- Air-Cored spiral -wound like a cylinder
- Wound on a magnetic core in the shape of an; E, or U either solid or built up from flat laminations
- Wound on a round pot core
- Wound on a rod, there are several ways to wind the turns too, and different types of wire
- a flat pack with a slit or mounted over a set of turns laid down on a flat substrate (e.g SMD inductors or inductors made of PCB material
- a cylinder with one, two or six holes or
- a ring/doughnut structure (which to be fair is the same as a cylinder with one hole.)
- DC,
- low frequency AC such as 50 or 60 Hz
- Slightly higher frequencies up to 100kHz and lately to nearly a Megahertz for switching power supplies and battery chargers.
- Also radio frequencies that are lowish, under 3MHz or
- In the radio bands we use for High Frequency communications (3 to 30MHz) or
- very and even ultra high - VHF/UHF as well as
- microwaves as found in mobile phones, satellite dishes and microwave (sic) ovens.
- I suppose the millimetre waves of 5G must use inductance, not sure we'd recognise them.
- Work out the DC operating point (no signal applied) - the values at various nodes around a circuit to help you fault-find and check the bias conditions have been calculated correctly.
- Work out "small signal" performances; gain and frequency responses are obvious examples, but also the input and output impedances (and how these vary with frequency) - this is important when you interface two modules together.
- Design filters and see how changing a components value to a standard value affects performance. (there are even tools that can generate LTSpice files by filling in a form about what type of filter you want - see http://www.tonnesoftware.com/elsie.html or other tonnesoftware programs)
- Work out how much noise a circuit produces. I want to experiment with this...I am a bit dubious about how accurate the noise data included in some transistor models by manufacturers are, whilst microwave transistor models may be accurate I think a lot of ordinary transistors have nominal or even missing noise parameters in their models.
- Work out what the maximum output can be before clipping, Usually either the "one dB compression point" or the third order intercept point (IIP3 or OIP3) or even THD.
- Calculate what harmonics are produced.
- Work out the power consumed or how much power each component dissipates.
- Calculate the temperature rises in a circuit and its effects.
- Of less use to amateurs perhaps - but you can vary every component randomly about its tolerance values - this tells you that if you build a thousand circuit boards, how many will pass or fail some output test. - known as Monte Carlo testing. You can use this to see if building a filter using 5% tolerance components is likely to give a reasonable performance.
Most of the time we use the .TRAN and the .AC analyses of SPICE.
A .TRAN analysis runs for a fixed time period from startup and uses numerical techniques to calculate voltages and currents at every node using large signal/non-linear models of the components. The .AC analysis works out the .DC operating condition and then creates a small signal/linear model of the circuit at that bias point. It applies a single frequency sine wave, calculates the V and I assuming linear operation, changes the frequency of the input signal and repeats the process. It sweeps the input across a frequency range and is useful to see frequency response. It is using an approximation of the actual circuit but is useful nonetheless.
Quote from groups.io LTspice forum
"Small Signal Analysis takes a system, and linearizes it about its operating point, and because the system it looks at is a linear system, there are a lot of powerful techniques that can be used to very precisely analysis the system, and often get closed form mathematical results. (But the precise answers are to an approximation of the system).
Large
Signal Analysis, uses a much more detailed model of the system,
including non-linearities. Because of this, you rarely get
significant closed form mathematical results, but the systems tend to
(be)
processed
with numeric methods. These answers tend to be approximations, but to
a much better model of the system, so can be more accurate."
And a further quote
"an .AC analysis in SPICE, ... SPICE converts those transistor and diode models into their small-signal (linear) equivalents at the DC operating point of your circuit, and all results from that point forward are based on that. Even if your simulation uses 10KV AC signals, the diode and transistor models in that .AC simulation are small-signal models because they've been linearized. But they were linearized starting with the original nonlinear (large-signal) models.
In a .TRAN analysis, the simulation uses the full nonlinear model so it works for small and large signals."
