Drip Governor - A most valuble first hack

[3dspider]

We need to watch what metals we place into the saltwater. Most neodymium magnets are nickel plated and will cause any aluminum parts (like the drip sensor) to corrode faster. see https://en.wikipedia.org/wiki/Galvanic_corrosion . To get around this, we could use plastic coated magnets or epoxy coated magnets like this.

[BrockMcKean]

To get around this, we could use plastic coated magnets or epoxy coated magnets like this.

Yes, and it really doesn't change the price much either. They should definitely be epoxy/polyurethane coated. Also, I did a bit of digging looking for some indication as to whether a cell phone could provide 5V output from micro usb and haven't found anything conclusive, except that micro and mini usb are indeed capable of passing 5V power. It's just a wire so I assumed as much, but if the ability to print from a phone is a feature the Peachy project is serious about, we should address this now and incorporate it into proposed variable speed drip systems.

Does anyone know if there is a standard for voltage output from phone usb or if that's something you can rely on at all...? I'd like to see the peachy's schematic if anyone can point me in that direction or if a beta tester can work one up in a SPICE program for us all. We can come up with a better guess at how to include an electrically controlled drip governor like this without disturbing the work that's already been done.

[jstrack2]

Thanks for the feedback! Again if anyone wants to see something more let me know. I know the videos were very short.


Brock McKean: I think that while it would be possible to use frequency modulation (or other techniques discussed in this forum) to directly control the drip speed with the speaker, in my opinion this is not the way to go. I think that having the speaker control the laser as well as possible should be the priority. Adding a bunch of complexity (or greatly limiting the drip governor's functionality) to save one or two dollars does not make much sense to me. I agree that USB power sounds good. What I did here was about 50 mA. I think any USB 2.0 including the micro does at least 500 mA. I think they are always at 5V. If it were made smaller and the coil was not poorly hand wound as it was here then I think it could work on even less than 50 mA. Since when the valve is open liquid rushes through fast I think that there is not a need to have it stay open when the coil is not powered. The coil will only be powered for a tiny fraction of a second anyway.

I agree that some schematics would definitely be helpful.


As for maintaining the orientation of the magnet this happens naturally, so no extra work is needed. The field makes the magnet's orientation stable.


There is something about the pressure worth thinking about though. This valve should be placed as high up as possible to minimize the pressure and hence the magnetism needed. Again having the valve be smaller will reduce the magnetism needed, since of course less area experiencing a pressure means less force. If the method of having the salt water drain out is used instead of poured in then the pressure level may vary significantly over the course of the print (depending on the setup the pouring method may also result in significant pressure change too). This may have to be corrected for in software (longer/stronger pulse when pressure is high). I don't think that spectacular accuracy is needed here though because that will be provided by the water level measuring methods.




3dspider: Great point, I forgot about that. I also agree prebought or coating afterwards of the iron/magnet will probably be needed. As you and Brock McKean said this shouldn't be a big deal.

[BrockMcKean]

EDIT: I just realized that Rylan has already taken pretty much all of this into account save specifics of electrical and programmatic operation, so perhaps we should focus on that. :S

Brock McKean: I think that while it would be possible to use frequency modulation (or other techniques discussed in this forum) to directly control the drip speed with the speaker, in my opinion this is not the way to go. I think that having the speaker control the laser as well as possible should be the priority. Adding a bunch of complexity (or greatly limiting the drip governor's functionality) to save one or two dollars does not make much sense to me.

If the point is to keep the cost as low as possible and it's reasonable to do something that would decrease the cost I don't see why it shouldn't be done. If the mirrors truly operate on AM, FM should not affect them at all. But if we did want to add a filter before it, it would guarantee that no FM would ever reach the laser assembly. Adding somethign like that will cost basically nothing, pennies. Placing a low pass filter before the peachy's mirror/coils would completely isolate it. Using an op-amp to, connected to the 5V for DC bias, and passing the frequency from the audio jack, the coil can be controlled directly. I guess it sounds a little complicated, but I assure you things like this (using multiple and separate modulation schema) are done all the time in many electronic devices.

