When will WAV files be functional? Right now WAV files skip like a bad CD. I want to be able to do some true A-B comparisons against source material.

Calvin

Hmm, I haven't tried WAVs since one of the earlier alphas.

Does your config.ini have the reducecache= line in it? Are you running Developer or Consumer?

Wav is working fine for me. I uploaded all of my test material so I could calibrate my system and it worked flawlessly. I'm useing 2.0b3

You using Mlord's kernel? Maybe that's the differing factor.

WAV files likely require lots more buffering than .mp3 files, and will be sensitive to available memory on the player unit.

There are three different players, each with different amounts of memory.
So each of you, which player do you have, and do WAV files play smoothly or not?

Hmm, I haven't tried WAVs since one of the earlier alphas.

My .wav files play just fine... but like you, I am using an interim beta that was released to the alpha group after 2.0b3. But it is unlikely that anything concerning playing .wav files would be different.

Has anybody else tried playing .wav files with 2.0b3?

tanstaafl.

I have a 16 Meg player. I've tried playing WAV files with skip using the 2.0b3 developer kernel when I got the terrible skipping. I installed one of mlord's kernels, and I just tried out WAV files and it seems to behave better.

When the playlist has 1 WAV song in it, it seems to play it fine. I've brought up the "vital signs" screen and the free memory is around 490k, dipping to as low as 100k. Free memory doesn't seem to be as much of an issue. The skipping doesn't happen every time now, but here's the way to duplicate it:

open up a playlist with a CD's worth of wavs. (i don't know if this has any relevance but this is what I'm doing). Then when the song is playing, briefly fast forward for 1/2 second and let it continue. This causes the disk to spin up, apparantly the empeg is reading ahead/caching the song and the whole time it is skipping very badly, making popping noises, pausing, etc etc. Really terrible. I don't know what i did before the mlord kernel, but it would skip badly no matter what i did.

Calvin

I don't think so, i haven't explicitly altered my config.ini yet. I also remember when I was running a plain 2.0b3 developer kernel, when the skipping stars, if I paused it and let the disk activity die down, it will stop skipping for 5 or 10 seconds and then start skipping again as soon as it spins up again. (!)

I spent some time with wav files again. Here is another way to cause the skipping. Play a wav, then pause it. reboot the empeg. when it comes back, it will be paused. now unpause it. it will skip like a 2 legged horse...

Calvin

Play a wav, then pause it. reboot the empeg. when it comes back, it will be paused. now unpause it. it will skip like a 2 legged horse...

I tried that several times, heard one tiny hiccup one time, but could not reproduce it. My .WAVs seem to play just fine.

Hmmm...

tanstaafl.

I'm putting out v53 of my kernel enhancements in an hour or so, and AS AN EXPERIMENT (warning to all?) it includes code to turn on multiple sector disk transfers. Which is mumblyspeak for FASTER disk I/O, typically double or better. It would be interesting to see if this helps with .wav files..

I doubt it will double disk transfer speeds; this is more throttled by the I/O cycle time - it's not an ISA bus, and cycles are slower - the fastest ARM-based hardware we have will manage about 6MB/sec from disk into thin air. Multiple sector transfers per IRQ helps a bit, but not a huge amount (in our testing, anyway). The ARM tends to have better IRQ latency than x86 boxes.

The car player's I/O speed configuration is a bit conservative, and can be inched up a bit - however, the player doesn't need this so it's best left alone really...

Hugo

Hugo wrote:
>I doubt it will double disk transfer speeds;
>...
>the fastest ARM-based hardware we have will manage about 6MB/sec

I have not timed anything on the empeg yet (no devel system installed),
but I did just try it on my gateway (SA110-based NetWinder).

The difference between "off" and "8" for multmode was only 5-10% throughput,
hovering around 3.8MByte/sec and 4.2MByte/sec.
Turning on DMA (Netwinder) has a much bigger impact: 10.53MByte/sec.
The Newinder uses a PCI-bus'd SL82c105 IDE controller with BM-DMA capability.

But Empeg is designed to excel at audio, not disk I/O.

{ by the way, thanks for bringing us such a fun toy, Hugo!

Yes, we'd love DMA (especially on this other arm-based hardware). Would come in very handy...

:)

Hugo

I installed mlord's kernel with the multi disk speedup adjustment, and the WAVs no longer skip! Whoo hoo!!

Calvin