QJack vs ALSA sound quality

[amc252]

Hi everybody,

I have recently been playing with bitKlavier Linux version, which functions very efficient with ALSA.
The other day, in order to put some plugins in line, I used QJack instead.
I have noticed a better sound quality, like a rounder, cleaner piano sound.
Is this possibe or am I kidding myself?
It seems strange that using Jack interface instead of ALSA direct connection to the system or a DAW would make a difference, yet that's the feeling I have and would like to hear other people's opinions.

[d.healey]

Is this possibe or am I kidding myself?

Does it null?

[amc252]

Is this possibe or am I kidding myself?

Does it null?

I'm not sure I understand what you mean.

[elcalen]

I'm not sure I understand what you mean.

Null testing audio means: when you have two recordings that are supposed to be identical, if you play them simultaneously but reverse the phase of one recording, the waves should cancel each other out and all you get is silence—only differences between the two waveforms would be played. Of course to properly test this you need two performances that are actually supposed to be identical, which depending on circumstances may be difficult to obtain.

But as for the original question, there really should not be any difference in quality, I think. In fact, Jack is just an intermediate layer between ALSA and applications to facilitate more advanced routing—so the audio with Jack should still technically be going through ALSA. So I don't see how sound quality could theoretically be improved by Jack (though I admit I'm not very familiar with it). If there is genuinely any difference for you, I don't think it would be due to the technology itself, but rather the particular settings and circumstances on your system. Like, were you using the same samplerates between the two, for instance?

[amc252]

Wow, that's really interesting. Thanks a lot for the detailed explanation.
I'll try to record a MIDI file with and without Jack to see how it goes.
I agree that in theory there should be no difference and that's why I was baffled.

UPDATE:
I ran this test in Audacity.
1) Recorded a MIDI track with ALSA and then with Jack.
2) Copied and aligned the same track to see if I was getting this right. Inverted one of the tracks and indeed got silence.
3) Aligned the ALSA track and the Jack track, inverted one of them. I still get sound.

So I assume the files are indeed different.

[GarryO]

If you look on Qjackctl's Settings --> Advanced tab ( I assume you meant Qjackctl when you wrote Qjack ) you will see there is a Dither option.
If this is set to something other than "none", this might explain the results you are getting.

[amc252]

If you look on Qjackctl's Settings --> Advanced tab ( I assume you meant Qjackctl when you wrote Qjack ) you will see there is a Dither option.
If this is set to something other than "none", this might explain the results you are getting.

Yes, I meant Qjackctl .
Dither is set to none. Also the bitrate is 48,000 both in the VST and Jack.
I haven't really touched anything.
I have uploaded my test files here:
ALSA
https://drive.google.com/file/d/1HDMYHg ... sp=sharing
jack
https://drive.google.com/file/d/13FYxYR ... sp=sharing
Just in case someone wishes to give it a listen.

[jean-emmanuel]

What about playing the same midi track two times with the same setup ? Does the instrument produce the exact same output everytime when the audio engine stays the same ?

[elcalen]

A fundamental problem with nulling and many synthesised instruments (as well as many effects) is that there may be elements of randomness in the sound processing. This means that even when using the exact same midi, two recordings may still vary slightly, and therefore won't null perfectly. Which makes it tough to say for certain whether a difference is due to recording technology, or simply inherent to the instrument.

jean-emmanuel's suggestion to compare two recordings with the same setup could give some insight to whether this is the case here.

[folderol]

As has been said, it is extremely difficult (if not impossible) to get identical results from a synthesiser - indeed, it's hard enough to get the same result for anything within a computer. Small time discrepancies can occur simply due to the machine swapping cores for thermal balancing.

[tramp]

I've read the source code of JackAudioConnectionKit and I found that it use golden plated wires internal and clears the buffers with snake oil. That may explain why it sounds better.

[Impostor]

Louder sounds better. A volume difference of .2dB, while imperceptible as a volume difference, can already color your experience.

[amc252]

All valid insight with which i totally agree.
I did use the same MIDI track and the same VI, simply changing the way i connected it to Audacity. Same bitrate etc.
I also very carefully manually aligned both the same tracks recorded with ALSA and the same tracks recorded with Qjackctl.
In both cases I got perfect silence.
When i did the same with the different recordings, I got music.
I also agree that a very minimal difference in volume could account for both Jack sounding better and the tracks not nulling.
It's weird, which is why I thought to ask.

[alex stone]

I've read the source code of JackAudioConnectionKit and I found that it use golden plated wires internal and clears the buffers with snake oil. That may explain why it sounds better.

And don't forget, Jack goes all the way up to 11......

[folderol]

Just one small point.
If you have a single audio source and destination, ALSA will always perform better. Jack goes through ALSA.