Since when the valve is open liquid rushes through fast I think that there is not a need to have it stay open when the coil is not powered. The coil will only be powered for a tiny fraction of a second anyway.

Yes, I agree. It shouldn't stay open. This means we need a switch between 5V DC and Ground, or some way of using the AC signal directly from the audio output that will result in the desired movement. Maybe we can use PWM over USB instead if we are going to use it for a DC bias directly from the device anyway. This would probably be easier and certainly separate if it can be achieved at the correct voltage/current level on the data line USB data line.

As for maintaining the orientation of the magnet this happens naturally, so no extra work is needed. The field makes the magnet's orientation stable.

Unless the magnetic field is on the entire time and the desecent of the magnet is forced and controlled, there's no way to guarantee that freefall will not result in the magnet ending up lop-sided, assuming there is pressure above the magnet still. However, if it has some sort of guiding rod to prevent it from turning to that kind of angle that extends from the magnet up/down the drip a bit it should be no problem.

There is something about the pressure worth thinking about though. This valve should be placed as high up as possible to minimize the pressure and hence the magnetism needed.

Yes, but then you also have to worry about the reservoir's resin level. I'm not really sure how the drip works now exactly (at all really), but any resin below the valve would have to be pumped to the valve somehow or the reservoir would have to mechanically change to deliver the resin. So if this is not a problem it should be placed as high as possible.

I don't think that spectacular accuracy is needed here though because that will be provided by the water level measuring methods.

If the drip governor can only move to one position we're limited to altering the period for variability, which is fine and probably the best approach. However, the dynamic range will then be determined by the minimum amount of time we can open the valve. Accuracy is not a problem here, but if we want to be economical about it, the software should take the drips into account and calculate the resin level without measuring afterwards (because the drip resolution and number of drips are known, the entire volume is known). This would require using a well defined right rectangular prism as a reservoir (all opposing sides parallel, all adjacent sides perpendicular) so that the area for each level is identical. I'm not sure if that's part of the peachy's design right now or not to use some sort of level measuring method, but calculating it this way would be much less invasive and not disturb the print surface/material in any way and would allow for more printable volume.

[jstrack2]

He has already worked on something very similar, however I don't think that he has gotten a super cheap and easy to make valve to work well yet. I could be wrong though.

If the point is to keep the cost as low as possible and it's reasonable to do something that would decrease the cost I don't see why it shouldn't be done. If the mirrors truly operate on AM, FM should not affect them at all. But if we did want to add a filter before it, it would guarantee that no FM would ever reach the laser assembly. Adding somethign like that will cost basically nothing, pennies. Placing a low pass filter before the peachy's mirror/coils would completely isolate it. Using an op-amp to, connected to the 5V for DC bias, and passing the frequency from the audio jack, the coil can be controlled directly. I guess it sounds a little complicated, but I assure you things like this (using multiple and separate modulation schema) are done all the time in many electronic devices.

Yes I know that using multiple modulation is common and relatively simple. However I am skeptical here that this will not cause issues by trying to do too much at once. Very little can throw off things when stuff is controlled directly by varying voltage levels as the mirrors are. Also if USBs are already used for power, why can't they just control this too? I think using the USB would be cheaper anyway. I am still not totally clear on the details of how everything is powered, but yeah I think using the USB is the way to go.

Yes, I agree. It shouldn't stay open. This means we need a switch between 5V DC and Ground, or some way of using the AC signal directly from the audio output that will result in the desired movement. Maybe we can use PWM over USB instead if we are going to use it for a DC bias directly from the device anyway. This would probably be easier and certainly separate if it can be achieved at the correct voltage/current level on the data line USB data line.

