I thought it might be interesting, for beginners to see how products are developed, by the unsung hero's of industry ... the Research and Development engineers. The image created in most peoples minds is that the average R&D 'rat' wears a white coat, is highly educated and intelligent, and never to be seen without a top pocket over flowing with coloured pens or with calculator in his hand. Ho, Ho, Ho !. In reality, he is more likely to be dressed like a refugee from an Oxfam skip and give the impression that he has just escaped from somewhere serious !. Companies usually hide the R&D department somewhere out the back, so the inmates cannot frighten visitors to the factory. Senior management live in fear of them, because their appearance usually heralds really bad news, that they will inflict slowly ... with a malicious relish. Gross misbehavior is usually grounds for instant dismissal of employee's ... except for R&D rats, who always behave like that !.
The Technical Director.
Technical directors are people who have already reached their own level of incompetence and now have to live out all of their hopes dreams and aspirations through others ... ie the R&D department, where all those horrible disrespectful rats hang out. The technical directors worst days are when he actually has to go put wonderful new ideas to them and ask for yet one more miracle to keep the company afloat. This is traditionally done after a well lubricated lunch on his expense account, because very few Technical Directors have the courage to face the rats sober !.
"Hi Chaps, there was a meeting of directors and we have a 'breakthrough' on a new product idea. As you know our competitors have launched their new Audio Spotlight range and the directors think that 'we' can do better. Sales are particularly excited about this 'breakthrough' because the Spotlight sells for £700 and our sales department thinks it would sell really well, if you guys could come up with a design that we could sell for £7 !. Oh yes the directors also thought it would be a good selling point if your version was , MP3, MP4, MP5 compatible and also played games !."
Having thus endured this humiliating experience, the Technical director will retire to his sumptuous office ..... to sleep it off. Two minutes later, the sales director appears in R&D, with a amended, revised specification list for the new product, that no-one has thought about yet, only to retreat very quickly when assailed by insults and assorted missiles. Didn't anyone tell him that R&D is hallowed ground ?. The last time a chap from sales got lost and ended up in R&D he was held down and snogged to death, for ten minutes, by the fat guy with bad breath, a beard and a great sense of humour !
The Audio Spotlight and HyperSonic Sound devices.
Although this a satirical look at product development, I decided to base the story on a real product, so that newbie's can see how we go about creating new products, in the real world. In this case we are not starting from scratch as two companies have already solved a lot of the problems and have patented their work. First let's have a look at these existing devices......
The original was invented by F. Joseph Pompei about 2000.
Another version appeared about 2002, invented by 'Woody' Norris.
Both of these devices sell for around £700 each.
As a personal message to Mr. Pompei and Mr. Norris, I don't think that you have anything to worry about in terms of competition !.
One way of finding out about a competitors product, is to buy an example and then try to 'reverse engineer' it. This allows us to take it to pieces to see how it works, measure waveforms etc., so that we have a complete understanding of the way it functions and it's limits. We would also buy up a copy of all patents related to the device, but patents do not always tell the whole story. Theoretically a patent should contain enough information to allow "One versed in the art" to build a clone, but that can be interpreted as having the same level of knowledge as the inventor, which they desperately try to conceal. Patent language also tends to be very vague, in order to also encompass new unexpected developments, as well as those incorporated in the device. So right from the beginning the R&D engineer is very aware that anything he comes up with should not infringe existing patents, or that it can legally circumvent them. Also one must bear in mind that even if a patent is issued, it does not prove that the underlying idea is complete or that it will work, for example Stanley Meyer was granted over 40 patents in both America and Europe, based on his claim that over unity power generation was possible and yet we are still dependant upon on fossil fuels.
The Internet has considerably simplified the next step and that is to gather together every published document and video on the development. For example one article may reveal that sine waves are used and in a video that frequencies in the 60 kHz range are used. A patent may indicate that one problem area with this type device is distortion and demodulation products. Health and safety requirements may limit the range of the device, for the threshold of pain is 90 db's and permanent damage to hearing may, or will occur at higher levels. It is during this stage that we start a list of things we are going to need, in our investigation. The first is a calibrated noise level meter capable of working up to about 100 kHz and a copy of the regulations concerning ultrasonic exposure times.
Although the product we are looking at is an ultrasonic projector of audio, similar techniques are involved in other fields , such as range measurement, crowd dispersal, HI-FI audio equipment, female sexual stimulation, ultrasonic weapons etc., so we have to also consider the latest developments in those fields too.
