[App_rpt-users] usbfobs

[Steve Wright]

On 07/07/16 09:13, Kevin Custer wrote:
> [...]  We have spent thousands of dollars on equipment - a lot of it 
> very high end for reliability.  Take the Dell servers for Node 2135 
> and other major hubs for instance.  These were built prior to the 
> Raspberry Pi 2 and 3. [...]  This used to cost thousands of dollars 
> per location, now it's hundreds.  I have several ACC controllers and 
> link stacks now excess to my needs, so believe me, I know. [....]   
> I've personally deployed dozens of $200 - 300 computers and $75 radio 
> adapters.  It's a hell of a lot cheaper than the alternatives, and 
> what it used to cost.

  Stings a bit doesn't it - looking back at obsolete installed gear, but 
that is the price of being an early adopter/constructor.  :) I've got 
shed-loads of WISP stuff I give away to clubs..


>   Building repeaters and maintaining them is a responsibility that 
> costs money.  app_rpt has cut those costs drastically.  Get over 
> spending money on deploying a system - it's not going to be cheap if 
> you want reliability.  I speak from experience - this isn't just some 
> wild guess. 

  You have done well to form a progressive group and do that, but you 
wouldn't do it with a $4k Andrew 23GHz link, you do it with some $89 
Ubiq link.  Sure you have to babysit the Ubiq, and the Andrews lives 
forever, but that's what it is.  Script up some monitoring for the Ubiq, 
bond some critical links for failover..

  The problem is - the signalling I/O is built into the audio driver, 
mandating and vendor-locking both.  That has never been 
ham-radio-suitable and was the whole reason for an open-source approach 
to begin with.  It was a clever trick to modify the CM119, and 
economical too, but in view of the new hardware(RPI3 etc), it is 
unreliable(USB), difficult(soldering), and slow(3 devices max).

  The I2S bus is demonstrably reliable in consumer gear - unproven in an 
RF environment.  USB is demonstrably UNreliable, in consumer gear AND in 
an RF environment.  I run high-horsepower digimodes, and any USB device 
will reset at least daily.

  If I was in a position to write a voting DSP channel driver (or hack 
the XIPAR driver) for six I2S audio devices on an RPI3 with separate 
signalling, I would go and do that. One box, six channels, voting, site 
I/O - under US$100 - I'd put my name on that..


S