Applied Science - 2014-06-08
I built a microbalance based on a design by Paul Grohe: http://www.youtube.com/watch?v=n90whRO-ypE It is has a precision of about 5 micrograms, and I measured a single eyelash at about 35 micrograms. The balance is built from an analog panel meter that is controlled by a servo loop which optically monitors the meter's position. Adding mass to the meter's needle requires that the servo loop add more current to maintain the needle's position. This additional current is read, and converted to a mass value.
Very cool, as always Ben. One thing to be very care with when using gutted meter movements like this is that it is extremely easy for the field magnet to attract any tiny metallic particles into the movement which can easily impede the motion of the coil.
I ruined my first meter this way when I went to drill mounting holes in the metal body!
What I find most remarkable in this video is your ability to do op amp circuit math. I have a computer engineering degree and basically got through fundamental electronic circuits by doing well enough on everything else that I was able to completely bomb any question involving an op amp. (I still had to take the course three times) I simply could never get the math to work through my brain, and never got the right answer.
Thanks, I really appreciate it, but I actually missed an important part on this one. As Chris Tacklind pointed out, I don't have a D term, so the loop will be unstable. It's possible that the binary nature of the optical sensor would still make this very difficult to stabilize, but adding a differentiator would certainly help, and is mathematically necessary.
+Applied Science Why not cut a diagonal into the mask so that it isn't binary?
+ExtantFrodo2 The shape of the mask doesn't mater. Once the light transmission level drops below a certain threshold the photodiode will stop conducting altogether. It's the photodiode that makes the sensor operate in a binary fashion, not the shape of the mask.
sivalley The shape of the mask will change the rate at which the light level changes. With a 90 degree cut the rate of change would as fast as possible. If you had (for example) a vertical cut the rate of change would be zero. Somewhere in between is a reasonable rate of change which crosses the "binary" threshold of the sensor smoothly. With an oscilloscope on the output you might even be able to calculate the necessary angle. In the real world most diodes don't change instantaneously. If you examine their out put at a sufficient rate of speed you can measure the minimum and maximum speed of switching. (It's not even necessarily the same switching on as switching off.) Similarly, there is a narrow, but definite window in the light level threshold the initiates the switching action. To the casual user the window seems like a very square precipice. It's not.
ExtantFrodo2
Yes, I understand the shape of the shutter affects the slew rate, but I was merely relaying that the shape of the shutter will not change the 'binary' output nature of the photodiode.
The fact that the incident light on the diode affects its effective conductivity more so than the time rate change created by the shutter by orders of magnitude, the slew rate becomes negligible. There are two ways to overcome this magnitude barrier; 1) Instead of measuring the voltage through the photodiode (as Ben is doing with the first op-amp), measure the current which varies much more predictably below the conduction threshold (although not entirely linearly). 2) Use a shielded photo-resistor whose output IS linear with respect to slew rate of the shutter.
3) Operate the system in a differential mode, rather than the direct proportional he is now.
Just to recap, I am not saying the shape of the shutter is unimportant, just that in the configuration shown it does not have any meaningful significance.
A microgram balance. Nice!
A simple idea that may add some linearity. Have you considered making the IR beam interruptor triangular.
Nice!
This is a 2nd order loop so it needs a PD controller (I is optional). You have only provided a P term. The needle could be counter balanced to minimize the DC current required.
That's a very good point. It might be possible to stabilize the system as-is with the proper amount of D.
Applied Science Well, that came out wrong...
Would it still be a 2nd order loop if the coiled spring wasn't removed?
I did not understand a word of that, except that he needs more D
I think in order to be precise you need to apply the measured mass to a single small point on that needle else the position on the measured sample will influence measured weight.
I used to work for CAHN instruments, we built microbalances, I used to even wind the coils
Were their microbalances based on the same operating principle?
I've been wanting to get a digital scope for a long long time, but all your recent videos are really making what little cash I have burn a hole in my pocket!
I had an idea for the oscillation issues caused by the tight gradient of the opto-interrupter: instead of using a flat horizontal edge that blocks the emitter use a diagonal one, so that the vertical motion causes a more gradual blockage.
All these videos are amazing. It's so rare to see such intelligence and mastery.
Always something absolutely fantastic. Amazing work!
First time I ever heard of making one with these was actually from a forum back in the early 2000's. It was much more basic where you turned a pot until it zeroed and then did a calculation to figure out how much weight was there. it ended up on hackaday too at one point.