In many companies, it is not R&D's job to design a final product, but merely to produce something that can be developed into a product by someone else. This usually results in a working 'proof of concept' model being produced by R&D and at this stage other awful people people get involved, so called 'designers' and button counters, all intent on undoing all of your careful development work and turning a potentially excellent product, into a pile of junk !.
Sales will also appear again with their revised, revisions of the amendments, to the issued revised specification ... which means everyone has to go back to square one and start again !. It is a good time to plan a visit to the Technical directors home pub one lunch time ... and give him a humiliating, public thrashing at snooker !.
It can be argued that the underlying principle used here, is that of the Superhetrodyne principle patented about 1931. It is based on the fact that if two frequencies F1 and F2 are mixed together in a non-linear device, the end products will be F1, F2, F1- F2, F1 + F2 etc, etc. The key to this application is "a non-linear device". In this application the non-linear medium is air. So if we beam a 40 kHz ultrasonic signal at someone's head .... and at the same time beam a 40.1 kHz ultrasonic signal at the same head ... if a non-linear device was present, then the head would hear a 40.1 - 40 kHz difference audio signal .... ie 100 Hz. If one of those ultrasonic signals was varying in amplitude at an audio rate ..... ie the latest top of the pops ..... that is what the ear would 'hear' !.
(There a small organ inside the head, which has the size of a pea called the Saccule, is reported as having the ability to demodulate ultrasonic signals with stimulation levels above 100 db's, but is not involved in the systems we are discussing).
If we look at the de-modulation products a bit more closely, I have over simplified things and many more products are produced that can cause the brain to subjectively assess the air demodulated signal as 'distorted'. It would be impractical to consider modifying the human target to correct, or eliminate this 'distortion' , so any compensation must take place in the device itself. Now since I do not have an example of the device to examine, I am going to have to simply identify distortion as a possible problem area, for which I want to provide as many options as possible, into my initial test rig. One option I want is to make possible the testing of different modulation techniques. I also want complete control of as many parameters as possible, so if I can work a PIC in I will. Unfortunately I will not get down to 60kHz in BASIC, I think I need to try it on a 60 MHz 18F25K22 and see how low PWM will go.
PWM produces square waves and my suspicion is that we will need to use sine waves for both beams. So right now, I decide that both of the channels will use square waves and not to convert them to sine waves until the very end. What I can do, is start off with a 5 volt to 0 - 300 volt variable switched mode power supply and use a simple push-pull stage to switch the HT voltage to give a variable 0 - 300 volt peak to peak square wave, which then needs to be converted into sine waves. I estimate one square inch board space for the SMPS. Next we need to consider the nature of the ultrasonic transducers.
The Ultrasonic Transducers.
The Audio Spotlight uses electrostatic speaker like transducers ?. Another option may be to use cheap piezo electric tweeters if they are capable of working at the frequencies we choose. So it is time to collect together the data sheets for available transducers. The electrostatic type transducers need several hundred volts to make them work. Piezo electric transducers also need high energising voltages to produce useful outputs.
A problem may be that piezo electric transducer can have very 'peaky' response curves and for greatest efficiency need to be operated on the top of one of those peaks. This has nothing to do with the bird in production with the big tit's, four undone buttons on her blouse and a face that none of the male workers seem to remember.
At the back of my mind is the thought that I may need to add resonance to any output stage, in which the square to sine conversion takes place, which means allowing PCB board space for an inductance, to allow the output tank circuit to be tuned.
The first test rig.
I am thinking in terms of two identical PCB driver boards, with provision for modulation on both of them. Each board will have it's own xtal controlled frequency generator (PIC), and 300 volt push pull output stage. Both boards will also have provision for the two square to sine wave conversion components. Lot's of test point pins for the scope etc.. I have done the 300 volt PSU already, to see if electrocution works on the Technical Director. 300 volts will be more than adequate for the commercial piezo electric transducers but something closer to 1500v will probably be needed for an electrostatic transducer.
If the demodulation does take place in air and the end result sound pressure waves, then it ought to be possible to use a microphone as a remote target and feed the received audio back to the operators control position, for equipment adjustment and analysis?.
End of day one.
Day one has mainly been about thinking and questioning everything that is presented to him. All kinds of ideas will be going through his head, most of which his experience will cause him to reject. Some of the things he does accept, others may not have thought of. Some may be money saving short cuts, some a realisation of killer problems and very often idea's for new products. Company culture does not allow him to voice his thoughts out loud and even if he did they would be ignored, only to re-surface at a later date, as the Technical directors latest brilliant revelation to save the company and the world in general.
