Software Defined Radio

 I recently became interested in a range of new fangled radios. Of course its been over 30 years since communications receivers and transceivers abandoned normal front panel controls and resorted to using a microprocessor. I suppose I should be able to understand all this because my professional career in electronics began in 1965 when I designed interface hardware for computer systems. Not exactly a microprocessor at this stage because the computer system concerned managed UK Air Defence and comprised some one thousand 7 foot high stove enamelled racks, however, as time progressed I moved on to custom microprocessors which were used to assist air traffic controllers to handle aircraft. This was way back, decades ago, and since then we've seen microprocessors become part of the integral design of every type of electronic equipment. Not surprising then, that radio experimenters have been given SDR (Software Defined Radio) because one can combine radio experimentation with ones home computer. This has really only been possible since processing speeds in home computers became fast enough to handle real-time events.

Below is a picture of my first venture into SDR, the SDRplay

 

 Despite the apparently large physical size above, the box measures roughly 100mm x 80mm x 32mm and is made of plastic. The computer interface is via a USB cable and the aerial via an SMA socket. The thing is said to cover from 10KHz to 2GHz and my experience with it is that it works exceedingly well right across that range. In fact with a simple aerial I can clearly hear ELF signals and a low power microwave beacon in the 23cm amateur band, some 30 miles away. Using my active loop aerial I get a signal strength for BBC Radio 4 on 198KHz of about -22dBm with a noise floor of -100dBm.

At this point I'm reminded of the early days of electronics advertising. The buyer must be aware of facts and figures and realise that these may not be defined in line with accepted standards relating to the application. Take for example loudspeakers. Power output should be quoted in what I call "Industry Standard" terms. I mean power output should be in the same units for all loudspeaker products, but alas this is not so. The customary term used to be "Watts RMS", but somehow this specification became corrupted. One can now find loudspeaker power output specified in numbers which to me are absolutely fraudulent. This practice started happening over many years and now one cannot believe virtually any loudspeaker power output figures.

Switching back to SDR specifications. I would expect any formally presented specifications to adhere to recognised standards and not to standards dreamed up by their advertisers or pseudo-technical staff. Take for example the frequency coverage of an oscilloscope. The upper frequency is often the key parameter in its choice. A 100MHz oscilloscope should mean that it can be used with confidence up to 100MHz, but if you present a signal of 1 volt to the oscilloscope input terminal what will you see on the screen? Of course it depends on how you connect it. Do you connect it to the 50 ohm BNC socket or the handy probe that came with the equipment? It won't surprise many to hear that the input is rarely 50 ohms (many scopes use a 50 ohm BNC socket connected to the scope's high impedance circuitry, but the BNC socket is used merely for convenience) and the probe specification has very little bearing on the scope's specification. The signal frequency is also important because its amplitude indicated on the screen may be nothing like its true amplitude. For example, and this is not always true, a 100MHz scope will display a one volt 100MHz signal as only 0.5 volts. If you use the scope's special probe you may not even see the 100MHz signal at all. All pretty confusing.

Returning to the SDR. I would expect the specified frequency coverage to be expressed in terms of 3dB points. This means that either a 10KHz signal or a 2GHz signal would appear as half its real level. Not only that, I would expect signals between these two limits to appear at the real levels once the rolling-off end effects have gone.

Now to my second purchase, a Lime SDR: first the basic board and then in a custom box.

 

 

 The Lime SDR was originally advertised for purchase a long time before it became commercially available.It's advertised specification looked impressive and compared really well with like SDRs that were listed in their advertising blurb, for the average listener it was head and shoulders better and initial sales figures show that it was indeed deemed attractive. It covers 100KHz to 3.8GHz an unbelievably wide range, in fact it is unbelievable, at least in terms of accepted standards.