This is in the realms of everything makes a difference. It really makes you appreciate wonder at how awesome meteorology truly is. Especial 100 odd years ago with there equipment and there finding still stand true.
Such a simple and yet incredibly elegant measurement method. Great video!
Must say, had a similar idea to do a servo measuring of acceleration. Schematic idea was a sourced from ELM home build laser. The servo part, but to be rearranged as linear actuator, not rotational, then the error part would be acceleration while position input would be constant.
Very nice idea, nicely done!
I think a lot of the temperature sensitivity is from the coil of the meter warming up, if were to make a current source drive the meter, or sample the current out of the meter with a resistor, you would eliminate this source of drift.
For even better resolution from the tiny scale place a reflective surface at the axes of the meter. Shine a lazor beam onto the mirror surface which will end the beam on the ceiling surface or for better resolution onto a half circle ribbon.
This concept comes from my Cousin Willard Buck back in the '60s. He wanted to measure very tiny changes in gas pressure. He had a man build him a pure quartz curved tube. He attached a mirror at the a axes of two strings attached to the tube and its base. Shining a tiny beam of white light to the mirror he got accuracy from his experiments. Please give my dear cousin many patients a look Willard E Buck. Early on he worked a Los Alamos.
Thanks so much for your great channel.
Making your own optical isolator with a cadmium sulfide cell and an LED so the resistance changes in a linear fashion as the beam is progressively interrupted instead of just an on/off arrangement may be helpful.
Maybe I'm missing something here but couldn't you just run a long average on the Oscilloscope? The noise seems relatively uniform, and even if it isn't, the bias should be consistent.
I don't know, maybe you wanted to just solve the problem entirely on the electronics level, which I'd understand - but from a practical point of view it just seemed like an obvious choice to turn on averaging on the scope... And just wait a little for every measurement, or reset the running average (which I can do with a push of a button on my old HP scope, so I'd expect your shiny new one to be able to do that easily as well).
Interesting video either way!
I guess there is a high chance he would like to use it without the oscilloscope later on, but yeah, it would be worth testing :)
you could damp the meter movement by wiring a series resistor to the meter coil and an electrolytic capacitor across, so it would slow down a sudden inrush of current and or sudden outflow of (discharge) current, and calibrate accordingly with this damping circuit. but indeed its the way to accurately weigh in micrograms. Resistor can also prevent coil from burning out due to massive overload. Also don't forget photo-interrupters can and do get influenced by external lighting, hence most of the noise can be due to mains light switching at 100hz on a 50Hz mains frequency, as each cycle crosses the zero line twice per cycle, starts with zero and peaks positively, then descends to zero again and goes to negative peak and back to zero.Black out the plastic cover and leave a small viewing window, follow good earthing principles as in audio amplifiers to get rid of hum induced noise.
That's a shockingly clever use of an opto interrupter. I'd love to see the system oscillate when unbalanced.
I once made a totally mechanical balance with a long thin wood rod. I could easily determine the mass of a hair with it. The balance was very simple, it rested on two very tiny ball bearing balls.
Regarding getting more gradual response from the photosensor, you can try turning the sensor 90 degrees and cutting the light blocker at an acute angle, so that the motions to open/close the light have to be larger.
Oscillation is the bane of feedback control systems.
Remembering way back to when I took the feedback control systems class... the system goes unstable when a pole goes to the right of the j-omega axis. In real life, that often translates to having too much gain on the feedback circuit.
So, reducing the gain via that feedback resistor pot will probably do the trick. Also, adding a capacitor at a strategic point will slow the system down.
For mechanical dampening, put a strong magnet by that aluminum foil that't near the light sensor. The eddy currents generated by movement will buck the magnetic field and dampen the movement.
The noise I see on your scope looks a lot like the noise that I used to pick up from fluorescent lights.
If you want to fiddle with your circuit more, look up the circuitry that's used for the front end of an EKG. It uses three op-amps (if I remember correctly.) It's designed to have extremely good common mode rejection, and is more noise tolerant than a simpler differential amp.
Try making that vane a slant, so that it interrupts the beam gradually over a small range. Add extra damping by using a resistor in parallel with the meter movement, lowering the series resistor to compensate.
Smallest masspiece I have is a 5mg masspiece, been verified that it is 5mg to 3 decimal places as well, during the last session at the cal lab. The 1g standard was measured at 1.00006g with an uncertainty of 2mg.
