Default Sample rate for Pulse Audio/Jack etc.

[Len Ovens]

There has been an interesting discussion on LAU this past week or so about
48k vs 44.1k sample rates.

I have in the past felt that an audio project meant for CD would be best
to start with 44.1k in and continue as such till CD print. It appears that
is not so for a number of reasons.

- Both 44.1k and 48k support 20 to 20k hz. In order to do so they have to
do some filtering. But in the case of the 44.1 the roll off has to be much
steeper than with the 48k. This done in two places, in firmware on the
audio card and in the analog circuitry as well. In the firmware the cost
is 4X as much in cpu speed, and memory for the 44.1k as for 48k. The
designers cut corners in this area for this reason. So 44.1k audio is not
as good quality wise as it could be. The greater roll off in analog
circuitry also demands more parts and if poorly done affects the in band
frequency response as well. Even in a high quality converter with good
quality filtering, the in band response will be affected to some degree.

- At some point many sound card manufactures started to design with a
single speed ADC/DAC... at (you guessed it) 48k. They already had a dsp in
place for such things as on board synths and such, so they could use that
for resample as well. Some of the resample setups were really bad and did
not even keep stereo channels aligned (some sound blasters a while ago).
The ac97 is designed this way too, But the Intel HDA, while using a 48k
centric bus and clocking, does have the ability to have a codec run at
44.1k. The parts count to do it right is quite high though, so I am not
sure if it is used in practice. In other words some Audio IFs are 48k
already and converting to 44.1k on the fly.

- Some audio hardware is 48k only. I have a built in mic like this, at
44.1k it pops and clicks it's way through. This is not quality HW
obviously, but the manufacture seems to feel 48k is standard.

- 48k has slightly lower latency (96k is much better, but adat IFs loose
half there channels for example and cpu use goes up. Live digital mixing
uses 96k for lower latency)

- Ardour's export to 44.1k wav from 48k project does less sonic damage
than using 44.1k in and less damage than one reverb use. The designer of
one of the better resamplers says that resampling uses the same stuff as
filtering. In other words a resampler is a digital low pass filter which
happens to have a different rate out than in.

- Ubuntu Studio supports video work where 48k is standard.

For all of these reasons I think it makes sense to set Ubuntu Studio's
default sample rate to 48k in Pulse Audio as well as Jack/Qjackctl. If I
had control over the codecs and other hardware, I would certainly suggest
96k and be done with it. I would suggest that our -controls app allow
setting the default higher (or lower) as well.

I am not sure, but I suspect that the Win desktop defaults to 48k and that
is why HW sometimes comes in 48k only. (though the HDA audio system is
designed to handle more than one sample rate at a time in separate streams
on the same device) That being the case, and as Linux and in particular
Ubuntu is designed to run on HW made for win.... maybe 48k should be the
desktop default (Pulse) in Ubuntu vanilla too. Things just might work
better.

Note to self... check how changing default rate in PA and Jack to 48k
affects our current problems with starting jack while pulse is streaming.

-- 
Len Ovens
www.OvenWerks.net