In retaliation, the rats spread all kinds of rumours amongst the Technical directors stooges, a sort of April the 1st thing that lasts all year. The competition being, to see who's idea induces a mental breakdown in management first. Of course, we must not forget the high light of day one, which is blindly picking out some completely outlandish and useless component, that has to be incorporated into the new product design !. You get extra points if it is terribly expensive !.
Before switching off the lights and going home ... one last check on all of the booby traps you have set for the technical director, who thinks no one knows he sneaks back at night, to find out what you have been up to all day !. One such plot involves those tiny little paper dot's that get punched out of paper tape recording machines. What you do is fold up a piece of paper containing several thousand of these dots, then enclose them in an open envelope marked ... "For the personal and discreet attention of the technical director" .... then hide it in the desk drawer of someone you don't like. If opened, it takes hours and hours to pick them all up, to conceal his sneaky underhanded ways and by then ..... the pubs will be closed !
Day Two. (1st February 2011)
The Rat saw the same two video's that you did ... and like you, missed the obvious, probably because he was sober on day one and this usually impairs his creative senses. A flat framed electrostatic speaker radiates from both sides of the diaphragm, and since the people in the video are holding the speaker a few inches from their head, are they frying their brains, or is someone lying about using a 110 db beam and electrostatic speakers ?. Get on the internet and find some data on the absorption properties of thin backing materials at 60 kHz !. At 50 kHz the beam attenuates at a rate of 1 db per metre.
R&D rats change their minds all the time and if they voiced their thoughts out loud, some of them would sound crazy, so let's see how the thought process goes. I am going to have to give the piezo electric transducers more than a fair chance of working, even if I have to drop the working frequency down to 40 kHz ... and all because of the cost of the ferrite ring transformer that would be required for the Electrostatic transducer PSU. Time to ring around the companies and get some sales people in with some free samples. Must get some confidentiality agreement forms made up. Ask purchasing to get a pair of piezo horns and two 4" piezo tweeters to play with, on the test rig. Cannot find a reasonably cheap sound level meter that will work up to 60 kHz, so may have to make one. Must download "The ride of the Valkyries". Playing that all day long ought to send the software guy nuts and he is the drinking buddy of the technical director !. Must read up on complex modulation systems. Probably need an equaliser that will work from 300 Hz up to about 60 kHz, with LOTS of bass boost !.
The switched mode 0 - 300 volt PSU for the test rig. The bottom pot is the voltage level control. The top pot set's the maximum voltage range. Works great and stabilises to within 20mV. Have to drop this now, because I have some important research and development work to do ..... on a rat trap.
Links of the day .....
Day 3 (2nd February 2011)
Yesterday I didn't have a clue on the best approach to this design problem and this is perfectly 'normal'. People like to think that you know what you are doing, but the truth is that you are just like everyone else, floundering from one perplexing catastrophe to the next. The first week is the worst because it takes time to develop a vocabulary of 'buzz' words, that you can use when people ask you technical questions about the project. The trick is to only give very vague answers and hope that they will go away,, or throw the ball back into the end of the court, by asking them if they had they investigated the companies legal product liability situation concerning health an safety on this product and had they considered the practical problems in addressing the third harmonic distortion problems that caused the Japanese to abandon research on this type of product in the first place.
Actually the third harmonic distortion problem, is a real one and the reason we don't yet use ultrasonic speaker systems for Hi-Fi. The solution is to use single sideband modulation on the beam, which while not a perfect solution, at least makes the idea work ... but we are not going to tell anyone that !.
The technical directors secretary just appeared to say our illustrious leader had an accident in the factory last night. The hospital don't not think that any of his fingers are broken, but they are too swollen for him to drive. How sad !.
OK, so on my secret list of things to do, I have to add ....'search for cheap balanced modulator chips'.
I probably ought to explain that all rat's in R&D has to maintain a bound work log book, It is basically a record of what you have tried and any developments that may lead up to the company making a patent application. From the rat's point of view, it contains the evidence that supports his work and his decisions. For example, if you develop a lamp housing, the all of the test results will be in your log book, so that when the designers subsequently screw up your work, you have the evidence that you were not responsible. All mine say's so far is get a really strong lock for my filing cabinet, so that I can safely lock up my evidence and keep everyone wondering what I am up to. It is also a good place to keep ones homer's. These are private developments for alternative products, just in case you get the urge to save or sink the company at a later date. I also keep my jam sandwiches in the filing cabinet.
