FM radio design resources

======================

Wot a raving lunatic......

..... Phil

VCO's are inherently nonlinear and drift due to temperature and time. This requires locking the vco to a crystal reference, which increases cost and complexity. It is difficult to correct for the nonlinearity.

PM can use a crystal-controlled reference, eliminating drift. Bandwidth can be extended using frequency multiplication.

Modern DSP techniques can produce any required type of modulation, for example the quadrature amplitude modulation (QAM)used in 802.11. This requires a crystal reference.

SDR can modulate and demodulate any type of modulation, from AM, SSB, FM, PM, QAM, etc. This technology is rapidly overtaking the old conventional technologies and should be a focus for new studies in electronics and communications.

Yep, but I am well familiar with digital techniques and the "haha let's just put it all in a computer" school of design and want to steer clear of it for this project. I'm not super worried about temperature drift since this is a personal project that won't be operated under a wide range of temperatures, but I did come across a crystal-stabilised design from Jim Williams here:

What is the difference between how the bandwidth is extended using frequency multiplication in a phase modulator as compared to how it needs to be extended in phase modulation? I thought they both used frequency multiplication (see diagram I posted).

Not necessarily. Real components do age and have some temperature dependence.

Locking an oscillator to a crystal reference can reduce some forms of drift. If you don't know what non-linearity you have to correct for, it's obviously going to be difficult to engineer a correction.

Crystal-controlled references still drift.

Only in the sense that if you multiply up a frequency range, it covers a wider bandwidth.

Quadrature amplitude modulation refers to the way the signal is constructed, not it's frequency stability. You probably wouldn't bother with such a complicated scheme if you didn't have a stable frequency source, but it isn't any kind of requirement.

Software defined radio may be able to do all of this, and lots more besides, but it needs an analog front end to convert the incoming signal into something that can be digitised and processed, or turn the digital bit stream being generated into a bandwidth limited signal without lots of higher harmonics.

A D/A converter does produce a stair-case waveform, unless you get very subtle.

In some areas. Software does run on hardware, though programmers don't like to admit it.

"Looking to build a sensible FM radio transmit/receive pair, designed for short range (10s of meters) and excellent audio quality. Making for personal use," "The project needs to be RF because I've arbitrarily decided to do an RF project :)" "the idea of this project was a fun and educational thing, not necessarily a practical one." "would like to have a high quality audio signal." "and is it possible to get audio that is more or less indistinguishable to the kind you get over a copper cable?" "I don't know what the efficiency of a FM radio transmitter is" "Discrete PLL's can apparently be a bit tricky to do in discrete components" "but I am well familiar with digital techniques" "I just want to build one system" "I'll drill a hole in my headphones and tap off the battery" "Do you think I could avoid standing wave problems if I just occupied a huge bandwidth" "I'm also no sure about the sidebands. I suppose I want a single sideband (SSB), but I have not quite figured out yet how to go about putting that in." "I failed to specify that my project must be designed using only classic RF components like our forefathers used, no all-in-one IC's allowed."

Now that we have a full requirement specification I think one of these would be most suitable for use in the receiver:

There are some practical circuits here:

Nonsense. You can phase lock to even rapidly drifting signals. Old timey entertainment radios use AFC, i.e. they dork the LO slightly to keep the signal centred in the IF passband. That's an example of a frequency-locked loop (FLL).

And making linear VCOs isn't as hard as all that. You can use an integrator/Schmitt trigger oscillator, which can get down to ~0.1% nonlinearity (the ancient LM331 does this, but it can be done much faster).

Alternatively you can use an LC linearized using an off-stage resonance. (series and parallel inductors plus a dual varactor).

Deviation can be extended, not modulation bandwidth. But you're complaining about ordinary VCOs being nonlinear, and still suggesting a _VCXO_? Frying pan, fire, etc.

All true, but pretty droll when the OP thinks that there might be such a thing as single-sideband FM.

Depends for what. The OP is probably not going to be using direct digitization at 900 MHz for a battery-powerd hobby project, and I'm not going to be using $20k worth of lithium niobate modulators to put FM sidebands on my diode laser, even if they worked at my wavelength.

Cheers

Phil Hobbs

Ain't no such animal as SSB FM.

For an educational project, it's worth going through the derivation of the FM spectrum. You'll need Wolfram Alpha or a copy of Abramowitz & Stegun to get the Bessel function expansion, but it's about a third-year undergraduate problem.

When you do it, you find that the FM spectrum is made up op of a forest of sidebands. The nth-order sideband amplitude is proportional to J_n(m), where m is the modulation index (see one of my earlier posts).

The math is not difficult, and it's really quite pretty.

Cheers

Phil Hobbs

That's just getting attention by "any means".

