piezo ceramic project

[hoola]

 an increased operating voltage beyond 3 KV will require some adaptation of the stacks. The entire stack assy. could be dipped into some sort of sealant that might prevent breakdown on the outside, but that leaves the insides with the need to seal them too. If the central post were smaller diameter, perhaps of 3/8" instead of the 1/2"  nylon bolts currently used, the bolt could be held in alignment to true center, then sealant could be pumped in to fill the interior cylindrical gap, insulating the interior rim of the stack.   In this way the entire stack might be usable up to the 10 KV potential I am capable of. The sealant would have to be flexible enough to not interfere with the torquing process, perhaps common clear silicone could be tried first. The stack might be torqued up to peak as the sealant is curing, preventing problems that could arise by torquing the stack if the sealant is fully cured and less compressible. An alternative method of sealing the stack might be to fill the interior completely with sealant, and not using the central post, but adopting the Woodward team method of clamping the stack with end plates, with torque bolts positioned around the exterior. This would perhaps simplify matters quite a bit by simply torquing the stack to a predetermined amount, then dipping the whole thing at once. Another mod might be not to use a common ground on the stack. The input lines of course are all separate, but I have tied the grounds together to simplify matters. If the grounds were separated also, the inputs could be switched to what were ground connections, and vice versa. This might make the stack respond with a negative thrust instead of positive thrust, as inverting the firing order has not done that, as I expected it to.

Edited November 22, 2021 by hoola

[hoola]

a simpler way of dealing with stack arcing is to make the ground return lines at HV potential, that is, the DC level of the high voltage  (instead of 0 volts),  filtered and direct from the power supply, and the signal input lines at that same potential modified by the sine wave signal, which will be less voltage on average due to tube conduction.  That should reduce the stack voltage differentials sufficient to exceed the 3KV level I am stuck at now with a quick and easy mod. I had considered using input line caps as DC blockers,  but caps with a large enough capacity / voltage rating were not chosen due to expense and practical concerns, so I went with the present scheme.  I put DC blockers in the drive unit,  as the HV that fed the  output transistors was only 600v and easy to do with cheap, off the shelf caps. I will begin building a new compact drive unit that contains the square wave oscillator, the 4017 IC decade stimulator,  the interface circuitry (basically just 5 gain stages from the 4017) and all relevant controls within one small prototype box. That leaves the HV supply and the tubes as the bulky components, which could be downsized considerably, with a switching type HV supply (much easier to filter) and perhaps high voltage transistors capable of kilovolt operation. The horizontal output transistors (since removed) in the main driver unit were the highest voltage transistors I could find online, and I still cannot find anything that can sub for tubes as of yet. Perhaps transistors cannot be made to handle much more that 1KV due to their physical structure requirements. I think radio and tv stations still rely on tubes for their final.

Edited November 22, 2021 by hoola

[hoola]

I am wondering if multiple ECG165 bipolar horizontal output transistors could be series strung  to give "tube like" voltage swings at the collector of top end of the series, with the bottom end transistor emitter at  ground, while parellel driving them using  caps to keep each base input DC isolated and allow for individual biasing, thus keeping each transistor running at it's normal rating?   I read that the basic construction design of power FETs is with series/paralleling many small FETs to give additive power and voltage in one case package. Couldn't that be done on a larger scale with metal bipolars?

Edited November 22, 2021 by hoola

[hu??]

You could drive the piezo at resonance and use an inductor to boost the voltage.

This is explained a bit more here: https://www.powerelectronics.com/technologies/power-management/article/21864451/use-resonance-with-a-highvoltage-piezo-driver

Keeping the electric resonance tuned to the mechanical might be challenging.

The downside is that you loose control over the exact waveform driving the piezo.

Separating the resonance driver piezo and the piezo that should do the actual work could be a solution.

