[Elecraft] Receiving and demodulation

[Brian Lloyd]

On May 13, 2007, at 11:03 PM, Don Wilhelm wrote:

> Brian,
>
> The 1st IF (hardware) AGC is highly dependent on the width of the  
> 'roofing filter'.  We do not have the relevant figures for the  
> relative dynamic range of the 8.215 kHz IF vs. the 15 kHz DSP  
> portion of the receiver.
>
> When those specifics are available, your question can be answered.   
> For now, I only assume that a narrow filter will be the best  
> solution for AGC 'pumping' when operating digital modes.
>
> The capability to demodulate PSK31 and RTTY signals and the AGC  
> 'pumping' situation are entirely two different animals.

In what we have come to see as a traditional receiving system I would  
completely agree with you. As we migrate to newer and more efficient  
modulation schemes I think we are going to find a need to couple them  
together in order to extract maximum performance.

With CW and SSB we have very simple modulation and demodulation  
schemes. We can use a relatively simple envelope detector to generate  
AGC. It gets a bit more difficult with more complex modulation  
schemes. OTOH, all we are trying to do is to ensure that the peak  
signal level does not exceed the range of our A:D. Perhaps a simple  
envelope peak detector will work for that too.

> One thing is for certain, the AGC cannot be totally controlled by  
> the DSP section,

Oh? That seems odd to me. It strikes me that the desired signal and  
only the desired signal should have an effect on AGC. Since the final  
bandwidth determination and demodulation take place in the DSP stage,  
the AGC needs to be derived at that point. That implies that the DSP  
stage needs to generate the AGC.

Hmm, no, we need the AGC to be derived from the signal arriving at  
the A:D and not post bandwidth limiting in order to reduce gain and  
prevent clipping at the A:D. So it still needs to be derived in the  
DSP after the A:D but before we do the final bandwidth limiting.

OK, here is the scenario I have in mind. I am going to use made-up  
numbers for convenience and these are not to imply that they are or  
should be specific numbers for the K3. I am going to copy a  
relatively weak PSK31 signal. He is going to be about 100Hz wide and  
high enough out of the noise for good copy but no higher. I have a  
500Hz roofing filter. My desired signal is toward the edge of the  
roofing filter's passband. A QSO between two very strong PSK31  
stations is going on 300Hz away but still inside the passband of my  
roofing filter. I have used the DSP to limit the final bandwidth to  
include only my desired signal. I cannot "hear" the undesired QSO but  
the receiver still has to deal with it as it is arriving at the A:D.  
If the AGC is entirely derived at analog before the A:D the undesired  
signals are going to modulate my AGC and hence the gain for the  
desired signal thus changing the S:N or Eb/No. If the AGC is derived  
after my final filtering in the DSP, there is a possibility that the  
strong signals will drive the A:D into clipping.

Back in the golden days of tube-type radios (my first receiver was a  
Hammarlund HQ-129x and my first transmitters were an ARC-5 and a 10w  
Central Electronics Multi-Phase exciter -- I was a poor kid in junior  
high and I couldn't afford anything better) I would solve the problem  
by turning off AGC and riding the RF gain myself to keep both signals  
in the "sweet spot" (dynamic range) of the receiver. It would be low  
enough that the strong signal would not overload the receiver and  
high enough that the desired signal was far enough out of the noise  
to be useful copy. I was, in effect, an intelligent AGC. (Well, in my  
case "intelligent" might be pushing it a bit. :-)

So getting back to the point, as I understand it, the reason for a  
very high IP3 and BDR is to prevent strong signals that do not affect  
the AGC from inducing compression in earlier stages or from exceeding  
the range of the A:D. I understand that, in order to keep the A:D  
from clipping and causing the whole receiver to go to hell, we have  
to provide some AGC on everything in the passband of the roofing  
filter. OTOH, we also need some awareness of the state of our desired  
signal in order to prevent undesired changes of AGC from making our  
desired signal uncopyable.

In a perfect world my demodulator would look at my desired signal and  
adjust the AGC to the highest point that would give me an acceptable  
Eb/No for a low bit error rate. Increasing the gain to get a higher  
Eb/No doesn't get me anything but it means that undesired signals  
will have more chance to clip in the A:D. So I see a new, smart  
receiver looking at what is in the analog passband, looking at my  
desired signal, and then "ride the gain" to do the right thing. If it  
is not possible to meet both requirements, i.e. sufficient Eb/No  
without A:D clipping, it could even give me a light on the front  
panel that says something like, "I'm sorry Dave but I don't think I  
can do that." (Apologies to Arthur C. Clarke.) I then know that I  
have exceeded the limitations of my receiver.

> so the real answer depends on the actual dynamic range that the A-D  
> conversion can handle - we don't have that answer yet.

I understand. Still, I think a systems approach and understanding the  
possibilities is reasonable. I look at the topology of the K3  
receiver and I know that what I am talking about is possible given  
the existing hardware. It just depends on how much more processing  
you want to do in the DSP to achieve it. As more demodulators get  
moved to the DSP I think this might become mandatory.

>
> 73,
> Don W3FPR

73 de Brian, WB6RQN
Brian Lloyd - brian HYPHEN wb6rqn AT lloyd DOT com