What’s All This SDR Stuff, Anyhow?

The Software Defined Radio (SDR) has become very popular in the radio hobby scene over the last few years. Many hobbyists own one, certainly most have heard of them. But what is an SDR, and why might you want one, over a traditional radio?

First, a very brief explanation of how the traditional superhetrodyne radio works. This is the type of radio you have, if you don’t have an SDR (and you don’t have a crystal radio).

Here’s a block diagram of a typical superhetrodyne receiver:

superhetrodyne block diagram

The antenna is connected to a RF amplifier, which amplifies the very weak signals picked up by the antenna. Some high end radios put bandpass filters between the antenna and RF amplifier, to block strong out of band signals which could cause mixing products and images.

Next, the signals are passed to a mixer, which also gets fed a single frequency from the local oscillator. A mixer is a non linear device that causes sum and difference frequencies to be produced. I won’t go into the theory of exactly how it works. The local oscillator frequency is controlled by the tuning knob on the radio. It is offset by a fixed amount from the displayed frequency. That amount is called the IF frequency. For example, the IF of an radio may be 455 kHz. Time for an example…

Say you’re tuned to 6925 kHz. The local oscillator generates a frequency of 6470 kHz, which is 455 kHz below 6925 kHz. The mixer mixes the 6470 kHz signal with the incoming RF from the antenna. So the RF from a station transmitting on 6925 kHz gets mixed with 6470 kHz, producing a sum (6925+6470=13395 kHz) and difference (6925-6470=455 kHz) signal. The IF Filter after the mixer only passes frequencies around 455 kHz, it blocks others. So only the difference frequencies of interest, from the 6925 kHz station, get paseed. This signal is then amplified again, fed to a demodulator to convert the RF into audio frequencies, and fed to an audio amplifier, and then the speaker. The IF filter is what sets the selectivity of the radio, the bandwidth. Some radios have multiple IF filters that can be switched in, say for wide audio (maybe 6 Khz), and narrow (maybe 2.7 kHz). Perhaps even a very narrow (500 Hz) filter for CW.

This is a very basic example. Most higher end HF radios actually have several IF stages, with two or three being most common. The Icom R-71A, a fairly high end radio for its time (the 1980s) had four IF stages. Additional IF stages allow for better filtering of the signal, since it is not possible to build real physical filters with arbitrary capabilities. There’s a limit to how much filtering you can do at each stage.

Now, onto the SDR. I’ll be describing a Direct Digital Sampling (DDS) style SDR. The other style is the Quadrature Sampling Detector (QSD), such as the “SoftRock” SDR. The QSD SDR typically mixes the incoming RF to baseband, where it is then fed to the computer via a sound card interface for processing. The main advantage of the QSD SDR is price, it is a lot cheaper due to fewer components. The sacrifice is performance and features. You can’t get more than about 192 kHz bandwidth with a sound card, and you suffer from signal degradation caused by the sound card hardware. Some try to compensate for this by buying high end sound card interfaces, but at that point you’re approaching the price point of a DDS SDR in total hardware cost anyway.

Here is a block diagram of the SDR-IQ, courtesy of RF Space, you can click on it to see an enlarged image.
sdr-iq block diagram

The RF input (from the antenna) goes in at the left end, much of the front end is the same as a traditional radio. There’s an attenuator, protection against transients/static, and switchable bandpass filters and an amplifier. Finally the RF is fed into an A/D converter clocked at 66.666 MHz. An A/D (Analog to Digital) Converter is a device that continuously measures a voltage, and sends those readings to software for processing. Think of it as a voltmeter. The RF signals are lots of sine waves, all jumbled together. At a very fast rate, over 66 million times per second in this case, the A/D converter is measuring the voltage on the antenna. You’ve got similar A/D converters on the sound card input to your computer. The difference is that a sound card samples at a much lower rate, typically 44.1 kHz. So the A/D in an SDR is sampling about a thousand times faster. It is not too much of a stretch to say that the front end of an SDR is very similar to sticking an antenna into your sound card input. In fact, for many years now, longwave radio enthusiasts have used sound cards, especially those that can sample at higher rates such as 192 kHz, as SDRs, for monitoring VLF signals.

