piezo ceramic project

[hoola]

I am working with piezo ceramics, 5.5 mm thick. I need to know the rate of a shock wave as it travels through the material.  James Woodward, et al are developing a piezoelectric Mach thruster device they are claiming develops propellantless thrust as a physical analog to the Shawyer engine. The materials to do a few simple tests are cheaply available so I figure it could be an interesting side project. Thanks for any help if you know that detail about ceramic piezos or have an interest in mach thrusters in general.

[hoola]

does anyone know of a rubberlike substance with a tailored characteristic that passes a mechanical force in one direction, and absorbs force from the opposite direction? In effect a non linear stiffness profile, or  "smart rubber". A material of this nature could act as a rectifier of physical forces, and this directionality might offer a thrust potential when placed between each piezo instead of a single conventional damper used in the woodward mechanism.

Edited January 27, 2020 by hoola

[studiot]

4 minutes ago, hoola said:

does anyone know of a rubberlike substance with the tailored characteristic that passes a mechanical force in one direction, and absorbs force from the opposite direction? In effect a non linear stiffness profile, or  "smart rubber". A material of this nature could act as a rectifier of physical forces, and this directionality might offer a thrust potential when placed between each piezo instead of a single conventional damper used in the woodward mechanism.

Substances don't 'absorb force'.

It is possible to build a mechanism that sort of answers your requirement using a dome shaped piece of spring steel foil set against a ring foundation.
But the foundation would have to be set against something to ultimately resist the applied force.

[hoola]

substances don't absorb force?  Wouldn't an absorbed force show up (at least)as heat within the material? The domed spring idea is interesting if it produces a diode like appearance to the system. The "thing" the spring is set against in this instance would be the inherent mass of the ring, but even then it seems the dome would couple as much energy overall in one direction than the other, although I see it might have a differing directional profile. If you had maxwell demons to do the job, such as tiny trap doors that all open in one direction allowing force through and close in the other, blocking such efforts with compressive release of heat from the stopped wave .

Edited January 27, 2020 by hoola

[hoola]

a purely mechanical system is what I am experimenting with now, using a double row of 24 contacts rotary switch spinning at 1800 rpm , sweep pulsing the piezos and scoping results with a 6th piezo at the end of the stack.  I am  in the process of assembling a purely electronic setup using an arduino  microcontoller to manage timing of the 5 power amps that will pulse each element individually, with the next pulse timed precisely to reinforce the previous one in the forward direction, or hit it at the next element's  "TDC". This hopefully will result in amplification of the pulse that after 5 repeated steps hits the end of the stack, and due to mechanical impedance mismatch, is reflected back in a way similar to a transmission line or an audio amp that has lost it's load. This reflected wave is the problem. It prevents thrust by equalizing forces. The driver schematic I am  drawing up has a full time negative feedback system, but with a hi Q 25khz trap in series, which is the resonant frequency of the piezos I am using.  This will suppress  the normal sideband noise but not the desired signal.  This could clean up the carrier signal and maybe even offer some added amplification due to reduced internal losses. At the end of the cycle, when the return wave is heading back through the stack, all feedback circuits remain on, but the 25khz trap is bypassed and all signals are suppressed  maximally within each element. This hopefully will cancel a portion of the  reflected wave energy within each element. So, the overall goal is to amplify kinetic forces in one direction and actively suppress it in the other. 

[hoola]

the question of how to detect micro newtons of thrust is going to be a tough one for a table top experiment in the kitchen. I am going to try swinging the stack as a long pendulum of approx 55" of length,  suspended from the ceiling with fishing line. I will time the duration of the swinging without power until the mass comes to rest. Then I will power up the stack and measure swing duration with the proposed thrust direction positive to the pendulum directions. Then thrust direction will be reversed to counter the pendulum motions and that will be timed. These 3 tests averaged over many trials might offer some evidence of thrust or lack of it.

Edited January 28, 2020 by hoola

[hoola]

The drive order at each limit of swing is switched in either instance. The stack will be suspended from the ceiling with two parallel fishing lines spaced equally to the distance between stack end attachment points to keep the assembly parallel to the proposed thrust vector.

