Measuring internal ballistics- first attempt ** NEW SET_UP in Reply #51

[Mole2017]

That piezo data is busy, isn't it? We might have a hard time with that one. Actually, looking at the data and thinking about the approach, I'm not sure it is needed. In fact, given a choice, I'd opt for more contact switches. I'm still sorting through that myself, but here's the thinking so far:

We take the time versus position data from the contact sensor trace and do a simple point to point slope calculation. That is velocity. We start with 6 data pairs (including "zero") and can calculate 5 velocity points. From there, we repeat to get acceleration, now leaving us with 4 data points to plot. Note: for the calculations, I took the time stamps at the rising edge of the pulse, i.e. the head meeting the switch and ignored the skirt part at that first switch.

When doing this, we hope to have meaningful trends in the data. In this case, with 4 points to plot, I see a nice curve in the acceleration that might suggest a friction term and pressure term interacting, but there isn't quite enough data to build a complex model. It is like curve fitting: I need two points to fit a line. More than that is just statistical icing on the cake (and very valuable if you can get it). In this case, we're right at 4 data points that, at the max, might help us pin down 4 unknown coefficients. However if you have a model for some factors already, that might reduce a few unknowns at the risk of assuming the model is correct.

The problem is like the guy with the air spring paper: a linear model showed what he wanted to show, but it was clear a nonlinear model would probably match the data exactly.

So what about the piezo? Well, it is the source of some uncertainty right now. Is it accurately indicating the start time? Is the pellet at "0" position when time is at "0" (or at any time)? I.e. is that first data point correct? If we have more switches, we might get enough confidence in our curve to leave that point out altogether.

[Mole2017]

PS: my calcs on the data suggest that if the barrel were more than about 4 inches longer, we might actually see deceleration: not enough air pressure to maintain acceleration against the friction. Actually, that might be interesting to see. Could we port a barrel to effectively drop the pressure to zero and let the pellet coast through a section of barrel, loaded with switches of course?

[lloyd-ss]

Stan, The time value at 23.46" is an estimate.The line was pretty straight and I doubt much changes in the last inch.

[rsterne]

Thanks Lloyd, and great data.... Depending on the BC used for that pellet (I found 0.019 -0.023) when you back up 4 feet from your 746 fps reading, the actual MV is 752-753 fps.... BTW, I think you have too small a transfer port volume in your spreadsheet, it should be 0.02 CI instead of 0.002, should it not?....

As we have discussed previously, there is a slight delay (on the order of 0.05 mSec) that it takes for the air to travel the roughly 1" distance from the valve to the pellet.... That will shift all the data in your Internal Ballistics spreadsheet about that far to the right, and make up for the lag you are noting with your new measurements....

Is the green line on the second chart from your Internal Ballistics spreadsheet.... or calculated from the time/distance data from the test shot?.... A quick distance/time calculation from your test data of the last two sensor positions gives the following average velocity....

Distance = 22.21 - 13.21" = 9" = 9/12 = 0.75 ft.
Time = 3.68 - 2.61 mSec = 1.07 mSec = 0.00107 sec.
Average Velocity is therefore 0.75 / 0.00107 = 701 fps....

I don't understand what the purple line is in the second chart....

Regarding shortening the barrel, don't forget that 750 fps with a 14.3 gr. pellet is a pretty "mild" shot for a Disco, and you could indeed shorten the barrel without losing too much velocity.... I checked the numbers using my version of Lloyd's spreadsheet (J4) and with an 18" barrel got 734 fps (instead of 746) and with a 30" barrel got 749, with the velocity starting to drop at 32".... That is using 2 lbs. of drag for the pellet.... Higher velocity shots will, of course, utilize barrel length better....

Bob

[rsterne]

For the timing of the valve hit by the hammer.... could you not use an electrical contact from hammer to valve stem to trigger that?.... You might even get a measurement of when the valve closes when the contact is broken again.... and therefore an accurate measurement of dwell....

