[hpsdr] General question - sampling rate vs bits on ADC

[Steve Bunch]

David,

On May 17, 2011, at 9:50 AM, Dr. David Kirkby wrote:

> ***** High Performance Software Defined Radio Discussion List *****
> 
> This is a general question, about SDRs and the use of a suitable A/D.
> 
> If one uses an audio card, sampling rates are limited to about 196 kHz, but 24-bit units are available.

With sound card maximum sampling rates, a mix-down approach is used to receive RF signals.  Mixing is an analog process, and can introduce distortion that can greatly reduce the actual resolution you will get.  But if you do a good job on the analog parts, this approach works well.  Sound cards are built for audio, so they generally have built-in DC blocking and sometimes other filtering, which is ok in some usage, but unwanted in this one.  Some sound cards work dramatically better than others for SDR (you can search for reviews of sound cards for SDR and identify some of the winners and losers, but the usual advice is to first try what you've got, in case it turns out to be good enough for your needs).  BTW,  "16-bit" doesn't really mean 16 *perfect* bits any more than 24-bit means 24 perfect bits, so you won't get quite as much performance from any A/D converter in a radio as you would in theory.

> In contrast, if one uses a faster card (say 100 MHz), the number of bits of the A/D decrease. Typically 14-bit, though there are some 16-bit cards.

Every additional good bit is good for another (up to) 6dB of dynamic range, so you can almost literally "do the math" and figure out what's going to happen in a given RF environment when you hook your A/D to an antenna there.  A/D converter sampling rate and accuracy haven't followed Moore's law like all-digital logic with an unlimited power budget has, but it's gotten to the point now with 16 bit converters where the dynamic range of high-80's dB is "good enough" for most practical amateur use (unless you live near megawatt ERP transmitters).

> What trade-offs does one get for an SDR? The audio card route is cheaper and gives more bits, so should I assume give better dynamic range. In contrast, I can see that with a faster sampling rate, the noise is spread over a wider range, so its lower in a specific bandwidth. But I would have thought the quantisation noise would be higher with the faster cards, as they have less bits.

There are some material advantages to direct RF sampling.  Once you've interfaced the A/D converter to the antenna (matching, protection, filtering, and possibly amplification), all the "hard" signal processing work can be done in high-level programming languages like Verilog and C/C++, and its accuracy is limited by your arithmetic and algorithms, not imperfect and imprecise analog components.  You can simultaneously demodulate multiple RF signals from distant parts of the wider sampled stream.  The receiver performance can be near the textbook level for the A/D converter's ENOB, minus a little to account for analog front-end functions and A/D converter imperfections.  With some cleverness, you can do an FFT of the entire sampling range of the A/D converter as a wideband bandscope.

When sampling with an RF A/D converter, the three most important things determining the quality of your receiver are what analog distortion or noise introduction you do to the signal before it gets to the A/D converter, the Effective Number of Bits of the converter, and the quality of the sampling clock.  When using a sound-card approach, you have those same factors, but the analog processing ahead of the A/D converter is more extensive and therefore adds defect opportunities.

There have been some comments about A/D noise figure recently, so I won't comment on that.

> I'm interesting in demodulation signal with bandwidth of <= 15 kHz, so even the audio cards easily meet the Nyquist criteria of a sampling rate of at least twice the bandwidth.

Yup.  If the analog parts of your down-converter (oscillator, mixer, amplifiers, filters) are good, and the sound card A/D performance is good, the results will be good.  You'd be amazed at how well a SoftRock receiver (q.v.) with a QSD mixer, crystal oscillator, and simple filter can do -- for $20 (when available as a kit, but you can make your own) -- and PowerSDR supports it.

73,
Steve, K9SRB

> -- 
> A: Because it messes up the order in which people normally read text.
> Q: Why is top-posting such a bad thing?
> A: Top-posting.
> Q: What is the most annoying thing in e-mail?
> _______________________________________________
> HPSDR Discussion List
> To post msg: hpsdr at openhpsdr.org
> Subscription help: http://lists.openhpsdr.org/listinfo.cgi/hpsdr-openhpsdr.org
> HPSDR web page: http://openhpsdr.org
> Archives: http://lists.openhpsdr.org/pipermail/hpsdr-openhpsdr.org/


 1305650411.0