Edited January 28, 2020 by hoola

[hoola]

I need to know the fastest switching speed an arduino can produce. After numerous utube tutorials and asking around, no one knows what that is. The reason I need to know is that the shock wave travelling through the stack will have to be " chased" in the manner previously described, and it seems likely that since the stack is on the order of 2" long and having a far faster speed of sound being a dense ceramic,  will need microsecond gradations, and not the millisecond pulse rates that the arduino might only be capable of.  If  the arduino speed is limited to a millisecond range,  a recommendation of a laptop controlled microcontoller with a microsecond control speed would be  very much appreciated.

[hoola]

I have abandoned the arduino as a trigger for the pulse delivery system due to it's apparent inability to deliver pulsed output in a microsecond range serially over the pwm ports. It does seem apparent that each one millisecond time division can be divided more finely with a hack I found online, but that seems unhelpful. I have been given a 4017 chip based design that could be capable of the needed output. It may prove easier to build what I need than to find something off the shelf.  I have also decided to skip the pendulum  idea of detection for the moment which has proven unwieldy, and will static mount the stack to start powered experiments with the goal of merging 5 small pulses into 1 large pulse, while keeping overall input power to a minimum. I have also had difficulty in the 25khz trap idea, but have found a circuit called the twin tee that does not use an inductor, only R and C components, that can be incorporated with a non inverting op amp to deliver accurate and deep rejection of a specific frequency. That is an option that can be easily added later and not necessary for the static mount tests. If the static tests prove positive, a simple pan balance setup to see if weight differences occur on a vertically mounted stack will be tried.

[hoola]

Actually the pan balance test isn't quite so simple. A laser pointer is aimed at a mirror placed on the scale with an acute angle, and the beam is reflected onto a distant surface to display amplified small movements.

[hoola]

early mechanical experiments indicated the possibility of scavenging the return energy by dumping the reverse shock wave into a dummy load resistor.  I did get some interesting waveforms  across this load that were trapezoidal (sloping square wave) and very regular and that didn't change waveform when slowing down the stimulation.  By adjusting tensioner nut on the center pin thus adjusting mechanical pressure on each element, a sweet spot was found indicated  by a moderate compression that produced maximum voltage output on the scavenged signal.  (The woodward crowd glues and clamps each element very solidly and I think that is a problem. The elements must be allowed to move, or the shock wave is largely converted into useless harmonics). But this dummy load was only a passive load to the unwanted shock wave, so I have the active neg feedback system which should be more effective in nulling the retrace wave.  A possible way of upping the efficiency might be to try to hit each piezo, then hold the volatage on each one to the end of forward scan, then drop the voltage serially, as the shock wave travels back, thus absorbing it as much as possible within each piezo, instead of neutralizing it with the neg feedback electostriction. As the  reverse wave hits each piezo, it's loss of voltage will not only shrink in the needed direction, but the wave may compress the element slightly as it passes, thus generating a voltage that could be recycled into the next forward scan sequence. This form of retrace period might require a higher scan rate, as the piezos will collapse even if the voltage is held steady and physically self-compress at the 25khz time constant (resonance) and so require a doubly  accurate pulse timing management. The first electronic experiments will drive the stack until the temperature stabilizes,  then turning on the neg feedback system. At that point an increase in temp would seem to indicate that the return wave is being converted to heat through electrostriction...thus the possibility of thrust in the direction of reverse scan.  Still waiting for the 4017 parts, hope the corona thing doesn't significantly delay delivery. 

Edited March 19, 2020 by hoola

[hoola]

well, there is no reverse scan in the present scheme...should have said " thrust in the direction opposite of forward scan". 

Edited March 19, 2020 by hoola

[hoola]

I ordered ten 7014 ICs from Newark Electronics (formerly MCM) and all were defective.  Under close examination the logo is blurry and indistinct. I have ordered most of the parts for this project from them with no other problems.  Last week I re-ordered ten more  4017s online  from Amazon, got them today with a Texas Instrument logo  that is clear, and the first three I have tried work correctly both in the solderless breadboard and the prototype board.

