0

I am currently designing a RF receiver for the APT protocol in order to collect NOAA weather images. This has been widely done on the internet, but I'd like to do it from scratch and not use any SDR.

I've come up with a superhet receiver followed by an in-phase and quadrature module. This is then going to be fed to a microcontroller.

enter image description here

The frequencies that I am interested in are 137.620MHz, 137.9125MHz and 137.100MHz with a bandwidth of 38kHz.

My question is: How precise do I need to get my oscillator? Ideally I would have it at those specific frequencies, but there is no oscillator available with these frequencies.

I am an embedded software engineer with some experience in digital electronics and my time working on telecoms at schools start to date.

PS: I have chosen an intermediate frequency of 10.7MHz with a low-pass filter to cutout the image response at around 147MHz.

EDIT: The answer is a fractional-N synthesizer that allows the generation of a precise frequency. I am going to go with the Si5351 IC that incorporates a PLL and a fractional-N synthesizer at low cost.

Updated block diagram

Clément Chaine
  • 33
  • 6
  • https://en.wikipedia.org/wiki/Automatic_picture_transmission – Bruce Abbott Aug 28 '21 at 22:38
  • If you're adept with software, you might take a look at Si5351, instead of the MAX2606 VCO. It is a bit complicated to program (187 internal registers), but can get quite close to those oddball frequencies. It uses a 27 MHz crystal as a frequency reference. Good enough, considering the added Doppler shift of satellite carrier...roughly 50 ppm – glen_geek Aug 28 '21 at 23:04
  • The Si5351 is indeed a good choice here as it abstracts the pll and fractional-n synthesizers in a small and cheap package. – Clément Chaine Aug 29 '21 at 07:50

1 Answers1

0

This was solved long ago using preset programmable fractional N synthesizers with a PLL. Thus 1 to 10 PPM TCVXO chips are preferred and now are programmable to the lowest common denominator and then divide the VCO match to get the carrier LO required to generate 10.7 MHz.

Then you have to have a good V or Yagi steerable antenna, low noise front end, low noise LO with a linear phase IF and then decode 256 levels into binary data from a modulated 2400 Hz signal like old analog modems which used a DSP.

p.s.

The lowest common denominator may be too low and other ratios may be required. The PLL Capture range must include the initial error frequency. You may also need a high Q RF filter in the front end to improve C/N for low level 5W Tx levels. Like a <2 @ 137.5 MHz double-tuned LC filter or maybe a 1/4Wave coax stub tuned to length into a diff. amp with Q = 100 and amplify the series current from a 50 Ohm shunt.

You must define the signal levels and gain and IP3 for carrier/noise ratio and interference ratios then before mixer. Otherwise, you can get blasted by other Ham operators with hundreds of watts vs this weak signal.

Tony Stewart EE75
  • 1
  • 3
  • 54
  • 182
  • 1
    Thanks for your answer. The fractional-N PLL was exactly what I was looking for! – Clément Chaine Aug 29 '21 at 07:37
  • 1
    For the 256 level demodulation, I'll probably get the raw IQ samples from ADCs and work on the demodulation on the DSP. – Clément Chaine Aug 29 '21 at 07:40
  • The lowest common denominator may be too low and other ratios may be required. The PLL Capture range must include the initial error frequency. You may also need a high Q RF filter in the front end to improve C/N for low level 5W Tx levels. Like a 2MHz BW linear phase @ 137.5 MHz or maybe a 1/4W coax stub tuned to length into an amp. – Tony Stewart EE75 Aug 29 '21 at 08:07
  • Sorry for my lack of knowledge on the matter. How do you define the signal levels, gain and IP3 here ? – Clément Chaine Aug 29 '21 at 09:34
  • https://en.wikipedia.org/wiki/Third-order_intercept_point When a strong signal mixes with another , distortion occurs which increases faster (steeper slope) with levels. This tells you the margin to an intercept point of that intermodulation distortion in Rx from strong unwanted signals. – Tony Stewart EE75 Aug 29 '21 at 12:51