Homebrew interdigital filter for 1420 MHz:  given the large gain required and the very wide bandwidth of low-cost SDR receivers, a narrow bandpass filter is essential.    I wasn't expecting this version in a die-cast box to work very well - but it is OK.  The BW is ~ 20 MHz.  Each of the 6 bolts is locked with the nut and they require careful adjustment! 

The image on the left was made by Phil after several hours of data collection in October 2016.  It shows the peak intensity of our 21 cm signal as the Milky Way passed through the region of sky to which the dish was pointed.  

More detailed view of the feed + LNAs.   So far we have used 2  LNAs at the feed, based on the PGA103+.  These amplifiers are excellent value for their cost,  but we would like an even lower NF in this application.  

Initial Experiments 

This shows the dish, now mounted on a tripod.  

2017 Results

The universe is 99%  Hydrogen. Much of that is ionised and is within stars, but there is a significant amount neutral hydrogen in interstellar space. Thermal jostling and absorption of radiation causes transitions of the single electron into higher energy states which spontaneously decay back to lower energy states. In doing so they release photons in a set range of energies. One of the lower energy transitions emits radiation at the radio frequency of about 1420 MHz. This signal is called the 21 cm line.  The neutral hydrogen in the arms and core of our galaxy provide a weak but detectable signal for amateur reception.  

First HI signal:  April 2016   Note the LHS scale!     Wanted signal is the bump above 1420 MHz.  

Phil's  2m dish previously used for satellite TV, with new feed for 1420 MHz.  Initially the dish was sitting on the ground, which probably didn't help the antenna noise temperature.  

In an effort in improve the system G/T we used the MGA-633P8 LNA from Avago. This device quotes a NF of 0.38 dB at 1400 MHz. (In contrast the PGA-103+ is typically 0.6 dB at 1 GHz and 0.9 dB at 2 GHz.)  Initially I designed a 4 layer KiCad PCB, only to run into trouble stitching vias to ground planes. I then resorted to Gerber files from the manufacturer for their suggested circuit, which we have loaded  (see image on the right). This is designed for 0402 SMDs, which are not easy to handle.   The board has worked and gives improved results (see below) but my loading has proved to be unreliable.  We are about to try a commercial LNA from Mini-Circuits. 

SDR receivers:   we have used RTL dongles and an SDRplay receiver.  The latter gives better results, as expected, but the low cost dongles are OK if used with care.   The standard SDR software packages are not very useful for this application, so we used custom software, mainly written in octave.  

This image clearly shows the variations in  shape, height and location of the H1 spectral peak,  over about 12 hours, with the dish in fixed Az/El.  The two main peaks result from the dish pointing to different parts of the Milky Way. Most of the frequency shift is probably  due to varying relative velocity due to the Earth's rotation.  

The spectrum on the right was generated from about 20 seconds of data, sampled at 6 MHz with an RSP SDR system, using custom C and matlab/octave software.  We obtained very similar results via an RTL dongle.   Hopefully longer recording durations will allow more spectral averaging and better resolution of the spectral shape.  The dish was pointed towards the Southern Cross.  The peak location corresponds to a net H1 velocity of about 20 km/sec  (away).  

​​Receiving the 21 cm Hydrogen line 

Results from late 2016

LHS shows our  H1 spectrum in May 2016,  with the SNR considerably better than the first effort.    Two copies of the spectrum are shown - a positive copy of the right and an inverted spectrum on the left.   In this approach we collected about 4 seconds of signal at Fs = 6 MHz, then shifted centre frequency by 1 MHz and collected another set of samples.  The spectral estimates (via Welch spectral averaging)  are subtracted, giving the two copies.  Ideally the noise floor will be flat.  

Receiving the H1 signal took a bit longer than expected, as care is required with antenna gain, system noise temperature, receiver calibration and spectral averaging.  This article provided the original motivation.  The following pictures give a brief summary of our project to date.