 

[hoola]

I have found that though the individual piezos have a 44 khz resonance, that has little to do with the overall picture. The stack is scanned at a frequency somewhat higher,  normally from 100k up to 200k, and I stay away from resonance,  avoiding potentially  over stressing the piezos. The thrust output seems no better at resonance anyway, and an additional bonus is that the stack runs almost silent at the higher frequencies, which is easier on the ears and seems to rule out simple acoustic waves bouncing off the table or into the ambient air and skewing results. I have a neighbor dog that comes by and she doesn't seem to hear any "dog whistling" coming from the shop, so ultrasonics seem unlikely.  Just as a speaker has a resonant frequency, you don't necessarily need, or want to peak that out with the crossover, sometimes you want to suppress that frequency a little (or tune the cabinet below it to give a linear output regardless of input freq.  I know that the Woodward team sweeps through resonance, which may be the way to go with their parallel,  pulsed inputs, but that I have found doesn't apply here...the stack can sit there quietly putting out 150 micronewtons all day without any observed problems. The voltage issue is not overly critical right now, getting the new support structure finished it the next goal.

Edited November 22, 2021 by hoola

[hoola]

,the new structure is working out well with weight changes in the 4-8 milligram range. The scale is positioned on the right side of the arm now, was previously on the left, with the 10" thread tied to the stack bottom, suspending a weight, which sits on the 3/8" felt pad, itself placed upon the scale. The scale now reads lower instead of higher, as the arm wants to torque in a counterclockwise direction just as before. Tomorrow I will re-wire the HV supply to remove the voltage sag under load and make the process less dramatic when the volts swing up and causes stack arcing when not being driven. 

Edited November 23, 2021 by hoola

[hoola]

I found that the driver transistors in the interface unit were being overdriven. I had adjusted them almost full gain before I took out the horizontal ouput transistors, and I thought the transistors were loading down the feed lines, but I think the diode action was distorting the signal, and not merely loading them down in a linear fashion. When checking  the interface module again, output transistor gone from their sockets, and with the piezos turned on one at a time, the sound coming from each one was louder and more complex, even prettier, as the interface gains were turned almost off, down to about 10% full rotation. This correction is coupled with a small output gain, which probably would be higher, but the corrected interface gain settings  cause the tubes to draw more current, which loads down the HV more, and thus limits output gain.

Edited November 23, 2021 by hoola

[hoola]

the re-wire of the neon transformer did not go as hoped. The conversion to full wave bridge and not using the center tap has made the output more uneven than before. I will have to return the supply to it's previous config. until  I can find a standard wire transformer.  The online description of the various current regulation methods used in older neons is not well explained, but acts as a sort of autotransformer, which I think uses a saturable core., but that is primarily to level an unsteady AC input voltage to equipment, not limit current.  Circuits referred to as "shunts" surround the windings and act as the current limiter. 

Edited November 23, 2021 by hoola

[hoola]

I have ordered a 5kv, 30 ma electronic neon transformer that is listed as dimmable. The case may be able to be opened to see if I can bypass the reg circuit if it doesn't work any better than the old one, which is also listed at 30ma max, but with a voltage listing of 10 kv. Hopefully the new one will regulate better at the 3kv level using the dimmer control. I have also ordered five .027@4kv caps to block the DC from the HV supply getting to the piezo inputs, perhaps solving the voltage issue if the caps can pass the signal from the tubes to the stack without excessive loss.

Edited November 23, 2021 by hoola

[hoola]

