Search This Blog

Friday, March 29, 2013

Power Suppy Refactoring

In this previous post I had the chance to present and tear down my old custom-made power supply. It worked flawlessly for about 10 years. I decided to buy a new linear lab power supply because I needed something better with more features such as constant current operation and digital display of both voltage and current.

After a while I realized it was a pitty to leave this old PSU at a corner considering that the transformer was of very good quality. A second unit would certainly come in handy for certain projects requiring dual power sources. Even though I used this power supply for quite a long time in many projects, its design had some flaws. For instance:

  • Electronics undersized for the transformer output power: the transformer could deliver above 8 Amps with 10% voltage drop in the secondary winding, however the circuitry could not regulate much more than 3 amps without entering overcurrent protection (probably the LM338 own overtemperature protection kicking in);
  • Only one of the two secondary windings being used. Both the output power and power used by support electronics would be sourced from the same transformer tap;
  • Poorly layed out board with power rails not being given proper size for the amps rating of the transformer/regulator;
  • Small heatsink for the maximim potential dissipation of the regulator: in a low voltage, high current setting, like for example 1.6 Volts at 5 Amps, taking into account that the input voltage cannot be varied (always the same transformer tap connected to the regulator), given the 24 Volts of input, it translates to a drop of 22.4 Volts. (24-1.6). Multiplying by the 5 Amps of the load, we get about 112 Watts of dissipation on the regulator. Well this is more power than a heatsink such as the one in the picture below is capable of dissipating, so taking into account the LM338 overtemperature protection we can never reach the 5 Amps at 1.6 Volts.
  • Unnecessary filter module on the AC input (extracted from a switching power supply, not needed for linear PSUs);
  • Mains voltage wires in poor contition. Power switch (carrying mains voltage) close to the low voltage DC part;
  • Low quality measurmenent instruments;
  • Cumbersome current shunt for measuring DC Amps;
  • Chassis not looking very solid.

Having identified all these problems I decided that I had to rebuild practically everything from scratch to obtain a proper PSU. Opened an account at Mouser, and started ordering all the parts I would need to complete this rebuild.

In this project the goal is to make the power supply adjustable both in current and voltage. The aim is to deliver up to 8 Amps from 0 to 20 Volts with the new circuitry. 

It will be based on the popular LM723 device, and for the power transistor I will use 4 x 2N3055. The two secondary windings of the transformer will be used, where one will be dedicated to power the LM723, the fan, the voltage and ammeter displays, while another of the windings will be used for the actual output voltage. This prevents  fluctuations in the load current from interfering with the operation of the support components.

I am currently waiting for all the parts to arrive. In the meantime I was able to advance with most of the mechanical aspects. The heatsink I salvaged from a broken LED light. I cut about 12.5 cm from this structure, yelding what I believe it will do a good job moving away the heat from the 2N3055 transistors, the driver transistor and the main bridge rectifier:

Attached to the heatsink is the fan which already existed in the previous version of the power supply.

In intend to have all of the main functions in the front panel, which include the power switch, the digital voltmeter and ammeter, knobs for coarse and fine adjustment of both voltage and current, and the binding posts for the DC output and ground:

The panel itself will be a galvanized metal sheet I salvaged from an old VCR. The basic shape is already cut, it only lacks the holes for the instruments, buttons and knobs:

Mostl likely later posts will be dedicated to advances of this project. After the remaining components arrive and as time allows, I expect to give this constructional project some progress, and perhaps finish it within a weekend or two.

Thursday, March 7, 2013

NOAA APT Satellite - Night time weather pictures!

My first and accidental attempt at receiving images from a weather satellite was surprisingly very succesful. However as I only started capturing the RF signal late in the satellite pass, it was only possible to retrieve a small portion of the transmitted image.

For this second attempt I took ephemeris data to predict a next pass with the satellite passing as overhead as possible. As such I aimed at receiving from the NOAA 18 satellite that did its pass at 04:11 GMT.

I programmed HDSDR to start recording about 1 MHz of RF between approximately 137 and 138 MHz for about 17 minutes, in order to give a confortable margin to cover the satellite pass.