Below is a picture of the HF performance of the Lime SDR. You'll notice that the red line, which is the performance figure before one has to modify it. It doesn't even show frequencies below 7MHz and at 7MHz the degradation of input signals is nearer a whopping 70dB rather than the accepted 3dB. In fact the performance at the advertised 100KHz is zero in real terms. Hang on a moment though... the graph has its vertical axis in a unit called dBFS not dBm. dBFS is a new term which is used for digital measurements and the 50% level is -6dBFS. So, in reality what is the degradation of a signal say at 200KHz? With the SDRplay I see Radio4 as -22dBm and with the Lime SDR, modified to the blue line below I see Radio 4 at -64dBm. Roughly speaking in practice a modified** Lime is 42dB down at 200KHz which is nowhere near 3dB. With an unmodified Lime SDR I cannot hear Radio 4 at all.

Even after modifications the figure at 100KHz is 80dB down in dBFS terms. It seems to me we are back into the dark days of misleading advertising. Admittedly the specification makes no mention of 3dB points, but surely in the field of electronics we should expect fair play?

** When customers declared their purchases were faulty, the manufacturer admitted that the product "tuned" the range specified but in fact was optimised for "modern radio"... a term loosely applied to UHF. A simple modification was suggested.. read on.

So, not long after the first production models were received in the hands of radio hams, keen to try it on the HF bands, there were complaints by many of it being faulty, that is, it was decidedly deaf on the bands for which it was purchased. It turned out that the design had been optimised for use in the UHF bands, ...a sort of oversight. The suppliers quickly responded with a simple design change. What was previously a vital part could be just detached from the circuit board. In fact, what was sold to most customers now became a piece of kit which should be readily modified to suit their own requirements. The choice of tiny components were now deemed large enough to be easily removed and, if necessary replaced with alternatives using one's soldering iron. In fact it is said that the tiny parts had been chosen with this in mind. It's a pity that the advertising blurb hadn't made this clear.... Many users will be quite happy to attack their purchase with a soldering iron and make it fit in with what they originally had in mind when they parted with their cash, but I suppose many will be a trifle annoyed. I, for one, will tackle any necessary modifications and report back on progress.

 

 Now, what else can I add? Well, when I bought my Lime I did so with the knowledge that it needed some rework. I was also attracted by the fact that it isn't exactly a receiver, it's a transceiver. It can be made to transmit as well as receive. The thing can develop an RF output signal of 0dBm. Not a lot, but hopefully enough. Whether the design of the output circuit is good enough to produce HF signals of sufficient level remains to be seen. For example, if the output has been tailored to UHF bands, will spurious outputs and harmonics be greater than the desired output RF signal?

Below is a picture of the Lime's RF input and output circuitry. You'll notice perhaps pretty quickly an anomaly. Output TXRF1_A is labelled 30MHz-1.9GHz but the RX inputs all show UHF bands. The values of the components in the circuitry are odd also. Taking the RF input labelled for the low range RXRF1_L. This uses a transformer type TC1-1-13MA whose details are given below and further down is the transformer used for transmit on the higher bands, but the other parts, such as the main capacitors are 510pF and 470pF whereas in the transmit circuit are all 100nF. Clearly the response of the receiver and transmitter output will be determined largely by these parts. In fact, once Lime's technical people had been made aware of the lack of HF receiver sensitivity the parts were revised. For example MN14 a tiny 8.2nH choke was shunting HF signals. It's value at 1MHz and 30MHz ranges from 0.05 to 1.5 ohm. Assuming a 50 ohm input, this little choke will result in a 30dB attenuation of aerial signals at 1MHz and at 100KHz this worsens to 40dB. So the choke was removed.

When I was experimenting with my upconverter (see later) I made some measurements inside the Lime case. It was then that I realised something was not quite as it seemed. If you compare the circuit diagram below with the transformer specs, shown below the Lime drawing, you'll see that the circuit diagram is somewhat misleading. Take for example "T5". This is shown as a primary winding (pins 6 and 4) and an isolated centre-tapped secondary widing (pins 1,2 and 3). In fact, in practical terms the primary and secondary should be rotated through 90 degrees. The primary (actually pin 6 and 1) connects C16 to MN33 and the secondary (actually pin 4 and 3) connects ground to MN37. The centre-tap connection, pin 2 shown wired to ground, doesn't exist except as an unused pad on the tiny transformer base.

All the transformers are, in fact Baluns converting balalanced to unbalanced connections. The tiny transformers are not designed to work at low frequencies, hence one of the reasons the Lime is deaf in the HF amateur bands.

