[PVRCNC] K3 speech compression

[Guy Olinger K2AV]

Yup. Sort of.  It has numeric digital speech processing which includes
peak and frequency band management.  It is never anywhere actually
clipped.  Wayne calls entirely digital processes by the rough analog
function because he learned instantly that his customers are very slow
to move forward to digital thinking.

Some of the discussion on the Elecraft repeater are legend where some
correspondents simply seem unable to think beyond the analog concepts.
 Customers are strictly expecting the analog behavior, even though
they don't know they are doing that.  The analog concepts are the
"memory pegs" used to remember and generalize how a radio works.

The mic gain and other audio controls are not potentiometers in analog
voice circuits.  They are encoders which report only that the knob was
turned, or voltage dividers on a standard voltage, which deliver the
divided voltage to an analog to digital converter (ADC), which in turn
sends digital advice to the CPU.  This digital numeric advice from all
the knobs on the front panel is used as input to program algorithms in
the processing of the internal number soup. This is why you can have
three DIFFERENT apparent resistance settings on the same "pot", e.g.
CMP PWR MON.  The little LED above the knob reports the saved state of
the control, which in turn determines what the CPU will do with any
numerical advice it receives from the knob.

There is a lot of different front panel knob information digitized and
then multiplexed on a SINGLE lead to the CPU.

Transmitting, voice from line in and from microphone go directly to
ADC's without any massaging.  An analog perfectly formed SSB signal
first emerges in the 15 kHz transmitting IF, completely carrierless,
as NO modulator was used to form the SSB.  Modulation is an analog
process.  Modulation is changing one waveform with a second to derive
a third. Nothing like that going on in a K3.  "Digital modulation" is
an oxymoron.  All aspects of the perfectly carrierless 15 kHz TX IF
SSB signal are in the digital stream driving the TX digital to analog
converter (DAC).  Everything past the DAC out to the antenna jack is
linear frequency conversion and linear amplification.  Looking like a
lot of other

On CW there is a similar complete lack of analog processes to generate
a CW signal in the 15 kHz TX IF. NOTHING is keyed. The direction of
bidirectional circuits is reversed to TX ahead of CW generation. After
a sufficient settling time a digital representation of a perfectly
shaped CW baud is put in the digital stream to the DAC.

You might say that the key line keys the CW, but you would be wrong.
The STATE of the key line is converted to digital advice and sent to
the CPU.  The CPU then knows that key down or up has been called for
and then does its own thing.  Note that the make and break waveshape
of a CW baud is STORED DIGITAL INFORMATION in firmware which is
applied via algorithm to the beginning and end of a created number
stream which turns into perfect CW following the DAC driving the 15
kHz TX IF.

Ain't your daddy's analog radio, for sure, and to understand it
completely, really have to get out of analog thinking.

RF gain on the K3, isn't RF gain.  Only variation of RF gain possible
is by PRE/none/ATT setting.  Variable hardware IF gain is there to
protect RX overranging of the 15 kHz ADC and the down-conversation
devices between IF's.   There IS an honest-to-god analog hardware AGC
that kicks in for very strong signals to keep from over-driving late
IF stages and keep from over-ranging the RX 15 kHz IF ADC.  But IF you
are setting your PRE/none/ATT properly, the hardware AGC is just about
never used.

CPU uses a DAC to generate the control voltage for that variable stage
to mimic AGC when it wants to. The variable IF is managed for
performance reasons.  But most AGC function immitation is done
entirely in DSP.

The S-meter is not a measurement of the voltage on an AGC bus that
controls the gain on all the up front IF circuits.  K3 S-meter is a
presentation of the firmware derived true signal strength, and has no
direct connection to any hardware.

Analog RF clipping is probably the best analog method for controlling
SSB envelope peaking.  But K3 does all that entirely in firmware, and
does not need to restrict itself to digitized versions of analog
methods.  Further, as time goes on, constantly improving digital
processing methods with no analog analogs are leaving the analog
performance in their dust.  Wayne is still working on the K3 firmware,
with nearly endless improvements of one sort or another possible
because of the unique positioning of digital processing in the radio.

An outstanding example of a purely digital method that can't be
replicated in analog is the K3's ability to silence CW key clicks
using the K3 NB and T1-7, T2-7, T3-7 DSP settings, NB IF setting to
off, in conjunction with certain selectivity settings. I have had the
K3 silence 10 over 9 key clicks down to a mushy S2.  That's a huge
advantage when a very loud Italian station with ghastly keying settles
up 400 Hz from your 40m run frequency.  I think that some of them run
clicky on purpose and just hope you will move in self defense and
leave them a clear frequency.  That doesn't work any more. It makes
the received CW sound "overly rounded", but that doesn't impact
readability, just a CW "note" that most people don't prefer.

"Think digital" will buy you a lot with a K3.  You can think of it as
an analog radio and operate it that way, perfectly OK place to start,
but you're going to have to understand its digital nature, and how
it's really put together, to really play it for all it's worth.

73, Guy.

On Fri, Nov 4, 2011 at 10:27 AM, Pete Soper <Pete at soper.us> wrote:
> To clarify what K2AV said last night, the K3 does have "true RF speech
> compression", but as Guy explained, it stems from a radically different
> implementation compared to conventional rigs that don't use DSP.
> Interestingly, another aspect of the K3 software-radio design is that
> it's ALC is splatter-proof: mic gain running way off the end of the ALC
> meter range appears to be without consequences compared to what it would
> create with a more conventional regular radio (unintelligble and six khz
> wide signals).
>
> -Pete
>
> ______________________________________________________________
> PVRCNC mailing list
> Home: http://mailman.qth.net/mailman/listinfo/pvrcnc
> Help: http://mailman.qth.net/mmfaq.htm
> Post: mailto:PVRCNC at mailman.qth.net
>
> This list hosted by: http://www.qsl.net
> Please help support this email list: http://www.qsl.net/donate.html
>