hello list,

i just bought a hammerfall DIGI96/8 card, and now i'm trying to configure pd to get the lowest possible latency out of it. i first tried with pd-0.36, then with pd-0.37. with both some strange things happened. (<ashamed to admit> using win ME</> on a DELL pentium III 64meg RAM)

pd0.36 seems to change the number of audio buffers i ask for, sometimes lowering them to 1!

the details:

when i try: "pd -asio -listdev -blocksize 32 -audiobuf 4:

pd first says "4 audio buffers", then "bufs 8" in stead

quoting the console: Number of devices = 1 ---------------------------------------------- #1 DefaultInput DefaultOutput Name = ASIO DIGI96 Serie Max Inputs = 16, Max Outputs = 16 Sample Rates = 44100.00, 48000.00, 88200.00, 96000.00, Native Sample Formats = paInt32, ----------------------------------------------

4 audio buffers

not using MIDI input (use 'pd -midiindev 1' to override)
using default input device number: 0
using default output device number: 0

nchan 2, flags 3, bufs 8, framesperbuf 32

as pd seems to prefer 8 buffers: " pd -asio -listdev -blocksize 32 -audiobuf 8"

now the console first says "4 audio buffers", then changes it to "bufs 16" latency (measured with the patch from the doc files) is +- 27ms - this works fine, but i'd expect about 12ms here - (32 * 16 / 44100 = .0116")?

(snipped the beginning as it stays the same..) 4 audio buffers nchan 2, flags 3, bufs 16, framesperbuf 32

another try: "pd -asio -listdev -blocksize 64 -audiobuf 4"

pd doesn't change the params here - latency 16ms (but the audio is fucked up, so i can't use this setting)

4 audio buffers
nchan 2, flags 3, bufs 4, framesperbuf 64

"pd -asio -listdev -blocksize 128 -audiobuf 4"

even weirder - pd sets bufs to 1, than says it's less than recommended..

4 audio buffers
nchan 2, flags 3, bufs 1, framesperbuf 128
warning: number of buffers 1 less than recommended minimum 2

(tried some other settings also, getting similar results as above..)

with 0.37 i first tried "pd -asio -listdev -blocksize 32 -audiobuf 8"

choosing media from the menu: everything seems ok. the delay (i suppose this the same as latency?) field is set to 8 ms, but the actual latency is about 25 ms

next i tried with a higher sampling rate, as according to the hammerfal docs this should allow for lower latency..: "pd -asio -r 96000 -listdev -blocksize 32 -audiobuf 8"

the displayed delay is again 8ms, the actual delay is 17.3ms (without jitter!?) but now metro works at about half speed and the osc~ freq is about two times (probably 2.18 times) too low, so somewhere something goes wrong with the samplerate.. when i set delay to 4 msec in the media window i get a latency of 12 ms. (with no -r 96000 flag this doesn't work, as the audio then is completely fucked up when i try to set delay to 4)

the 12ms seems to be the lowest i can get.. is there any chance i can get lower if i'd try it on linux with ALSA? i need to put real time effects on sharp-attack-no-sustain-instruments such as harpsichord and lute, so i really want the latency to be as low as possible..

thanx for reading this far - now i hope someone can answer my questions..

kind regards,

kristof

kristof.lauwers@logosfoundation.org

http://kristoflauwers.de.vu ____________________________________

...

the 12ms seems to be the lowest i can get.. is there any chance i can get lower if i'd try it on linux with ALSA?

A hammerfall is not a digi96/8 so which do you have ?

it's the digi96/8

Because if it is a 96/8 you can't get much lower, at least not reliably ( well

hm. it's ASIO settings dialog tells me i should be able to get near 5ms (don't remember how much exactly - i'm at another computer now..) anyone getting better results than the 12ms?

greetings,

xof

Hi, with the -pa -blocksize 64 -audiobuf 6 parameters I get something between 10 and 12 ms in/out latency, means about 5 ms one way with the rme9652 on winXP with 0.37test10, which is a great improvement to former versions.

on linux with guenter geiger's oss 32-bit driver its 1,6 one way.

Greetings Michael

How about putting these into the CVS? (yup, I'm the CVS nag.) Then they become part of the distributions that are built, like the guenter's debian pkgs, and now, my MacOS X distro.

.hc

On Sunday, Aug 31, 2003, at 08:33 America/New_York, CK wrote:

I read:

...

Can someone point me to the documentation for the objects in cyclone (if

this is a little out of date, but I hope to catch up soonish:

http://test.pilot.fm/pd/patches/help/

HTH

x

-- chris@lo-res.org Postmodernism is german romanticism with better http://pilot.fm/ special effects. (Jeff Keuss / via ctheory.com)

---

PD-list mailing list PD-list@iem.at http://iem.at/cgi-bin/mailman/listinfo/pd-list