[Elecraft] Regarding the K3 and high QRN levels

[Wes Stewart]

I'm sorry, but you can't have it both ways.



"RF" (i-f) gain as implemented in the K3 operates on the post-filter
(8 MHz) i-f amplifier.  The control voltage is derived at least two
(perhaps three) ways.



The post 2nd mixer i-f amplifier (15 KHz) output is detected and if a threshold
is reached, so-called hardware AGC is applied to the post-filter
amplifier.  The manual i-f gain control is derived from an encoder,
processed by the DSP and summed to the same post-filter amplifier.  If I
remember what Lyle told me the last time I saw him, there is also a
contribution to this control voltage from the signal processing DSP.  I
could be mistaken on this last point, but no matter, for the sake of this
discussion it's immaterial. 



So as far as the 8 MHz if amplifier is concerned, it can't tell manual r-f gain
from hardware AGC.  So when you worry about AGC degrading SNR (it does) a
reduction in i-f gain does exactly the same thing!



In fact, as I demonstrated to Lyle over a year ago, before the "R-F gain
calibration" routine, my K3 suffered noticeable SNR degradation even on
S9+50 dB signals, with the slightest engagement of the "r-f" gain
control.  This is because of the wacky gain control characteristics of the
FET(s), which can have Gm variations of 2:1, device-to-device.



I submit that a modern DSP radio with all of the "smarts" this
suggests, should not even need an r-f (or i-f) gain control.  I've used
spectrum analyzers for over 30 years (even before computer control) that slaved
the input attenuator and the i-f gain control to maximize dynamic range without
overloading the front-end mixer. If you uncoupled the controls, the test for
overload was to change the input attenuator 10 dB and see if the display
changed 10 dB.  If it was less than 10 dB then the mixer was in
compression.



If this trivial calculation can't be made in a DSP radio, then I want an analog
receiver back.



To be sure, AGC development isn't trivial, particularly with the latency that
seems to be associated with DSP.  (I'm not a digital guy, so I'm guessing
here)   But even in analog receivers, group delay is an issue with
control loop stability.  In a homebrew received I did in the 70's I had to
pick off i-f for AGC detection before the tail-end-of-the-if-crystal filter to
stabilize the loop.  But with attack time set by i-f and decay time set by
audio, it worked flawlessly.



I know it can be done digitally, if my TS-870 is any measure.  I never touch the manual gain control on that
radio (or my K3 for that matter.)  If a
human has to intervene, there is something wrong with the ACG design.

 

Wes

--- On Fri, 7/9/10, Guy Olinger K2AV <olinger at bellsouth.net> wrote:
AGC is instant and dynamic, and does not vary RF gain, PRE/ATT settings which are static unless the user changes them. RF gain, PRE, and ATT result in numeric "advice" to the CPU, rather than being in direct control of a circuit or device as in an analog receiver.  The CPU in turn drives the actual circuit devices from it's many outputs.  There are many uses for this indirect linking.  


At issue is that the main smart AGC is digital signal processing, and is *AFTER* the analog to digital conversion (ADC).  The RF gain (which is really IF gain), PRE and ATT are before the ADC, and improperly setting those can squeeze the noise into the high numerical range in the conversion or even engage the defensive hardware AGC which is better off not being engaged.  The defensive AGC is analog, has no smarts, and unfortunately must reduce signal-to-noise in doing its job of preventing ADC input overload.


Those who understand how the rig works will throttle back the pre-digital controls to where noise is at most moderately loud to obtain best operation of digital features.  

What I am suggesting is that in an "AUTO" novice mode a slow (rate of human turning the RF gain) throttling back can be set (and then left alone) by the radio for those who are not into the theory and do not recognize the source of their complaints as being caused/worsened by their misadjustment of RF gain, PRE and ATT.  


Leaving PRE on and RF gain at max on all bands regardless causes the digital AGC to increase or reduce the noise to the same level as the wanted signal in pauses.  This reduces the effectiveness of the NR algorithms. 


Those who ride their own RF gains would not be using this.  

73, Guy.