The SDRplay device provides a signal strength for our local long wave transmitter of something like -22dBm at my location with my active loop aerial but only -64dBm with the Lime (and that's with MN14 removed).

 
 

  What exactly can you do with a Software Defined Receiver? The main difference you'll see if you use a decent software package is a visual display of what's going on. Audio will be produced at your computer's loudspeakers and, if you tune to a TV broadcast, you'll even see pictures. Virtually any type of modulation can be decoded. Software makes an SDR into a basic spectrum analyser so you have a "panadaptor". You see, not only a view of the station to which you're listening, but a display of adjacent stations also. You'll also see something a little disconcerting, that is interference. It isn't surprising to see interference because a computer includes in its design a rather large number of oscillators, both narrow band using crystals, and rather wobbly wideband oscillators used for things like generating DC power. The further away from the computer and its peripherals the better, but of course you still need a USB lead and this must be kept short. You need a decent aerial and this has to be sited as far away as practically possible from the nearest mains wiring, ethernet cabling and telephone wires.
 I did mention the fact that PC controlled SDRs were only possible once real-time processing speeds became the norm. Don't bother running an SDR unless you have a newish computer. My i5 computer is just about OK but when I tried to use an SDR on my workshop computer I had no end of trouble. The processor is an E2200, an Intel dual core D type thing and it's just not man enough. It can manage a receiver bandwidth of about 1MHz but anything greater than this and you get stuttering or audio breakup. You also get computer lock-ups and really annoying side effects of this. In fact, even with a fast computer the odd lock-up can be very troublesome, often causing a blue screen of death.

See the new computer for SDR reception

 

 I finished making the workshop computer and started to make tests on the Lime SDR. This example has had its "shorting" coil removed from the receiver "L" input modifying this from its 700MHz to 900MHz to something starting lower in frequency... but exactly what?

Well, the performance is pretty poor. Certainly, compared with the SDR Play, it's rubbish. One major problem is that although it can certainly receive signals at low frequencies, if one tunes say to 300KHz and injects a signal of 100mV at this frequency, besides the fundamental you'll see a string of harmonics (see table below). Why an input of 100mV ?... because the receiver is very deaf at 300KHz.. that's why.

 

 FREQUENCY in KHz

 AMPLITUDE in dBm

 300

 -50

 600

 -60

 900

 -85

 1200

 -90

 1500

 -100

 From the above table you can see that reception at lower frequencies, say the medium waveband, will be marred by spurious responses, and as all the action is taking place in the Lime front end an aerial filter won't do much if anything to suppress these types of spurii.
 I had another idea... perhaps because of the view by some that the Lime's circuitry inside the custom chips is entirely unsuitable for the HF band and lower. Although modifying the external circuitry almost certainly would improve matters, another idea is to use the Lime fronted by an upconverter. This general method has been used before by radio amateurs where, for example a perfectly good MF receiver made an excellent IF amplifier was preceeded by a crystal controlled converter and of course double suprerhets and triple superhets usually have superior perforance to single and double types. So, the next project is to make a mixer driven by a crystal oscillator and RF amplifier. I'll make a double balanced diode mixer using schottky diodes, a 50MHz xtal oscillator and a simple single transistor amplifier. I'll see how this performs on a test bench then after adding or modifying the circuit use it to front-end the Lime SDR. Thus 0-30MHz will translate to the band 50 to 80MHz.

