[hpsdr] Mercury Noise/Sensitivity

[Richard Stasiak]

Hi Greg

I finally got around to reading this post.  Thanks for the explanation  
of this aspect of the Mercury design.

Last night on teamspeak, Phil mentioned that one of the latest changes  
to Mercury was to incorporate circuitry to allow the preamp to be  
switched in and out. Was this done to increase the strong signal  
handling capability of Mercury ie giving it 17db more overhead?

73

Rick ve3mm

On 19-May-08, at 4:45 AM, Greg - ZL3IX wrote:

> ***** High Performance Software Defined Radio Discussion List *****
>
> Hi Gerd,
>
> Sorry for my late comment on this question, but I have been away on  
> holiday.  I did not see a comment from any other list member, so let  
> me share some thoughts with you.
>
> You are correct that the quantisation step for this ADC is 22uV.   
> There is a formula that states that the RMS quantising noise is  
> equal to the quantising step divided by root 12 (3.5), so in this  
> case the quantising noise would be about 6.6 uV.  This however, is  
> spread over the whole Nyquist bandwidth, or half the sampling rate.   
> To find the noise floor in uV per root Hz, we would therefore divide  
> by about 8000, giving just less than 1 nV/root Hz.
>
> There is a catch to this theory though.  The ADC is not a perfect  
> converter, and other noise sources have to be taken into account.   
> These noise generators are in the front end of the ADC.  Some  
> manufacturers use a parameter called ENOB or Equivalent Number of  
> Bits.  This factor is calculated using the actual observed noise  
> figure of the device, and is always smaller than the resolution of  
> the device.  For example, a 16-bit device may have an ENOB value of  
> 12.5 bits.
>
> The practical noise figure of the LTC2208 is around 29 dB, which  
> means that about 17 dB of gain is needed in front of it to achieve  
> the 12 dB noise figure normally considered adequate for HF  
> operation.  This limits the dynamic range at the top end of course,  
> as strong signals are 17 dB closer to causing an overflow.
>
> ADC's suitable for HF direct conversion have had this kind of noise  
> figure for about 10 years now, little progress having been made in  
> this area.  Any improvement that will be achieved in Rx BDR will  
> need an equivalent improvement in noise figure, or in the ENOB  
> parameter if you prefer to think of it that way.  What has improved  
> quite considerably in the last 10 years, is the Spurious Free  
> Dynamic Range (SFDR).  The LTC2208 is the best of the bunch for  
> SFDR, leading to its choice for the Mercury project.
>
> I hope that goes some way to answering your question.
>
> 73, Greg, ZL3IX
>
> Gerd Loch wrote:
>> ***** High Performance Software Defined Radio Discussion List *****
>>
>> Hi,
>> there is a fundmental calculation for the receive sensitivity in  
>> the Mercury
>> Wiki based on noise figures for the ADC LTC2208. There is no  
>> dependency on
>> the bit resolution of the ADC in the calculation.
>>
>> I wonder which influence the bit resolution has on the sensitivity.  
>> 16 bit
>> resolution of a full scale signal of 1,5V means 22uV. Does this  
>> mean that
>> the minimal discernible voltage is 22uV?
>>
>> What would be the result if the ADC had only i.e. 10 bit resolution?
>>
>> Gerd, DJ8AY
>>
>> _______________________________________________
>> HPSDR Discussion List
>> To post msg: hpsdr at hpsdr.org
>> Subscription help: http://lists.hpsdr.org/listinfo.cgi/hpsdr- 
>> hpsdr.org
>> HPSDR web page: http://hpsdr.org
>> Archives: http://lists.hpsdr.org/pipermail/hpsdr-hpsdr.org/
>>
>>
>
> _______________________________________________
> HPSDR Discussion List
> To post msg: hpsdr at hpsdr.org
> Subscription help: http://lists.hpsdr.org/listinfo.cgi/hpsdr-hpsdr.org
> HPSDR web page: http://hpsdr.org
> Archives: http://lists.hpsdr.org/pipermail/hpsdr-hpsdr.org/