Still no luck in finding dirt cheap ultrasonic transducers that will work at 40 kHz, with sufficient power for the application. At this stage if anyone asks me about transducers, I will simply say that unless we can come up with a miracle, we will probably have to buy them in from America, which is guaranteed to ruin their day !. Secretly I am coming to the conclusion that a single transducer per channel may not work. If you asked me to explain why, I simply couldn't say at this stage. I do have a gut feeling that the power handling is in proportion to the total area of the ultrasonic 'emitter'. Against that, larger area transducers are moving away from the natural wave length's at 40 kHz, ie .... 8.6mm . The larger transducers are also more inefficient because the diaphragm's weigh more, have a greater resistance to air etc.. So rather than one single large transducer, we are probably looking at an array of much smaller ones. (A slotted electrostatic array comes to mind).
This example has the transducers mounted on a parabolic shaped backing which helps focus the beam. The problem here is that this type of transducer has a radiation angle of about 80 degree's so it is barely 'directional'.
Probably a better approach is a slotted grating using a single sheet of piezo electric film. These transducers come in pairs, marked T (TX) and R (RX). The maximum driving voltage for the TX device is 20 volts peak to peak, which gives an output of > 115 db's SPL. I don't think this is the transducer for the project BUT .... it may be suitable for an alternative product idea, I have on my homers list !. So I can kill two birds with one stone and put some of these transducers on 'life test'. That is I will connect a batch of these up to a 40 kHz, 20 volt peak to peak signal and run them 24/7 until they self destruct. Life testing is a very important because customers expect a product to last forever and manufacturers hope that the will last at least the warranty period !. Life testing results will be entered into my log book. I think I will point the life test transducers at the software guy's cubicle and see if they produce any useful adverse effects, that I can use on his boozing buddy ... the technical director !.
A parametric loudspeaker, that presents remarkably narrow directivity compared with a conventional loudspeaker, is newly produced and examined. To work the loudspeaker optimally, we prototyped digitally a single sideband modulator based on the Weaver method and appropriate signal processing. The processing techniques are to change the carrier amplitude dynamically depending on the envelope of audio signals, and then to operate the square root or fourth root to the carrier amplitude for improving input-output acoustic linearity. The usefulness of the present modulation scheme has been verified experimentally.
Day 4. (3rd February 2011).
Over my computer I have a poster which say's first we make it work ... and then we make it simple !. Making something simple invariably makes it cheaper and because simple designs have less parts, they tend to be more reliable. However that is for the future. I have scrapped my previous idea for two PCB's and decided that to put each electronic stage on it's own 'board', but in a way that the boards can plug into each other in cascade (use headers). It will make design changes simple and hopefully make the test rig small enough to fit into the filing cabinet at night. As my granny used to say "Mis-placed trust is a self inflicted injury".
Idea's continue to run around in my mind and up to now the block diagram for one channel of the test rig looks something like this ......
If I start off using the Futurlec transducer, it has a maximum rating of 20 volts peak to peak, so I can use the SMPS to provide that, but now have to consider a linear class B power amp. Lots of variables, lots of options. The mic amp also needs a signal generator input and a music source (MP3)
I am also going to need a dummy head circuit. Audio microphone, amp and audio level meter. Also I must get a dummy polystyrene head, because bull-shite baffles brains !.
OK there is a germ of an idea in the back of my mind, that started off as a thought about telecom CODEC chips, which I need to think about more. I think I will go for a walk around the factory and annoy people.
Day 5. 4th February 2011.
I woke up with two thoughts this morning, one was objective measurement of the distortion and the other was patents.
Actually measuring the distortion, is straight forward. Even if you don't have the required instruments, you can make your own and I will probably end up doing that in this case, because the technical director hates spending Company money, on anything else apart from his extended boozy lunch breaks !. Incidentally the 'pub' is where the technical director and his slimy buddy, the software guy, make all the companies important decisions. The company premises, is where we totally ignore them.
Having worked out how I can measure the distortion, means that I now have an indicator of which direction I need to move in ... concerning any element of the project, so I am quite pleased with that and am rather looking forward to getting started. It also means that providing I do not suffer a case of 'premature ejection ' from the company, I will have plenty to do, to fill out my days for the next nine months.
Patents. A product can be successful on price alone, providing a mass market exists, or can be made to exist. However it is always a nice selling point if you can prove your product is 'leading edged' by securing patents. The patents may be for something radically new or simply something that caps the competitions existing patents. Parametric speaker systems are still rare beasts, in a high value market and there is plenty of room for improvements. If you think that you can produce those improvements then, and you want to apply for a patent for them, your work has to be done in secret, from the very beginning.