Exactly. Although it doesn't have to be pulse width modulation necessarily, just a pulse whenever it wants! Although I assume this is what you mean. As for the voltage and current I am very confident this can be made to work.

Unless the magnetic field is on the entire time and the desecent of the magnet is forced and controlled, there's no way to guarantee that freefall will not result in the magnet ending up lop-sided, assuming there is pressure above the magnet still. However, if it has some sort of guiding rod to prevent it from turning to that kind of angle that extends from the magnet up/down the drip a bit it should be no problem.

I disagree. I think this is a non-issue for multiple reasons. First if the solenoid gets pulsed for a tiny amount of time then the magnet rubber ball object plug won't have time to rotate. However let's say that instead it is pulsed for a long period of time, like in the videos that I showed. Say then downstream of that a narrow opening is used so that the tubing still has drops come out. In this case there still will not be an issue because the rotation forces caused from asymmetries of the water passing through the valve will not be enough to rotate it, even will some pressure. Also if there is minor rotation this is fine. The ball is a sphere of course, so even if it is rotated some it will still plug the hole the same way. If you are still worried about it then if the magnet/iron is thick enough and close in diameter to the PVC adapters inner diameter it won't be able to rotate. And in addition to this if there was still problems then people could figure out what is causing crazing rotational forces and maybe have the solenoid semi powered when the magnet/iron is falling. So yeah there are a lot of reasons this is not worth worrying about.

Yes, but then you also have to worry about the reservoir's resin level. I'm not really sure how the drip works now exactly (at all really), but any resin below the valve would have to be pumped to the valve somehow or the reservoir would have to mechanically change to deliver the resin. So if this is not a problem it should be placed as high as possible.

I should have been more clear here. When I said "as high up as possible" I meant taking this into account. If the salt water is dripped in then the valve will have to be below any water that it going to be dripped in eventually. If the salt water is slowly dripped out (as would be done if the salt water + resin + fresh water printing method is used) then the pump will have to be below the lowest point that the resin gets when the print is finished (which is when the resin is in its lowest point).

If the drip governor can only move to one position we're limited to altering the period for variability, which is fine and probably the best approach. However, the dynamic range will then be determined by the minimum amount of time we can open the valve. Accuracy is not a problem here, but if we want to be economical about it, the software should take the drips into account and calculate the resin level without measuring afterwards (because the drip resolution and number of drips are known, the entire volume is known). This would require using a well defined right rectangular prism as a reservoir (all opposing sides parallel, all adjacent sides perpendicular) so that the area for each level is identical. I'm not sure if that's part of the peachy's design right now or not to use some sort of level measuring method, but calculating it this way would be much less invasive and not disturb the print surface/material in any way and would allow for more printable volume.

I think that measuring after the fact is pretty important. I am skeptical that only counting drips will ever be that accurate (at least with low cost equipment). Drip sizes vary some. It is pretty hard to get around that. Now if you don't care about print accuracy this can be fine. For example you could print some object that is real detailed but is ten percent too tall and it may still look great. However for other applications very high accuracy is needed. Gears would be a very clear example of that.


I submitted a video in a different thread using a caliper I took apart to measure the water level with up to 0.02mm accuracy: http://3dprintboard.com/showthread.p...on-idea/page11 Check out the improved video, the first is really bad!


It is quite cheap. Unfortunately the video may not do much to calm your worries that it will take up space, since I used a huge sheet of styrofoam. However I assure you this is not at all necessary. With a properly designed float it can have almost no effect on the maximum size of the print. Additionally, a slightly bigger container for the printer could always be used! Also it will not cause an ripples. The float steadily rises with the liquid if built properly (which laser cut parts would be). As I stated on that thread this with a Kalman filter will allow for very very high accuracy and resolution in the Z-axis. The USB can be used to receive the data here too! haha

[BrockMcKean]

Exactly. Although it doesn't have to be pulse width modulation necessarily, just a pulse whenever it wants! Although I assume this is what you mean. As for the voltage and current I am very confident this can be made to work.

