Here, we explore concepts of digital signal transmission. We'll see different digital modulation schemes like Frequency-Shift Keying (FSK), Phase-Shift Keying (PSK) and Quadrature Amplitude Modulation (QAM), and explain the difference between bit rate (bit/s) and Baud (Bd).
Let's start with Morse code and “transform” it until we can use it as a basic digital system.
Say we (arbitrarily) represented dots by 0s, and dashes by 1s, the letter “A” (•−) would then be represented as “01”.
The first thing we need to fix is that the dashes last longer than the dots. In fact, to speed things up, “in 1910, Reginald Fessenden invented a two-tone method of transmitting Morse code. Dots and dashes were replaced with different tones of equal length.”1)
For example, here's how “VE7” (•••− • −−•••) could sound like if dots were tones of 600 Hz and dashes tones of 800 Hz (with some dead space in between each note for us to hear the breaks):
The second thing we need to fix is that the number of 0s and 1s (number of bits) that a letter needs is different for different letters. For example: “V” has four bits (0001) where as “E” only has one (0) and “7” has five (11000).
Since the biggest number of bits in Morse Code is 5, it might be tempting to try and fix this problem by adding leading zeros to all the other letters, but this doesn't work because letters with less than 5 bits might map onto existing letters once we add the extra zeros. For example: we can't make “V” 00001, because that's already “4”, and “E” can't be 00000 since that's already “5”, and so on.
So in addition to having every bit be the same length, we'd also prefer an encoding method where every character has the same number of bits. To that end, we could use the Baudot Code, which can encode 32 characters of 5 bits each, or the ASCII Standard, which can encode 128 characters of 7 bits each.
The third thing we need to fix is to make the notes as short as they can possibly be. With morse code, the speed someone can receive is dependent on their skills, but if a computer had “perfect skills”, would there be a theoretical limit to how short a note can be? It turns out there is! The shortest note has to be long enough to include a full cycle of the lowest frequency used. For example, as we'll see below, APRS uses 1200 Hz and 2200 Hz tones for its transmissions. One cycle of 1200 Hz is 1/1200th second (0.000833... s) and 2200 Hz has a cycle of 1/2200th second (0.0004545... s). That means that each “note” has to be at least 1/1200th second long.
The fourth and final thing we need to fix is to eliminate the spaces between each note so that the transmission is one long continuous two-note song. This is easy enough, but it leads to a technical issue that we'll address in a moment: how will the receiver know when a note starts and ends if they're all smooched together?
Before we answer that, let's have a look at a real example2) of a digital transmission that uses 1200 Hz and 2200 Hz tones:
If we download the audio file and open it in Audacity, we can see the wave form visually:
The first half of the transmission contains the same pattern of one long wave (1200 Hz) followed by 12 short waves (2200 Hz) repeated 76 times. I separated these in red. Since each of these sections lasts 1/150th of a second and each note lasts 1/1200th of a second, each section contains 8 notes (separated here in yellow).
But why spend so much time sending essentially no information? Remember when we removed all the spaces between notes and wondered how the receiver would be able to know when a note starts and finishes? This part acts as a sort of clock to synchronize the receiver to the transmitter. By sending a series of “ticks”, the receiver now knows where each group of 8 notes starts.
From here, there's a few ways to decide how we should encode the information: do we want each note to literally represent a 0 and a 1? or do we want the change in note to represent a change in bit? See this short but very instructive page for more details.
What we have created here is a Frequency-Shift Keying (FSK) system.3) The characteristics of (A)FSK are that:
Now, there's a distinction we need to make that's not clear yet (but will be soon). There are two ways to describe the speed of the transmission:
So what's the difference? Well, with FSK, the bit rate is equal to the Baud so the distinction is not clear. But imagine there was a system where each note could somehow encode more than one bit. The bit rate would then be greater than the Baud. We'll see this soon...
Where as with FSK, the information was encoded in the change of frequency (using two different tones to represent 0s and 1s), PSK encodes the information in the phase. This is a little trickier to conceptualize because while the human ear can detect differences in frequency (as different notes) and differences in amplitude (as different volumes), it can't detect differences in phase.4)
Supposing we use the blue wave as a reference wave, the green wave is 90° out of phase, while the red wave is 180° out of phase. Notice that all three waves have the same frequency (the number of cycles per second) and the same amplitude (the height of the wave), but they start at different points. To the human ear, all three notes would sound exactly the same, but a computer can be made to detect the difference, which means we can use two of those waves to represent different bits.
A useful way of representing the different “states” that the waves can be in is using a constellation diagram.5) For example, using the Blue and Red waves to encode a message, the constellation diagram would have the two dots on the horizontal axis (one on the right at (1,0) and one on the left at (-1,0)).
The distance between a particular point and the centre of the graph represents the amplitude of the wave. In this case, both dots have the same amplitude. The phase is represented as the angle the point makes with the right side of the horizontal axis. So the blue wave has a phase of 0°, while the red wave has a phase of 180°. Finally, we can assign each wave a bit (0 or 1).
Notice that this diagram doesn't tell us anything about the actual frequency of the waves because they are made to be the same. What changes here is the phase, not the frequency.
Before we continue, let's briefly review the difference between a bit rate and Baud:
Again, in this case, the bit rate is the same as the Baud since as with FSK, each note encodes a one bit. But this is about to change!
If instead of only using two different phases (0° and 180°) we used eight different phases (0°, 45°, 90°, 135°, 180°, 225°, 270°, and 315°) each “note” can now encode three bits instead of just one! This particular scheme is called 8-PSK.6)
The advantage is that while we might only be able to transmit 1200 notes per second (1200 Baud), the data rate is now triple (3600 bps). That is, each note encodes 3 bits.
But why stop at eight states? Well, in reality, the more states we use, the closer together they become, which means that it can get more and more difficult to tell them apart, which leads to more errors.
In 8-PSK, the eight states were all on the same circle because their amplitudes were all the same. One way to add more states while keeping them apart as much as possible is to spread them around over the area of a “square” instead of on the perimeter of a circle. This is called Quadrature Amplitude Modulation (QAM).7)
QAM is kind of a mix between PSK and ASK (Amplitude Shift Keying). That is, in addition to being able to vary the phase of the waves (where they start), we can also vary their amplitude (how strong the signal is). In the case of 16-QAM, each note can encode 4 bits. So a 1200 Baud signal has a bit rate of 4800 bps.
ADSL technology for copper twisted pairs uses a constellation size of up to 32768-QAM, equivalent to 15 bits per tone.8)