Project Logs

Collapse

• The search for bandpass filters

  Reginald Beardsley • 09/06/2014 at 01:06 • 0 comments

  I've been working on the spectrum analyzer design.  Murata  offers a 1 MHz BP SAW filter centered on 137.5 MHz (FS2059B1).  It claims -40 dB at 7 MHz BW.  Combined w/ a 120 MHz LP filter and an AD8307 log amp on the frontend and a 500 KHz SCAF at the ADC input this would make for a 0-120 MHz double conversion spectrum analyzer.  The STM32F429 PLL would provide the tunable first LO w/ a  137.5 MHz fixed clock for the 2nd LO.  Unfortunately Digikey lists the device, but has no stock and no one else seems to even list it.  I've been looking for other options for a 0.5 - 1 MHz BP at a frequency low enough to generate the clock w/ the STM32, but so far all I've found are at much higher Fc.  That's doable but requires another clock and the high frequency Si507 parts sort of blow my target of staying under $100.  I'm trying to avoid anything that requires test equipment to construct.  I have a failed 40 M DC receiver front end lying around.  It failed because I had no way to measure the passband of the filter  when I built it.

  FWIW The same design minus the AD8307 would make a nifty HF SDR module w/ 1 MHz BW.  When I got my STM32F429 board  also  got a TM4C1294 which has a complete ethernet interface. If I can find a 1 MHz BP filter that I can feed using the Tiva PLL I'll divert course and build an SDR that connects via ethernet and then build the rest of the instrument around that.  So if you know of a cheap 100-120 MHz BP filter module, please let me know.

  
  

  Potential parts options so far:

  120 MHz LP filter   - TDK MEM2012V121RT

  137.5 MHz  1 MHz BP filter  - Murata FS2059B1

  600 KHz LP SCAF - LTC 1069  (also serves as antialias for scope)

  I'm a bit unsure what's realizable w/ a SCAF.  Reading the datasheets doesn't fill me w/ confidence :-(

  BTW If anyone reading this wants to get involved please contact me.  The main reason for doing this here is to find someone to talk to.  The one off cost is drifting towards $100, but the function is increasing significantly.  I'm also investigating adding equivalent time sampling 'scope functionality to 100 MHz.

• Reality is analog. Digital is just a concept

  Reginald Beardsley • 08/22/2014 at 21:33 • 0 comments

  The low pass filters needed to avoid aliasing are proving to be a significant pain.  I've found switched capacitor filters which will work to 400-600 KHz, but not higher.  That leaves the problem of building high order filters w/ limited test equipment.  One needs 6 dB per bit of filter rolloff at Nyquist.  For a 12 bit ADC w/  Fc at 1/2 Nyquist that's a 72 dB/octave (aka 12 pole) filter.   Component tolerances make such things difficult to build.  One option is to aacept less dynamci range. My Rigol is 8 bit so it only needs a 48 dB/octave low pass filter to prevent aliasing.

  On a happier note, I realized that a 20 MHz spectrum analyzer can be implemented using a comparator and some processing.  By collecting sign bit data from the comparator at 80 MS/s and then doing an autocorrelation and an FFT one gets an amplitude spectrum w/ the dynamic range and frequency resolution governed by how long you collect data.

• Lots of fun!

  Reginald Beardsley • 08/10/2014 at 02:45 • 0 comments

  I  burned most of a day figuring out how to make TIM4 function on an STM32F4Discovery board.  I'm a novice at programming at bare hardware level, but finding the information I needed was maddening. "Selects internal clock", which clock?  There are lots of them!  There seems to be good support from STM for C programming in the form of libraries that hide the complexities, but at least for now I'm committed to using Mecrisp.  Which in turn implies I'm coding peripheral operations in assembly.  Scope traces and code in a few days.  I've still got PWM output tol do.  But hopefully I learned something applicable today.

View all 3 project logs