Yes, which should be a constant PWM per layer. The only time the PWM should change is when a layer will take more or less time than the previous layer. The only time the PWM will be off is when the print is finished or paused. I think we are essentially saying the same thing.

I disagree. I think this is a non-issue for multiple reasons. First if the solenoid gets pulsed for a tiny amount of time then the magnet rubber ball object plug won't have time to rotate. However let's say that instead it is pulsed for a long period of time, like in the videos that I showed. Say then downstream of that a narrow opening is used so that the tubing still has drops come out. In this case there still will not be an issue because the rotation forces caused from asymmetries of the water passing through the valve will not be enough to rotate it, even will some pressure. Also if there is minor rotation this is fine. The ball is a sphere of course, so even if it is rotated some it will still plug the hole the same way. If you are still worried about it then if the magnet/iron is thick enough and close in diameter to the PVC adapters inner diameter it won't be able to rotate. And in addition to this if there was still problems then people could figure out what is causing crazing rotational forces and maybe have the solenoid semi powered when the magnet/iron is falling. So yeah there are a lot of reasons this is not worth worrying about.

I think that measuring after the fact is pretty important. I am skeptical that only counting drips will ever be that accurate (at least with low cost equipment). Drip sizes vary some. It is pretty hard to get around that. Now if you don't care about print accuracy this can be fine. For example you could print some object that is real detailed but is ten percent too tall and it may still look great. However for other applications very high accuracy is needed. Gears would be a very clear example of that.

These two issues are related. If we want lower cost, we want a smaller magnet, smaller coil, and less current. A larger magnet and stronger field will be less susceptible to fluctuations caused by the viscosity and pressure of the fluid, but shrink it all down and the forces might become relativistic. In either case, the solution is to create a guiding element with the epoxy/polyurethane. If you look at Rylan's original design concept in his video at the beginning of the thread, it is "encased" in a right circular cone on each side. This will keep the magnet coming back to where it should be regardless of minor fluctuations and will allow the magnet to be much smaller, require less energy to operate, and potentially increase the drip frequency.

Additionally, the drip is not precise, you are correct. However, it is accurate. That is to say, it does of course deviate a little, but this magnet can only come up and go down in so many ways. The drips will be of a size that will all fit under a normal PDF curve, and the drips are so small in comparison to the level (assuming a large enough print volume), the error in calibration and calculation should be no more negligible than the error in measurement and calculation. You have to either assume the drip and container volume or assume the resolution and the zero position of the measurement device. Standard PLA printers do calibration tests. Inkjet and Laser 2d printers do calibration tests and alignment prints. Doing an alignment print on the peachy and assuming you are using a right rectangular prism for the print volume is no different than assuming you have assembled and properly calibrated a level sensor within the print area.

In shrinking the magnet and speeding up the drip frequency, the constant surface tension and viscosity should hit a critical point that will simply not allow liquid to pass. If it is designed such that it's relatively close to this critical point the surface tension and viscosity should maintain a very constant drip volume. Another approach to this would be to use some sort of telescoping system for the magnet that only allows a specific volume by design to enter a chamber. Magnet goes up, chamber is released, magnet goes down, chamber is refilled. Not sure how to do that right of the top of my head, but I assume something like that could be done.

Calibration and assumption is used often in electronics for many purposes, but yes, there is usually some feedback mechanism to determine that the calibration and assumption still holds true. I would agree that measuring is important, but adds another layer of complexity for users that just want the $100 printer To print some really cool objects or don't care about it being accurate to a few microns. So, for simplicity and costs sake I say why bother? On the other hand, measuring afterwards allows you to have basically any printable volume container shape. Whichever solution is more important for the most basic $100 users and cost lest to implement, we should pursue. Hell, why not just do it both ways?