And
"One common way to determine whether a signal is small enough to be “small signal” is to do a transient (.TRAN) analysis using a sine signal source. Then, you can do a Fourier Analysis of the output. One of the metrics of this analysis is a harmonic distortion number. You can then vary the amplitude of your input signal until the harmonic distortion is “small enough”. At this point, for your application, it is then Small Signal."
LTspice can calculate THD (total harmonic distortion) which suits audio amplifier designers. RF designers work with third order distortion rather than the total distortion as it is usually the biggest and our circuits have filters or we are only interested in outputs around a single frequency or frequency band. The distortion due to third order products should increase rapidly as the input rises (three times more rapidly). You can test for this by applying two similar tones f1 and f2 at a certain level and use the FFT function to see the distortion (2f1-f2 or 2f2-f1). i.e if you apply 0 dBm tones and get 2f1-f2 at -46 dB below that and then increase the inputs to 3 dBm you should see f2-f1 at 9dB higher or -40dB below the tones. My initial run of LTspice did not do this. I was not setting the FFT parameters correctly and others have noted "funnies" in the FFT results. Gunthard Kraus simulates an entire 137MHz satellite receiver (convertor) using LTspice at http://www.gunthard-kraus.de/LTSwitcherCAD/SwitcherCAD-Tutorial_English/pdf-File/ When I noticed this I went on to assess IMD a different way - If you apply a signal and then increase its strength, there comes a point when the amplifier cannot amplify enough and the gain starts to drop, a standard value is when the gain drops one dB from its theoretical value. the "one dB compression point" POUT(-1dB). The OIP3 figure tends to be 13 to 15dB above this value in BJT amplifers and about 10-12 dB in FET circuits. This gave me an alternative way of estimating OIP3.
In parallel with this work I asked for help on the forum, Tony Casy, Andy, Vlad and Dana all replied (it is a really good group) and gave me FFT settings that finally gave believable results - I could see third order products rise three times faster than rises in the input - which is correct behaviour and also the OIP3 were about 15dB higher than the 1dB compression point which further reassures me the results can be trusted. Tony also created a version of one of my models to present automated results of compression, it uses a lot of LTspice commands but a lovely example of what a "power user" can do
See my Youtube videos when they go up as there is a lot of fiddly bits to get it all working - spread over 8 different LTspice simulation runs. They are a good example of how to use simulation as a design tool when designing, or before building, homebrew transistor circuits to be used a amplifiers as part of receivers and transmitters.
Of course LTspice is also very very useful for filters and other radio circuits. When I get around to simulating the uBitx I will be simulating mixers, filters and audio circuits - hopefully everything!. I have a simulating model of a ring mixer already and it works - the computer gives the same answer as the theory and the same results as a real circuit. One wrinkle to the model is that all 4 diodes have the same model parameters - absolutely identical. In real life, if you pick up 4 diodes they all differ slightly, we often manually match them by picking 4 from a larger batch so that they have the same forward voltage drop maybe within a millivolt or two (at a chosen fixed forward current - perhaps what the diode test function on your DVM uses!). This is important - as a first approximation, but of course the question is - are they still matched at differing currents and what happens when the ambient temperature increases? This is an example of when you need to do multiple runs at slightly different conditions. In LTspice this can be done by using Monte Carlo simulations and these can take a lot of time - I shall be running my computers overnight or longer to get results for this one. It will have to wait until I am actually building my uBitx however and I will create detailed diode models from the few hundred diodes in my junk box by measurement.
No manufacturer publishes the data that I need to see the statistical tolerance spreads on the parameters - they usually just publish a single figure, or at best (but very rarely) the three figures of minimum, typical and maximum. I need the standard deviations of each published parameter (assuming a normal bell curve spread of values). This is way beyond what a hobbyist radio amateur needs but I find it interesting and the current situation is giving me time to be thorough. (Not sure if it is the lockdown or the fact I am recently retired)