[hoola]

Using the 7014 chip with a +10V supply,  am using odd numbered outputs, 1,3,5,7,9 to drive the stack as this allows an equal time gap between pulses. The neg. feedback ckt. needs to be activated in an equal time to the total forward scan, so I have output 10 hooked to the enable pin with a .002 capacitor in series with a control set approx. at 2.7k ohms. This gives a controllable pause till you want the pulse order to begin again. I find it convenient to have all outputs hooked to LEDs, as with a fast scan one can easily tell  pulse duration of each output by relative brightness.  Individual switches to each output transistor allow base drive  to go to the transistor, or to be shunted to ground through a 1.8k ohm resistor. I have found that through adding and subtracting piezos, that a clear overlap of pulses occur at a drive frequency of 340 khz. The  stack  is one and a half inch, center to center, including 4 poly washers approx 2 mm thick and 10 thin electrode attachments of stainless steel. There is a sixth piezo at the end only used to scope the waveforms, and so not included in the length measurement but is the same 5.5 mm element. Although the sixth piezo  is passive, it does absorb and reflect the pulse, so does contribute to overall pulse speed calculations. As the neg/ feedback circuit is switched off, then back on, a clear display of removal of most harmonics is observed. The  neg. feedback circuitry is normally on only between scans. The next step is to build a non-inverting op amp twin tee circuit to insert into the signal path, which will be active during both the scan and retrace modes, with a 340 khz  notch. This hopefully will  remove  harmonics as they are being produced, without overly attenuating the 340 khz. This might make the overall trace more clear and perhaps add to pulse amplitudes by the additional removal of "noise" in the system. A possibly unavoidable source of noise is that the drive is square wave and the Woodward team says piezos like sinewaves better, but it may not make much difference at a drive frequency so much higher than the 25khz resonance. 

Edited April 2, 2020 by hoola

[hoola]

It has seemed apparent since the beginning that the sixth piezo doing the monitoring of the waveform is a problem in it's placement. I have found that putting another piezo  touching edge to edge of one in the stack offers a strong signal. The placement is like two car tires touching treads instead of sidewalls. To my surprise, the 5.5 mm thicknesses  offer a good enough physical transfer of energy to give a strong signal output, observed from face contacts. My next alteration will be to remove the sixth piezo from the stack and place it outside the stack for monitoring from the edge of the center element of the stack, offering a more linear output of overall energy flows. This way, only active elements are in the loop, and the monitor element will become a negligible factor in stack behavior. This is one more reason not to clamp the assembly too tightly,  and to use slippery poly washers between each element, allowing lateral physical movements. I have also changed the .002 cap in the enable circuit to .0039 to center up the sweet spot on the 10k delay control. Another planned change is to raise the output transistor voltage supply from +150V up to approx +250V.

Edited April 5, 2020 by hoola

[hoola]

re positioning of the 6th piezo has raised the  frequency of interest to 1.8 mhz,  much higher than I expected. Before the latest change, frequencies above 1 mhz were trivial scope fuzz. I am getting 6V p-p from this sensor arrangement. It seems productive to add more sensor piezos in this way, adding four more for complete coverage. Unless I had four more scopes, I would need to matrix the results. A decrease in the retrace delay was of course needed with this sensor change, but much less than expected.

[hoola]

Further refinements in the stack and mods on the NFB circuits have given a new maximum of measured output at 71 khz. Scope traces have stabilized on certain damped waves that look very interesting. Rates of damping shows an increase when NFB  gain is increased. Also have installed a 4PDT switch to reverse the order of stimulation of the piezos. That will prove useful as I have placed asymmetrical  masses at ends of stack, as the Woodward team has done. I will reverse the masses and repeat test of forward/ reverse differences.

Edited April 11, 2020 by hoola

[hoola]

I have noticed that the damped wave trace at 71 khz is cleaner and has a higher output if the fifth piezo is turned off. This may be because the main shock wave travels through the first four piezos in one direction on the drive phase, but the 5th is now the turn around point as the wave reflects back towards the first piezo, quickly passing back through itself just as it's being stimulated, causing these instabilities and lowered output. I previously removed the 6th piezo  from the stack thinking that would clean up the signals, and the opposite has happened, at least at this new 71 khz frequency. I  will keep the side mounted sensor, but will install  another piezo  back into  the stack as a sort of "dummy load" as well as a second sensor. From the beginning, I had been having issues with instabilities with the 7014 chip as the stack  sometimes emits strong EMF hash that is picked up by the chip and confuses it's count process.  The chip circuit board is now shielded in an aluminum box at the opposite end of the chassis and that has  resolved the issue.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

for a total of 5 active elements, and 2 independent sensors. It may prove that the last piezo element needs to be electrically neutral, as the reflection point in this serially stimulated scenario.

