2011-01-30
555 Super-Regenerative IF HF Spectrum Analyser
This is another of those crazy ideas that have been rolling around in my head for a while now. I have previously used super-regenerative circuits as logarithmic detectors, and built super-regenerative super-heterodyne receivers - it wasn't a huge leap to build a swept front-end and turn the combination into a simple spectrum analyser. The recent announcement of a 555 Contest by Chris Gammell and Jeri Ellsworth had me playing with 555 circuits and this was all the encouragement I needed to finally go ahead and do it. The end result is a toy SA adapter for your CRO, but is still useful instrument (and it fits in an Altoids tin).
Architecture
The circuit is a high-side injection single-conversion super-heterodyne receiver mixing up signals in the HF region to a VHF IF around 150 MHz. The IF is amplified and detected directly in a self-quenched super-regenerative detector. Log-linear signal strength information is extracted from the detector quench frequency.
RF Part
The VHF local oscillator is a JFET Hartley with varactor tuning allowing it to be swept by a ramp generator. The front-end consists of an LC low-pass filter with a cut-off around 55 MHz, followed by a JFET cascode mixer being fed from the filter and LO. The mixer drain inductor is loosely coupled to the super-regenerative detector tank by proximity (but is itself untuned). The detector provides the only selectivity of the circuit, but achieves about 100 kHz 3 dB selectivity. The skirts are poor however (as one would expect from a super-regenerative detector), and the final resolution bandwidth is at best 250 kHz. This is still a fair bit better than what might be achieved (say ~1% BW) with a much bulkier LC filter at the IF frequency.
LF Part
The linear ramp generator uses a LM555 with a PNP current source to linearise the capacitor charging. This helps linearise the RF tuning sweep (but some non-linearity remains due to varactor properties). As the 555 in astable mode normally keeps the capacitor voltage between 1/3 and 2/3 Vcc I added a green LED to the threshold line to shift the high-water point closer Vcc and give more voltage swing for varactor tuning. With the particular supply and varactors and LO design about 22-25 MHz of sweep range is achieved, a little bit more would be nice, but adjustment of the detector or LO offset tuning capacitor allows tuning right up to (and beyond) the input LPF cut-off.
The super-regenerative detector quench frequency is around 35 kHz when quiescent, but rises to beyond 40 kHz when detecting large signals. The frequency is almost linearly proportional to the log of the signal strength due to the basic exponential nature of the signal growth during regeneration. To convert this FM signal to a DC voltage suitable for display with the CRO a 7555 is used in monostable mode to generate constant width pulses on each quench cycle. These rectangular pulses are integrated in a capacitor and the resulting voltage shifted and amplified with an op-amp for easy integration with the CRO Y channel.
The Circuit
Yes I could have just biased control higher rather than using a LED dropper on threshold, but it costs less current and just happens to work almost perfectly at the 9V Vcc with the drop of the green LED I had sitting on the bench. Two diodes instead of a diode + resistor is probably rather evil, temperature stability wise. Replace one or both with a resistor if this becomes an issue. I should probably use the other half of the LM358 as a follower to buffer the output and stop external loads affecting it; don't put low-impedance loads across the X output. Hopefully your 10 uF electrolytic capacitors are low leakage...
The CE-NPN inverter is needed because of the -ve level triggering of the 555. I did at one point remove the inverting transistor and run the time constant longer, effectively inverting the F2V slope, but the extra sensitivity of the inverting stage doesn't hurt, and it is more obvious how it works. You can switch the time constants in the integrator if you want to control the video bandwidth.
The mixer is a bit of a hack. I measured the Idss and Vp of the J310s in question and picked the source resistor to give about 2 mA. The 56 ohm gate resistor terminates the filter in something close to its design impedance. Was going to use pads either side, but this works reasonably well as-is. The gate capacitance makes the match worse at higher frequencies, not a major deficiency. Tuning the mixer drain upsets the detector a bit (as it is moderately coupled to the detector tank), with less coupling some selectivity at the mixer drain couldn't hurt?
Half-wave input filter is not as sharp as possible of course, and perhaps adding a notch at the IF is a good idea? I tune the IF to move the region of interest (instead of tuning the LO set-point as this changes its tuning range because of the relative capacitance change), so a notching trap would need to be tuneable, but you can fix-tune the IF if you like.
The 5p6 value suited my particular circuit best. I selected it using a trimmer try-it-and-see... Perhaps not the best approach. Other topologies should work, as long as you can get them to squegg at something near 40 kHz.
The coils are 90-100 nH, the LO coil stretched out a little more than the mixer and detector one so it is perhaps closer to 80-90 nH. You could use toroids, which might give you better control of the mixer/detector coupling.
Performance
Here is a video I shot shortly after finishing construction, it has a few shots of my DSO screen while using the SA adapter and a signal generator.
IP3 has not been measured yet but it is not expected to be very good because of the simplistic mixer used. LO leakage should also be measured but should not be too bad because of the LPF in the input. Precise log-linearity has not been measured either, but I did test it with a step attenuator visually confirming its better than 50 dB dynamic range, good linearity and 10 dBm input handling. No doubt it is not especially sensitive and probably has a terrible noise figure.
None of this makes it a completely useless instrument however, and it is quite good at quantifying harmonic energy from homebrew transmitter projects. With a noise source or a tracking generator it might be useful for sweeping HF filters.
It pulls 21 mA from its 9 volt supply.
Some Screenshots
2.5 MHz square wave from generator.
5 MHz square wave from generator.
DC-24 MHz from random-wire shortwave antenna.
Notes
The circuit was originally envisioned using a discrete ramp generator and tachometer one-shot. The use of 555s simplifies the design - and makes it a candidate for the 555 contest utility category of course. ;) Honestly though, I doubt I would have been able to get it going in the 2 days of the weekend if I had not used ICs.
There is no retrace blanking circuit. I did consider adding one driven from the output of the ramp generating 555. If you are using an analogue CRO you might like to add one, either railing the Y output during retrace to take it off screen or providing a Z output. A simple transistor switch pushing the level shifter intercept point (or the output directly) should do the trick.
While the vertical output intercept is variable the slope (gain) is fixed. You may wish to change part of the feedback resistance to a pot to allow slope adjustment for calibration against your CRO graticule. This would make the instrument more useful for absolute measurements, but as the mixer frequency response is anything but flat this would be frustrating. I use a step attenuator and power meter on my generator to confirm measurements.
Naturally more complicated VCO control schemes can allow width and centring controls. It is debatable if this is worthwhile with the relatively narrow sweep and wide resolution bandwidth. I find its "HF overview" nature is what makes it less of a toy and useful for quick assessment of harmonic energy and filters.
The sweep is quite slow, especially if you are using an analogue CRO with short phosphor persistence rather than a DSO. The time constant of the vertical tachometer can't be reduced too much, as it is essentially sampling the RF signal level at 35-40 kHz. The process is also quite noisy and the grass will grow rapidly without heavy integration. This kind of detector doesn't really have the capability to provide high video bandwidth, and increasing the quench will just compromise the selectivity and sensitivity.
Some of you are no doubt thinking "Why not use the xtal super-regenerative detector to get excellent resolution bandwidth?". You could, in fact I have tried this in a lash-up with a DDS LO and a DBM. It works wonderfully, except for the image problem which could be dealt with using an extra conversion step. However the xtal super-regen detector is very slow, it samples at about 100 Hz max. This is not a major problem for a scalar test instrument, but for a general purpose SA it makes wideband scan times frustratingly long. It also pretty much mandates a digital interface - which is a feature IMO, but you may disagree. It is quite easy to use an MCU and USB interface to make this PC-based.
The IF frequency is not limited to VHF, you could build UHF or higher RF systems controlled the same way. It might be fun to have a DC-1GHz version, maybe in USB-stick format? But perhaps at that point you may as well build one using an Analog Devices Log chip and a real resolution filter.
The IF detector might integrate quite well with an old analogue TV tuner front-end...
18 comments.