The output of the A/D converter, which at this point is not RF but rather a sequence of voltage readings, is fed to the AD6620, which is where the actual DSP (Digital Signal Processing) is done. The AD6620 is a dedicated chip for this purpose. Other SDRs, such as the netSDR, use a device called a FPGA (Field Programmable Gate Array), which, as the name implies, can be programmed for different uses. It has a huge number of digital logic gates, flip flops, and other devices, which can be interconnected as required. You just need to download new programming instructions. The AD6620 or FPGA does the part of the “software” part of the SDR, the other part being done in your computer.

The DSP portion of the SDR (which is software) does the mixing, filtering, and demodulation that is done in analog hardware in a traditional radio. If you looked at a block diagram of the DSP functions, they would be basically the same as in a traditional radio. The big advantage is that you can change the various parameters on the fly, such as IF filter width and shape, AGC constants, etc. Automatic notch filters become possible, identifying and rejecting interference. You can also realize tight filters that are essentially impossible with actual hardware. With analog circuitry, you introduce noise, distortion, and signal loss with each successive stage. With DSP, once you’ve digitized your input signal, you can perform as many operations as you wish, and they are all “perfect”. You’re only limited by the processing power of your DSP hardware.

Since it is not possible feed a 66 MHz sampled signal into a computer (and the computer may not have the processing power to handle it), the SDR software filters out a portion of the 0-30 MHz that is picked up by the A/D by mixing and filtering, and sends a reduced bandwidth signal to the computer. Often this is in the 50 to 200 kHz range, although more recent SDRs allow wider bandwidths. The netSDR, for example, supports a 1.6 MHz bandwidth.

With a 200 kHz bandwidth, the SDR could send sampled RF to the computer representing 6800 to 7000 kHz. Then additional DSP software in the computer can further process this information, filtering out and demodulating one particular radio station. Some software allows multiple stations to be demodulated at the same time. For example, the Spectravue software by RF Space allows two frequencies to be demodulated at the same time, one fed to the left channel of the sound card, and one to the right. So you could listen to 6925 and 6955 kHz at the same time.

Another obvious benefit of an SDR is that you can view a real time waterfall display of an entire band. Below is a waterfall of 43 meters at 2200 UTC (click on it to enlarge):
43 meter band waterfall

You can see all of the stations operating at one glance. If a station goes on the air, you can spot it within seconds.

Finally, an SDR allows you to record the sampled RF to disk files. You can then play it back. Rather than just recording a single frequency, as you can with a traditional radio, you can record an entire band. You can then go back and demodulate any signals you wish to. I’ll often record 6800 to 7000 kHz overnight, then go back to look for any broadcasts of interest.

For brevity, I avoided going into the details of exactly how the DSP software works, that may be the topic of a future post.

And yes, I borrowed the “What’s All this… Stuff, Anyhow” title from the late great Bob Pease, an engineer at National Semiconductor, who wrote a fabulous series of columns under that title at EDN magazine for many years.

How many watts do you need?

Let’s say you’re a ham radio operator, or even a (gasp!) pirate radio broadcaster. How many watts of transmitter power do you need to reach your target(s)? Well, if you’re the typical ham, the answer is easy – just crank up the transmitter RF output knob to max. If you’re the typical pirate, you may do the same, although you’re a little more cognizant of the risks involved. Higher power is more likely to cause RFI issues with the neighbors’ TV, and possibly get you some unwanted attention from the FCC.

