Subject: Re: [linux-audio-dev] syncing soundcards
From: Iain Sandoe (iain_AT_sandoe.co.uk)
Date: Thu Mar 16 2000 - 12:24:20 EST
> has this issue been continued private?
> I can't find the messages you are replying to.
Sorry, not intentionally,
as someone said - this list has an auto-reply-to-poster...
I only missed the cc on one comment... here it is put back together...
> On 16-Mar-2000 Iain Sandoe wrote:
>> what about syncing 2 soundcards (sbawe, sb16) the hard way?
>> I mean if I would remove the crystal from the sb16 and, with
>> some soldering, share the sbawe's crystal as clock-source.
>
> Provided the "crystal" is actually an integrated oscillator (with
> Vcc, GND and a TTL or CMOS logic level output, that is), this should
> work. (Pure crystals depend on some additional components on the
> board, so you have to dig some more to insert an external clock, or to
> extract a usable signal.) You may have to use coax cable and/or some
> buffering to get a fine clock signal to the slave card.
>
> (Note: Of course, the two cards have to use the same oscillator
> frequency in their original shape, or you'll end up with one card
> behaving very strangely... :-)
cool, I'm gonna do some more testing now.
>>> On Sun, 12 Mar 2000, Iain Sandoe wrote:
>>>> You also have to be aware that they will *probably* not end up phase-synced
>>>> (as they would if you had a W-CLK) - because there might be several steps
>>>> between oscillator & sampling - which could be different on the two cards.
>>>> -- of course 2 off the same model, same version, same issue should be OK :)
>>>
>>> Then again, unless the drivers start the two cards at the same time
>>> (within less than a sample period), you'll get an additional phase
>>> error. You probably will get the cards within a sample or two, but
>>> you might have to use RTLinux + hacked driver to get beyond that. It
>>> might be possible, but personally, I don't think it seems to be worth
>>> the effort, unless it's for some system that is to go into mass
>>> production...
>>
>> My comment was really aimed at the sub-sample issues.
>>
>> I would hope (sincerely) that the linux sound implementation would aim to
>> allow multiple cards (with differing latencies ;) to be accommodated
>>
>> I realise that this is quite hard (and therefore fun :)...
>>
>> once (in the big, bad, closed-source commercial world) we had to sync up
>> eight cards - all with on-board dsp... to solve a real problem. For many
>> people several inexpensive cards (although I would hope with W-CLK at least)
>> - is a much better option than one biggie $$$ super-whizzbang.
>>
>> one of the main limitations of the V** system at present is that it doesn't
>> (yet?) allow this.
>>
>> perhaps this is really for PBD/et. al's discussion on stop/start issues
>> related to cards/channels.
>>
>> Iain.
This archive was generated by hypermail 2b28 : Thu Mar 16 2000 - 20:21:10 EST