more on DDS
Tue, 11 Jul 2000 17:23:38 -0400
Klaus von der Heide wrote:
> From: "Johan Bodin" <firstname.lastname@example.org>
> To: <email@example.com>
> Subject: SV: LF: What is the best RX for 136 kHz?
> Date sent: Mon, 10 Jul 2000 00:23:35 +0200
> Send reply to: firstname.lastname@example.org
> Johan, SM6LKM wrote
> > At first sight, the DDS may look as an ideal frequency source
> > for LF. This may be true for transmitters but not for receivers!
> > In-band spurious signals are generated due to the limited
> > DAC resolution.
> Yes, that's the reason why I simply divide the 24576 kHz of the
> oscillator of the A/D-converter on my DSP-boards by 64*3 to get
> 128 kHz as an LO for shifting down the 136 kHz to AF.
> The LF-band is so narrow that a DDS is not neccessary.
> > Another option is to make the receiver almost totally digital.
> > The only analog parts would be a preselector filter and a
> > preamplifier.
> That was just my first answer to Petr's question on this subject.
> But amateurs like mixing their signals hence and fourth, and so
> this discussion came up.
> > A suitable ADC for LF is the AD9260. It is a
> > 16-bit oversampling ADC (20 MHz) with Fs/8 decimation
> > (decimation filters are on-chip). Input BW is 1 MHz and
> > output Fs=2.5 MHz. The output from the ADC can be fed into
> > an Intersil (formerly Harris) HSP50016 downconverter chip.
> > The HSP50016 has a quadrature DDS LO (the actual tuning),
> > digital I/Q mixers, programmable decimation, main I/Q filters
> > (FIR) and a serial data output. The output can be fed to a DAC,
> > for direct audio output, or a complex signal can be fed to a DSP
> > for further processing. The RX BW depends on the selected
> > decimation ratio and the filter skirts are STEEP!
> > The HSP50016 has a lot of different output formats, both
> > complex (I/Q) and real (Weaver method).
> I let one of my students design a similar chip, but fully
> programmable with a finite state machine and additionally
> DDS-generators for frequency shifting on board. Using a
> 0.25 micron technology and 5 metal layers the chip covers
> the entire HF-region as input. The programmability enables
> reception on many frequencies at different bandwidths at the
> same time on the same tiny chip (3*3 mm).
> But, such a chip in our case is not neccessary because:
> (1) The LF-band is only 2 kHz in bandwidth. Therefore it is
> possible to undersample at very low rate directly, say 32 kHz,
> presumed an input filter supresses any aliases sufficiently.
> A simple DSP then could analyse the signals without the use
> of an Intersil/Harris-chip. I am searching for an 16 bit-A/D-
> converter without alias filter or with an disable for it.
> Another possibility is to throw away 79 of 80 samples when
> using the AD9260.
> (2) Using sophisticated multirate technique it is possible
> to run any desired bandpass filter on a modern DSP at an
> input rate of up to the high output rate of the AD9260.
> 73 de Klaus, DJ5HG