the basic difference between mine and woodwards' idea, is that they sitimulate all piezos concurrently, and shut them off, and see a result. Mine are lined up in a stack and stimulated separately and run continuously. I offered the "phononic gain" idea fairly early in the going, so let me reiterate it now. There are 5 active elements, two passive. The overall action is like a rail gun, that is, the shock wave is generated in the first which is placed at the top of a vertically arranged stack. This shock passed to the next one down, wherein I stimulate that next one,,,,,at just the right time....and so on down the line, as in the rail gun analogy where  each magnetic kick raises the velocity of the projectile  more. The shock wave scope traces showed this early on with steep vertical up deflections at the end of scope traces.   Another main diffference is the way the piezos are held. They clamp them down in an apparently vise like manner, excepting the rubber material, whereas I torque my stack lightly, on a central nylon rod, as when the individual piezos expand axially, they contract radially. With an overly restricted mounting, the shock waves diminish due to this radial restriction. The elements must have some physical freedom to move within a certain range. This indication was shown on the scope traces, and also the sound of the singing decreases either tighter of looser of a certain torque, in an obvious sweet spot of shock transfer and phononic gain. The nylon rod I use allows an even torque pressure within the ring elements (SMR3515T55 from steminc) of the stack, and is simple mechanically. I now use 1/2" rods, but have some 1/4" ones on order to try a more pliable rod, but with a somewhat higher torque to see how that responds. Hopefully, that will broaden the "sweet spot" and lower instantaneous stress on the elements. Today I am working on the series string transistor idea using 4 seven hundred volt transistors (NTE165) to see if I can eliminate the tubes. The used tubes I bought first are failing, perhaps in part due to cathode stripping from the excessive DC plate voltages needed. The can take a large pulse, but not designed for much more than 800 volts continuous. There is an occasional heavy arc across the stack which needs eliminating, as transistors are less forgiving of such insults than tubes. I hope the .027 caps on  order  will show a good transfer of energy, if not they can be easily paralleled to the needed value. I hooked the stack directly to the tube plates to maximize transfer of pulse, and was curious to see how much voltage can be safely placed across the elements. This appears to be 3KV.  The AC pulse signal is likely only 50% of the supply running as a class A amplifier with that inherent limitation. The piezos are a high impedance, low inductive load, so no "flyback" raising of the voltages. The tubes were designed with running into a 15K tuned load, in early television and could induce upwards to 30KV to accelerate the electron beam to the screen. With a supply voltage of 6-7KV  being blocked by the caps, I hope to get a 3KV AC signal pulse to the stack, as a pracitical maximum for now.

Edited November 26, 2021 by hoola

[hoola]

the "super transistor" idea seems encouraging. I hooked up  5  nte165 power transistors in series with a 2.2meg resistor from collector  to base, then a 27k resistor between base and emitter on each, and am getting 10ma current through the string, which is in series with a 100k 10 watt load resistor, being fed  3kv from the HV supply. I ported in a 600 hz signal to the paralleled base circuit, which DC isolates the bases from each other with .022 mfd caps. and got good signal drive to a sample piezo  placed across the string (hooked at the collector of the top transistor and the emitter of the bottom transistor), which is common ground. Overall current through the string is 10ma @ 3kv, no drive signal

Edited November 27, 2021 by hoola

[hoola]

I had earlier remarked that there was no flyback voltage gain due to it being a non inductive load, but there is a feedback from the elements coming right back down the feed lines. In an inductive load this would be a more or less constant set of waveforms, but in this scenario, it seems there would be a somewhat randomized set of waveforms, depending on drive parameters, with nulls and some major peaks sometimes occurring within various piezos. About 4 days ago, the machine was humming along nicely with an occasional arc that I had gotten quite used to, and a firecracker loud snap occurred of short duration, more of a pop sound than a discharge, and I didn't see any light coming from the stacks or anywhere in the overall layout. I wonder if some "random sunami" wave had occured at that point, when all possible waves converge at the same piezo element that could have produced a brief high voltage pulse  that I missed seeing being discharged.  This seems related to the anomalous solder pencil incidents where a tiny amount of heat caused a discharge several times.

Edited November 28, 2021 by hoola

[hoola]

for higher voltage tests, I will add external spark gaps across each input, then between each input and ground, so as to keep the arcing visible and not allow stray charges to build up excessively . I think the loud pop was internal of one of the stacks, and so I couldn't see it. With external spark gaps set properly, internal ones that could cause damage will hopefully be eliminated and loud ones reduced in frequency, if not in intensity.

[hoola]

while the supertransistor idea seems valid, there was some blurring of the output waveform that may prove undesirable, and I think that might be due to the signal pulse to all five transistors arriving essentially instantly, while the actual time lag of the transistors may be blurring the output signal somewhat, due to their staggered output responses, as the waveform goes up the chain, travelling through transistors that had previously fired. It seems possible that the emitter resistor pulse could be capacitivly coupled to the base of the next transistor up the chain, using that in a sort of chained trigger function, thus keeping response delays in sync with signal triggers and clean up the output trace, which could increase output with the increased efficiency.