As others have said, try using a light blocking element with an angled edge or a transparency with a gradient printed on it. If the photo-diode still has too much of an on/off sensing characteristic, maybe try a different type of photo-diode, or maybe redesign the circuit to use a light-dependant resistor (LDR).
This is a very cool project though. Great ingenuity.
You could low pass the output that goes to the oscilloscope, with a big cap, small resistor (to minimize resistor noise). There seems to be some oscillation in the signal in the scope, probably servo loop caused; so the RC low pass filter with a cut frequency 10x smaller than that freq you see in the scope can help a lot to get a better "weight value".
I think a lot of the noise you are getting is from instability of the platform. eg the fan motor running in your oscilloscope of whatever that thing is and even traffic outside etc.
I wonder if printing out a gradient on a piece of transparency film would give the interrupter a more "analog" response?
Or cut the foil diagonally so it's a triangle instead of a square
this is what i was going to suggest
Or put a minute section of thin mirror on the arm, shine a laser off of it and put several sensors on the wall to establish an analog range with the sensitivity of a literally perfectly rigid, weightless, 10 foot long lever arm.
id put an rc low pass filter on the photo diode line and filter the gain with feedback from the input the to will cancel out (microphone feedback cancellation/ clip limiter ) and give a better reading. another idea is servo the led output or bounce it off the bottom.
@@ceriand the slit is so thin, the response is practically still ON - OFF.
Maybe you could modulate the servo with a let's say 1 kHz sine wave and then measure the dampening different masses would cause. Also with an 1 kHz notch filter you then could get rid of most of the interferences caused via EM, sound and static charge.
Instead of a pot for variable gain, use an R+C feedback network, with a time constant on the order of the movement time constant. So, probably in the seconds. Pick impedance to match the source -- which isn't very well defined (a photodiode operated photovoltaically has an impedance dependent on intensity...nasty nonlinearity!), so a shunt resistor to set that might come in handy. Offhand, I would guess, 100k across the diode, and 10k-1M in series with 0.1-10uF for feedback.
Better alternative: operate the diode in reverse bias and use a transimpedance amplifier (check out photodiode circuits for ideas) to get a stable (and precise) signal, then do the error loop and stuff.
For better averaging, just zoom the scope out more. It'll take the average of whatever data you're viewing. More data = better average (and slower updates..).
Why would you be building a microgram balance? :)
A capacitor across the feedback cap will kill the HF gain and help with stability, as will a series RC shunt to add a differential term in the output drive to the meter movement. Cutting the interruptor flag at an angle might help give it a less steep position sensing gain too. Electrostatic forces will become a major problem at microgram resolution, you should probably build it completely out of (moderately) conductive materials and bond them together.
I built a microgram balance years ago using a polystyrene cantilever with a chip of ferrite glued to it. The position was sensed with a permeability tuned oscillator beat against a reference crystal oscillator and the difference term triggered a monostable which was integrated to get a DC voltage. It wasn't very linear at large deflections, but it was exquisitely sensitive. Unfortunately it was a better electroscope, thermometer and seismograph than a balance, but it worked well enough to resolve grains of salt.
I think this nulling servo approach is likely to have far better dynamic range than my hack.
This is great because most people can do it! (Of course I also like to see the projects that I couldn't possibly repeat.)
I have an idea. I'm pretty sure it has been done before, but I wonder if it is possible to make a scale that works on resonance frequencies. For example, placing an object on vibrating platform will change resonant frequency of that platform. Then, by adjusting frequency to make it resonant again, it may be possible to calculate the mass of an object by comparing these two frequencies. Thank you very much, I enjoy your channel!
I think you could gain stiffness in your balance by forming a circle from very light plastic and centering it on your meter movement. Attach a platform to the face near the rim. If it was clear plastic, you might print a graduated shade on the disk near the opposite edge and use the disk itself as the beam interrupter. The addition of the graduated beam splitter and the slight mass of the disk may make it easier to damp out.
You can extend that lever arm to arbitrary length by adding a minute mirror to it and shining a laser off of it. Using sensors to pick up the laser on a wall can easily put your lever arm at 10' and you can easily add an analog range and still shrink the point its servoing to.
Tens of micrograms!!!! Very interesting indeed!
My first idea is to imagine microgravity experiments with your device. Most probably each experiment should be specifically design, to annihilate any electromagnetic noise and to reach the maximum sensitivity for the microgravity event.