Often R&D work in ad hoc teams, perhaps someone from software along with a hardware engineer, plus an expert in the subject field. In this case it is becoming obvious that the technical director wants to turn the whole project into a 'dog and pony show', with him as the visible head of it. By the end of the first R&D project meeting everyone in the factory and the county, is going to know what 'we' are working on and everyone is going to talk. At this stage no software is involved and there is no tame subject expert. Basically, only one R&D rat will be required to see the potential product through to the 'proof of concept' model stage and it should mean that apart from that person and the Technical director, everyone else should be kept 'out of the loop'. It is not going to work, because the technical director will reveal all, at his very next pub 'briefing' with his drunken cronies and the local yokels. The only way to stop this requires a very bold, courageous, strong willed individual ..... who does not rely on his job, for his long term survival.
I suppose I had better bring my CV up to date !. I wonder if they still sell bear traps ?.
Day 5 1/2. 5Th February 2011.
Today is Saturday, when every other sane person in the world, forgets about work and it's problems. Also you may wonder why I appear so aggressive while I am in the working environment. It really does not seem so long ago that we had to sack 75% of our work force, because the companies leading product failed. It was a pile of crap and everyone knew it, but no-one did anything about it. The company ended up being sued on two continents and the workers paid the price for this, not the perpetrators, who both kept their well paid job's. When I said goodbye to all those people, I swore that I would never let it happen again. No real surprises here, why Britain is in the state that it is in. Some people have some funny ideas about the concept of 'work', for me it simply means trading 'some' of my time and 'some' of my expertise, for 'some' of their money. I am not there to prop up anyone's ego, or to do anything that conflicts with my own integrity. I also believe that the guilty should be punished .... as often as possible !.
Since I am going to do this objectively I will not need the microphone amp at this stage, so I will be feeding the AF signal generator straight into the 12 db's per octave filter. However the actually cut off frequency for minimum distortion, has to be found empirically. It therefore makes sense to make the filter tunable over the expected range, ie for speech (300 Hz - 3300) Hz so decided to make the filter cut off frequency variable. This circuit is a Sallen and Key, low pass filter, with a cut off slope of 12db's per octave or, in other words 20 db's per decade. The actually cut off frequency can be varied over the range 250 Hz - 2.5 KHz, by adjusting the twin ganged potentiometer. This range may prove to be a little on the high side but if I need to move the control range, up or down the frequency spectrum, I can do it by changing the values of the 6.8nF and 5.8nF capacitors both together. Eventually we will be looking at expanding the required spectrum to include music etc..
The upper side band modulator is starting to look like a problem area, because there are so few types available and some have prohibitive costs (up to $22 each !). For a 'proof of concept' model I don't have to worry about cost, but I see it as a possible project killer, so I might as well deal with the problem now, rather than leave it to surface in nine months time. There is one chip that retails for 50 cents and I need to get some to play with, to see what I can squeeze out of them without a high component count. Related to this is the fact that double balanced mixers, in addition to suppressing the carrier, do unfortunately generate both the upper and lower side bands. My gut feeling is that the unwanted lower side band is going to contribute the the distortion problem, to one degree or another, so I do have to consider ways to eliminate the lower side band right at the beginning, (just in case!). There are methods to do this, but all involve added complexity, or cost. This is one of those problems that will solve itself with time.
I had a little problem, in that I was concentrating only on the one channel and it took a bit of time before I wondered exactly what the other channel was supposed to be doing ?. The clue was that the distance between the hardware transmitter and the human receiver is exactly the same on both channels.
Day Off. 6th February 2011.
Spent most of the day trying to find balanced modulator data sheets on the internet. I also listened to examples of other peoples experimental equipments and they most have the distortion problem. It was very difficult to judge by ear, exactly what was causing the total distortion, but did make a mental note to include a speech level compressor, on the microphone amplifier board, to prevent over modulation of the carrier.
Most experimenters had made up their output transducer assembly from the common Futurlec style 40 kHz transducers. Just as a matter of interest these do not peak at 40 kHz, something worth bearing in mind when thinking about xtal controlled oscillators.
I also did some more thinking about the need to make the sine wave carrier oscillator xtal controlled, yet variable in frequency !. The frequency steps do not have to be extremely fine ... about 10 Hz. The software can set the centre frequency to about 39,000 Hz and then we can step up, or step down from there. A lot depends upon the parameter spread of the transducers (2).