Edited November 29, 2021 by hoola

[hoola]

the chained trigger function set up as  previously described, functioned fairly well, but the transistors would occasionally create a cyclic feedback between each capacitively connected pair, which makes sense as they have enough gain to do so, but the smeared traces were not present as with the common drives.  A more promising approach is to do a grounded base chain, where each emitter is driven from the collector from below, with the bases hooked to AC ground through .47 mfd caps, excepting the first one, which has the base driven by the oscillator and emitter to ground. I have a good clean signal now, no oscillation and good output, and can draw a small arc off the top collector. The biases have been adjusted to provide a similar temperatures of each transistor, and am using a 100K 10 watt load resistor from top collector to the HV supply. Although the generator  input is square wave, the output is a clean triangle wave, which is fine as both have been used and found to produce the same result. The test piezo is quite loud at 1KV input and hopefully the 5 tubes may soon be replaced with the 5 sets of these arrays of 5 transistors each. I think I can drive the piezos at up to 4KV and not exceed the specs on the nte 165s, as the volatage swings seems fairy well divided between them.

Edited November 30, 2021 by hoola

[hoola]

while the DC voltage drop component across each transistor is roughly equal, the AC component is not. The peak instantaneous voltage develops mostly across the top transistor, as the load resistor allows expression of this to develop. I have not stated for simplicity that there are 100 ohm fusible resistors in the connections between the collectors to emitters  of the chain, placed there to pop in the prototype boards under development, two of which did pop with the previous arrangement with the instabilities developing occasional large current swings. The grounded base has eliminated that issue, but I may have to keep the interconnecting resistors and increase their value somewhat in order to allow some AC signal to be more evenly distributed, so as to not exceed the top output's max peak voltage rating. I think the chained transistor idea is a valid replacement for the tubes, but may not be able to exceed it by much.  To get much above this level of drive may be to do what the Woodward team is doing and use transformers to up the peak voltage, thus keeping the driver transistors within design limits. There are plenty of what are described as " 15kv high frequency" arc producing transformers on ebay for cheap, so will try using a single transistor as driver for each, thus keeping drive voltage requirements down, and simplifying the overall system. An added benefit of the transformer drive would be isolation for the drivers from  arcing within the stacks.

Edited November 30, 2021 by hoola

[hoola]

I must report that the device has dropped in apparent thrust to the level of insignificance. The readings peaked on the 15th of last month and slowly dropped every day, quickly unable to create any visible movements, to only being addressable via the scale, then within the week, not registering on the scale. There was some discussion on another site, a youtube  comment I recall, concerning that if thrust could be attributed to a piezo device, the energy required would degrade the atomic structure domain walls, causing the eventual failure to produce thrust. I have 16 piezos coming this week and shall rebuild the balance arm stacks with new elements. The old piezos will be set back until I can devise some sort of response test to serve as a comparison to new ones. I will have two extra piezos that will not be included in the rebuilt stacks #2 and #3, the arm stacks, to hold back for that purpose. I have replaced the neon transformer with two microwave transformers wired in series. Now i have a 3KV supply that only drops about 100volts under load, so voltage overshoots when not driving the piezos are eliminated. I do wonder if the loud bang and the near constant smaller arcings could have contributed to the present condition. 

[hu??]

18 minutes ago, hoola said:

 There was some discussion on another site, a youtube  comment I recall, concerning that if thrust could be attributed to a piezo device, the energy required would degrade the atomic structure domain walls, causing the eventual failure to produce thrust.

Indeed, no discussion needed: that is just specified in piezo specs:

from http://www.piezo.ws/piezoelectric_actuator_tutorial/Piezo_Design_part2.php

 

"The maximum allowable field strength in piezo actuators is between 1 and 2 kV/mm in the polarization direction. In the reverse direction (semi-bipolar operation), at most 300 V/mm is allowable (see Fig. 10). The maximum voltage depends on the ceramic and insulation materials.

Exceeding the maximum voltage may cause dielectric breakdown and irreversible damage to the piezo actuator."

 

So there is no use to increase the voltage above this. I think it is better to optimize for the frequency and the waveform.

