2007-05-19
FM Detection Experiments
Some time ago I found some 4.5 MHz filters on eBay designed for TV audio IF stages. When they arrived I just put them into the junkbox and forgot about them until now. This weekend I was digging around for a filter for another experiment and found these. This triggered the usual "What can I build with this?" thoughts, and eventually lead to these experiments - I decided to build an FM detector IF strip.
Foster-Seeley Discriminator
Working backwards from the detector, I first constructed a Foster-Seeley Discriminator on the solderless breadboard and got it working with a centre frequency of near 4.5 MHz. This circuit was then soldered into its final implementation and tested. The resonator is a 7.9 uH inductor which is bifilar wound on a T50-2 core, resonated with 156 pF. I eventually replaced one of the fixed capacitors with a trimmer to allow precise centring of the detector slope. The diodes are unmatched 1N4148 garden variety rectifiers.
The big problem with this detector is its input amplitude dependence, meaning ideally I need to limit the IF before application. For now, I decided to ignore this and enjoy the fact it offers a much larger output signal than a Ratio Detector which is essentially amplitude independent.
Ratio Detector
I tried building a traditional Ratio Detector, but I had a lot of problems getting it to work. I am unsure exactly how the third winding is meant to be coupled to the other two windings. It is also a lot more touchy to tune up, you have to resonate both the primary and secondary, the diode matching seems pretty critical too, the DC swing across the slope wasn't very symmetric in my experiments, but that might have something to do with my lack of understanding of how to couple and phase the third winding. Any feedback on how to get this working would be greatly appreciated!
With the solderless breadboard and a bit of evil empiricalism, I eventually made something that seems to be a Ratio Detector, in fact it works quite well and seems to have good (but not perfect) AM rejection. The topology is similar, but not quite a classic ratio detector:
It does not exhibit the classic "S" shaped discriminator curve, rather a slope through resonance without a noticeable peak at the bottom (but a small one at the top):
As I do not completely understand this circuit I decided to leave it aside for the moment. It does work and is elegantly symmetric and fairly easy to build, so I will come back to it. I believe it may be acting as a resonance bridge, and producing a DC signal proportional to the reactance of the secondary, but I haven't studied it mathematically yet.
IF Amplifier
I proceeded with the Foster-Seeley Discriminator, now I had to build an IF amplifier that matched the output impedance of the filter. I had some trouble finding the specs on the filter, but it appears to be 1 kilo-Ohm in and out, and prefers a little capacitance on the output side for best ripple. A two-stage basic common emitter amplifier was constructed using BC549 devices and, for a change, ADT1-1 minicircuits transformers in the collectors. (Laziness, pure and simple, the ADT1-1 is an SMD device, rated from 150 kHz to beyond 400MHz, but I got a reel of them of eBay for about $40. You can use any old ferrite bifilar transformers giving a winding reactance of at least a kilo-Ohm or so, ferrite beads or FT23-43s will do, as would BN-x and FT50-43 or larger if you don't mind the size.)
With the filter in place this amplifier gives a transducer gain of about 38 dB at 4.5 MHz. The input impedance at the actual base of the transistors was measured at around 380 Ohms and somewhat capacitive, a little low for the filter, but not too bad. The output is "low", and with the collector transformers the inter-stage match is actually quite acceptable, but the final device will be driving the high-Z discriminator. I shunted the output with an 820 Ohm resistor to improve its stability. For now, the losses and noise figure issues of this IF hack will be ignored.
The assembled IF and detector modules were tested, and proved capable of detecting the 1 mW output of my signal generator across the bench using just my finger on the input-side of the filter as an antenna, and those Archer mini-amplified speakers as the AF stage. (The input impedance of the amplified speaker is actually far too low, and worsens the distortion of the detector a little. I've left space on the detector board for a higher-Z buffer at some point.)
Mixer
Now I needed a mixer to convert the RF down to the IF. I considered using my traditional diode mixers, even contemplating using a TUF-1 or SBL-1 rather than spending half an hour building a diode mixer from ferrite beads and wire. Instead I did something completely different and new for me, I constructed a cascode FET mixer with a pair of J310s. My dislike of the cascode mixer is well-known, but in this particular application it works wonderfully. It works to beyond 200 MHz and is extremely simple. I would have used the dual-gate FETs that John VK2ASU gave me, but I wanted to save them for a more critical application.
The drain circuit is a 4.5 MHz resonator to select out the IF and try to reject all the garbage that the cascode multiplication doesn't balance out like a gilbert cell will.
Just for a laugh, I connected a 3 meter (FM broadcast) band folded dipole to the bottom FET and my signal generator to the top one. To my complete surprise I could tune in FM broadcast stations, but the filter was too narrow to allow their full deviation through. Oh a whim, I pulled out the filter and replaced it with a 22 pF capacitor, amazingly this allowed me to receive FM stations with quite impressive clarity! This actually blew me away, the selectivity was quite acceptable with absolutely no filtering beyond the drain resonator and the discriminator resonator. It appears that a similar circuit could used for an FM broadcast receiver.
Local Oscillator
Thoughts of a VHF Autodyne Converter filled my head - but instead I built a varicap tuned common-base oscillator for the LO injection. This worked well and produced a quite acceptable FM broadcast receiver. (Curiously the floating LO port of the mixer was sensitive enough to weakly function with the leakage from the LO stage a few inches away, but I opted for direct connection.)
Discussion
Attempts to use this assembly to detect 2 metre narrow FM transmissions has been unsuccessful so far. The LO is not stable enough for this purpose (it is fine for wide-FM though), so I am considering building a pre-mix VFO based LO, probably with an LC VFO around 8-10 MHz and a high-side crystal oscillator module or multiplier chain. A VXO and multiplier chain might be a possibility, but a suitable crystal frequency I don't appear to have in the junk box.
The filter is also too wide for good selectivity of narrow-FM, but this is easily fixed, I can build a new filter using computer crystals and adjust the resonators to match. Using the current filter I have listened to TV channel audio signals, for this purpose it works very well and which indeed it was originally designed.
To complete it as a usable receiver it needs a front-end band-pass filter and LNA. A J310 common gate amplifier and series-tuned filter would be sufficient for experimental use, but a coupled parallel resonator filter would be a better idea for the final receiver. The IF is just high enough to give reasonable image rejection on 2 metres with just a BPF, but for broadcast band use the front-end would need to be track tuned.
5 comments.