 See the upconverter I'm making

See some pictures of received signals

 After playing around with my two receivers I began to notice problems with them. Going back to the first examples in radio design might shed some light on these problems. The first receivers were TRF or "Tuned Radio Frequency" types. These sets had lots of amplification at the frequency of the station being received and this demanded, at least in the earliest days of sensitive receivers, two or more valves, each having tuned circuits. These circuits used a coil and a variable condenser which were used to peak the received station. Two valves meant two tuned circuits and the pair needed to be kept in alignment for optimum reception. Some really early sets had three tuned circuits which meant the operator needed to align all three for best results. See the Polle Royal.. as an example. Later, when valves such as the new screened grid tetrode were introduced, sufficient gain enabled the number of stages of signal amplification could be reduced which eased the problem of aligning the tuned circuits. Improvements in the manufacture of coils helped also because a twin gang tuning condenser enabled the user to listen to lots of stations with the confidence that his set was performing well across the whole waveband. Roughly at the same time as these improvements to the TRF sets were being made, an alternative design the superhetrodyne receiver was being introduced. This type of set which had been invented many years earlier but hadn't gained many followers solved the problem of aligning RF amplifier stages. It achieved this by converting whatever signal was required to an intermediate frequency (known as the IF) where it was amplified. Two advantages presented themselves.. first the IF being fixed didn't need to be tuned and therefore several stages could be used without these requiring twiddling by the operator. The second advantage was that the IF was chosen to be well within the acceptable frequency range of low cost amplifying valves. In fact a sensitive superhet receiver might only require a single tuned circuit at signal frequency and a single IF amplifier. The superhet however had a major disadvantage, not that this really much bothered a listener. The problem was due to the mixing process which was required to produce the IF. Say the IF was 500KHz. To tune a broadcast at 200KHz required a local oscillator to run at 500KHz plus 200KHz or 700KHz. One could also achieve the same result by mixing 500KHz with 300KHz because the mixer simultaneously adds and subtracts the incoming signals. In a typical receiver the choice of frequency for the local oscillator is made by the availability of suitable components, essentially the tuning condenser, so for long wave reception it was much better to have the local oscillator higher in frequency than the incoming broadcasts. In fact it was well nigh impossible for the local oscillator to be lower in frequency in a typical design. Take then the example above. The local oscillator running at 700KHz mixes with our 200KHz broadcast resulting in an IF of 500KHz. Now comes a problem. The mixing process is essentially a non-linear process meaning that harmonics or multiples of incoming signals will be generated within the mixer. The strongest output from the mixer will be a 500KHz signal derived from tunig the RF amplifier to 200KHz. The local oscillator of 300KHz will also produce at the same 500KHz IF a signal at 800KHz. This is called the image frequency. There may be nothing being broadcast at 800KHz and all will be well, however if a there's a broadcast at say 805KHz then detuning our receiver by 5KHz to 305KHz will result in us hearing the broadcast at 805KHz against a dial reading of 205KHz. In other words a listener tuning his receiver would hear two adjacent broadcasts, one at 205KHz and the other at 200KHz. Admittedly the tuning condenser for the RF input coil will be set to 205KHz for the image and it will therefore be miles out of tune, but if this broadcast at 805KHz is a very strong local station it will be strong enough to produce interference to our station on 200KHz. What's the solution?
 The first thing a designer will do is to carefully consider exactly what IF to use, especially if there are strong broadcasts in the band including our typical choice of IF. Simply put, in my simplified example, 500KHz was the international shipping distress frequency and any receiver operated near a port might receive interference because of this. Frequencies such as 465KHz were chosen because not much was broadcast around that frequency. Now back to problems with a typical mixer. A local oscillator running at 300KHz would produce harmonics at 600KHz, 900KHz and 1200KHz etc. Each of these will mix to produce our 500KHz IF. So our receiver will be open for reception at 600KHz-500KHz = 100KHz, 600KHz + 500KHz = 1100KHz, 900KHz - 500KHz = 400KHz, 900KHz + 500KHz = 1400KHz, 1200KHz - 500KHz = 700KHz and 1200KHz + 500KHz = 1700KHz. In fact our superhet receiver is not looking quite as good as we first thought. In the noisy electrical environment of 2017 where broadband noise from chopper power supplies and computers is present at high RF levels a quiet broadcast band turns into a carcophony of spurious signals. The solution, especially in terms of professional listeners was to use a double superhet receiver or even a triple superhet receiver where additional filtering is available to reduce spurious signal reception. Another solution is to employ an IF which is very much higher than the received signal. At least this places images where its very much easier to filter them out.
 I'm getting away from the SDR but, having described typical receiver design problems, it will be more meaningful to look at SDR problems. All SDR designs use complex chips carrying most of the circuitry required for a receiver, however in order to minimise external parts clever methods are employed within these complex chips. The clever methods are great for many applications but are pretty useless for a listener that's been brought up on classic receivers. One such method is to use what used to be known as "Autodyne" reception. Many moons ago clever entrepreneurs used autodyne techniques to turn out exceedingly cheap receivers. One such receiver was for VHF FM reception and was fine once you got used to its idiosynchrasies. However, use an autodyne receiver for lower frequencies and all sorts of problems emerge. Another name for this receiver is a zero-IF receiver. The local oscillator runs at the signal frequency. This means that for our broadcast at 200KHz the local oscillator will run at 200KHz. The IF will be zero and our amplifier will be effectively running at audio frequencies. The mixer will produce harmonics of the local oscillator at 400KHz, 600KHz and 800KHz etc. Looking at the potential signal responses of our receiver, these will be 200, 400, 600 and 800KHz etc. There's another problem also.. because of the absence of filtering within the complex chip any strong signal at our aerial will also produce harmonics of itself within the mixer. These will also be 400, 600 and 800KHz. Looking at a spectrum which is wide enough you'll see signals at 200, 400, 600 and 800KHz all of which will carry the same modulation as our broadcast station on 200KHz. Strong signals at basically any frequency will have the same result and the spurious responses will be generated not only by addition, but by subtraction as well. In fact all the signals generated within the mixer will interact and generated mases and masses of spurious signals.
 Back to the complex chip in the SDR. Because there are lots and lots of different modulation methods a new technique was desirable in order to be able to resolve these. This technique is to use not one but two mixers, one fed with the local oscillator signal and the second with the same local oscillator signal but shifted by precisely ninety degrees. The mixer outputs are known as "I" and "Q" signals, standing for "In-phase" and "Quadrature". Without delving into the niceties of these signals and their processing, there are problems arising from using this technique. The most significant is any DC voltage in the I and Q outputs will be amplified and cause decoding or processing problems. To get around this one could AC-couple the mixers to the following circuitry but the capacitors required must cope with audio frequencies and will therefore be large enough in value to present difficulties to the complex chip designer. Another method is available. That is to use feedback to automatically neutralise the DC. Again, because of the low frequencies involved this neutralising method presents problems in terms of response time. Additional problems are met if strong signals other than the one you want, or from outside the band displayed get into the mixers. For example you may see the effect of modulation from an out of band signal changing any displayed signals or even the level of the noise baseline. The most objectionable problem encountered with a typical SDR using a zero-IF will have what I'd refer to as an "artifact". Typically, right in the middle of a display you'll see a spike or a bulge which is affected by the receiver gain. The spike or bulge is a no-go area for any specific signal you'd like to resolve because it will produce distortion on the recovered audio of a broadcast. So, to resolve a clean signal you may have to ensure that it's not right in the centre of the display. Sometimes the spike is quite tiny and looks insignificant but sometimes it's very pronounced (click to see an example).