Bob

[lloyd-ss]

David, The piezo is indeed pretty "busy" and hard to extract good time data from. More sensors would definitely be better, in the barrel, and on the valve stem. Some sort of contact or whisker mechanism. Using the resistor network for multiple successive sensors per channel, and with 2 channels,  recording the data shouldn't be a problem. Getting all the sensors to fire successfully will be a problem, though.

Bob, on the last chart with the 3 lines, the uppermost green line (velocity) is taken from the calculations from the J4 spreadsheet.  The lower purple line, the time vs distance line, is also taken from the data in the J4 spreadsheet, although I normally don't graph the time vs distance.  The red line is the actual data from the data taken by the oscilloscope... the "real" data. I changed the t-port volume in the J4 spreadsheet as you suggested, and it did indeed raise the purple line up to more closely coincide with the red line. 

It seems like determining the actual zero time point in the process is important, and a bit tricky to do, also.
Lloyd

[Scotchmo]

...
Before and after pressure readings of the 128cc air reservoir will be measured with a digital pressure gauge....

It would be interesting to see those values. We might be able to determine if the lost energy was in the exiting air (reservoir pressure fell to what you calculated it should), or still remained in the 128cc reservoir (reservoir pressure fell a little less than you calculated).

[WhatUPSbox?]

Lloyd,
If you think the Disco is repeatable shot to shot, getting the zero time is fairly easy and cheap (price of wire) using the barrel probe method George describes in his CP-2 testing in the tail end of the other thread.

The Piezo data does get busy but in past tests (piezo microphone was used) it correlated well with a hammer probe to repeatably detect the initial hammer strike on the valve,  as you've also noted. After that discerning other events was more difficult. It may be interesting to fire the Disco without a pellet and see if there is anything in the trace that is distinct to identify valve closing. On a 2240, a hammer probe with a light trap seemed to capture valve open/close (1.6 to 1.9ms duration) but it needed rear access to the hammer through a modified power adjuster.

Great work

[lloyd-ss]

...
Before and after pressure readings of the 128cc air reservoir will be measured with a digital pressure gauge....

It would be interesting to see those values. We might be able to determine if the lost energy was in the exiting air (reservoir pressure fell to what you calculated it should), or still remained in the 128cc reservoir (reservoir pressure fell a little less than you calculated).

Scott, Shot #1, for which the data is posted was 1455psi before and 1421 after, 746fps, 14.5 gn.
Shot #2 was 1414, 1377, 754 fps, 14.5 gn.
Shot #3 was 1376, 1341, 558fps, 25.4gn.

I still need to post the data for shots 2 and 3.
Also, the power level needs to be increased on the gun (eventually) to get a broader range of data.

Stan, I am working on a method to detect valve stem movement, but don't know what it will be yet, LOL.
Lloyd

[lloyd-ss]

Another sensor has been added to the test rig, and it seems to work. It is an analog position sensor for the valve stem to show the actual shape of the valve open and close pattern, along with the position of the pellet as it travels through the barrel.
Again, simple stuff, not pretty, made from odds and ends.

I had to make a new valve poppet with a non-hardened steel stem so that I could drill a hole through it to put a spring steel pivot wire to transfer the valve stem travel out of the air tube to the outside world. The valve has seen better days, LOL. It is bent to pass thru the window in the air tube at the correct angle.



Thsi is the actual position sensor.  The stiff wire from the valve stem sticks thru the window in the air tube, and then all the way thru a hole drilled in the black Delrin pivot block. As the valve stem moves back and forth, it causes the pivot block to rotate in an arc of about 30 degrees. It rotates on the little 2-56 screw. Then, a separate wire rod with a loop soldered onto the end of it extends out of the pivot block, so as to swing a radius around the inside surface of the piece of pellet tin. This is the wiper for the potentiometer (rheostat) that will vary the voltage with the movement. (Measured across the yellow and gray wires) The wiper rubs along piece of mylar film with a resistor material deposited on it. It came from the inside a throttle position sensor, and is two-face-taped to the inside of the tin can. The 2 faced tape is foam backed so it has some "give" for the wiper movement. There is also an anti-backlash spring (hairpin spring) on the delrin pivot block to keep everything tight and jiggle free.
When I push the valve stem in and out of the valve, the wiper sweeps on eh resistor, and the resistance varies from about 1.5 to 3.0 k-ohms. With a second resistor in the circuit to make a voltage divider, the varying voltage read across the resistor is proportional to the distance traveled by the valve stem. In this set-up, a change of 32 millivolts is equal to 0.01"of poppet travel.
   


