Registered users: Bing [Bot], Exabot [Bot], Google [Bot], MSNbot Media, Trauts
Tech: Never thought of a capacitive divider. That might work pretty well. You do still need to worry about the high pass filter in the soundcard. At low frequency (actually low dV/dT) the soundcard will just ignore the signal.
I wonder if you could use the piezo (amplified, or at least impedance matched) to control an AM oscillator? Set the oscilator at say 1KHz so the sound card is happy with the changing signal then modulate the amplitude to encode the pressure data. That should let you get low frequence past the soundcards high pass filter. Have to be careful that you don't loose the sign of the data but in this application there might not be any sign data, at least not at any time that you are actually interested in.
Another thing to consider is that there are other types of piezo based pressure sensors. We've been talking about piezoelectric devices where the piezo is an EMF source. There are also piezoresistive devices where pressure changes the resistance of the element. A pair of elements is generally configured in a Whetstone (sp?) bridge. It is a heck of a lot easier to deal with resistance than it is a low impedance voltage source. Piezoresistive elements are often used to do things like measure the chamber pressure in rifles. Only problem is I've never found a really cheap version with suitable characteristics. I've considered the thin film types that are just glued to the chamber. Basically using a strain gauge to measure the flex of the the chamber as the pressure changes. I believe there is a company that sells the sensors and electronics to rifle reloaders so they can measure the shot to shot changes in their loads. The sensor is just glued to the outside of the guns chamber. Not exactly sure how to go about calibrating one of those sensors.
Good topics. First regarding the bandwidth of a sound card. Yes it is a concern. Is it important? To find out we need to look at our lowest frequency component and see if it is well within the bandpass of the sound card. On my large cannon, the entire shot is completed in under 50ms. If this were a repeating waveform, then the frequency 1/X would be 20 hz. It is pushing the lower limit of a sound card so some roll off in response is to be expected. With a good schematic of a typical sound card and standard component values, it wouldn't take much to modify a standard sound card for extended low end response.
As far as using 1 KHZ FM, the upper bandwidth limit is a concern. What is the highest frequency component you need to capture. In a combustion, it would be the pressure rise peak and the pressure drop as the projectile passes the end of barrel sensor.
For example attached is a capture of a t shirt launched from a 4 inch barrel fed with a 2 inch valve at 70 PSI. The sensors are magnetic pick up coils spaced 1 foot apart. 1 KHZ FM would not have worked to capture this waveform.
The sensors glued to the outside of a barrel are often calibrated by a test fire of a known projectile and charge and verifying the shot with a chronograph. Once a calibration shot is taken, then unknowns can be compared to the reference shot. The reference shot is an average of multiple shots for statical average accuracy.
Has anyone considered using an A/D converter and the PC's parallel port?
"If at first you dont succeed, then skydiving is not for you" - Darwin Awards
It is easier to pick up a USB pro sound capture device for extended frequency range. If needed, they are easier to modify for extended low end. Flat response with no modification from 20 HZ to 22 KHZ is possible with 48 K samples/sec.
There is a known issue with Sound Blaster compatibility where the real internal sample rate is fixed and the software "transposes" sample rates. Avoid SB compatibles if you are doing measurements. The software sample rate may not reflect the true sample rate. Your software may see 96 K samples 24 bit, but the card is only doing 48K 16 bit. This software manipulation of your scientific data is not desirable.
Info here.. http://www.pcpro.co.uk/reviews/78843/creative-sound-blaster-xfi-xtrememusic.html
Last edited by Technician1002 on Fri Jun 05, 2009 2:22 pm, edited 2 times in total.
I don't think that analysis is correct. The problem is that the freq isn't constant, it rises as combustion progresses. On my generic combustion gun it is about 50mS from the click of the sparker to the "bang" of the spud exiting the barrel.
In the graph below the black trace (labelled dP/dT) is a raw piezo recording of the gun firing.
