Conventional wisdom holds that the best way to learn a new language is immersion: just throw someone into a situation where they have no choice, and they’ll learn by context. Militaries use immersion language instruction, as do diplomats and journalists, and apparently computers can now use it to teach themselves Morse code.
The blog entry by the delightfully callsigned [Mauri Niininen (AG1LE)] reads like a scientific paper, with good reason: [Mauri] really seems to know a thing or two about machine learning. His method uses curated training data to build a model, namely Morse snippets and their translations, as is the usual approach with such systems. But things take an unexpected turn right from the start, as [Mauri] uses a Tensorflow handwriting recognition implementation to train his model.
Using a few lines of Python, he converts short, known snippets of Morse to a grayscale image that looks a little like a barcode, with the light areas being the dits and dahs and the dark bars being silence. The first training run only resulted in about 36% accuracy, but a subsequent run with shorter snippets ended up being 99.5% accurate. The model was also able to pull Morse out of a signal with -6 dB signal-to-noise ratio, even though it had been trained with a much cleaner signal.
Other Morse decoders use lookup tables to convert sound to text, but it’s important to note that this one doesn’t. By comparing patterns to labels in the training data, it inferred what the characters mean, and essentially taught itself Morse code in about an hour. We find that fascinating, and wonder what other applications this would be good for.
In part one, I compared the different Analog to Digital Converters (ADC) and the roles and properties of Delta Sigma ADC’s. I covered a lot of the theory behind these devices, so in this installment, I set out to find a design or two that would help me demonstrate the important points like oversampling, noise shaping and the relationship between the signal-to-noise ratio and resolution.
Check out part one to see the block diagrams of what what got us to here. The schematics shown below are of a couple of implementations that I played with depicting a single-order and a dual-order Delta Sigma modulators.
Basically I used a clock enabled, high speed comparator, with two polarities in case I got it the logic backwards in my current state of burn out to grey matter ratio. The video includes the actual schematic used.
Since I wasn’t designing for production I accepted the need for three voltages since my bench supply was capable of providing them and this widget is destined for the drawer with the other widgets made for just a few minutes of video time anyway. Continue reading “Tearing Into Delta Sigma ADCs Part 2”→
It’s not surprising that Analog to Digital Converters (ADC’s) now employ several techniques to accomplish higher speeds and resolutions than their simpler counterparts. Enter the Delta-Sigma (Δ∑) ADC which combines a couple of techniques including oversampling, noise shaping and digital filtering. That’s not to say that you need several chips to accomplish this, these days single chip Delta-Sigma ADCs and very small and available for a few dollars. Sometimes they are called Sigma-Delta (∑Δ) just to confuse things, a measure I applaud as there aren’t enough sources of confusion in the engineering world already.
I’m making this a two-parter. I will be talking about some theory and show the builds that demonstrate Delta-Sigma properties and when you might want to use them.
My DSL line downloads at 6 megabits per second. I just ran the test. This is over a pair of copper twisted wires, the same Plain Old Telephone Service (POTS) twisted pair that connected your Grandmother’s phone to the rest of the world. In fact, if you had that phone you could connect and use it today.
I can remember the old 110 bps acoustic coupler modems. Maybe some of you can also. Do you remember upgrading to 300 bps? Wow! Triple the speed. Gradually the speed increased through 1200 to 2400, and then finally, 56.6k. All over the same of wires. Now we feel short changed if were not getting multiple megabits from DSL over that same POTS line. How can we get such speeds over a system that still allows your grandmother’s phone to be connected and dialed? How did the engineers know these increased speeds were possible?
The answer lies back in 1948 with Dr. Claude Shannon who wrote a seminal paper, “A Mathematical Theory of Communication”. In that paper he laid the groundwork for Information Theory. Shannon also is recognized for applying Boolean algebra, developed by George Boole, to electrical circuits. Shannon recognized that switches, at that time, and today’s logic circuits followed the rules of Boolean Algebra. This was his Master’s Thesis written in 1937.
Shannon’s Theory of Communications explains how much information you can send through a communications channel at a specified error rate. In summary, the theory says:
There is a maximum channel capacity, C,
If the rate of transmission, R, is less than C, information can be transferred at a selected small error probability using smart coding techniques,
The coding techniques require intelligent encoding techniques with longer blocks of signal data.
What the theory doesn’t provide is information on the smart coding techniques. The theory says you can do it, but not how.
In this article I’m going to describe this work without getting into the mathematics of the derivations. In another article I’ll discuss some of the smart coding techniques used to approach channel capacity. If you can understand the mathematics, here is the first part of the paper as published in the Bell System Technical Journal in July 1948 and the remainder published later that year. To walk though the system used to fit so much information on a twisted copper pair, keep reading.