[hpsdr] MDS and relatet architectures

[jeff millar]

My email connection crashed and I had to forward a copy to the list.

jeff millar wrote:
> I have some comments on this.  But rather than assuming mixers and 
> multiple conversion, the comments apply to the Mercury Board with 
> direct sampling of the HF spectrum.
>
> k3bu at optonline.net wrote:
>> To add to the picture of super front end:
>>
>> 1. In case if there is a problem controlling the gain of RF preamp, 
>> the step and programmable attenuator should be used on the front as 
>> outlined by N1UL. 
>>
>> 2. Use the bandpass filters. They could be shared with TX in the 
>> transceiver arrangement, switchable/selectable when needed.
>>
> Good idea to share filtering with the Tx.
>>
>> 3. Use of tunable band filters on the front and behind of RF preamp. 
>> Hi-Q circuits or variation of Q multiplier/notch. Selectable if needed.
>> They are needed especially on low bands 40/80/160, when very strong 
>> signals within the band are present and one is digging for weak ones 
>> in close proximity.
>>
> Tunable filters will do nothing for strong signals in close 
> proximity.  We have to rely on a combination of very strong signal 
> handling capability in the A/D and the front end buffer/attenuator.  
> On 160/80/40 meters, the band noise rises to the point that the front 
> end to the A/D can has no gain or even negative gain.
>>
>> 4. RF preamp switchable in/out. In case signal levels are sufficient 
>> (high gain antennas) switch it out to improve IP3. Have items 1, 2, 
>> 3, available as needed.
>>
> The A/D needs a low noise buffer amplifier between it and the antenna 
> for ESD and lighting protection.  The buffer need very high IP3, >50 
> dBm if the A/D has 100 dB SFDR.  Then put a digitally programmable 
> attenuator in front of the buffer amplifier.  With the programmable 
> attenuator set to minimum attenuation, the buffer and A/D combination 
> has low enough noise figure to work on dead band 10 meters (and 6M?).  
> With some attenuation set, the IP3 rises to meet worst case low band 
> conditions.  It probably doesn't need more than about 20 dB 
> attenuation with modern A/Ds such as the LTC2208.
>>
>> 5. AGC continuously adjustable - time constant (attack, release) and 
>> gain (DSP derived?).
>>
> Suggest that the receiver have two AGC loops.  First a loop to control 
> the programmable attenuator.  The receiver starts out with minimum 
> attenuation.  When the DSP detects A/D output that approach the clip 
> point, about -2 dBFS, the digital loop insert 6 dB of attenuation and 
> compensates by adding 6 dB of gain in the DSP.  The end to end gain of 
> the programmable attenuator, A/D, and front end DSP remains 
> constant...which simplifies later DSP, AGC, metering, etc.  The first 
> AGC loop simply adjusts the dynamic range band automatically to keep 
> the signals within the optimum range.  The attack time of the AGC 
> occurs in 10s of nanoseconds because as soon as a single sample 
> exceeds -2 dBFS, the attenuator bumps up 6 dB.  The AGC release occurs 
> after about 10 microsec of no signals approaching -8 dBFS.
>
> The use of 6 dB increments in the attenuator enables the digital gain 
> compensation to use a simple shift left by two to restore the net 
> system gain.  Using an attenuator with finer step size reduces IP3 and 
> adds complexity.
>
> The design needs to consider two overload scenarios. 
>
>     1) A single strong signal causes clipping in the A/D.  But that
>     single signal has to exceed 0 dBm on the low bands.
>     2) Multiple strong signals add together creating voltage peaks
>     that clip. 
>
> Case number 2) probably will cause the most instantaneous A/D clip 
> events because thats the way hams operate, pileups, etc.  The peaks 
> occur momentarily and rarely as the phase of many carriers happens to 
> coincide at the A/D.  It's not easy to calculate the statistics, but 
> with 2 carriers spaced 10 KHz, then the voltage peaks occur at the 
> beat frequency, 10 KHz, and for a small fraction of the beat interval, 
> about 10 usec.  As more more strong carriers get added to the the 
> overload scenario, the probability of a peak drops and the interval 
> between overloads gets longer and longer, but the period of the 
> overload remains inversely proportion to the frequency spread of the 
> strong carriers.
>
> Watch out for commercially available digitally switchable 
> attenuators...they have their own problems with strong signal 
> handling.  With the design outlined above, the front end attenuator 
> can use 3 stages of 6 dB each, possibly implemented with discrete 
> components, PIN diodes, etc. for the very high IP3.
>
> The second AGC loop operates at traditional modulation rates.
>>
>> 6. High level mixer with clean injection signal followed by 
>> selectable Xtal filters - steep skirted and/or low ringing banks. 
>> Option to take the wide band or filtered bandwidth for 
>> bandscope/waterfall or filtered signals.
>>
> With a 90-100 dB SFDR A/D converter, HF receivers can't really use 
> mixers and buffer amplifiers.  Partly because it doesn't need them and 
> partly because mixers and buffers with 100 dB SFDR don't practically 
> exist.   The modern 16 bit high speed A/Ds have dynamic range 
> performance that exceeds all but the most extreme buffer design.
>>
>> 7. Wonderful world of I-Q processing and controlling the above.  
>>
>> Unless we have super overload mixer that can tolerate and eliminate 
>> the above, the preamp and filtering use if needed is highly desirable 
>> and should beat anything out there. Right now we have to use 
>> variations of outboard gadgets. Incorporating them into the RX front 
>> end design and be able to control switching and settings as needed by 
>> various antennas and situations would make the ride ultimate and not 
>> available presently in any of the rigs AFAIK.
>>
> Mixers, Roofing filters, and low IF frequencies help present day 
> receivers reach high dynamic range by filtering out at many strong 
> signals as possible before presenting a narrow band to a low dynamic 
> range 2nd IF.  When the 2nd IF moves to DSP, dynamic range limits go 
> away, the whole reason to use mixers goes away, and the reason to use 
> filters changes from steep skirts at IF to preselector at RF.
>
> Present receivers have dynamic range specified at about 20 kHz and 2 
> KHz, with state of the art SFDR around 95 dB and 75 dB respectively.  
> The close in dynamic range measurement shows the limited performance 
> of the 2nd IF within the passband of the roofing filter.  With a 
> Mercury style SDR, the close in dynamic range never changes, the 
> design gets close 100 dB all to way to zero offset. 
>
> The new SDR specification will look more like 130 dB SFDR at 10% off 
> frequency (due to the preselector) and 100 dB SFDR any offset from 0 
> to about 1-3% of frequency (in the passband of the preselector).
>
> jeff, wa1hco
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openhpsdr.org/pipermail/hpsdr-openhpsdr.org/attachments/20060816/a94ecbe7/attachment-0004.htm>