[Elecraft] Understanding the KX3 Block Diagram

[Alan Bloom]

The advantage of using a PC for the SDR engine is massive economy of
scale, which drives prices down.  The processor in a modern PC is many
times more powerful than the processors in a ham transceiver.  On the
other hand, it has to do a lot more to support the bloated operating
system, which, as you note, was not designed for real-time processing.
In real-time applications the effective processing power is only a small
fraction of the theoretical.

There are companies that make real-time operating systems (RTOS) that
are specifically designed for use in "embedded systems" (that is,
systems that use a microprocessor but are not general-purpose
computers).  An RTOS delivers far more computing power with the same
hardware because it is not trying to support word processors, web
browsers, video games, etc. etc. at the same time.

Or you could go all the way and eliminate the operating system entirely.
With a little extra work you can write real-time software from scratch
to squeeze the maximum performance possible from the available hardware
resources.  That's what I did in the P3.  A single inexpensive,
low-power, 40 MIPS processor does all the user interface, hardware
control and digital signal processing, yet it achieves excellent
real-time performance.

Just think if you did the same thing with a 3 GHz, quad-core Pentium
processor!  The possibilities are mind-boggling.

Of course it would also draw as much power as a whole K3 - don't even
talk about a KX3.  But that wouldn't be a problem as long as you plan to
always use it with a line-operated power supply.

Alan N1AL


On Thu, 2012-03-08 at 09:28 -0600, Tony Estep wrote:
> On Wed, Mar 7, 2012 at 10:20 PM, Wayne Burdick <n6kr at elecraft.com> wrote:
> >...This is a substantial departure from other SDRs.....
> ----------------
> Yep, for sure. And the biggest departure from the typical ham SDR is
> the fact that the computer is inside the radio box. There are various
> commercial and non-commercial SDRs in use on the ham bands, but nearly
> all of them rely on the PC to do filtering, detection, and other
> stuff. At first this had a certain irresistible gadget-lover's appeal,
> but when you stop to think about it, it's not surprising that it opens
> up a can with a nearly infinite number of worms.
> 
> The processing inside the PC introduces latency which is aggravated by
> delayed procedure calls, an operating-system process that happens when
> certain interrupts happen at certain times. This can mess up
> performance both on receive and transmit.
> 
> Moreover, the programming involved is complicated. To replicate all
> the functions of a competitive ham transceiver, a lot of code is
> required. When that code has to run in an environment where an unknown
> mix of other resource-hogging code is running, in an operating system
> not designed for real-time application, the complexity multiplies. The
> result can be an endless bug-squashing process, which has to start
> over every time there's a big change in the running environment (e.g.
> a new OS version).
> 
> I sorta think that in the medium term the future of SDRs in ham radio
> will be those like the KX3, wherein the computer is completely devoted
> to the radio, using its own real-time OS, and handing off ancillary
> functions like the pan display to an external client. I don't think
> the KX3 is the last word by any means, because with a bigger CPU and
> different A/D gizmos you could obtain much wider bandwidth and more
> gadgetry. But at the moment, it's the thought-leader in the SDR
> sweepstakes.
> 
> Tony KT0NY