48_LR.wav
0
1.0
1.0

Instead of:
C:\Users\Matt\Documents\Acourate\Filters\20150406-Aligned\Cor2S48_LR.wav
0
0.0
0.0
C:\Users\Matt\Documents\Acourate\Filters\20150406-Aligned\Cor2S48_LR.wav
0
1.0
1.0


But it should not matter in reproducing the bug.

[mattkhan]

Well spotted, too much copy pasta for me

[jjazdk]

*bump*

Will this bug be fixed for JRiver 21 ?

[jjazdk]

*bump*