 An SDR is good if you want to see a specific signal or look across a small bandwidth, but once you open up the bandwidth you'll see an enormous number of radio signals. These comprise of course real transmissions on their actual frequencies, but in addition a whole raft of other signals. You'll see patterns of signals with strong peaks spread at regular intervals, originating from electrical products using chopper power supplies such as low energy electric lamps, computer monitors, TV sets, clock radios, your computer together with its backup UPS if you have such a thing. There are also SDR mixer products, various harmonics from crystals used in your SDR, your computer and anything nearby using a microprocessor. As you reduce the bandwidth most of these spurious signals will vanish but the base noise level will still be affected. You'll also see artefacts from software and if you're using a USB lead this will provide a pathway into your computer carrying interference from this.
 The answer is to minimise unwanted signals. A good start is to use a bandpass filter between your aerial and your SDR. This will provide a degree of protection. Back in the 1920s and 1930s such filters were very desirable and bearing in mind many SDRs have no filtering at their aerial input, I should say a good bandpass filter is essential. Your aerial is also important too. A resonant aerial chosen for your frequencies of interest will reduce undesirable signals and lower the background noise. A long wire aerial is jolly useful for general purpose listening but is absolutely the worst type of aerial because any strong signals will get into the SDR and produce huge numbers of spurious signals as well as modulating just about everything else.

Return to Reception