I have also been thinking about the output amp board, to get a ball park figure for it's output power, so we can get a couple ordered. I am thinking that we ought to take advantage of the fact that the output impedance of the amp will be 8 ohms, while the Z of the transducers will be much higher and require a Z match. It also means that maybe we will not need to have a 'high' voltage PSU and can run the output amps off the common supply rail. Again a compromise will have to be found because of the relative cost of any solution.
The technical director will probably be in tomorrow. I wonder if I ought to ask him for my rat trap back. I hope he didn't damage it !
Day 7. 7th February 2010.
It appears that I have to attend a disciplinary meeting today, to discuss my future with the company. That should not take long. So at the moment I am not quite sure what I will be doing next week, perhaps gardening.
As you can see I resisted the temptation to dive in until I had thought out all of the foreseeable problems leading up to the final product, so maybe I can dwell on some of those, just to fill in the time before my planned execution today.
With a selling price of around £700, it makes for a very expensive toy, if all you want to do is create voices out of thin air, which seems to be the theme of most video's on these devices. I think it is pretty clear, that no UK company is going to produce one for £7 ... so a reasonable question, would be could we design a device that could sell for £70 and if we could, would there be a market for it ?. The answer, I think is yes. Next question is what is the nature of that market ?. That leads us to the question of what possible applications can we find for such a device, that would sell at £70. The really BIG question is can we find a mass market for such a device at that price ?. Again, I think the answer is definitely YES, but I am certainly not going to discuss the answers at this stage with anyone.
Now if we consider the above block diagram, it will work as a toy, but is incomplete for a serious mass market application, three things in particular are missing. Add those three things and there will be no market competition and sales could well be into the millions bracket.
In most of the video's showing the audio spotlight in use, the operators are invariably whispering !. In some other applications they may have much higher levels of audio input, for example music at teenage levels. The purpose of a signal compressor is to amplify the lower signal levels, so that they can be heard, while compressing higher levels that could cause distortion. Of course you can design a circuit to do this, but specialised chips exist for this task. A typical type is the SSM2165-1, which can give up to 40 db's of compression. It is available in an easy to use 8 pin DIP package. The end effect is since the modulation level can be set higher, without running into amplitude distortion, the human target will hear the demodulated signal 'louder'.
08 February 2011.
The story ends, but I take the red pill anyway, because the the subject of parametric speakers interest me. They interest me because I know nothing about them and I find the best way to learn, is to write about things. I did wonder how many other people in the world were actually working on parametric devices and so far found only one gentleman in Japan. Last night I sent him an email with a link to this page and he very kindly sent me the following reply ...
"This page is interesting.
He collects known information well.
But there is not important factor for building speakers.".
Right at the beginning of the story (Getting started) I said that I was going to do the whole design digitally and then I got lost and looked at the analogue approach instead. Well all I can say is that I am 70 years old and can get lost if I want to Hi!. I think I got interested in the analogue route because I knew nothing about balanced modulators and the subject of the distortion interested me. A digital version does have a lot of advantages over the analogue, it is simpler, cheaper and much more efficient. Before considering what a digital version looks like let's have a look at the ultrasonic speaker system. Please bear in mind I have never seen one or heard one and know very little about them.
In the story I chose to use 40 kHz ultrasonic transducers to make up an experimental speaker system. I did that because they are relatively cheap and easily available. They are also 'scalable' in that I can use a single one, or make up a large and powerful array using many. For the private applications I was thinking about I only wanted a range of about two metres and would probably have started off by trying a single transducer on each channel. Let us first take a look at the specification of the Futurlec transducer. You can find the complete data sheet at ....
Starting at the top I note the manufacturers centre frequency tolerance of plus or minus 1 KHz, To see if this is going to be a problem we need to have a look at a graph of output against frequency ...
The first thing I see is that this particular one peaks just below 39 KHz, so I am thinking maybe I ought to make the reference oscillator variable. Otherwise the peak looks flat enough for the application. However, I am also mindful that in a multi transducer array, this spread will produce a 'plateau' between 39 KHz and 41 KHz.
The sound Pressure level is next on the list and that is given as 115db. That is fine with me as I don't need a lot of power for a test rig. I also note the effect of temperature and humidity on the output. What does jump out at me is the radiation angle of the transducer, ie 80 degree's.
It is directional, but for many applications a more directional transducer would be required. Even if we mounted the array of transducers onto a parabolic surface we are still going to end up with a very broad beam and that is not what we want for several reasons. Ideally we do not want the beam to be much wider than the target, because anything outside of this will be a waste of power and in some applications can affect objects adjacent to the target. For example in crowd control, you want to take out the ring leader, but not the little old lady standing next to him !.