Actual test results and charts of the valve poppet travel coming shortly.
Lloyd

[rsterne]

Rube Goldberg would be proud.... 

The best thing is that as crude as it looks, it WORKS GREAT, I have seen the data.... 

WELL DONE, Lloyd !!!

Bob

[lloyd-ss]

And you can't beat the cost, LOL.

[lloyd-ss]

Here are scans of of the o-scope pics of 2 shots with the various sensors attached and recorded.

The first shot is from the Disco with 24" barrel, 48 cc reservoir,1472 psi start, 1419psi end, 53 psi used, 14.5 gn, 778fps, .035 wire spring set for a light strike.
The horizontal axis is time, 1/2 millisecond per large division.
The 5 spikes along the bottom of the scan show the timing of the pellet as it passed each of the 5 position sensors. 
The humped line at the top shows the valve as it opens and closes.
The valve starts to open about .15 millisecond before the pellet passes the sensor at .11" of travel.
At 3.4 milliseconds, the pellet passes the last sensor, but that is still about 1.25" before the end of the barrel.
The poppet only  rises .029" off of its seat, and that is shown by the height of the hump in the line.
Sensor #4 is 13.19" down the barrel, and you can see that the valve is closed before the pellet passes that point.



Here is a scan of the second shot; a more aggressive shot.
The main difference in this shot is that a .040 wire spring was used (stiffer than the other shot), and the striker was set for a medium strike rather than a light strike. This shot used much more air, 88 psi, and achieved 840 fps.
The valve dwell is about 3.1 seconds and the lift height is much higher at .100". Because this chart uses a vertical scale of .1V/div, instead of the previous .03V per division, the extra poppet height is very obvious, but it is 3 times as high as the previous shot.  Also note that the valve closed just barely before the pellet passed the last sensor, that is 9" farther than the previous shot.



More work to do!
Lloyd

 

[WhatUPSbox?]

Lloyd, great job, lots of data in those two curves to chew on. Looks like there may be a little ambiguity on the start time. Have you run the valve and the piezo signals in one run to see if the piezo can identify the hammer strike?

Again, great job, I think you win the MacGyver award

[Mole2017]

Nice! Getting that pot in there really gives a nice trace!

[lloyd-ss]

Thanks guys! 
I noticed the start-point ambiguity also, and I think I've figured it out. If you look at the trace for Disco 10, you can see a little bit of movement when the hammer hits and the valve poppet is just barely starting to move.  This is with a 30mV/div vertical scale.  This apparently is where the plastic deformation in the poppet material is recovering, but the poppet is not yet moving briskly.  Once the poppet cracks open sufficiently, the air starts to pass thru and the opens quickly.  For shot 12, the vertical scale is 100mV/div, and the scope trigger had to be set higher, and it therefore missed the start of the hammer strike.

The time and voltage scales are masking the start time, so I ran a few more shots, but this time with a .2msec per horizontal division, and with a 10 millivolt per div vertical scale. With the 10 mV vertical scale, I could set the scope trigger close enough to the idle voltage so that the vague initial movement of the stem was detected. The couple of shots that I took showed a time of about 0.4 milli sec between the  hammer hit and when the pellet reached .11" of travel in the barrel.

I will probably try and install a sensor that will detect the first movement of the pellet. That, along with the hammer strike detection will give the lag between strike and movement.
I think the piezo discs are a little too excitable for the task at hand.