(Ignore the annotations, they are likely incorrect. The "one flame front" "two flame front" transition is probably something like the spherical to two-domes flame front transition.) The igntion event is recorded in the data stream since the piezo makes a distinct click.
Notice the fairly significant lag between "click" and a significant pressure rise.
The assumption that the lowest frequency is 1/(time from click to bang) is incorrect since the "frequency" changes as combustion progresses, indeed the "frequency" rises significantly during combustion. So, the early frequency is significantly less than 20 Hz which is going to be well below the roll-off frequency of the soundcards input high pass filter. I figure the 1/RC of the soundcard's filter is probably 100 to 200Hz and the 3DB point is 15~30Hz so it attenuates "60Hz" hum. As I mentioned previously, you probably can recontruct the true wave form if you knew the RC of the filter but only if you are actually getting a reading. If the signal is below the minimum resolution of the sound card (or the attenuation has dropped it there) then there is nothing to reconstruct from. A 0V, 0V, 0V, 0V data stream can't be used to reconstruct the actual waveform. I tried it way back when but now that I look back at what I did I realize I didn't take into account the capacitance of the piezo itself. It might be worth redoing it even if the early part of the data is unrecoverable.
It is pretty easy to measure the RC characteristics of the soundcards input. I've done it before for the LINE input of a PC. If anyone is interested I can't give a procedure. Actually measuring the characteristics would be better than going by the schematic. Particulary since the error on the filter cap spec is probably pretty big, at least 10% and it may be more like 20%. Besides, locating the schematic for a particular sound card is nearly impossible.
As to AM modulation (not FM modulation), a 1KHz base frequence would essentially give a "pressure" data point once every 0.5mS. Roughly 100 data points accross the 50mS time frame from "click" to "bang". If you boosted the carrier to say 4KHz (roughly 1/10th the sound cards max sample rate) the resolution would be 0.12mS and you would get 400 data points from a firing. That should be adequate to reconstruct the pressure versus time curve. You might loose some detail, like the spike in dP/dT when the flame front transistions from spherical to two domes.
This is fun.
The problem is cost. You can get low freq AD converters for less than $100 but they won't sample fast enough for this application. The link Tech posted is for a $350 converter.
We've already got a 48KHz, 16-bit, dual channel A/D converter that is basically free in the soundcard. Just need a way to generate a suitable signal. The $350 converter is probably too low impedance to be used directly, you will still need an amp for the piezo. The A/D converter has two advantages; (1) it may not have an input audio band filter and (2) it probably isn't doing any hidden processing of the signal after the A/D. (That's another worry with a soundcard, sometimes there is filtering done to the raw data stream in software on the soundcard.)
I wonder if I can rummage up an old computer and rip out (or bypass) the input filter on the soundcard. Sure would be simpler if there wasn't a high pass filter to worry about. I recall seeing a web page some years back where they did just that. They wanted to be able to do DC measurments with the soundcard.
EDIT: The link Tech posted is for an audio A/D system. So it may well have the high pass filter on it's inputs.
Wow, looking at the raw signal shows severe loading on the piezo. The huge negative spike as the pressure peak passed is a good indication. In the graph, I think assumptions were made regarding the apple exit and apple starts to move. Resonance of the container is evident after the apple left. Working back along the raw wave shows where that frequency started. Speed of sound and time for sound to travel will affect this some as the sensor is not at the exit point but sees the exit effects at a later time. On the graph of the raw signal the start of the bang is most likely the apple exit delayed by sound travel time, so the apple did exit just before this.
The apple starts to move is most likely in error. In examining the black raw signal I see 3 prominent peaks followed by 3 valleys. This is most likely a resonant compression wave transversing the chamber. The apple will start to move on any significant pressure rise. The apple most likely started to move before the first echo pressure wave returned or at the end of the 2 wave fronts time.
Mixing a barrel pickup with this data would tell more. Apple acceleration could be used to verify chamber pressure. Delta time between pick up points is directly related to acceleration. On my prior posted t shirt graph, delta time between all samples will clearly show the acceleration graph of this launch. The piezo could really use a low leakage buffer to remove the integration of the signal.