Having said all that, this transducer does allow a workable speaker array to be made, but cost and the beam angle would cause me to look for a better solution, for long distance applications.
The next thing on the list tells me that the transducer has a capacitance of about 2100 pF. Now this is one of the very few things I knew about these transducers. So I make a note to work out the value of the inductance that will give resonance at 40 kHz and also I need to know what Q that combination will give me. The reason I need to know that, is because the Q of a resonant circuit means that in the series case we could get a lot more voltage out of the tuned circuit than we put into it and the specification for the transducer, limits us to 20 volts RMS or 28.3 volts peak to peak. So somewhere in our driver circuit we need to have a level control. As a little illustration of the danger here, take a look at the graph below.
I have taken a five volt square wave and fed it into a simple LC filter. At resonance I get a pretty good sine wave out at 30 volts Pk to Pk. (The oscilloscope probe as set to divide by ten). This tells us several things and the first is that we do not have to take the analogue route. Second, is that a first order filter works pretty good, but I will make a note to look at this sine wave on a spectrum analyser.
You will notice that the specification sheet does not give the transducers impedance at 40 KHz, so that is something I need to find out before I can work out exactly how I am going to wire up a transducer array and drive it. Otherwise I can make up a PCB the size of the array and that material will increase the back to front ratio, and reduce the level of radiation to anyone holding it. OK I checked and found that a typical transducer of this size and power has an impedance of about 1000 ohms at 40 KHz. Two samples had impedances of 500 and 1500 ohms, but most were 1000 ohms. The importance of this will be seen later.
Now, I have a problem going around in my head and that is because I know nothing about the behaviour between two ultrasonic beams and the effect on the air demodulation. Imagine I separate the two transducers, air demodulation will only be possible at the point the two beams intersect. At that point, some of the energy in the two beams are converted into a sound pressure wave, which is what the ear hears. So my crazy thought is that at the interception point air is absorbing energy from the two ultrasonic beams, the question is how much ?. You might wonder ... so what !. However consider the case where the two transducers are close together, which is the way most commercial types work. Since the two beams overlap along their complete length, logic suggests that energy is being extracted from the beam along the entire beam length. If that is so then it may be a limiting factor for long range operation, but not a problem in my application.
I just noted that the Futurlec ultrasonic transducers are marked "T/R", which is interesting, because most are marked with either a "T" or an "R" (sometimes with an "S" and an "R". If theirs are dual function then it solves the problem of them being sold in matched pairs, one for transmitting and one for receiving, otherwise you end up with a pile of unwanted receive only transducers. Futurlec also sell other ultrasonic transducers including a 60 KHz water proof version capable of being driven up to 140 volts Pk to Pk !. Yes the above are dual purpose !.
A look at the design of a digital parametric system.
Any digital design is going to be based on the class D amplifier. The very first class D amplifier I ever saw was Clive Sinclair's X-10 amplifier which was belatedly marketed in December 1964. It was actually designed by a chap called Gordon Edge and caused quite a stir at the time because it was so 'different'. At that time the 'standard' Hi-Fi amplifier was the Mullard 5-10, a 5 valve, ten watt job, built like a tank !. Compared with the Mullard amp the Sinclair X-10 pcb looked positively small, simple and cheap. Unfortunately it did not quite live up to specification claimed for it, although the next model, the X-12 did. The main lesson learned from this is that although the class D amplifier looks simple, it is much trickier to design a good class D amp, than it is a good Class B amplifier. Pop, click, shoot-through, radiation and quantitisation distortion being the main problems. However when you have a good design, they can be very impressive on several counts, ie size, efficiency, cost, weight, and performance.
If you GOOGLE "Class D amplifiers you will find some excellent articles written by people who actually know what they are talking about. As usual I know very little about them, so lets have a go at designing a practical class D amplifier for use in a simple sound spotlight. Our starting point is a microphone and the simple circuit uses an Electret microphone capsule, which requires power provided by the 47 K ohm resistor. Next we need to amplify the small audio voltage from the microphone, up to a useable level, by means of the simple single transistor, working in the classic common emitter mode. At this stage I am going to tell you a little story, the circuit is actually part of a burglar alarm I made many years ago.
