Subject: Re: [linux-audio-dev] Measured latency
From: Karl MacMillan (karlmac_AT_peabody.jhu.edu)
Date: Wed Jan 10 2001 - 23:44:26 EET
Erik,
> Thats about right. The A-D and D-A converters probably use sigma-delta
> technology so that would add about 30-60 samples delay at each
> conversion
> stage. Thats a total of 60-120 samples delay just in the converter.
>
This is very good to know - I had no idea that the latency for the converters
could be so high. Do you have any good sources for information about the
latency of converters?
> If you're using a buffer length of 192 samples, you must remember that
> you need a minimum of 2 buffers; one for the input and one for the
> output.
>
Well, this is fragment based io so the read/write is 64 samples. Your right
that the capture will add at least 1 fragment, I think, which would make the
total delay 4 - 6 fragments (256 - 384 samples). The number of capture
fragments used depends on when the app gets woken up by poll(2).
Thanks a lot for you comments,
Karl
> (30 to 60) + 192 + 192 + (30 to 60) = 444 to 504
>
>
> Hope this helps,
> Erik
> --
> +----------------------------------------------------------+
> Erik de Castro Lopo nospam_AT_mega-nerd.com (Yes its valid)
> +----------------------------------------------------------+
> Percussive Maintenance: The fine art of whacking the c**p out
> of an electronic device to get it to work again.
>
This archive was generated by hypermail 2b28 : Thu Jan 11 2001 - 00:32:08 EET