A sound card with only 100-200 hz low end is not good for music. It's only a speech grade device. The telephone considers speech to be 300 HZ to 3KHZ. For music you want 20HZ to 20 KHZ. For instrumentation, DC is nice, but the bandwidth should be enough to encompass your signal bandwidth.
It does.. However look at the BW spec. The cannon launch is fully within the bandwidth of this capture card. The high pass filter is not an issue as you are not going to measure any steady state DC off the piezo. Most likely the sound capture will have a lower passband than the buffer you use on the piezo.
You don't need to go super high end to get a decent capture card. My favorite is quite inexpensive and the specs are not too bad.
This one is under $30
Spec'ed at 10 HZ to 20 KHZ +- 1 db.
I have one.
Here's a cheapo A/D converter that can be cascaded to give you 9-bits of resolution:
http://ca.mouser.com/Search/ProductDeta ... UxAtq9g%3d
It isn't 16-bit like a soundcard, but you don't have to worry about any audio circuitry interfering with your signal. It can operate MUCH faster than the sound card too, with a 392 kSPS max sample rate.
I don't know if the bit depth is high enough, but you can't really squeeze much voltage into a sound card anyhow.
"If at first you dont succeed, then skydiving is not for you" - Darwin Awards
Or, it's the high-pass filter on the sound card "springing back" to what it thinks is a fairly large DC offset.
Yes, the "resonance" is literally the "bang" the gun produces. Working backwards to the location of the piezo, which is nearly at the guns breach, is a bit problematic since we have zero idea of what the speed of sound is. Working backwards in the "bang" to the muzzle exit event is also problematic because of the uncertainty in the speed of sound over this time range. It's certainly greater than 1100 FPS and could be more than 3000 FPS.
The brown line is the integral of the raw piezo signal. It looks much more like you would expect the P vs. T trace to look like. The spikes in the raw signal turn into humps and inflection points. This is exactly what you would expect if the raw signal was actually dP/dT vs T instead of P vs. T. The integral makes a much more sensical estimate of the time when the apple exits the barrel. The integral also shows a bit of offset in the "bang" due to loading of either the piezo or the response of the soundcards high pass filter. But even on the integral trace there are clearly some odd things. The roll off will make the trace cross the x-axis early. If the axis crossing point is at least close to the true exit time then the trace suggests the pressure was dropping pretty rapidly before the apple exited. It is likely that that is actually the case. This gun is pretty close to a C:B of 0.8, that means it is pretty close to the most efficient ratio. If the barrel was made a bit longer the performance would be expected to drop a bit. The most energy efficient combustion gun has the projectile exiting the barrel right at the time when the chamber pressure is equal to atmospheric pressure. A longer barrel starts to exhibit some "suck back", a shorter barrel wastes energy because the chamber pressure is still above ambient when the ammo exits the barrel.
Of course, it is also possible that the actual exit point is between the point labeled "apple starts to move" and "apple leaves barrel", or perhaps "apple starts to move" is exit point. If the latter is true than a longer barrel should improve the performance of this gun.
Yep, the start of movement is questionable. You would expect a small change in P vs T and a much more noticeable change in dP/dT but both may be too small to see since initially the apple is moving very slowly.
Looking at HGDT (and my own combustion modeling) you do expect a spike in dP/dT when the flame front transitions from spherical and expanding as something along the lines of a3 rd order polynomial. Once the flame reaches the wall the "form functions" changes to nearly a constant and all that is left is flame front acceleration due to the rising temperature and perhaps turbulent flow.
