[Asterisk-Users] Problems with TDM400P card

[Rich Adamson]

> > Thanks for the info.
> > What hard drives are you using ide or serial ata. Does it make a
> > difference. Thanks
> 
> There have been some references recently regarding disk drive types
> relating to tdm400 noise problems.
> 
> Has anyone established there is a correlation between drive hardware
> and noise?
> 
> If this is the case, it may be indicative of marginal interrupt timng
> performance
> on that hardware.
> 
> FWIW the system described earlier has a single sata drive attached to
> the Intel 925XCV mb on-board controller, from dmesg:
> 
> atapci1: <Intel ICH6 SATA150 controller>
> 
> The drive is a Seagate ST3250823:
> 
> ad4: 238475MB <ST3250823AS/3.01>

One of the easiest ways to determine whether any disk drive is
impacting audio quality is evaluate the system in a no-load
environment. (eg, process a single call with nothing else going
on in the system including no swapping.)

Then compare the audio to the same test repeated while generating
large amounts of disk activity. (If I recall, 'hdparm -t' generates
lots of disk activity.)

Tests conducted this weekend (but incomplete right now) suggest the
OS is doing something that impacts the TDM card specifically. Not
sure what that is as yet, but likely to have something to do with
the pci bus and/or interrupt handling.

Seems the TDM card implementation (at least in the RHv9 distro) is
not being serviced in reasonable timeframes. I modified the zttest.c
app to display the length of time required to receive 8192 bytes
of data from the card. In all cases tested thus far, the 8192 bytes
are received in about 1.021000 seconds (21000 microseconds to late).
That would suggest the data arriving from a TDM card will miss a
frame of data roughly every ten frames. That has a serious impact
on trying to run things like spandsp, but less of an impact on pure
audio.

The tests on this single system indicate that playing with the 
pci latency values had zero impact on the TDM timing. Also, suggestions
involving 'udma2' on the drive had zero impact. That only confirms
that if there isn't any disk activity, those parameters would have
no audio impact.

To help identify the source of the delays, I built a new system this
weekend from scratch. When that is complete, I'll use it to compare
the differences in motherboards, OS distro's, and maybe kernel versions.

Rich