[jstrack2]

If you look at the video I posted there really isn't any rotation. If it were made smaller I agree the rotation effect would increase, but not that much. I think it would scale linearly with the radius of the plug, if not even sublinearly. So even if the whole thing were four times smaller in all dimensions I think that this effect would still be pretty negligible. Even with my 27 mm wide design the current draw was not much. At a fourth this width the current draw would be absolutely tiny. Also as I said the PVC tube itself already guides the plug some. Finally I don't know what you mean by relativistic, maybe you mean quantum? But that is irrelevant unless you get down to the scale of a few nanometers! Anyway since I don't understand the theory nor did my experiment show this to be a problem, I think that you should make a video of the valve if you are still concerned about it. If it has that problem then I will try to help and address it.

It is neither super precise nor accurate. You can not expect it to behave with the same predictable probability density function every time (such as a normal distribution for example). It will always drift. This is why for example gyroscopes and accelerometers are used together, and not just gyroscopes. Gyroscopes measure rate of change of the angle with respect to time. So then angle is found by integrating over time. But there is always drift, so accelerometers which directly measure angle are also used to correct for this limitation. Another example is how battery capacity remaining is calculated. In theory you could integrate current use over time. However because of the drift problem a direct measurement method is also needed (such as checking voltage here). This is why the threads discussing making a direct measurement were created. It is not a matter of just calibrating it once. That will not get very accurate results. Again if you want to save a few dollars on the printer and have not very good at all accuracy in the z axis that is fine. But this problem can cause the z axis to be off multiple percent (not microns) and direct measurement should only add a few dollars cost (probably less than 5). I do not mean to say that drip counting shouldn't be done however. It gives amazing resolution, just not accuracy.

[BrockMcKean]

If you look at the video I posted there really isn't any rotation. If it were made smaller I agree the rotation effect would increase, but not that much. I think it would scale linearly with the radius of the plug, if not even sublinearly. So even if the whole thing were four times smaller in all dimensions I think that this effect would still be pretty negligible. Even with my 27 mm wide design the current draw was not much. At a fourth this width the current draw would be absolutely tiny. Also as I said the PVC tube itself already guides the plug some. Finally I don't know what you mean by relativistic, maybe you mean quantum? But that is irrelevant unless you get down to the scale of a few nanometers! Anyway since I don't understand the theory nor did my experiment show this to be a problem, I think that you should make a video of the valve if you are still concerned about it. If it has that problem then I will try to help and address it.

I'm just saying it is something that should be given some attention as it's scaled down more. All I mean by relativistic is that the forces from viscosity is basically negligible beyond a certain size and magnetic field strength. I don't know what size magnet and field strength you would need to go down to to make these two relatively close, but there is a point that it would happen. We can probably just use more coil windings if in shrinking it that did become a problem, but that is my point. Just something to think about and be aware of as a factor that could possible have some influence on a smaller scale.

It is neither super precise nor accurate. You can not expect it to behave with the same predictable probability density function every time (such as a normal distribution for example). It will always drift. This is why for example gyroscopes and accelerometers are used together, and not just gyroscopes. Gyroscopes measure rate of change of the angle with respect to time. So then angle is found by integrating over time. But there is always drift, so accelerometers which directly measure angle are also used to correct for this limitation. Another example is how battery capacity remaining is calculated. In theory you could integrate current use over time. However because of the drift problem a direct measurement method is also needed (such as checking voltage here). This is why the threads discussing making a direct measurement were created. It is not a matter of just calibrating it once. That will not get very accurate results. Again if you want to save a few dollars on the printer and have not very good at all accuracy in the z axis that is fine. But this problem can cause the z axis to be off multiple percent (not microns) and direct measurement should only add a few dollars cost (probably less than 5). I do not mean to say that drip counting shouldn't be done however. It gives amazing resolution, just not accuracy.