It would be interesting to look at an HGDT model and see if there are indeed spikes in dP/dT when the flame front transitions and when the ammo starts to move. (see below)
I have my doubts about detecting any resonant compression waves. The problem is that the pressure isn't constant, it's rising so there is no coherent wave to detect. In other words, yes, there are waves but they are created at different times and with different frequencies. As the pressure rises along a more or less smooth curve any echo would also be a smooth curve. The strongest echo would occur sometime after the maximum pressure is reached since that is the only time when you have a sharp change in the pressure versus time. I suppose you might get an echo of any dP/dT spikes but the echo would be of similar or lower amplitude than the amplitude of the dP/dT spike. I don't think resonance is going enter into things too much. It can't. Yes a closed cylinder resonates at a predictable frequency but only when the air is at a constant temperature. In this case the temp in the gun is swinging from ambient up towards 3000K or so then back down towards ambient very rapidly. Furthermore, the temperature is not the same everywhere in the chamber. I don't think that set of conditions is very conducive to resonance. Of course, echo and resonance aren't the same thing and echo's would certainly be more likely.
Perhaps, of course you would have to have a reasonable estimate of the static and dynamic friction for the apple. My wag puts the static friction in this 2" barrel at 30 to 50 pounds (10~20 PSI), and a really WAG of half that for dynamic friction. This barrel has lots of friction, combustion guns tend to work better with significant friction. That's what a double beveled muzzle knife is for.
That depends. The 1/RC is value is the elbow in the response curve. At a freq of 1/RC there is essentially no attenuation. The 3DB point for 100-200 Hz RC is in the 15 to 30 Hz range. (I believe that is correct, the 3DB is 1/(2PiRC) and 3DB is an attenuation of a factor of 2.) So some but not that much attenuation at 50 Hz. In addition, it depends a lot on how insensitive you want the system to be to 60 Hz hum from the AC mains.
Well, actually, you are going to be measuring what an audio A/D converter considers "DC". That slow initial rise in pressure is a real signal but the effective frequency is well below 50 Hz.
But that looks virtually identical to the soundcard. Heck, to me it looks like a standalone soundcard with a USB connection. That capture device might be a great candidate for ripping open and removing the input high pass filter.
I modeled this gun in HGDT. Below is the HGDT window showing the graph of chamber pressure versus time. I exported the data and used Excel to calculate and graph the dP/dT trace. you'll notice several difference and some similarities between what HGDT says and what the piezo says. HGDT doesn't predict a spike in P or dP/dT when the projectile starts to move. HGDT does predicts the chamber pressure is actually dropping rapidly when the apple exits (not that HGDT "knew" it was an apple )
Clearly though the overall shape of both the P and dP/dT curves are significantly different than either my raw piezo or the integral of the piezo traces. HGDT wouldn't calculate any echoes or resonation, not will it calculate the bang.
One small note. The attenuation of 3 DB is a power measurement, not an amplitude measurement. 1/2 the voltage (a factor of 2) is a 6 DB attenuation.
With a response down to 10 HZ with a maximum of 1 DB of attenuation, the adaptor is good for most any waveform you wish to capture with a interval of under 100 ms. All of my launcher tests have involved launch events well within this limitation.
The typical external sound card.. the one specified has the 10 hz to 20 KHZ range which is flat within 1 DB. Most netbooks have a mic in and no line in. Most SB compatible soundcards have poor mic in response. The USB device provides a known response.
On the graph, the arrow points to Projectile Moves.. Why does it wait until the pressure reaches 10 PSI at 15 ms? Do yours stick that badly? I have done 5 PSI launches on the ABS DWV test cannon. I would believe the start of motion is under the 5 PSI pressure than it is at the 10 PSI point. An apple in a 2 inch barrel has over 3 square inches of cross section. At 10 PSI friction, it would need over 30 lbs of force to get the apple in there.
Ooops, you are right about the 3DB versus 6DB, it depends on if you are talking about voltage (or current) or power.
Yes, the ammo sticks that bad. But that is a good thing. 30 Lbs of force is actually probably a bit low for an apple in that particular 2" barrel. The muzzle is double beveled so the ammo is cut a bit big then compressed in the barrel. When loading I have to lean hard on the ramrod, even with the cleanout cap removed.
High static friction gives you some of the characteristics of a burst disk. It allows the chamber pressure to build before the ammo starts to move. Of course, that also means you probably have higher dynamic friction which robs you of some acceleration once the ammo starts to move.