[hoola]

Sorry about that last bit on the previous post, I thought I had deleted that.  Today I mounted the sixth sensor piezo back into the stack, and as expected the #5 piezo helps the signal now, and the  frequency of interest has moved up to just over 100kc. The reflection of the energy has always been the issue to be dealt with and that has me thinking that if there were several more "dummy load" piezos at the end of the stack, wouldn't that absorb the energy before it's reflected back towards the front of the stack? I had tried to scavenge energy off the reflected wave in the early mechanical tests, and actually got enough to light up an LED weakly.  Being a passive load, it was a very inefficient transfer . If the neg feedback circuits were moved to the extra piezos, the electrostriction effect should remove the wave much more efficiently, so fewer elements needed, lowering complexity, weight and cost. The  large steel mass would be replaced by the "absorber piezos",  and the  smaller front mass may no longer be needed. Another useful aspect is that the frequency of the drive stimulations may become more arbitrary as the feedback circuits will be on full time and not tied to the  newly obsoleted  pause circuity and therefore operate independently to eliminate whatever signals are presented to them. If any measurable thrust is developed, the stimulation frequency might act as thrust control. I would have to add separate output transistors for the absorber piezos. I don't think I would need all five channels, so will start by adding 3 more absorber piezos behind the 6th sensor piezo, for a total of 9 elements, and 3  attendant output transistors.

Edited April 12, 2020 by hoola

[hoola]

The signal that I am now tracing seems as likely to offer potential thrust as any observed so far at a new 188khz frequency. The five scan pulses have been merged into a single parabolic waveform that pops up smoothly from baseline, increases linearly  in amplitude until an upward spike at the end, signifying a possible amplification of the pulses as they progress down the stack. After the last drive pulse, the trace drops quickly, and continues a further smooth decrease  to almost zero until the next scan cycle begins , signifying a workable  NFB system that has proven reliable and not needing further adjustment at this time. This pause circuit cannot be eliminated, as no useful candidate traces were observed with that turned off. The two NFB circuits have been re connected back to P4 and P5  and the extra output transistors removed, as they added no interesting  traces, and the installation of an aluminum shield cover over the 7014 chip board has also helped overall stability. I found the while the  newly reconnected NFB circuits allowed a trace that would occasionally show up that looked promising, the inherent noise in the scan mode now seemed to prevent that trace from becoming stable. I had left the two extra piezos in the stack unused, until I found that I could hook them up passively with various components and get significant reduction in scan noise. After much switching around on values, found that the trace stabilizes with only three components hooked up between the  2 absorber piezos,  the 5 drive piezos, and ground.  A 12k resistor, a 36 pf cap, and a 72pf cap. I believe these passive components cause a phase cancellation of scan noise and have  allowed it to run a stable trace on the bench for hours. One of the useful things about the 7014 chip is that it's frequency output varies a few hertz upon raising and lowering the drive voltage to it. This has amounted to a vernier adjustment that has proved quite useful. One method of checking progress on  this " drive noise phase cancellation" as I was subbing in dozens of parts, was to raise and lower the drive voltage, and seeing if the trace would fail and dissolve into chaos. After a few hours, a stable trace that varies in only trivial ways is observed as drive voltage is raised from 12v to 20v.  Both of the asymmetric masses have been removed and no longer necessary. I will build another stack and try to use only one absorber piezo this time and use heavy aluminum foil as element contacts, as the phosphor bronze sheet metal  has proven difficult to work with and the foil should conform better to element faces. I will also build a continuously variable power supply for the output transistors to raise the voltage slowly from the 300v I'm now using up to approx. 1kv.  Some of the Woodward team have speculated that a potential thrust could increase at the square, or possibly the fourth power of the drive voltage increase.