Gyroscopes measure orientation and accelerometers measure acceleration. Sampling your accelerometer will give you a function of acceleration. Integrating that gives you velocity and integrating that will give you displacement. Match that up with your gyroscope samples, and you have a time correlated displacement and orientation pair that can be used to explain where you are in space from a given reference point along this vector. So, if you were to depend on only the gyroscope you would not be able to determine if you were moving at all. The gyroscope does not know the difference between facing a direction for a specific amount of time and traveling in the exact same direction for the same specific amount of time. Likewise the accelerometer doesn't know direction at all. It only knows force due to acceleration, so you can move it in any direction with identical force and it doesn't know the difference between any direction in particular. However, even using an accelerometer and gyroscope together, you cannot be sure of how much time and distance was covered exactly, due to drift.

Drift occurs in sold state devices and crystals with respect to time. The crystal structure's and electron affinity changes overtime and becomes misaligned internally. The same can be said for any kind of doped semiconductor. The depletion regions do not stay the same and therefore the output of diodes and transistors is not consistent over long periods of time. Any clock dependent measuring source also experiences this drift depending on the precision of the clock source.

This proposed drip governor does not contain any of these elements. The mass of the magnet will not change. The mass of the coils will not change. The strength of the magnet and the strength of the field induced by the coils will not change. They will operate the same way every time, given a specific voltage. The voltage/current and modulation will be SLIGHTLY different each time, probably by a few uV/uA and ns/ps. That is to say, using PWM or any other clock dependent form of modulation (anything coming from a computer would be), is susceptible to this. Computers adjust their clocks because they experience drift, but they do it only frequently enough for everyday units of time. microsecond, nanosecond, and picosecond drifts pile up until the clock decides millisecond drifts are too much, so this would happen. However,since our drip rate shouldn't be near the frequency to have a period of microseconds, the computer's compensation for this drift should actually keep it pretty steady.

Beyond that we're talking about the consistency of the voltage output from the USB. If one system uses 5V and another puts out 4.99V and another puts out 5.01V, that will certainly result in some amount of variability in the drip, but not much, and it can be accounted for by a calibration print that assumes differing amounts of variability within a range measured in development for small "identical" objects. Whichever prints the best is the correct amount of variability. All other printers 3d and 2d have some form of calibration. Measuring digitally would be ideal, but software is free and the resin is pretty cheap. So, unless it costs less than a first calibration print of a few mL and it can be accurate to 100 micrometers or so, I would rather make sure the kit costs as little as possible. Now, for an assembled kit or a higher quality commercial version, I agree with you 100%. I'd be a bit confused if there was no measuring of the fluid level.

Also, I'm very busy with some other things right now but I'll be wrapping them up today and have a lot of free time for the forseable future, so I will most certainly be doing my own experiments and contributing as much as possible. I just REALLY want to get working on this and I just don't have the time until tomorrow on, so I'm talking about it instead.

[jstrack2]

Oh alright you mean relative. Relativistic means that relativity is applicable. Like for example the water is flowing close to the speed of light. If that is the case then I have no problem believing that there could be some rotational effects on the plug! haha Seriously though I definitely think in general that trying to find problems before something is built is a good thing, but here I think that rotation won't be an issue.

As for the gyroscopes I was again not clear. Sorry about that. It is true that these gyroscope sensors can not measure linear acceleration, so an accelerometer (or something else) would be needed for an application requiring linear acceleration measurement. They measure angular velocity. Accelerometers measure acceleration, but since there is gravity pointing the same direction they can be used to measure angle as well. So anyway my point of the example was that if you tried to know the angle by just integrating the gyroscopes data about angular velocity you would over time be more and more wrong. Just using the accelerometer may not be optimal either, because linear accelerations like vibrations will in the very short term give weird readings. If the information is combined then you can get very good orientation data. This is similar to the drip and direct measurement system for the Peachy Printer. If only the drip counting is used then the accuracy will not be great. If a direct measurement method is used and not the drip counting method then the accuracy will be great, but it likely will not have the potential submicron resolution that the drip counting offers. If both methods are used and the software uses the Kalman filtering method then the printer will have both great accuracy and resolution. This is why I was excited to make the videos showing the caliper working. Having submicron resolution and extremely good accuracy with a 100 or so dollar printer is pretty ridiculous!

