[hpsdr] Question re: RX attenuator

[John Marvin]

  Phil,

I don't know if I can construct a "totally convincing" argument. I'm a 
software engineer who dabbles in hardware with just enough knowledge to 
be dangerous. :)
I originally thought that I was not overloading the ADC on my Hermes, 
but then discovered that it was occasionally "going into the red". 
Enabling the -20 db attenuator eliminated any issues there.  However, 
Looking at the wideband input, my guess is that I would probably be OK 
with just 6db of attenuation. So I originally thought the argument was 
obvious, i.e. I would be unecessarily throwing away 14 db of signal on 
the low end in order to avoid the distortions introduced by overloading 
the ADC. But then I looked a little deeper and noticed that the noise 
figure of the LTC6400-20 is only around 6db, whereas the noise figure of 
the ADC is much higher (Larry Gadallah was claiming something on the 
order of 25-30db). So, given that the difference in noise figures is 
about the same as the attenuation level, theoretically (and I may be 
completely wrong about this), I'm not losing anything near the noise 
floor because anything that gets shifted below the noise floor of the 
LTC6400-20 (which then has a 20db gain) would have been lost by the ADC 
anyway. Note, I do semi-pro recording as one of my other hobbies, and it 
just doesn't feel right to attenuate the signal more than necessary to 
avoid clipping.

What about people who have levels that still overload the ADC with the 
-20 db attenuator enabled? If the cause of the problem is known and 
fixed (i.e. a local high power AM radio station) then the right solution 
would be to insert an appropriate filter. But if the interference is 
intermittent or just happens to show up at the wrong time it might be 
nice to have the option of adding up to another 11 db of attenuation. 
That would lead to loss of useable signal, but depending on the 
situation that may be better than having to live with the distortion 
introduced by overloading the ADC.  It's always nice to have more tools 
available to solve problems.

Anyway, perhaps I haven't convinced you. Maybe someone else can make a 
better argument.  Just out of curiosity, are there any reasons not to do 
this, other than it takes time away from doing something else that may 
be more useful? I'm just wondering because some day I'd like to play 
around with FPGA programming, and this looks like it might be a 
relatively simple first project.  I may never get to it, and I'm not 
likely to get to it anytime in the near future, but just wanted to know 
that if it isn't done by anyone else whether or not the change would be 
accepted if it was done correctly by me.

Thanks,

John
AC0ZG

On 10/28/2012 4:57 AM, Phil Harman wrote:
>
> It would indeed be possible to implement 1dB attenuation steps over a 
> 31dB range for Hermes.  If you can construct a *totally convincing* 
> argument as to why this is necessary then you are in with a good 
> chance of having it implemented in the future :)


 1351492651.0