Edited December 6, 2021 by hu??
accidental commit while not ready

[hoola]

today I received the 16 new piezos and began replacing the #3 stack elements. Upon removing them I thought I noticed a difference in the weights of some of them. Upon checking, the end piezos, that didn't have a voltage applied to them as they were passives, weighed slightly more than the five active ones. The weights of the passives were 33.712 and 33.727 grams.  The five actives were all slightly less. I will not give the first two digits, as they are all 33. The weights are .684, .704, .545, .532, .676. I will rebuild the stack and record each weight for further reference. It is too much to imagine that the lost weights indicate a loss of mass, that was converted into thrust energy, but it is an amusing thought. Almost certainly the heavier ones randomly got placed on the ends, but weight is a parameter I need to start keeping track of to falsify any conjectures along that line. Upon weighing a random sample of the new ones, I got.....(33)  .919, .765, .780, .707,  .632,  ,620,  ,591 so a rather wide range of values that likely accounts for the differentials in the #3 stack.  After rebuilding, this stack will be designated as stack 3.1

Edited December 7, 2021 by hoola

[hoola]

 hu??, I did not see your entry last nite when I described the weight issue. Thank you so much and I will  check out that webpage today. This is perhaps a good thing, as the piezos are 5.5mm thick, so perhaps are useful for the 3KV DC across them. I must see what is meant with polarizable and reverse polarizable terms in relation to the rings. I will have both stacks done today, designated as 2.1 and 3.1 and proceed with tests keeping the HV down below what seems prudent with the new info from the webpage you listed. I have not researched piezo properties specifics, but it is certainly time to do so.

a quick readthrough of the webpage is rather daunting, and shows how subtle the processes are that take place in construction, with many new terms used in controlling applications. The most obvious term that applies to my situation is "creep" wherein a piezo loses some displacement over time. I will know today if a simple replacement of the elements will restore the desired responses noted last month.

[hoola]

a quick check of stack 2 shows no obvious correlation to position within the stack as related to weights.

[hoola]

while initial tests with the new stacks are negative, some odd things need attention before a good outcome might be achieved. The first was a lost ground to the left side stack, and upon checking, no problem could be found and no reason for an intermittent to have happened, and now indicates correctly. The next issue is with peaking out the new stacks.  Neither have a definitive sweet spot show up on the scope as with all previous stacks, but I can hear the increase/decrease as the spot is neared, which is inadequate and must be addressed. I have ordered 3 NOS tubes and the high frequency transformers that should be here this week. I will address current technical problems at hand and  replace the weaker  tubes before giving up on the tube drive idea for a while.  I am drawing up a schematic for a solid state final drive module using five nte 165 transistors driving five high frequency transformers while using the same interface module output DIN socket , so a swap out from the tube final drive with an SS final should be easy. I am aware now that the elements are "poled", that is, given a domain preference similar to lining up domains in a magnet, and that poling is accomplished with an electric field. I have no idea if the DC that my signal rides upon is in accord with the initial poling polarity during manufacturing, or I am counter to that orientation and am scrambling the domain structure more each time I turn it on. With the transformer drive, only the AC signal will be present, eliminating that as a possible issue. I will send steminc an email and request what polarity and voltage is used in poling my particular elements. The piezos all have a red line on the side near the top of each disc. I face all red marked elements up in accordance to that to keep them in phase. Traditionally, the red mark is denoted as the positive input to a device, and hopefully that is also used in the poling process.

Edited December 7, 2021 by hoola

[hoola]

I have decided to not test again past the initial negative test as I have no further information as of yet as to the poling polarity used, so as to not potentially damage the new stacks, will wait until the SS driver unit is ready to proceed.

[hoola]

I got the arc transformers in and they are smaller than pictured, being not much larger than your thumb, but do seem useable for the next series of testing.  At some point I will build another 4017 driver assy, and an oscillator on two small project boards and dispense with the now year old driver unit which has the original 4017 board and the wavetek 182 oscillator, which are both large and cumbersome, and have all controls, components and power supply on one open breadboard other than the piezo stacks and their support structure.

[hoola]

I found a 10KV high frequency transformer online that is three times the size and has 6 separated windings for the secondary that have dividers between them, so that the wire loops up over the dividers, allowing me to cut the line between the two middle secs and rewire for two 5KV secondaries in parallel with the additional current. The nomenclature involved does not identify it well, only listed on ebay as " 10KV transformation ratio 60 high-frequency coil transformer" There is an ID# but a google search showed nothing. They should be in in a week or two. Meanwhile I have built up a test breadboard for getting the impedance match as close as possible with the nte 165 drives. I will use a emitter follower circuit at first as that seems the best transfer using the smaller ones as a guide. The smaller ones are somewhat amusing in that they are literally held together with scotch tape, yet work fairly well, and can develop a considerable arc.