As for why this will have drift it is for different physical reasons then a MEMS gyroscope sensor. Here I think the main source of inconsistency will be how the salt water flows through the valve. Maybe the changing pressure could be correctly compensated for in software, but that won't be enough. Especially if the valve is only open long enough to allow for a drop or a few drops to pass through at a time. The exact amount that gets through will be very complicated. There are a lot of factors at work here which will lead to the amount of water getting through each time varying. Also when you consider that there is salt here to which can change the properties and that the valve and rubber ball aren't flawlessly manufactured you will have inconsistencies. This will happen way way before you get the plug rotating. Also if you perfectly calibrated it (which would not be that easy) and then did a couple prints the accuracy might be half way decent (but still not that great). But then could you say that a couple weeks later without more calibration it would be? I don't like the idea of having to do intense calibration all the time for only decent accuracy. I mean I would still buy the Peachy Printer without this feature, but I would rather spend an extra five dollars or so to have this direct measurement. On a side note though it is weird to be someone arguing for a more expensive solution, usually I am fighting to spend less! haha

But yeah I look forward to seeing the videos you make!

[BrockMcKean]

Yeah, it probably wont' be an issue but I like to build things so that it's not possible for them to be an issue. Kind of like how you suggested PWM instead of further complicating the AM signal with FM. Yes, you are right that error compounds in calculating something like the integration of the gyroscope, but that is due to the device's clock drift and breakdown of the semiconductor over time. And yes! I meant relative, not relativistic. Brain fart. I hope the drip doesn't get close to the speed of light either, although it would be quite interesting.

I guess it's just my experience that rigorous testing in r&d phases can result in a true value calculation. That is to say, you can usually measure the system in some way that can guarantee accuracy one time in r&d, one time in production, or extremely infrequently in use. If we were to vary the frequency of the drip, using the same voltage and current each time, shorter pulses for higher frequency, longer pulses for lower frequency, we would see that the drip size will be fairly similar across a band of drip frequency in the middle, but at a critical low frequency, the drip volume begins to increase much faster until it becomes an intermittent stream and at a critical high frequency the drip volume decreases until it does not drip at all. Those critical frequencies will be based on the viscosity of the mixture and the cross sectional area the mixture has to flow through. If you observe the band between these two critical frequencies for "long" periods of time at a frequency resolution twice or greater than the frequency resolution the governor will use to vary the drip speed, you will see that the drips will be a little larger or a little smaller than the previous drip, but the drip volume at a specific frequency will fall under a normal curve and over a large number of drips, this is calculable.

I'm not saying it's the perfect or best solution, I'm just saying if we want to keep the cost down, this should be an option on the table and it should be tested. These kinds of things cannot be proven to be viable or not one way or the other without a wealth of related experience with this specific kind of system, specific technology, specific materials, etc. Doing it should be a matter of keeping manufacturing error smaller than the measured PDF that governs the drip volume for a "perfect" version of this drip governor and proving (or disproving) that the manufacturing error is indeed small enough with testing. In a DIY kit, it would be useful to keep the cost down. In a pre-assembled printer or a professional quality printer, I would want a measuring device on the other side and I would expect a self cleaning and calibration function just like every other printer. I would not expect the cleaning or calibration to be as trivial as calibrating a 2d inkjet or laser printer. I would expect it to be relative to the print time of a 3D printer, and I'm fine with that just so long as it calibrates and cleans itself! So, I'm not arguing either way. I'm just saying both are viable and one makes more sense than the other in different contexts.

Still waiting on a new camera. Maybe we can Google+ sometime and get a hangout going or work on parts of this together. Virtual peachy makerspace!