With a very low friction ammo, and particularly if it is also low mass, the ammo will exit the barrel long before combustion is complete.
Attached is a graph showing what HGDT thinks the affects of varying the static friction is on the muzzle velocity and muzzle KE for my "reference gun". HGDT only uses a single friction value, which is the static friction, it assumes the dynamic friction is half the static friction. As you can see, the difference in performance between say a 1 PSI friction and a 10 PSI friction is pretty significant; 40 FPS (17%) increase in velocity and about a 30% increase in muzzle KE.
I think you are probably right about the frequency response, especially if you can get "flat" response down to 20Hz or so with a better sound digitizing system. I suspect that the generic MIC input of a PC probably uses a 1/RC of a hundred or so Hz because the designer figures it's only going to be used with cheap $5 microphones and not hundred dollar studio grade mics. Better to attenuate the inevitable 60Hz hum of a cheap mic than it is to worry about flat frequency response below 100Hz.
I need to think about the freq response some more. Combustion of propane in air starts off as an incredibly slow process. The initial flame front speed is only something like 14 inches/sec (< 1 MPH). That is so slow that if you filled a hanger sized building with propane and air and lit a match in the center of the building you could turn and leisurely walk away from the flame front and out "run" the front at a walking speed. Well, at least for a while, as the flame front speed accelerates it'll catch up to you eventually.
I won't bother with quotes..
Many sound card inputs have well documented speech bandpass on the mic input for intelligibility in VOIP and voice applications. This is common. To make matters worse, many newer laptops do not have a line in at all. Here is a graph of a typical sound blaster mic input.
The graph and explanation is from this page;
http://www.epanorama.net/documents/soun ... value.html
Good call on the stuck apple acting as a burst disk. I'll have to keep that in mind when I build a combustion.
Now all that is left to do is cut down the loading on the piezo so it is not integrated by a high pass network of the unit's series capacitance and parallel resistive loading. A good sound card has low enough passband for little issues with the high pass filter.
That MIC response curve is about what I figured my laptops MIC input was. Rolling off below ~100hz. I wonder how much variability there is in different manufacturers. My laptop soundcard isn't made by soundblaster.
Shouldn't that be "differentiated"?
The piezo's output is the derivative of it's internally generated EMF. The soundcard high pass filter is also a differentiator, particularly at low dV/dT.
I spent about an hour looking for a spice model of a typical piezo transducer with no luck. There are a lot of references to publications, but nothing available for free online. I wonder if a simple minded model like an arbitrary behaviour voltage source through a ~10nF cap would be a good enough model?
There's too much to reply to so I'll just say that I read most of it all. I might reply to more in the future.
Makes sense. Odd that they never mentioned this once in my electronics class (which was supposed to have a focus on instrumentation).
Sounds more than adequate for this and very good for any audio recording. I'll buy one.
I had mentioned a "piezo technical manual" available from here earlier: http://www.meas-spec.com/product/t_reso ... spx?id=540
They have two simplified equivalent circuits and the simplest one is a voltage source (that is proportional to the force on the piezoelectric element) in series with a capacitor. See page 39 of the PDF. I used this approximation in my calculations. I used the capacitance values given in the piezoelectric element's data sheet and assumed it was constant for all frequencies.
If I understand correctly, a circuit like the one below is what we should be shooting for.
Note that while I'm using a generic JFET symbol I'd use a IGFET per Technician's suggestion. I couldn't find the correct symbol in gschem.
A question about this sort of buffer: if the piezo voltage goes negative (from if the pressure drops below atmospheric), will the IGFET buffer voltage also go negative? From my understanding, it would not.
All spud gun related projects are currently on hold.
The circuit is a typical single ended AC amp. It does pass an AC waveform including the negative half. The output can swing between the supply rails of ground and +. It operates as a class A amplifier.
Who is online
Registered users: Bing [Bot], Exabot [Bot], Google [Bot], MSNbot Media, Trauts