Actually he is a crooked business-man, who got into politics because he didn't have enough money bribe his way out of the hole his incompetence had dug him into.

Clive Palmer is neither kind-hearted nor charitable. He's not all that convincing when he tries to act as if he is - maybe convincing enough to fool you, which is a pretty low bar.

And I'm not the last surviving Communist in the western world - that species became extinct here quite a while ago. Any time now you will be calling me a Tasmanian Tiger (or thylacine) which is no less extinct. You really don't have any grasp of reality, do you.

IC-based quadrature detector ICs for FM stereo have to deal with a signal bandwidth of 60 kHz or more (and an IF bandwidth of more than that, per Carson's Rule). Quite a few well-spoken-of commercial FM tuners and receivers use these ICs. So, if all you need is mono audio with a 20 kHz frequency range, I don't think you'd be pushing the envelope.

The quadrature coils ("transformers" although they aren't, really) usually have a swamping/damping resistor in parallel with each LC tuned circuit, to reduce the Q and thus increase the width of the linear range. For example, the LA1235 data sheet calls out a couple of different models of quad-coil can (one Toko, one Sumiko) and the damping-resistor values are different (I presume to compensate for different impedances of the LC circuit components). [Both of these particular coils are essentially unobtanium now - I can't even find data sheets for them - so you may have to wind your own quad-coil (or add a cap to a standard single-coil can) and do some experimentation to tweak the design.]

For what it's worth: some years ago I helped debug an audio-quality problem in a commercially-build amateur-radio repeater (2-meter band) which uses a quadrature detector. The audio was distorting, and then clipping, on normal voice-speaking IF deviations... the distortion was bad enough that DTMF tones couldn't control the repeater properly. We tracked the problem down to the quad coil - the manufacturer had changed coil types when the original model went out of production, and hadn't realized that the new coil's higher Q required more damping (smaller R in parallel). A simple change to this resistor changed the audio from harsh, to clean and pristine. We notified the manufacturer and they changed their next production run, with very happy results.

You might want to skim through some of the FM stereo tuner descriptions at fmtunerinfo.com, look for models that mention quadrature detection and discrete components, and look at the schematics for inspiration. Green and Bourque's "The Theory and Servicing of AM, FM, and FM Stereo Receivers" might also be useful, although they give scant coverage to quadrature IF detectors.

A friend of mine (now in his early 80s) did a college project which involved building an FM receiver... this was back in the tube era. If I recall correctly he designed and built a tube-based pulse-count FM detector, and wowed his instructor with clean reception from an FM station about 40 miles away (he described it as "sounding as if it was wired"). He's been threatening to dig through his attic to see if he still has the board; I'd love to see it, if so.

My impression is that getting really-low-distortion FM is a harder problem on the transmitter side than the receiver side. Most modulators use varactors, and their response isn't surpremely linear. You might want to take a look at the schematic for the venerable Sound Technology 1000A FM alignment generator - they managed to do an impressive job with a varactor-and-single-transistor RF oscillator.

A wide-range VCO built with a hyperabrupt varactor has pretty nearly linear tuning, and as I noted upthread it can be linearized further by using series and parallel inductors.

Cheers

Phil Hobbs

What I used to do IIRC (and I'm going back a hell of a way here) is to take a crystal oscillator's output and pass it through a Schmidt Trigger (or whatever you can find that switches very fast). That gives an output that's rich in harmonics. You then make a filter for the highest stable harmonic to pass on to your power amplification stage(s).

So your wonderful Wikipedia is wrong, then? Whoever wrote the article about him was fooled and Wikipedia published a falsehood! Just imagine that... Yet no doubt you'll still be citing from it in future like it's a trustworthy source LOL!

It's ingenious how they do all that with just one transistor in some designs. I think they use a BJT with a suitable BC junction capacitance and apply the audio control voltage to it. Saves using a varactor diode!

On that subject (forgive me if it's already been suggested) the Chinese sell cheap but powerful lasers and just a 50 cent 5mW one ought to be able to convey a modulated signal over 20 meters or so that the OP requires, surely?

Daft question about an almost impossible to do and almost impossible to do legally project with more and more ludicrous restrictions being added rapidly with each message posted.

This is clearly a very nasty troll of just about the worst kind.

killfile it.

Not really. More a classical dumb newbie, of the more than usually incompetent sort. Trolls are malicious.

Just ignore it.

Presents a bit of a pointing problem, of course.

Cheers

Phil Hobbs

Impossible? Why? He probably needs about 10 mW of RF for a job like that, and 0.1% THD ought to be quite doable with a decent VCO design.

Whether the OP can make it work remains to be seen.

Cheers

Phil Hobbs