To my surprise (and initial confusion) I found that in the recording there were two distinct signals with the pattern of the APT transmissions, in practically opposite positions of the spectrum portion of the recording:

At first I thought this could be image artifacts of the receiver, but by better looking at the ephemeris I found that NOAA 19 would also pass nearby at the same time, but with a much smaller maximum elevation of 15º.

I decoded the small fragment of signal for this satellite (at about 137.100 MHz), and still managed to obtain this segment:

Next I "offline" tuned to the other signal at 137.915 MHz, recorded the audio, resampled to 11 KHz and fed it to WXtoImg, and obtained a much more interesting result:

Still this picture is about 400 lines shorter than the full transmitted frame. Anyway, given my antenna setup I am still surprised that the results were so good. Besides being a regular vertical antenna, it is mounted in one side of my building, in practice only being exposed to less than half of the sky, the other side being blocked by the building itself. During much of the satellite pass I could not receive a detectable signal, in particular during half of its N-S flight direction.

Interestingly, given the full spectrum recording of the signal, by analysing the doppler shift it becomes very clear what portion of the flight I was able to receive. We know that the frequency increases as the satellite moves towards us, and decreases as it moves away from us. As I could see from the plot a initially flat curve followed by a smooth drop in frequency, it is obvious that I only caught the part of the flight in which the satellite moves away from my location:

So this tells me that at least in theory, if I could raise the antenna above the building in order to get full sky coverage, I should be able to receive a complete transmission (perhaps with a slight ammount of degradation in the edges). And this not to speak of a helical polarization antenna with a reasonably good gain..

Below a few more pictures of the capture using the various filters from WxtoImg:

  • A colorized continent and political borders overlay:

  • Thermal image:

Sunday, March 3, 2013

Receiving weather satellite images with cheap hardware

Before the Internet, anyone willing to independently obtain satellite imagery from the source would need to buy expensive equipment capable of decoding the analog slow scan video images transmitted by weather satellites  such as the NOAA APT ones. Today there are four of these NOAA satellites still operational, the NOAA-18, 17, 15 and 19. All of these are sun-synchronous satellites, which means these orbit the earth at around 800 Km of altitude and cross every latitude at approximately the same mean local solar time for each pass. This kind of orbit is useful because of the consistent illumination (by the sun) of the target upon each pass.

I have recently bought a DVB-T USB stick named "Ezcap". This receiver cost me a little over 10 Euros, shipping included. Not long ago some people realized this device was actually a part of a SDR (Software Defined Radio), where the demodulation and decoding of the TV signal is actually performed by the host PC. With a tunable bandwidth in the range of approximately 22 to 1200 MHz this makes it a very interesting wide band receiver device:

The SDR community took the very useful task of integrating this device to existing SDR tools that work with other hardware devices such as the USRP (Universal Software Radio Periferal).

Currently there are applications such as SDR#, HDSDR, gqrx among others, which are capable of using this USB dongle as a software radio device:

I have personally been experimenting with some of these applications for the reception of several kinds of transmissions. In general I have been very well impressed by being able to receive and decode HAM radio, CB, WFM Stereo radio (88-108 MHz, RDS decoding included), Aviation band, 2-meters HAM (also APRS packets decoding), and also digital DMR transmission (thanks to the DSD tool - see

More recently, while scanning in the 137 MHz band, I found a relatively wideband signal that seemed likely to be from a weather satellite. Initially I was a bit scheptical because I was using a simple vertical antenna optimized for the aviation band (108-137 MHz), that I though it wouldn't be enough to barely detect the carrier signal of one such transmission.

To my surprise, I compared the demodulated signal (WFM demod.) with a sample transmission obtained from the web, and both had identical patterns, such as for example the characteristic 2400 Hz subcarrier. After a bit of fiddling around with the appropriate audio format (which I later realized it had to be resampled to 11 Khz/16 bit), to my surprise I managed to obtain a picture through the WXtoImg application. As I could not capture the complete transmission (only about 3 minutes), and the signal started to fade away at the end, I could not obtain a full picture, but still the result was well above my expectations, given the fact that there was ZERO effort dedicated to prepare the hardware to receive this transmission:

The picture show is the raw data containing the two sensor channels of the satellite. Normally the two wavelengths of these frames can be combined to provide a false color image that is easier for the meorologist to analyse. One such example is the picture below:

With these results, I might think of improving the reception by building a more adequate antenna such as the quadrifilar helix antenna (e.g.