The idea was when a neighbor went on holiday, they could lock this device inside their home and it would listen to any noises inside the house, ie burglars etc. The resultant audio was then used to pulse width modulate an 80 KHz carrier wave which was fed, via a small transformer, into any mains socket, between Earth and Neutral. Whoever was keeping an eye on the house would have a small receiver, that plugged into a mains socket in their home and any disturbance in the target house would be heard, identified and the police called, to apprehend the intruders. Since the signal was fed between the Earth and neutral cables, it did not matter if the two houses were on different phases. Worked quite well except the circuit needs a level control because it was very sensitive.
( Warning, Not all house sockets are wired correctly, be safe not sorry!).
OK now let us look at a class D amplifier driver circuit ....
Exactly the same circuit !. The audio from the microphone is amplified by the transistor and then fed into the control pin of the 555 integrated circuit, where it is converted into a 9 volt Pk to Pk pulse width modulated square waveform at 40 Khz. The carrier frequency is set by the potentiometer. You will note that there is one resistor without a value and that has to be selected experimentally to give a 50/50 duty cycle at the required frequency. I would start a 22K ohms and work up or down from there. The transistor was a BC109C. The pulse width modulated 40 kHz output is available on pin 3 of the 555 timer.
Now in our application we now need to change the 40 Khz PWM signal into a 40 KHz amplitude modulated sine wave that can be used to drive the ultrasonic transducer. For that we can use the simple, first order filter made up of one inductor and one capacitor that are resonant at 40 Khz. The digital approach is as simple as that. Ok, this particular circuit will not win any Hi-fidelity contests, but it is cheap and cheerful.
Class D amplifiers are appearing in more and more domestic products and therefore the applications are well supported by special function chips, that do just about everything for you including avoiding the common class D design problems listed above. They are also available in a range of output powers, commonly from less than one watt up to several hundred watts, and one of them is going to be ideal for our final design. Before we can go any further we have to consider how the Class D amplifier is going to be coupled to the transducer !.
Class D amplifiers are typically intended for use as audio amplifiers and have out put impedances of either 4 or 8 ohms, let us assume we are using one with an 8 ohm output. The big problem is that a typical ultrasonic transducer has an impedance of around 1000 ohms ... an enormous mismatch !. The problem also involves driving the transducers for maximum output and as we have seen from the data sheet, the drive voltage needs to be close to 20 volts RMS, so we need to bear that in mind. Another factor is that we will not be using a single transducer in the spot light application, but many of them assembled into an array. This means that we can wire them into any configuration, ie series and parallel, to give various resultant impedances for the assembled array. To get started let us look at matching the 8 ohms output to a single 1000 ohm transducer. We could, of course do that with a relatively small transformer . The formula is Zp = N²Zs where Zp = the primary Z (impedance), Zs = the secondary Z and N = the turns ratio. So if we transform this to give us the turns ratio N = The square root of Zs/Zp = The square root of 1000/8 = 11.18 : 1. Using that and the impedance of one side we can work out how many turns of wire we need on a ferrite ring core to make the transformer.
We could reduce the resultant Z of the transducer array by wiring them in parallel. For example if we had 128 transducers wired in parallel, the resultant Z = 8 ohms .... BUT we also need to consider this in terms of the wattage of the amplifier and amplifier output voltages. If we used a 10 watt amplifier that with an output Z of 8 ohms, the amplifier output would be about 9 volts RMS and we already know that we need about 20 volts RMS to get maximum power from the transducers, which means we would need a 50 watt amplifier to give this output voltage with the 8 ohm load. Now go back and consider the case where we used a matching transformer. The formula for calculating the output voltage is dependant solely on the turns ratio, ie Vout = N x Vin. Or transformed Vin = Vout / N. Therefore to get 20 volts out of the transformer to drive the transducer we need 20 / 11.18 volts = 1.7 volts RMS. Now to find the amplifier wattage required to give that voltage , = 4.8 watts RMS. Now compare that to the wattage required with the direct connection to the parallel transducer Array and it can be seen that the transformer solution reduces the required amplifier power, by a factor of ten. Against that, is the cost of making the two ferrite ring output transformers.
A latest development in Class D amplifier chips, is that the square to sine conversion can be done inside the chip and not require the external LC filter. Practical chips that have 5 watt outputs are the AD1990 and the AD 1992. The AD 1994 chip has a 20 watt rating. National Semiconductors and most of the other big players also have a range of suitable class D amplifier chips.
Ah yes, I see I have not said anything about signal phasing and wave lengths, which are terribly important, except at pub opening times .......
In conclusion, I apologise for any errors in the above.
9th February 2010. All rights reserved. John Kent