[LAU] strange jackd zalsa_in behavior

[Chris Caudle]

Someone on the Ardour forums reported  very strange zita_a2j behavior recently, where 
the signal level is much higher when using zita_a2j or the zalsa_in internal client than it is 
when using jackd directly.
I suggested that poster describe the problem on the LAU list, but he was still not able to 
get a list join confirmation from attempting to add his address to the list yet.

I have also managed to duplicate the problem, so I will link to the original problem 
description, then describe my setup and observed behavior.

Originally posted as "Zita-a2j is way too hot!" here:
https://discourse.ardour.org/t/zita-a2j-is-way-too-hot/108040/6[1]

My setup:
My setup:
Lexicon USB interface as primary
Focusrite Solo (also USB) as second interface

My typical settings for primary interface:
48k sample rate, 1024 frames/period, 3 periods/buffer

Process to duplicate:
Started Lexicon interface with qjackctl, 48k/1024/3 settings.
Using JK meter to measure levels, K20 scale.
Connected both channels of Lexicon to JK meter, set level so that
average from synth note was around 0, peaking around +10.
Stopped jackd on Lexixon, started jackd on Focusrite, and set level so
that single channel of synth connected to Focusrite input  was
approximately the same.

After stopping jackd on Focusrite, starting jackd on Lexicon again,
verified that left channel was still at same amplitude as previously verified.

Loaded zalsa_in for Focusrite Solo interface with:
jack_load Solo zalsa_in -i "-d hw:USB -r 48000"

Connecting Focusrite input from zalsa_in to JK Meter showed just idle output
noise level at around -5dB, with peaks around +8 dB. Synth note is still
around 0dB on Lexicon channel, but is now at least +18 dB in zalsa channel.

I verified the same behavior when explicitly giving a period size with:
jack_load Solo zalsa_in -i "-d hw:USB -r 48000 -p 512 -n 3"

I see exactly the same behavior when using zita-a2j as the client with:
zita-a2j -d hw:USB -r 48000 -p 512 -n 3

I have the latest released jackd version:
$ jackd --version 
jackdmp version 1.9.21
The jackd build is from the Fedora repository, although they usually do a pretty good job 
not breaking anything in the build.  The original report was using Ubuntu 22.04.  I think 
that release of Ubuntu has jackd 1.9.20, which seems to indicate this is not a distribution 
specific problem.

My previous experience with Fons's software is that it is pretty bullet proof, so I'm 
surprised to see the same strange behavior out of zita-a2j.  

I don't see anything unusual in the way the devices are opened, looks like standard format 
S243LE for the older Lexicon device and S32LE for the newer Focusrite device:

Card 0 (Lambda): 
 * Playback Device 0 (USB Audio): 
   - Subdevice 0 (hw:Lambda,0,0): 
     used by: jackdbus (PID 121752) 
     access: MMAP_INTERLEAVED 
     format: S24_3LE 
     subformat: STD 
     channels: 2 
     rate: 48000 (48000/1) 
     period_size: 1024 
     buffer_size: 3072 

 * Recording Device 0 (USB Audio): 
   - Subdevice 0 (hw:Lambda,0,0): 
     used by: jackdbus (PID 121752) 
     access: MMAP_INTERLEAVED 
     format: S24_3LE 
     subformat: STD 
     channels: 2 
     rate: 48000 (48000/1) 
     period_size: 1024 
     buffer_size: 3072

Card 2 (USB): 
 * Playback Device 0 (USB Audio): 
   - Subdevice 0 (hw:USB,0,0): 
     closed 

 * Recording Device 0 (USB Audio): 
   - Subdevice 0 (hw:USB,0,0): 
     used by: zita-a2j (PID 130423) 
     access: MMAP_INTERLEAVED 
     format: S32_LE 
     subformat: STD 
     channels: 2 
     rate: 48000 (48000/1) 
     period_size: 512 
     buffer_size: 1536

Any thoughts on what is going on, or how to debug?

-- 
Chris Caudle


--------
[1] https://discourse.ardour.org/t/zita-a2j-is-way-too-hot/108040/6
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.linuxaudio.org/archives/linux-audio-user/attachments/20221225/b8669cb3/attachment.html>