Neat project. To prevent oscillation, put snubber circuit parallel to the (-)input/output (where is 100 kohm trimmer pot for sensitivity) of the first 'servo' opamp. It consist capacitor and resistor in series, then both paralel with this pot. It should gently dump all rapid changes in voltage or current. If single RC is not enough, then put one more than one RC pairs, but with various different values. For example, 1nF/1 kOhm, 10nF/10 kOhms, 100nF/100 kOhms...
Edit: This RC snubber should prevent high excursion of the 'servoing' back, while maintaining relative good feedback speed.
For air cover, make glass protective hood. Glass can be electrostatically charged too, but to much less extent than plastics of any kind.
For temperature compensation, you may use NTC resistor and some measurement electronic, just to tell you overal temperature inside measuring device, and how much needle may change length (thus leverage length, which may affect precise and consistent reading of the weight). Later, all parameters can be feed into some u-controller including compensating value depending of different device temperature.
Callibration of the temperature difference can be done measuring the same weight at different temperatures, then plot error correction values, and so on...
EDIT2: Watch out for temperature changes of the trimmer pots - they are very prone to change in resistance with the temperature (I experienced big problems at one attempt to stabilize temperature of the laser for holography porpose - trimmer pot[s] gives me bad behaviour fo the whole circuit, indicating false temerature rising or drop of the laser head, especially when room temperature changes). This small change in resistance is more evident with increased feedback amplification of the opamp(s).
EDIT3: I wonder, whether this optocoupler has oppening in form of dot, or line? If line, then it is perhaps possible to roate whole setup 90 degrees, so that shutter doesn't cut beam of light rapidly, but rather gradually.
EDIT4: Oh, well. Such opto switch may have slot, but inside is both: photodiode and LED encapsulated with focusing lens, which may produce very narrow beam of light. Maybe making your custom opto switch?
I hope this helps.
You're amazingly cool. Thank you for sharing.
I have seen this idea many times before, but few had a nice circuit and involved heavy use of a micro processor. They also lacked the nice video you have. Nicely done.
The AC gain (more stability) can be decreased by putting a capacitor between the negative feedback and the output of the first opamp (in parallel with your resistor). The time constant of the filter is roughly the feedback resistor x the cap I am suggesting you add. To add damping for stability you can add a second cap, only this one is in series with a resistor that is approximately the same as the original one used to adjust the gain (exact value is not critical). Adding both the damper and the integrator caps at the start may save allot of grief, as there may be certain combo's that are unstable in very annoying ways.
Put a diffuser over the led slit to make the detector more proportional.
The flag should be cut at an angle (45 deg.) to give you some linear range in the servo.
I did this in 1975 for a motor servo positioning system. It damped it out quite nicely. Just don't use plastic. A lot of plastics are clear to IR light.
"I hope you found that interesting."
I always do.
This design is actually originally from the amateur scientist column by Shawn Carlson in Scientific American. He credits the idea to his friends George Schmermund and Greg Schmidt. http://web.archive.org/web/20020606134612/http://www.sciam.com/2000/1000issue/1000amsci.html
You made static device, balances should be astatic. For this, sensor (optycal sensor and amplifier) should be a zero-detector which works on integrator. Integretor drives current.
+ input of photo amp should be connected to GND. What is a purpose of first potentiometer? Photodiode should work on short-circuit to be stable, but you apply some voltage. Device used should be on strings, not needles. Also I see periodic nature of noise - to stop mechanical oscillations you need differentiator working parallel to integrator (parallel AC correction).
You could adjust the gain by moving the photo interrupt further along the needle.
You could damp by an alu or copper disc between a magnet., or provide a low impedance to the meter. (just short to meter and move it , it will behave different then when left open)
A bit of averaging and star point ground should do it. Also choose an op-amp with a really low 1/f noise corner frequency.
If you want the best accuracy put the whole rig in your vacuum chamber. You'll also have to make sure all external vibrations are dampened (such as from the vacuum pumps).
Just got done watching all of your videos! Gotta love jobs with a lot of chair time.
@DavidRichfield - 2017-01-07
If you ever do microgram balance work again, I'd recommend using fine copper wire as a quick & dirty mass piece: if you know the diameter you can get the weight per unit length, so there's only one dimension to cut, and it's less susceptible to humidity than paper is.
@mahmutaksit7715 - 2022-04-22
it oxidizes