rme96 and adat, problems opening for capture when clock-master

rme96 and adat, problems opening for capture when clock-master

[vanDongen/Gilcher]

Hello,
I am using a rme 96/8 card.
I encounter a problem whenever I try to open the adat device, which is the 
second device on the card, for capture. I get an input/output error.
The problem occurs when I switch from spdif to adat (from hw:0,0 to hw:0,1 
in my case) _and_ the clock mode is master. If I set the card to slave mode 
and change the settings on my adda converter, there is no problem.
Playback only will always work.
So f.i. if I start pd, with hw:0,1 as audio device, I have to start once , 
with an error message that it can't open the card for capture. Then I quit, 
and start pd again. 
The second time it works for capture as well as for playback.

Is this a bug in the driver?

regards

Gerard


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

Re: rme96 and adat, problems opening for capture when clock-master

[Martin Langer]

On Mon, Jul 01, 2002 at 12:29:09PM +0000, vanDongen/Gilcher wrote:
> I am using a rme 96/8 card.
> I encounter a problem whenever I try to open the adat device, which is the 
> second device on the card, for capture. I get an input/output error.
> The problem occurs when I switch from spdif to adat (from hw:0,0 to hw:0,1 
> in my case) _and_ the clock mode is master. If I set the card to slave mode 
> and change the settings on my adda converter, there is no problem.
> Playback only will always work.
> So f.i. if I start pd, with hw:0,1 as audio device, I have to start once , 
> with an error message that it can't open the card for capture. Then I quit, 
> and start pd again. 
> The second time it works for capture as well as for playback.
> 
> Is this a bug in the driver?
> 

First a queston for all clockmode users:
Can you set the clock mode to master without specifying a frequency? 

I have no experiences, but I've specified three master modes for the rme32
(for each sample rate). Maybe that's a mistake, but it looks more plausible
for me, than one master mode (like rme96) with an undefined sample rate.

Maybe we need a real experiment here.


martin




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

Re: rme96 and adat, problems opening for capture when clock-master

[Anders Torger]

On Monday 01 July 2002 14.29, you wrote:
> Hello,
> I am using a rme 96/8 card.
> I encounter a problem whenever I try to open the adat device, which
> is the second device on the card, for capture. I get an input/output
> error. The problem occurs when I switch from spdif to adat (from
> hw:0,0 to hw:0,1 in my case) _and_ the clock mode is master. If I set
> the card to slave mode and change the settings on my adda converter,
> there is no problem. Playback only will always work.
> So f.i. if I start pd, with hw:0,1 as audio device, I have to start
> once , with an error message that it can't open the card for capture.
> Then I quit, and start pd again.
> The second time it works for capture as well as for playback.
>
> Is this a bug in the driver?

I'm in a mess of work right now, so I cannot look into this for a 
while, but it may be a bug in the driver. ADAT is the least tested mode 
(by me at least).

/Anders  Torger


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

rme96 and mmap, bug?

[vanDongen-Gilcher]

Hi all,

When I try to use mmap with this card I get the following behaviour:
(BTW, I give an explicit snd_pcm_start)
In stereo-mode (device 0), I can preload the buffer with 2 periods of audio 
data, start and it works well.
In ADAT-mode (device 1), trying to preload with 2 periods, gives a very 
distorted sound. It only works with preloading 3 (or more) periods before 
starting.

When I use regular writei, I don't see this limitation. I only need 2 
periods of audio in ADAT-mode.


Is this a bug of the driver, or a property of the card?


regards

Gerard



-------------------------------------------------------
This SF.net email is sponsored by: Get the new Palm Tungsten T 
handheld. Power & Color in a compact size! 
http://ads.sourceforge.net/cgi-bin/redirect.pl?palm0002en

Re: rme96 and mmap, bug?

[Anders Torger]

On Tuesday 26 November 2002 00.24, you wrote:
> Hi all,
>
> When I try to use mmap with this card I get the following behaviour:
> (BTW, I give an explicit snd_pcm_start)
> In stereo-mode (device 0), I can preload the buffer with 2 periods of
> audio data, start and it works well.
> In ADAT-mode (device 1), trying to preload with 2 periods, gives a
> very distorted sound. It only works with preloading 3 (or more)
> periods before starting.
>
> When I use regular writei, I don't see this limitation. I only need 2
> periods of audio in ADAT-mode.
>
>
> Is this a bug of the driver, or a property of the card?

It was a while I looked at the driver last, and I don't use it in memory 
mapped mode nor ADAT in my daily work, so I really don't know about the 
state of the driver in those parts.

The hardware does not support memory mapped I/O, so it is copied by the 
CPU in the interrupt handler. Thus for that card you gain nothing from 
using memory mapped I/O, it was only implemented for compability.

I shall save this mail into "stuff I should look into", but right now it 
is unfortunately not among my top-ten items on the to do list.

/Anders Torger


-------------------------------------------------------
This SF.net email is sponsored by: Get the new Palm Tungsten T 
handheld. Power & Color in a compact size! 
http://ads.sourceforge.net/cgi-bin/redirect.pl?palm0002en

Re: rme96 and mmap, bug?

[vanDongen-Gilcher]

Out of curiosity,

What makes a card support mmap directly in general? And what would be 
missing?

Gerard



Anders Torger said at "Re: rme96 and mmap, bug?."r[2002/11/27 18:13]


> The hardware does not support memory mapped I/O, so it is copied by the 
> CPU in the interrupt handler. Thus for that card you gain nothing from 
> using memory mapped I/O, it was only implemented for compability.


-------------------------------------------------------
This SF.net email is sponsored by: Get the new Palm Tungsten T 
handheld. Power & Color in a compact size! 
http://ads.sourceforge.net/cgi-bin/redirect.pl?palm0002en

Re: Re: rme96 and mmap, bug?

[Anders Torger]

On Friday 29 November 2002 00.05, vanDongen-Gilcher wrote:
> Out of curiosity,
>
> What makes a card support mmap directly in general? And what would be
> missing?

I'm sure I'll get corrected if I say something wrong here:

If the card itself can transfer its samples from its onboard buffer into 
computer RAM, without CPU involvement (this is called DMA, direct 
memory access). Then one memory maps that memory, and the program can 
process samples directly in that memory area.

For RME96 which has no DMA functionality, the CPU must also copy to/from 
the card's onboard buffer into RAM, and thus there is an extra memory 
copy, and the gain from having memory mapped I/O in terms of reduced 
amount of memory copies disappears (depending on how the driver is 
designed).

There is also some user-space/kernel-space stuff concerning this, but if 
I'm not mistaken, it is not relevant for this discussion.

/Anders Torger


-------------------------------------------------------
This SF.net email is sponsored by: Get the new Palm Tungsten T 
handheld. Power & Color in a compact size! 
http://ads.sourceforge.net/cgi-bin/redirect.pl?palm0002en