Skip to contentSkip to conversation

Question regarding the alsa's audio latency benchmark

DISCUSSION Ivica Bukvic 2 participants 2 messages Oct 7, 2003

lore
Share & tools
lore
#0 lore raw

Hi all,

I've had an interesting discussion with a professor and a distinguished member of the electroacoustic music community regarding audio latencies which made me realize that I did not understand the issue in its entirety. Hence, I looked around the net in order to educate myself.

I soon stumbled across the following site:

http://old.lwn.net/1999/0916/a/latency.html

Admittedly, it's quite old but that, if anything speaks only in Linux's favor in terms of its pro-audio readiness. At any rate, I was checking out the benchmark data and was wondering as to how did this person/software app get to the 0.73ms buffer fragment that is equal to 128bytes? In other words, what sampling rate was used?

128 bytes in 44100Hz sampling rate = 3ms
128 bytes in 88200Hz sampling rate = 1.45ms
128 bytes in 176400Hz sampling rate = 0.725ms (this one being obviously closest, but at the same time, what kind of hardware supports this sampling rate, especially in 1999 when this test was done?) 128 bytes in 192000Hz sampling rate = 0.3ms

So what gives? It seems like it is some kind of a 176k-ish sampling rate that, AFAIK does not exist.

Furthermore, my question is what app was used to produce those graphs/results and whether these latency tests take into account hardware latencies (i.e. DSP converters, PCI->CPU->PCI->output etc.), in other words, is this latency that is achievable with the following setup:

Input->soundcard->cpu(with some kind of DSP)->soundcard->Output

Your help on this matter is greatly appreciated!

Ivica Ico Bukvic, composer & multimedia sculptor
http://meowing.ccm.uc.edu/~ico

-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf

#1 lore raw
Quoted 5 lines

Admittedly, it's quite old but that, if anything speaks only in Linux's favor in terms of its pro-audio readiness. At any rate, I was checking out the benchmark data and was wondering as to how did this person/software app get to the 0.73ms buffer fragment that is equal to 128bytes? In other words, what sampling rate was used?

its based, IIRC, on the frequency of the RTC interrupt, which is always in power-of-2 Hz.

Quoted 7 lines

Furthermore, my question is what app was used to produce those graphs/results and whether these latency tests take into account hardware latencies (i.e. DSP converters, PCI->CPU->PCI->output etc.), in other words, is this latency that is achievable with the following setup:

Input->soundcard->cpu(with some kind of DSP)->soundcard->Output

It isn't measuring that kind of latency. Benno wrote a custom program to measure the ability to satisfy audio interface requirements. Benno's test program is measuring how rapidly after an audio interface interrupt it is possible to service the card. this is the "bottom line" for latency performance - if you can't service the interrupt quickly enough, nothing else matters. what his test was showing that we can, from user space, service the card, although his test was based on using the RTC, which for some hardware will allow you to get even lower latencies than using the interface's own interrupt. its a hack, though, and i wouldn't encourage people to do it.

--p

-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf