Pick up sounds above human hearing range and learn a little about the heterodyne principle along the way - by Daniel Koch
Have you ever been sitting on the lounge when all of a sudden, the dog bolts up and throws you aside in its scramble to get to the door and assert dominance over something that it has heard and you have not? Or have you noticed a presence in a room that you just can't quite put your finger on? In the dog's case, sometimes their hearing is just more sensitive to ours in terms of volume, but they can also hear a greater range of sound frequencies than we can. Our own ears can also detect sound pressure that our brains cannot quite 'hear' as such, like noises just outside our hearing range. Our ears pick these up as sound pressure but do not identify them as sound in the traditional sense of the word, and we can sense this as the presence or a feeling of pressure like the room or space has become more full.
The circuits we are working with here were originally intended as bat detectors. Bats use echolocation, sending pulses of sound out into the environment, and timing the return as well as processing the direction the return came from. These sounds are a mix of frequencies. Some are in the upper levels of the human hearing range, but most are above it. The specifics vary by species. The bulk, however, are in the ultrasonic range above 20kHz, and most of those are around the 50kHz mark. A few bats have been recorded as emitting signals closer to 100kHz, while rarely some species get down to the 20kHz range.
On that note, human hearing is not the traditionally quoted 20Hz to 20kHz either, unless you're a newborn. Sound response declines straight away, albeit less so at first than in later years. By their twenties, most people cannot hear effectively above 15kHz. During bench testing, my own hearing turned out to cut off at about 15.5kHz but another team member could only hear to 14kHz while another could hear to 17kHz. Note that this was not a properly controlled hearing test, just a frequency generator and a good quality monitor speaker, with the user facing away from the controls while the manual sweep took place so there was no unintentional bias or placebo effect.
While these circuits may have once been bat detectors, their applications go far beyond that, beginning at extending hearing for finding noises in the upper sections of the range you used to be able to hear in. This might be something like high-pitched squeaks from machinery that you can tell are there from sound pressure sensing but can not actually hear. Its applications continue from finding leaking gas and even water pipes, to finding out what your dog hears and of course, detecting bats.
We are going to do this using the heterodyne principle, and the frequency divider principle. When two frequencies are mixed, the output contains more frequencies than went in. The frequencies interact with one another, known as 'heterodyning' or 'beating together'. The result is two frequencies out: One is the sum of the two frequencies, while the other is the difference between them.
Some designs also port the original frequency out, too. For example, if frequencies of 20kHz and 25kHz went in, the frequencies out would be 45kHz, and 5kHz. This principle is used in radio receivers quite a lot, although in that role it is being displaced by software-based digital signal processing. In our case, however, we are going to use the heterodyne as a frequency scaler.
The other option is the frequency divider. These tend to be digital, so rather than actually stretching out a waveform to reduce its frequency, the frequency divider circuit uses the high points of a sine wave as a clock signal to advance the counter in a binary counter IC. This way, we can choose the amount of division. A common one in bat detectors is a division of 16, but we will add a switch to choose four, eight, sixteen, thirty two, and sixty four. The advantages of a system like this is that there is no background noise and no interference. The disadvantage is missing detail, because the analog systems like the heterodyne pass more information.
MICROPHONE CHOICES
Finding a microphone or sensor for any ultrasonic detector is not easy. Many makers are familiar with ultrasonic sensors and rangefinders, however, almost all of these work with a centre frequency of 40kHz. That's fine if what you want to hear is a 40kHz frequency, but for our purposes, that will rarely be the case. Leaking high-voltage electricity makes a pretty broad-spectrum noise, as does leaking gas, but most other sources have a narrower bandwidth. Bats in particular do not emit a 40kHz sound, although many species are said to emit at 50kHz or so. That might seem close enough, but the response curve of these sensors is very sharp.
Many sources say an electret microphone can pick up frequencies beyond the audible range, and some suggest piezo sensors, too. So, it was time for an experiment. The first half of the experiment was to hook up the function generator to a piezo transducer with a stated centre frequency of 4kHz. We wanted to see just how different the volume was across a frequency band, for a rough approximation of its frequency response curve. We only have volume to go off here, because, without a sensor known to respond to such a range of frequencies, we had no way of using the oscilloscope to measure the response. Ideally, a microphone with a very flat response would be ideal but we have none. So, it was ears only and that made it a subjective, perception-based test.
This test revealed a stark rise in volume at around 3.5kHz which was louder, perceivably at least, that the 4kHz sound the unit was supposed to be centred on. However, there was almost no noise past the 6kHz range. This did not fill us with hope for the sensor's ability to work across other frequencies in the audio range, let alone anything in the ultrasonic range. However, we would revisit the piezo later after realising that the right inductor might flatten the resonant curve adequately. However, at this point, we had not thought of that. This all but ruled out piezo sensors but we would build the circuits in such a way that the sensor can be easily swapped out anyway.
The next test was to use two of these units, stuck together with Blu Tack face to face, with one hooked up to the function generator and one to the oscilloscope. This did not yield any new information but was a useful exercise in what the varying amplitude of the signals looked like, and in how strong the signal may be from the piezo if we were to use it. As noted, a flat-responding microphone would have been better, if we had one.
Following this, two 40kHz ultrasonic transducers were tested. These are not the wire mesh-covered ones common on the ultrasonic proximity sensor Arduino modules, but the individual ones from Jaycar in which the unit can be both transmitter and receiver depending on how it is used. We connected one to the oscilloscope probe, and the other to the function generator. This had a better response curve, for our purposes at least, but was still very narrow. However, it was at least ultrasonic!
With no viable ultrasonic transmitter beyond this narrow band transceiver, we set about testing the electret microphone with it. A quick circuit was built to power the electret's FET amplifier, using a 10kΩ resistor from +9V to the positive electret terminal, a 100nF capacitor to decouple the output, and the other terminal of the electret connected to the power supply ground along with the ground probe of the oscilloscope. The transducer was connected to the function generator, which was set to 40kHz. On the 200mV setting, the oscilloscope showed a slight waveform when the transducer pressed against the electret. More than about 4mm away, and the transducer no longer had any visible effect on the electret. To verify we hadn't done something wrong, we hooked the piezo transducer to the function generator, set the frequency to 4kHz, and measured nearly 3V peak to peak on the electret across the same distance.
That all but nailed the coffin for the electret, but other people report success. There are a huge variety of electret microphones, differing in quality, sensitivity, and frequency range. While the basic over-the-counter ones we had were out of the question, at least from this testing, we wanted to include the option. This was both so that others can use different electrets accessible to them to explore with, and also for thoroughness: Maybe environmentally, in a different context with a different frequency source, performance would be adequate? Is the signal in the single millivolts that we expect with environmental sounds enough with the right amplifier? We don't know yet, so we included the option in the end.
It is also worth noting that there are some dedicated ultrasonic electrets out there, and a fair few more that work in the audio range but have acceptable ultrasonic performance. We did find some listings, but transit time was going to be an issue. Price was also starting to add up when we weren't even sure if this was going to work. Many bat aficionados spend a lot of money and many iterations developing ultrasonic microphones, and serious researchers spend even more. However, for us, this is a curiosity that will not see a whole lot of use after we play with it for a while, save for the occasional fault-finding mission. So, while we have added some links in the 'Reading and Resources' section, we have not purchased any of those microphones beyond the ones you see in the builds.
PROXIMITY SENSOR
It was around this point that we decided to strip down an ultrasonic proximity sensor used for Arduino, and use the receiver from it. We also used the transmitter for bench testing, but this has no role in the final use situation. Testing with a range of frequencies centred on 40kHz found that indeed 40kHz gave the greatest response, but this sensor had a better low-level response to a wider range of frequencies either side: It was still giving a useable waveform at 50kHz, albeit at a much smaller amplitude. Even so, it was a much stronger signal than that from the transducer for the same frequency.
ADDING INDUCTORS
Most of these elements are capacitive sensors. They have a tuned resonance but after hooking them all up to our LCR meter, we found that the transducer had a capacitance of 1.89nF, the proximity sensor unit 2.15nF, the piezo 22nF, and the electret, which has a FET amplifier stage in the way, has no externally measurable capacitance.
Just like matching inductors with capacitors to make a tuned circuit, it is possible to add the 'wrong' inductor to detune a circuit, too. That is why, in some of the photos, you can see 20mH worth of inductors set up next to each transmitter and receiver, as we repeated the test with each type of sensor, with the inductors. We found some observable difference, but could not identify either a pattern or a significant enough difference with either 10mH or 20mH to bother pursuing this concept any further. However, we later saw some designs which did include inductors, so some of our prototypes leave connection points for future experimentation.
MEMS MICROPHONES
One of the more promising options was the MEMS microphone. MEMS is an acronym for Micro Electro-Mechanical System, and it describes a range of very small but capable devices like microphones, speakers, and similar options which have both electronic and mechanical aspects. They are the microphone in most mobile phones, earphone and headphone sets, and the like. They are not available over the counter but are available from some online retailers and definitely from trade suppliers like Element14, where we bought ours. They generally have a very good frequency response and are very sensitive, making them useful for our project.
Stock is a bit of an issue at the moment, as we found later with the inductors, but we found some with a transit time of five to eight business days, held in a warehouse overseas (not China, they are often in the UK, Singapore, or the US if not in Australia). We ordered some options as well as a different type of ultrasonic sensor. Careful study of the datasheets is needed. Many MEMS microphones have a digital output in I2C or SPI: We do not want that, as only analog will work for our circuits. In addition, all have different voltage requirements and connection methods.
HETERODYNE OVERVIEW, THE FIRST ATTEMPT
This design ended up being a blend of several others because, as is often the case, many people have done this before and there are only so many ways to do it. One logical approach to this is to process the signal from a sensor with preamplifiers, high-pass filters, and then through the
mixer to add the signal from the oscillator for the heterodyne. That's exactly how most heterodyne microphone designs work, although some vary, maybe using gating or diode mixing rather than an op amp mixer. We based ours on parts we can get over the counter at retail stores, so we ended up with the LM324, a very common and reliable quad op-amp that happily works from a single supply. The rest of the parts are fairly basic common building blocks.
At the front end of our circuit is a microphone. Because this element requires careful choice and sometimes modification, we added a header so that it can be easily removed and changed out. That way, we could explore different options and find one that responds well to the exact bandwidth we wanted to listen to.
Along with the microphone is a preamplifier. This boosts the very weak signal from the transducer and makes it usable for the following stages. Next in line is a second pre-amplifier stage, and both this and the previous one have a form of high-pass filter on the inputs, which passes frequencies over roughly 20kHz, and grounds frequencies below that point.
After this stage is a mixer. This takes the signal from the pre-amplifier/high-pass filter combination and mixes it with the feed from an adjustable oscillator. This is where the heterodyning happens. All of this so far is built around the four op amps within the LM324 quad op amp IC. It's not the quietest or the fastest op amp IC around but that works in our favour by providing an upper bandwidth limit and therefore reducing the need for low-pass filters to cap the frequency the circuit responds to. There is an oscillator built around the same IC, tunable to produce frequencies from 16kHz to 107kHz (measured) to be fed to the mixer. Finally, following all this is an adjustable power amplifier stage to boost the signal to headphone levels. This is based on an LM386 amplifier IC.
Before committing to any soldering, we prototyped the individual sections on a regular solderless breadboard and used a frequency generator and oscilloscope to check the performance of each block against the theory. The first thing we discovered was more of a rediscovery: Solderless breadboards are terrible, particularly for audio frequencies and above and when they get a little old. Poor contact was the root of many of our deviations from theory during the process. We also experienced lots of capacitive coupling and leakage. However, we did verify that the preamplifier arrangement had very different amplitudes at its output for only a small change in input frequency, with all other parameters being left unchanged. By and large, the blocks of the circuit worked as intended, although we changed a few component values from our original plans as a response to testing performance.
With the breadboarding done, it was time to apply some solder. With a full breadboard layout completed, we transferred to the solder version and built up the circuit. The potentiometers are external, with a mix of 24mm and 16mm depending on what we had in the parts box. We soldered the circuit up and used jumper wires to connect the external components. It is worth noting that this circuit does not include the power option for an electret microphone, as yet. We hot-melt glued the board, battery snap, and potentiometers onto a piece of foam-core board to hold it all while we experimented.
PERFORMANCE AND LESSONS LEARNED
It was terrible. Using a known source by way of the function generator and another ultrasonic transducer, we tested the unit with a lithium-ion USB-rechargeable 9V battery. We immediately discovered huge amounts of noise coming through the headphones. We had been careful in soldering but any time a potentiometer was touched, we lost our hearing. That immediately told us that future builds would need to feature grounded potentiometers and plastic knobs to avoid adding interference. We also decided to use a shielded cable to connect them.
We also realised that the DC-DC converter inside the battery was probably cheap and noisy. Sure enough, swapping out the USB-rechargeable version for a regular alkaline 9V took a decent slice of the noise away.
THE ELECTRONIC ENGINEER
We tend to look for articles and discussion boards when researching circuits like this one, but sometimes videos hit the spot too. Such was the case as we tripped over Mark Donners, known as The Electronic Engineer, on the Element14 Presents YouTube channel. Mark designed a heterodyne type circuit that was also based on an LM324, but using an electret microphone as the input. It also used a different high-pass filter arrangement, and some component values around the amplifiers that are a bit different to what we calculated. It is worth noting that not only do we not know everything (the only people who do are kidding themselves) and are therefore open to expert opinions, but in addition, the relationship and ratio between the components in many parts of this circuit is what makes it work, rather than their exact values. Besides this, the circuits were similar, having been designed with the same criteria as ours.
Above is Mark's schematic, as published on the YouTube channel and in the supporting Element14 Presents project page. Mark also has produced a PCB which can be ordered from a Chinese manufacturer, PCBWay, to order. We really liked the idea of a dedicated PCB, to help validate the concept, and we didn't have to design this one.
So far, some of the problems we encountered had been attributed to issues with prototyping boards, as even the solder variety can generate some problems. Using someone else's tried, tested, and published design also helps us benchmark our expectations and compare problems. For
example, would the lithium-ion 9V battery with its DC to DC converter also induce noise in that circuit? So, we ordered some.
We built Mark's design largely as-is. We stuck with his component values, but omitted the speaker and opted for the headphones socket. Other differences were in the potentiometers, which were the wrong format and size to mount on the board, and the headphone socket, which had the same problem.
These are attached using shielded cable, and the potentiometer cases grounded. Incidentally, the pots used in Mark's design are grounded types, which we felt validated our experience with and assessment of our first prototype. We eliminated the switch because we are only using this as a test bed, and used a plug and socket for the battery instead. Also, the intended headphone socket is a switched type so we had to bridge it to use the speaker header for our headphone socket.
The other difference between Mark's build and our execution of his PCB is the microphone: We did not have the same part and used the header so we could use both the ultrasonic transducer and our electret microphone. We also planned to use this PCB to test MEMS microphones before moving on with our own design. It's worth noting that this design uses the voltage divider at the input of the first preamplifier stage to power the electret, something that is suitable for some but not others.
Using the MEMS units will require a separate power supply as most are 5V or 3.3V, but the electret should still work with the voltage divider.
OUR EXPERIENCE
The PCBs for Mark's design arrived before our MEMS units did, so for now, we tested what we had. The amplifier is quite sensitive, and with nothing connected to the microphone header, there is quite a bit of noise present in the circuit. Connecting anything across the header eliminated a fair bit of this. Initially, we connected the function generator to the microphone input and set it to 40kHz, at 50mV. This allowed us to test the rest of the circuit with a known source, and adjust the oscillator until a comfortable frequency was heard in the headphones.
For this test, we had available a generic electret microphone insert, an ultrasonic transducer, and an ultrasonic receiver desoldered from an Arduino proximity sensor module. For use as a test source, we set up the Wimshurst machine so that there was strong leakage between the electrodes but no sparks. This provides a solid source of ultrasonic noise and, if the gap is not too small, no audible noise. The first thing we discovered is that the electret microphone picks up almost nothing. The transducer was next in line but was surprisingly weak. The best performance, by a significant margin, was the ultrasonic receiver desoldered from the proximity sensor module. This one was the clearest, loudest, and captured the most detail.
OUR NEW DESIGN
With what we had learned from our first design and then Mark Donners' published project, we embarked on something different. We are going to build this circuit around separate ICs, both to choose characteristics that suit us and to enable easy swapping of sections of the circuit if something needs to change later. The functional block is similar to the first design in some ways, except that the two-stage preamplifier is now unfiltered, and there is a high-pass filter on its own followed by a unity gain buffer.
The schematic shows more detailed differences. The original set-up of a power supply resistor for an electret is still there, along with the set-up for the passive sensors, too. However, this time, the preamplifiers are based around the LM833, a fairly good low-noise dual op amp capable of working up to 16MHz, far above anything we need. The LM324 is still used, but it becomes the buffer, the oscillator, and the mixer. Then, the reliable LM386 amplifier is used to give good volume for the headphones.
There are more potentiometers on this version, too. All of them are aimed at customisation as different sensors are tried. We noticed in the initial build earlier that the amplitude of the oscillator output waveform reduces significantly as the frequency increases. Because of this, we added a potentiometer between the buffer and the input of the mixer.
This allows the signal input to be reduced to match the oscillator waveform amplitude, although it will have to be a guess because there is no reliable way to show the information. There is a potentiometer for the variable frequency oscillator, of course, and the volume control for the final amplifier stage. However, we also added adjustable gain, via a trimpot. This is so that, once you have an end-use, sensor, and other variables dialled in, you can adjust the gain to give better range from the volume potentiometer.
AROUND THE CIRCUIT
Working from the left, we have the battery and power switch, followed by a power LED and resistor R1 to limit its current. You may wish to increase this resistor value significantly if you want to use this unit mostly at night, to keep the LED very dim. R2 feeds current-limited supply voltage to the electret microphone connection, if it is used. C1 decouples the same. R3 and R4 form the voltage divider to set the bias for the non-inverting input of the first op amp, IC1a. Note that there is a bypass capacitor C2 which sets a filter, helping keep stability. Because of the capacitor, R5 is needed to avoid the capacitor absorbing all the input signal current. The inverting input of IC1a is connected to the output for feedback via a voltage divider formed by R7 and R6. To set a gain of 20, the voltage divider ratio should be around 1:19 for R6:R7. For a gain of 10, for example, the ratio will be 1:9 for R6:R7. You can change these values to suit your own experience.
C4 and C5 are power rail filters. They should be mounted as close to IC1 as practical. C6 decouples the output of IC1a, and ensures no DC reaches the input of the second preamplifier stage, IC1b. This has its non-inverting input biased by R8 and R9 to half the supply voltage, and R11, R10, and C7 form the feedback network. Both C7 here, and C3 in the previous block, ensure that the feedback circuit remains AC coupled and does not pass any DC. C8 and R12 form a high-pass filter set up for a theoretical value of 19kHz but with a measured value of around 17.8kHz after component tolerance was factored in. This helps ensure the following stages only deal with the target ultrasonic frequencies. That will be more important when using electrets and some MEMS units than with an ultrasonic transducer.
Following the high-pass filter, IC2d is set up as a unity gain buffer. You could add extra components here to give this buffer some gain, either negative or positive. However, we did not feel it was necessary. The buffer's main job here is to pass the signal from the filter, to the mixer, without any change of feedback from the mixer or any loading of the filter to change its characteristics. C9 decouples this output and feeds it to potentiometer R17, which is used to reduce the amount of signal sent to the mixer when the oscillator's output is reduced in amplitude at the higher frequencies.
IC2c is set up as a basic oscillator. R13 and R14 set its inverting input to half the supply voltage while R15, potentiometer R18, R16 and C10 are the feedback components that cause the oscillation. The output is sent straight to IC2b, which is set up as a mixer. The output of this is decoupled by C11, and fed to potentiometer R19. The wiper of R19 feeds the input of amplifier IC3, the LM386 power amplifier. The inverting input of this is grounded, while the gain is variable due to the 5kΩ resistor R20 in series with the 10µF capacitor C12. The bypass pin is unused. Capacitor C13 is another capacitor to help cope with supply fluctuations, while C14 is the output capacitor where the headphones feed is taken off.
BUILDING IT
| Parts Required: | ID | Jaycar | ||
|---|---|---|---|---|
| 2 x Solder Breadboards | - | HP9570 | ||
| Packet of Wire Links | - | PB8850 | ||
| 1 x 470Ω Resistor | R1 | RR0564 | ||
| 1 x 820Ω Resistor | R12 | RR0570 | ||
| 1 x 1kΩ Resistor | R16 | RR0572 | ||
| 2 x 5.6kΩ Resistor | R6, R10 | RR0590 | ||
| 1 x 10kΩ Resistor | R2 | RR0596 | ||
| 9 x 100kΩ Resistor | R3, R4, R7, R8, R9, R11, R13, R14, R15 | RR620 | ||
| 1 x 470kΩ Resistor | R5 | RR0636 | ||
| 1 x 5kΩ 25-Turn trimpot | R20 | RT4648 | ||
| 1 x A10kΩ 16mm Potentiometer | R19 | RP7610 | ||
| 1 x B10kΩ 16mm Potentiometer | R17 | RP7510 | ||
| 1 x B50kΩ 16mm Potentiometer | R18 | RP7516 | ||
| 1 x 680pF Ceramic Capacitor | C10 | RC5334 | ||
| 2 x 3.3nF MKT Capacitors | C3, C7 | RM7033 | ||
| 2 x 10nF MKT Capacitors | C8, C9 | RM7065 | ||
| 3 x 100nF MKT Capacitors | C1, C5, C11 | RM7125 | ||
| 2 x 1µF 63V Electrolytic Capacitors | C2, C6 | RE6032 | ||
| 1 x 10µF 16V Electrolytic Capacitor | C12 | RE6066 | ||
| 2 x 100µF 16V Electrolytic Capacitors | C4, C13 | RE6130 | ||
| 1 x 220µF 16V Electrolytic Capacitor | C14 | RE6158 | ||
| 1 x LM324 Quad Op Amp IC | IC2 | ZL3324 | ||
| 1 x LM386 Amplifier IC | IC3 | ZL3386 | ||
| 1 x LM833 Low-Noise Dual Op Amp IC | IC1 | ZL3833 | ||
| Sensor of Choice | See Text | AU5550/XC4442 | ||
| 1 x LED | LED1 | ZD0150 | ||
| 1 x 9V Battery | - | SB2423 | ||
| 1 x 9V Battery Snap | - | PH9232 | ||
| 3 x Plastic Potentiometer Knobs | - | HK7770 | ||
| 1 x Headphones Socket | - | PS0134 | ||
| 1m Single-Core Shielded Audio Cable | - | WB1500 | ||
| 1m Four-Core Shielded Audio Cable | - | WB1510 |
* Quantity shown, may be sold in packs. You’ll also need a breadboard and prototyping hardware. † The mounting hole for this switch may be smaller than the specified 20mm.
We built this version on two separate solder breadboards. There are longer versions of these around but the ones we could get over the counter do not have full-sized rows, with some of the holes missing. Building this way also allows the amplifier to be separate, which is handy for people building this who do not need the amplifier: If the output from the mixer is strong enough for sensitive headphones ias is, for example, or if the mixer output is to be fed into a recorder line in.
The build process was very standard, where low-profile components are placed first, then others in order of height. However, we built in stages, around each IC one at a time, because we do not have the benefit of a screened overlay like on a PCB. It's very easy to get a resistor or wire link one or two rows to the side of where it should be without the IC placed as a guide. We used PCB stakes for all of the external wiring points. You can see some near the power LED, which are for the electret microphone if chosen, or any other input which has a DC bias. Note that if you are using something that does have a DC bias but has its own power supply, like the output of some sensor modules, then remove the 10kΩ resistor. R2.
After building the individual boards, they were joined with wire links for the power rails, and the connection
between the mixer output and amplifier input. Note also that we have soldered screened cable to the bodies of the potentiometers, the other end of which is grounded to the circuit. After building and double-checking connections, we glued the board, potentiometers, and headphones socket onto another piece of foam-core board. In the long term, we will put this in an enclosure, but we still have much experimenting to do for now.
Initial testing was performed using a frequency generator connected to the transducer input, and an oscilloscope connected successively to each input and output along the chain. Once this proved satisfactory, the ultrasonic receiver that had previously been desoldered from the proximity sensor was soldered onto the end of a piece of screened single-core cable, the other end going to the transducer input on the board.
This was for future use, as we want to put the sensor in a sound-deadened tube so it can be used for locating point sources of sound. For now, the sensor was held by the cable, a little behind the body, and pointed at the Wimshurst machine set up with a gap slightly too big for sparks to form. While nothing is audible to the human ear (although that sense of presence or pressure mentioned in the introduction was definitely noticeable), donning the headphones and pointing the sensor at the gap yielded a thunderstorm of cracks and clicks. Tuning the level and frequency potentiometers cleaned up the sound and gave more definition.
SUMMING UP THE HETERODYNE VERSIONS
So far, this is our favourite of the heterodyne designs we tried. It was more fussy and difficult to design and build than Mark Donners' version, and may be needlessly complex. However, we wanted distinct blocks in the design to leave room for experimentation. For example, the high pass filter components can be very easily swapped out because they are not part of the amplifier feedback network and therefore changes are independent. This enabled a high degree of experimentation, with confidence that a small change would not affect the entire circuit. The overall product was quite sensitive and very tunable. It still suffers from some stray nose and a lot of background hiss, but by all accounts, most heterodyne ultrasonic receivers do.
Unfortunately, the sheer number of hours, and amount of energy, plus the experimenting, number of iterations, and the amount of fiddling, are not reflected in the size of the article. It would be rather boring to see every version, every experiment. We tried many amplifier configurations with just the frequency generator to see how they behaved, before looking into sensors. We also tried a great many oscillator configurations to see if any gave a decent waveform output or, even more desirably, a consistent amplitude across the range. However, none were found to be suitable. We did have something at one point that had a reasonably clean sine wave output with a fairly consistent amplitude, but it had more components and op amps than the whole final circuit combined!
FREQUENCY DIVIDER
We somewhat deliberately chose the harder option first in building the heterodyne ultrasonic microphone. We knew we were going to need a mental break after it!
WRAPPING IT UP
That's all for this month. We're going to keep developing this design, and work on a frequency divider version, too. They work on a different principle and have different advantages and disadvantages. We'll also explore the MEMS modules to see how different they are with both circuits compared to the plain sensors, and also see what we can achieve with the addition of some inductors to detune the sensors. We'll also make an enclosure for at least one of the designs, and try them out in the field. That is, if the weather is more bat-friendly than it has been this month!