[hoola]

As I have finally got a stable enough trace to begin to fine tune things,  it seems that the  interelectrode capacity between the stack lead wires and the contacts themselves are influencing stack behavior. A perfect waveform is being achieved with spacing of certain wires and interleaving of others just before they attach to the contacts. The contacts themselves are round to match the ring size with the necessary bump out for solder terminals. These contact pads appear to offer some additional effects of concern,  as they form small variable capacitors, being directly attached to the piezos. I had attributed  certain odd behaviors to the contact material (phosphor bronze) and was going to make the next stack with heavy aluminum foil, but now I think that wouldn't help with the conductivity issue, but might make the variable cap. issue worse, with them vibrating more randomly than the rather inflexible bronze.  A possible minimizing of the issue might be to stagger the contacts in a spiral around the perimeter, keeping capacitive interactions to a minimum. This seems to answer the question of why a  36pf  capacitor has a strong effect at the relatively low frequency of 188khz.

[hoola]

have found that dressing the lead wires into loosely wrapped pairs of P1-P2 and then  P3-P4, and keeping P5 and ground separate has contributed to a higher peak output and more consistent and "resilient" waveform. Before this change even careful handling of the stack would greatly perturb the image. Now the contact issue seems to be gone too. That was always a sensitive spot on each element and now I can poke them with a nylon probe and see no disturbances. I suspect that arranging the leads in this manner has simplified the "conversation" going on between the elements (less noise). Also, have seen that just holding a piezo disc edge to edge with any of the five active elements has a further calming effect. There were no electrical connections, so I added them back on and tried shorting them out and saw no difference. Reversing the polarity of the element (simply flipping it over) also seems to make no difference. Have tried other items, such as metal screwdriver blades, and other metallic and non metallic objects which have no effect. It seems as if the piezos are "talking" to each other, edge to edge even under casual pressure. I have two ways of monitoring the stack now, the edge connector and monitoring the voltage expressed to the #3 piezo through a 10 megohm resistor. In the voltage monitoring, the voltage gain and the strong negative feedback pulse sequence are clearly evident.

[hoola]

upon retesting the piezo simply held up to the stack with no connections, I still see no trace difference  but the sound that the stack makes does change a little. One of the diagnostic tools is to monitor the noise, a  medium level chirpy hash  mixed with various whistles and clicks,  sometimes a rather pleasant sound  like a sleep machine.

[hoola]

I built a simple passive twin tee notch filter for the 188khz pilot signal and placed it in the nfb loops one at a time and observed a minor additional suppression of noise, and very little suppression of forward pulse ( as the trap is active during drive sequence). Under higher voltage tests these additional noise suppression traps may be helpful in maintaining waveform stability,  but it would leave me stuck at the 188mhz freq unless they are also changed in tandem. 

Edited April 22, 2020 by hoola

[hoola]

Today I got in five new nte 165 horizontal output transistors and installed them and operated the machine at a 700V driver supply voltage for the first time.  I had the machine in it's normal position on the bench inside the air current isolation box and transparent cover over the top. After about ten minutes of attempting to get a familiar waveform on the scope, shutting the unit on and off to make adjustments in drive frequency and pulse duration, I powered down as the new transistors were running warmer than I had anticipated. Taking  what I was anticipating as a short break, I  opened up the chocolate container that was sitting on a shelf about 2 feet from the stack, and noticed that the chocolate had melted, and more so on one side that the other, the side toward the unit was nearly liquified. The serendipedous discovery of the microwave oven was a chocolate bar melted in the tech's pocket, and indeed I could be replicating that here. The temperature here is mild, outside is mid 70s, and the house is opened up  and at the same temperature, and nothing hot was placed near the container.  So I have discovered a major issue to address, and fortunately quickly enough to avert overexposure to the microwaves. I certainly did not imagine this to be an issue, but it may prove to be. Further tests will be needed to be positive of the cause of the melted chocolate.  I need to get a microwave detector and  appropriate shielding laminated onto the isolation box before any further testing can be safely done.

Edited May 27, 2020 by hoola