Lloyd

[Machinist]

During the first low power shot, the valve closed when the pellet was roughly halfway down the barrel.    But the next slightly hotter shot had the valve open much longer, with much more lift.  To me the data from the slightly more powerful shot was surprising.  If the second shot was 950 fps such a change woul be expected.

Steve

[lloyd-ss]

Steve, Yes, some of the data makes me scratch my head, too. For these 5 shots (graphs for only two are displayed) a wide range of settings was used to give a broad picture of activity within the barrel. The stock .035 wire spring, a heavier .040 wire spring, 14.5 and 25.4 gn pellets, and different force setting on the spring using an RVA. But all the shots were taken with pressures below 1500 psi, so maybe that accounts for the lower velocities.

Here is a chart of the 5 shots taken during the first session with the test rig.  Wearing-in of the test sensors might account for some anomalies in the readings, but overall, the readings are not unexpected.
The position contact sensors are at .11", 1.91", 6.15", 13.19", and 22.16" of travel. The end of the barrel where the tip of the pellet leaves (as if contacting a sensor) is 23.41" and there is no contact sensor at that point.
If you look at the last 4 columns in the chart, a lot is revealed about the final acceleration of the pellet in the barrel. In particular, look at low powered shots 10 and 11. In both of these shots, the average velocity between sensors 4 and 5, the last 2 sensors in the barrel, was very similar to the MV as measured about 4 feet past the end of the barrel. I think this shows that the pellet was decelerating before it left the end of the barrel. Think of it this way, the 4-to-5 distance is 8.97" and the vel would normally be rising as the pellet passed thru that section. But if the average vel is the same as the MV, the vel must have risen past the level of the MV, and then slowed down to the MV before it left the barrel. I realize that there is a drop in vel from the barrel to the chrony, but I don't know how else to account for those numbers.

The higher powered shots 12, 13, and 14, all show a 4-5 velocity noticeably less than the final MV, which is what I would expect.



There is certainly a margin of error in the test equipment, and in my physically measuring the distances on the charts, but a consistent pattern seems to be showing up in the results. Yes, a lot to chew on.
Lloyd

[rsterne]

Lloyd, when I set up your J4 Spreadsheet with the parameters for the first shot (#10), which requires an efficiency factor of 59% and a dwell of 0.00144 sec.... the time increment where the pellet has moved 0.11" is 0.23 mSec.… you got 0.4 mSec. so the difference must be the poppet decompressing and the time for the air to get from the valve seat to the base of the pellet.... It calculates a velocity of 95 fps for the pellet at that point....

Shot 11 worked out to 0.0011 sec. dwell, 65% efficiency....

Shot 12 worked out to 0.00215 sec. dwell, 53% efficiency....

Shot 13 worked out to 0.0027 sec. dwell, 50% efficiency.... exactly as measured !

Shot 14 worked out to 0.0025 sec. dwell, 52% efficiency....

Just based on the big difference on the 2nd shot (#11) I think the measured data for that shot is flawed.... I don't know where, but it just "doesn't fit", IMO.... It is interesting that your measured dwells are greater than the calculated values.... That could be the difference between the "mechanical dwell" and the "effective dwell when flow occurs" as we discussed previously.... Still, all information aids the learning and understanding process....

Bob

[lloyd-ss]

After some clean up and refinement of the fixture, and adding a few more sensors, I will be able to collect some cleaner data.

Regarding the seemingly long dwell times as we've discussed, I agree that a portion of it might be in the plastic deformation of the valve poppet. In this situation, the valve is being held closed by a force of about 65 pounds. The hammer hits it from an equivalent drop height of 4.5 feet (17 fps.).  The hammer is basically trying to punch the .156 dia valve stem thru the Delrin poppet head.  I believe that the Delrin head will deform more than a few thousandths before it starts leaking air past itself. Once the pressures start to equalize on both sides of the poppet head, the force trying to close/hold closed, the poppet, drops from 65 pounds to about 12 pounds. Then the poppet can more or less fully open. I think a fast moving high energy hammer will deform the valve more than would be revealed by static testing with a similar dead weight force.