The alternative is to run low power. In ham lingo, this is called QRP. Most transmitters let you adjust your power level, so you can just dial it down. But to what level? How low can you go? What you’re trying to accomplish is to be heard by your listener(s). That is, the received signal is large enough to overcome noise levels, both from other signals and static, as well as receiver noise. The latter is a concern at VHF/UHF frequencies, but essentially a non-issue for HF, where atmospheric noise always dominates.

The signal to noise ratio (SNR) is defined as the ratio between the signal and noise levels, and is usually expressed in decibels (dB). 0 dB means the ratio is 1, the signal and noise power levels are the same. a 10 dB SNR means the signal power is 10 times the noise power, 20 dB means the signal is 100 times (it is a log based scale). These are for power values, for voltage ratios the SNR is twice the power value. A SNR of 0 dB would just be barely detectable, in practice you need a few dBs for even a weak signal, and a SNR of 30 or 40 dB is considered an excellent quality signal.

Noise levels vary tremendously, of course. Atmospheric noise varies with the frequency (higher at lower frequencies) and time of day (higher at night, when static from distant thunderstorms is more easily propagated). Then there are the potential man made sources of noise, such as other stations, as well as unintentional noise from the multitude of TVs, computers, switching power supplies, and so on, which have all contributed to a rise in the noise floor over the years.

There are many software tools to estimate received signal levels, based on transmitter power levels and propagation conditions, such as DX Toolbox. Plug in the numbers, and you can get an estimate of the received signal level. It might even be close – there are a lot of factors to consider, and many of them are unknowns, or at least estimates, such as solar effects on propagation.

Another way is to actually measure the received signal level. The good news is that most shortwave receivers have an s-meter, to tell you how strong a signal is. The bad news is that most of the time, the s-meter is wrong.

First, there is no concrete definition of how an s-meter should work. The ARRL suggestion is that an S9 signal is 50 microvolts at the antenna input, and that each S unit represents a 6 dB change in input voltage and power (that is, the voltage doubles, meaning the power level is 4 times higher). Tests on common receivers and transceivers show about a 1 to 5 dB per S unit change. That is, each increase in indicated signal level on the s-meter actually represents a smaller change in received power level, as compared to the theoretical 6 dB/S unit standard.

All other things equal, a change in transmitter power level causes a corresponding change in received power level. So if you double your transmitter power, the received signal will also double. According to the ARRL standard, increasing the transmitter power by a factor of 4 would add one S unit, in practice with most receivers it would add several S units, depending on what the original received signal level was. Again, the s-meter is just an indicator, the actual received signal level is what is important. Doubling the transmitter power will increase the signal, and SNR, by 3 dB. Likewise, cutting it in half will reduce the SNR by 3 dB.

So, let’s assume we have a transmitter running at 150 W (a pretty reasonable value for a good old fashioned tube rig like a Johnson Viking II). And let’s assume that the received s-meter reading is S9 dB, a very good signal, and it’s nighttime with a noise level, as indicated on the s-meter, is S4.

Here’s the Icom IC-730 S-meter sensitivity values from the previous link I gave:
S1 - 2 1.4 dB
S2 - 3 1.3 dB
S3 - 4 1.6 dB
S4 - 5 2.3 dB
S5 - 6 1.8 dB
S6 - 7 3.2 dB
S7 - 8 3.1 dB
S8 - 9 4.0 dB
S9 - S9+10dB 5.6 dB
S9+10dB - S9+20dB 7.3 dB
S9+20dB - S9+30dB 6.6 dB
S9+30dB - S9+40dB 10.5 dB
S9+40dB - S9+50dB 11.3 dB
S9+50dB - S9+60dB 13.5 dB

Ok, so let’s see what happens as we reduce the transmitter power. Each time we cut it in half, we reduce the received signal by 3 dB. Reducing it to a quarter would be 6 dB, an eight would be 9 dB, and a tenth would be 10 dB. Got it?

