Re: [Discuss-gnuradio] synchronizing sound cards in a cluster (fwd)

[Eugen Leitl]

---------- Forwarded message ----------
Date: Fri, 14 Mar 2003 08:08:13 -0600
From: Gerry Creager N5JXS <address@hidden>
To: Eugen Leitl <address@hidden>
Cc: address@hidden
Subject: Re: [Discuss-gnuradio] synchronizing sound cards in a cluster (fwd)

Except for a reference to a low-cost clock kit which should prove 
useful...  http://www.tapr.org/tapr/html/Ftac2.html

For those of you in the NASA world, this was developed by Tom Clark of 
GSFC fame and history (VLBI program).  I've seen data indicating 5ns 
short term and 15ns long term stability with 5ns bias using SiRF 
receivers...

gerry

Eugen Leitl wrote:
> This is getting too off-topic, so this should be the last post.
> ---------- Forwarded message ----------
> Date: Thu, 13 Mar 2003 18:57:33 -0800 (PST)
> From: Chris Albertson <address@hidden>
> To: Eugen Leitl <address@hidden>, address@hidden
> Subject: Re: [Discuss-gnuradio] synchronizing sound cards in a cluster (fwd)
> 
> 
> This is to long, skip to last paragraph.
> 
> I notice on your sig line you are at JPL.  I'm cross town at
> The Aerospace Corporation.  I think the way they handle this problem
> is to get a time souce and have it broadcast time signals on the
> network. The systems would pick up these UDP packets and mix them into
> the analog samples.  
> 
> You could also use NTP to keep the PC clocks in sync to the "few micro
> second" level.  NTP would work very well, to the microsecond level
> if you could distribute a pulse per second signal to each PC.
...
> My opinion is the sooner a stable clock gets mixed with your data
> the better.  More of their paths will be the same.  but also the harder
> to do. The best an hardest is to mix a time code signal with the analog
> data before it is sampled.  Worst case is to distribute the clock over
> Ethernet, use that to keep you local clock in sync using ntpd and then
> mix the clock with the data using software in the PC to read from the
> sound card and local close alternatly. Botom line is, your sampled data
> has to be mixed with a clock.  
> 
> 
> --- Eugen Leitl <address@hidden> wrote:
> 
>>---------- Forwarded message ----------
>>Date: Thu, 13 Mar 2003 11:56:19 -0800
>>From: Jim Lux <address@hidden>
>>To: address@hidden
>>Subject: synchronizing sound cards in a cluster
>>
>>Anybody have any good ideas on how to synchronize the sampling from 
>>multiple sound cards in a cluster using Ethernet as the interconnect.
>>The 
>>application would grab data from the sound card (notionally at 100 
>>ksamples/second total, for two channels) and do a ton of signal 
>>processing.  At some point in the processing, the streams of data
>>need to 
>>be shared between processors (i.e. to do beamforming), and so, needs
>>to be 
>>time registered.
>>The bandwidth isn't a real challenge here (with, say, 16 processors,
>>that's 
>>only about 32 Mbps total), nor is latency, but synchronization is.
>>
>>One can fairly easily synchronize to a millisecond over Ethernet, but
>>this 
>>application needs sync to, at worst, 1 sample time (20 microseconds) 
>>although order of a microsecond would be nice.
-- 
Gerry Creager -- address@hidden
Network Engineering -- AATLT, Texas A&M University      
Cell: 979.229.5301 Office: 979.458.4020 FAX: 979.847.8578
Page: 979.228.0173
Office: 903A Eller Bldg, TAMU, College Station, TX 77843