[Elecraft] Re: K1 AGC time constant [and how to radically improve K1 AGC]

[Tom Althoff]

I think what Ron and I were shooting for was not a really fast AGC but more
of a "Not-so-gosh-darned slow" AGC.

Wayne...doesn't 9 to 15 seconds for the S-meter to drop seems excessively
long to you?  Especially for a QSK rig.

When I changed C31 from 2.2 to .22 the hang time seemed to approximate that
of the "slow" AGC setting of the K2.

The pop exists on strong signals with both the stock and the modified K1 but
did not seem any worse after the mod.

Is there ANY chance that a decimal point was moved between the design specs
and production runs so that C31 is supposed to be a .22 instead of a 2.2
ufd?

Those of use who see 9 seconds or more for the sensitivity to recover really
like the 1 to 3 second recovery time after the cap values are changed.
That's one great thing about Elecraft kits...you can make them perform just
as "badly" as personal preference merits.

I love my K1.   And if I hadn't of changed the AGC time constant I would
still love it.   I just love it a litte bit more now. 8-)

Tom K2TA

----- Original Message ----- 
From: "wayne burdick" <n6kr at elecraft.com>
To: "Elecraft Reflector" <Elecraft at mailman.qth.net>
Cc: "Ralph Tyrrell" <w1tf at yahoo.com>; "Tom Althoff" <althoff at verizon.net>
Sent: Wednesday, April 05, 2006 4:05 PM
Subject: Re: K1 AGC time constant [and how to radically improve K1 AGC]


> As the principle designer of the K1, I just love these AGC circuitry
> discussions, and thought I'd better put in my 2.5 cents.
>
> The K1's present AGC circuit, being AF-derived, is a compromise between
> attack time and recovery time. This topic has been the subject of
> scholarly efforts by Haward, DeMaw, and others, so I'll just summarize
> by saying that the slow rise time of an AF-derived signal limits how
> quickly you can respond to a large signal using a simple diode
> detector. Nearly all low-cost QRP transceivers that have AGC at all use
> this technique, including all of the ones I've designed (KX1, Sierra,
> NC40, SST).
>
> DSP can be used to emulate a faster response using various techniques,
> including post-processing of the signal as it propagates through the
> DSP's pipeline. But assuming one wants faster AGC without having to
> write DSP code, a relatively simple hardware-based improvement is
> possible. Actually implementing it is left to the reader, and here's
> what you need to know.
>
> If you look closely at the K2 Control board schematic, you'll see that
> the K2's fast AGC is obtained using an "auxiliary" I.F. of around 150
> kHz. This is 100 times higher in frequency than the audio signal that
> the K1 uses to drive its AGC detector, eliminating the rise-time
> problem. The same technique could be used in the K1. You could start
> with the K2's AGC circuit (mixer, amplifier, and detector), perhaps
> breadboarding it on a proto board. You could pick off the 4.915 MHz
> I.F. signal from the output of the K1's crystal filter, routing this to
> another '602/'612 that has its oscillator running at 5.068 MHz (a
> common crystal frequency). As in the K2, you'd then need to amplify and
> detect just the 150-kHz difference product coming out of the mixer.
> Various circuit simplifications may be possible relative to the K2's
> circuit, which also includes manual RF gain control, T/R swiching, and
> AGC on/off control.
>
> An important subtlety: optimal results might require gain-controlling
> the auxiliary I.F. mixer at pin 2, using the same derived AGC signal
> that drives pin 2 of the K1's RX mixer and product detector. This would
> ensure that the aux I.F. mixer's gain is scaled downward at the same
> rate as the product detector as signal voltage goes up. This was not
> necessary in the K2 case, because the I.F. amp (MC1350) is the only
> gain-controlled stage, and it is ahead of the auxiliary I.F. mixer.
>
> If anyone tries this and succeeds in creating very fast AGC for the K1,
> we'd be happy to publish it as an application note on the web site. Who
> knows? It might even make a nice little option module. The entire
> circuit, if done using SMD parts, could fit on a board placed
> underneath the K1's RF board.
>
> I'd do it myself, but I'm a bit busy  :)
>
> 73,
> Wayne
> N6KR
>
>
> ---
>
> http://www.elecraft.com
>