Looking at the chart, going from S8 to S9 is a 4 dB change. That would correspond with reducing the transmitter power by a factor of 0.40, or down to 60 watts. Going from S8 to S7 is 3.1 dB, a power reduction to 0.49, or 29.4 watts. S7 to S6 is 3.2 dB, a factor of 0.48, or down to 14.1 watts. S6 to S5 is a factor of 1.8 dB, 0.66, or 9.3 watts. And finally S5 to S4 is 2.3 dB, a ratio of 0.59, or down to 5.5 watts.

So what does all this mean? Well, if we dropped our transmitter power from 150 watts to 5.5 watts, the received signal would drop from S9 to S4. We stopped there because the noise levels were S4. At this point, the signal is barely audible. At 9.3 watts, pretty close to the magic 10 watts that most grenade type transmitters put out, the received signal is S5, one S unit above the S4 noise, 2.3 dB above, so an SNR of 2.3 dB. Something you could listen to, but it would really be down in the noise.

What about the original 150 watts that produced an S9 signal? Well, let’s just add up our dBs. 2.3 + 1.8 + 3.2 + 3.1 + 4.0 = 14.4 dB. So in this case, the SNR is 14.4 dB. Not the 20 or 30 dB you’d expect from say the BBC, but certainly pleasant enough to listen to.

Obviously this is just one example. With different assumptions, especially noise levels, the results will be different. Much lower noise levels would allow weaker transmissions to be heard. If the noise was S1 instead of S4, that’s 4.3 dB of SNR right there. Likewise, higher noise levels intuitively imply more transmitter power is necessary. But I think these are reasonable assumptions for nighttime noise levels on 43 meters, and typical pirate transmitter power levels.

The numbers speak for themselves. The difference between the received SNR for a 150 watt and 10 watt transmitter is huge. Of course, as the difference between getting the knock and not is also huge. Assuming transmitter power levels have an influence on FCC enforcement activity…

A 30 mW unlicensed CW beacon was busted last year:
Last summer the F.C.C. DFed the Echo beacon that was on 11002 Khz. It had been running 30mW.
The FCC agent was kind and considerate and dropped further investigation. In fact he was respectful of the fact that it was
completely Homebrew and will under 100mW. There was no complaint
and the only one who cared about the beacon was the DF site in Maryland. The little beacon was disconnected and dismantled.
The operators pirate beacon days are over.

20 to 30 mW is extremely lower power, even for CW. This tells us two things: first, the FCC has very good ears, and can pick up weak signals. If they can pick up a 20 mW beacon, they can easily pick up your 10 watt grenade. Second, and more importantly, they are probably more concerned with the frequency the unlicensed station is using, than the number of watts. If you look through recent busts of pirates, you will see that they are mostly due to choice of frequency, as well as the unusual bust of Weather Radio, which appears to have been motivated by their use of the National Weather Service’s DECtalk speech synthesizer voice.

This is not to say that transmitter power plays no role. But it may not be the FCC’s primary enforcement trigger. The FCC is Complaint Driven. A scan through their Enforcement Bureau confirms this. Busts are mostly for FM pirates, likely based on complaints from licensed stations, as well as for other offending transmissions, such as those that interfere with cellular phone service.

Busts of HF pirates would also likely be due to complaints from licensed services, especially the military, which does use parts of 43 meters. Those BLEEP BLEEP BLAPPP sounds you hear are the TADIL-A/Link 11 system. I can imagine that 70s pop music or cut and paste audio loops interfering with them don’t go over well with the men in uniform. One call is all it takes for the offending pirate to suddenly be #1 on the FCC’s enforcement list.

Back in the 1990s, The infamous pirate Voice of the Night, operated by Lad, was QRMing a Havana/Moscow CW net on 7415 kHz. The operators could often be heard sending strings if FUFUFU in CW in response. Apparently it also annoyed certain radio monitors employed by No Such Agency, as Lad was quickly busted.

So worry about how many watts you’re sending into the ether, but also worry about your choice of frequency.