[Elecraft] K3 AGC improvements (MCU rev 4.50 / DSP 2.74)

[Dave Hachadorian]

To my ears, this AGC change is a MAJOR improvement!  It didn't 
take any extensive testing to reach that conclusion.  The 
difference is immediately obvious.

As mentioned previously, with low values of slope and high values 
of threshold, loud signals "blank out" with the current firmware 
release, so the most aggressive combinations can't be used now. 
Threshold 15 and slope 0 seem to work pretty well though.  I am 
listening to 20 meter cw now with two K3's.  Rig 1 has the new 
firmware, with threshold 15 and slope 0.  Rig 2 has threshold 8 
and slope 0 with the current mainstream firmware.

Rig 1 sounds MUCH cleaner, both on the loud W1AW code practice 
signal, and on the 7O6T pileup.  The pileup is not too deep on 
any one frequency, so it's not the ultimate test, but signals do 
sound much cleaner on rig 1.

If more aggressive values of slope and threshold can be 
accommodated without the blanking effect, the improvement will be 
even greater.    Elecraft is definitely on the right track here. 
Thanks for working on this issue!


Dave Hachadorian, K6LL
Yuma, Arizona


-----Original Message----- 
From: Wayne Burdick
Sent: Sunday, May 06, 2012 4:06 PM
To: Elecraft Wayne Burdick - N6KR
Subject: K3 AGC improvements (MCU rev 4.50 / DSP 2.74)

Hi all (everyone is BCC'd on this),

As I mentioned, we've made some improvements to the K3's DSP AGC. 
No
hardware changes are being made this time. This is strictly a 
firmware
change to the MCU and DSPs.

One change will benefit the AGC at all settings: elimination of 
the
interpolation errors in the AGC onset curve. This by itself adds 
some
clarity to signals.

The other change allows the AGC THR parameter to be set to a 
higher
value -- up to 20. The original default (5) put the threshold for 
AGC
onset to around S2 or S3, and the max was 8 (about S4-S5). A 
setting
of about 11 now puts the threshold around S7, and the new max of 
20
puts it around S8-S9. Any signals below the threshold should not
activate DSP AGC at all. (Hardware AGC is not affected by the 
changes;
it still kicks in around S9+20 dB.)

Two interesting side-effects of raising the threshold: (1) noise
blanking may seem to be much more effective than before; (2) QSB 
on
weak signals may be more noticeable.

**********
IMPORTANT:

Please set AGC SLP (slope) to something above 0, for now, as this
appears to cause problems.

Setting AGC THR above 8 will *temporarily disable* auto-notch and
noise reduction. The switches and NR level control will still 
behave
the same, but they will have no effect unless you set the 
threshold
back to 8 or lower. We may make the threshold per-mode in a later
release so that, for example, you could set a high threshold for 
CW
but still have a threshold of 8 or lower for other modes, 
allowing NR
and auto-notch to still work in this case.

The reason autonotch and NR must be disabled for high-threshold 
AGC is
that the dynamic range of signals can be an order of magnitude 
higher.
We'll be experimenting further to see if NR and autonotch can be
modified to still allow their use with high thresholds. But 
generally
speaking, pileup conditions in CW (the principle target) don't 
require
the use of NR, and autonotch doesn't apply to CW.
***********

Keep in mind that it may be difficult to reproduce the conditions
under which in-band IMD has been reported. Most operators have 
claimed
it was caused by "dozens" (maybe even hundreds!) of signals in 
the S5
or lower range, all practically on top of each other. This effect 
has
gotten worse in recent years due to the prevalence of spotting. 
Short
of actually heading for a DX QTH, you may not be able to 
reproduce the
conditions accurately.

But it is true that some operators prefer a higher AGC threshold 
in
general, while allowing AGC to remain turned on. You'll now have 
that
capability with the K3. This has been touted as a difference in 
favor
of some other radios. Not any more :)

The new firmware is attached. As with a beta release, you'll need 
to
unzip (extract) the files and put them somewhere easily located,
usually on the desktop, then tell K3 Utility where to find it. 
You
must load all of the new code, so use the "send all" option.

Make sure after loading that your MCU is at rev 4.50 and both DSP 
and
AUX DSP at 2.74.

Please test the new code in a variety of conditions. If you have 
two
K3s, please put the new firmware on one of them, set them both up
identically, and compare them by switching the antenna between 
the two
radios.

Assuming it passes your own tests, feel free to pass it on to 
anyone
who you think might benefit from it, including those planning
DXpeditions (or even *at* a DXpedition, if they're willing and 
able to
try a new firmware load).

Also, by all means post your observations to the Elecraft 
reflector
(or elsewhere). If all goes well, we can release this as beta 
code
next week.

73,
Wayne
N6KR