RE: [linrad] PC soundcard alternatives

[Leif Åsbrink <leif.asbrink@xxxxxxxxxxxxxxxxxxxx]

Hello Niels,

> 1. 14 Mhz analog IF input from 144 - 14Mhz transverter
> 2. Analog to digital conversion using an AD6644 or similar 
>    sampling at 65Mhz
> 3. Digital down converter (HSP50016, GC4016 etc.) to shift 
>    from 14Mhz down to 0Hz and produce complex (I and Q) data. 
>    Local oscillator in DDC provides frequency coverage e.g. 
>    14.00 - 14.500Mhz. The DDC decimation rate determines the 
>    output bandwidth - for example 100kHz for monitoring and 
>    down to 20Hz for CW.
> 4. Complex (I and Q) data used as input to Linrad - maybe using 
>    USB to move the data from the DDC into the PC.
> 5. DDC control from PC based application.
>
> There seem to be a number of commercial boards available that 
> follow this approach but they are all very expensive. It may 
> be a bit daunting for an amateur to produce the PCBs for this 
> type of system but I don't think it is impossible.

I am waiting for someone to do what you propose here:)

1) Is of no interest (at the moment)

2) I think it is trivial to make the input circuit wideband
so anyone can use the board for their own purpose. You will have
to add a roofing filter at 14 MHz if the selectivity you
provide at 144 MHz is not good enough, but I think that should
be placed outside the design.

You may also route 144 MHz directly to the AD6644. It may produce
equally good performance. The data sheet says the analog bandwidth 
is 200 MHz if I remember correctly. It is a matter of making the 
65 MHz clock really low in phase noise. 

3) I am happy to see you suggest GC4016 rather than AD6620. As far as I
understand you can not get 100 kHz bandwidth with AD6620 because you 
would then have to send a far too high data rate into the PC. The GC4016
can sample up as the last stage so you can series connect the filters
and there are four DDC blocks inside a chip. This way it should be
possible to design a good enough filter to have an output that
does not have an excessive sampling rate.

4) Yes. How to move the digital data stream into Linux is something
I know nothing about. Look for something that can handle a high data
rate. One receive channel will need 40 bits per data pair at sampling
rates below 100 kHz or so. The day I can have a board like this I will
want the largest possible bandwidth up to about 1 MHz. That might call
for 32 bits per complex data point so the need could be up to 50 megabit
per second. And I absolutely need 2 receive channels minimum!!!

5) It will be very desireable to have provision for synchronising the
DDC chips and to run from an external clock. I will want two boards that
run synchronously :) If the input is wideband it will be possible to 
play many games with such hardware. 

Unfortunately I am too ignorant in digital design to go into this
myself. I hope you will find it interesting enough to actually get into.

Maybe you should discuss it with Hans, HB9CBU who builds a receiver 
ADR-100A with AD6644 and AD6620 E-mail: hans.zahnd.sm5bsz.@xxxxxxxxxx 
(remove .my call. before you use this address, I do not want to expose 
Hans to spam.....)

Also Pieter, N4IP and Moe, AE4JY are working with the AD6620, AD6644 
combination. Their project, SDR-14 is a design to use with the PC computer.
They have started selling it. http://www.rfspace.com/ I do not know
what it takes to interface it to Linrad. Device drivers are beyond
my software skils...

Sampling at VHF frequencies is the obvious solution for the future.
The AD6644 should give about -150 dBc/Hz if the sampling clock is 
good and if the NCO dithering is disabled. (A few spurs can easily 
be avoided by a good operator but wideband noise would be everywhere) 

Actually what you propose is the hardware for which I am writing
the Linrad software:)

The WSE converters are special. They provide much better dynamic range
for signals outside the 95 kHz passband but I do not see them as a
good solution for general usage. They are too big and consume too much
power. As far as I know they make by far the best spectrum analyzer money 
can buy if you are happy with the limitations in frequency coverage.

An AD6644 solution should be adequate in most situations. I really
wait eagerly for it to use with Linrad:)


73

Leif / SM5BSZ

 



LINRADDARNIL