As a retired Electrical Engineer with 45 years of experience, I must say that I admire your procedures! You are absolutely correct to document faliures, that's how we learn!
Here are a couple of tips from Deuteron Technologies Ltd, Jerusalem Israel. First, you can get at least an order of magnitude higher magnetic field if you close the magnetic circuit with a laminated iron or ferrite core. Secondly, use a lock-in amplifier, or equivalent to look for your signal. In other words you should process the signal by continuously multiplying it by a number proportional to the applied current, and then use a low-pass filter to extract the flow signal. It's easy to see nanovolts of signal that way. Your biphasic waveform is good, but better if that was the current waveform, not the voltage waveform, so you should make a controllable current source rather than applying a voltage directly to a coil. This becomes much more important as you increase the inductance of the coils and you increase the frequency.
the issue here may not be electronic. it might be more practical. your pipe setup allows for air in the pipe. this will show up as noise. raise the discharge so the pipe stays full. long straight runs get a more laminar flow which reduces noise (10*diameter at least before meter). also make sure your bucket is full so pump does not vortex and suck air. I troubleshoot these occasionally, air in pipe is one of the most common issues.
Joshua Phillips thanks I was thinking about using one on my efi engine swap but I'm not spending big money on one. So are there pre-made ones that are under 600$ and are pretty accurate for things like water oils gas ? Or is it cheaper to make one?
Rotate the pipe into a vertical orientation. This ensures the pipe is full of liquid. This is a practice used in industry when installing mag flow meters.
Dear +Practical Engineering, Please use the "math" function on your 'scope to multiply channels A and B, where A is the uV signal from the electrodes and the B is the current through the field-coils. Next, "measure" the average (mean) of the product-signal (i.e. the output of the 'math' channel), normalize for the excitation-amplitude and voila, you have filtered the needle out of the haystack :) The DC (average) level that you measure should be proportional to your flow. You can even measure bidirectional flow. For best results, put channels A and B on AC-coupling, HF-reject, heck you can even do triggered averaging if you trigger off of the current (channel B)! This is my favorite version of a low-budget-synchronous-detector :)
Yep, this is the reason for using the chopped Signal in the first place, so you can synchronously detect it in York noisy signal. This basically behaves like a really small bandwidth bandpass filter, thus reducing the noise voltage extremely.
A "synchronous detector" is a subset of a "lock-in amplifier" - this particular use-case doesn't warrant a (way more expensive) lock-in amplifier, i.m.o., especially since the reference excitation-signal is readily available.
I really love this video, really shows how electrical engineering and engineering in general is riddled with difficulties and some times failure opposed to just the façade of perfection some people imagine.
The most significant advances I've made in life have often stemmed directly from initial failures. Your frank and refreshing embrace of this notion is admirable. Tellingly, your unsuccessful first attempt has led directly (through the comments as well as your own self-reflection, I should think) to the basis for a new experiment that will undoubtedly improve upon version 1.0, as well as the broader enlightenment of your audience. Kudos!
Thank you, we are currently doing an engineering school project with a homemade magnetic flow meter, and we couldn't figure out why we ended up with so much noise on our signal. But you resumed everything ! Thank you very much.
When you "spin your wheels" it helps connect the dots for me as well as it "spins" off answers to problems i cant quite grasp a way to answer the issue. This is why i am completely enjoying your channel. The parallels of how to and why are shown in a way where i can apply the therory and have found solutions through the things you explain in the method you ecplain them. I cannot thank you enough for "spinning your wheels" especially on a roof top. That one long tube from your roof top solved 4 issues that i couldnt find a way to figure out what the root cause for. Please dont stop. You keep my wheels spinning to find the answers to the issues i encounter. Thank you.
Thanks for the videos. I tried to be an electrical engineer but i didn't realize how important trig identities are. Your videos allow me to live vicariously through an engineer.
Having implemented many magnetic flow meters, two of the main issues to consider when installing the meter is to make sure there is no air in the line near the meter (maintain a full pipe at all times). If possible install the meter vertically, but if not, make sure it's not in the highest place in the piping or anywhere else where air might accumulate. Looking at your testing fixture I would assume that there is air going past the meter. I would have put some kind of trap in the end of the line before running the pipe back to the bucket. The other factor is to install in a place so many pipe diameters upstream and downstream to avoid turbulence and try for laminar flow. That part of the test rig looks ok to me.
You want the SNR to be as big as possible so you have to do to things: Increase the signal: Easiest way would probably be to increase the windings on the coil and/or the voltage you're feeding them. This increases the magnetic field and therefore the induced voltage. You could also try to switch to a fluid that is less conducting than water. Decrease the noise: Use shielding! You basically built an antenna! I would shield the set-up with a non-magnetic material like aluminium foil. You also have to use coaxial cables to the electrode in order to decrease fields feed into the cables. Hope this helps.
If you don't have coax cables, go with twisted-pair wiring. It's easy to work with, and Cat5e cabling is available at most big-box hardware stores with 4 conductor pairs inside.
The quality of these videos is crazy. I like how you pour your heart into these projects. I'm living with my parents still so i can't exactly do these kinds of experiments and understand hands on but these videos really help a budding engineer see what engineering is like in the real world. Keep the videos coming!
Great video and very thought provoking for someone like myself who has been working with electronics all my (Relatively long) life. I was too lazy to go through all 500 comments, particularly on a 2-year-old video, however here are a couple thoughts... First off, without an iron/ferrite core in your coils it’s going to seriously limit the amount of signal you get out of it, a core helps to localize and focus the field, which is why you’ll see so many coils with cores regardless of whether they’re generating or trying to receive fields, aside from very high frequency, low inductance coils. Second, you might want to check into isolated, high gain amplifiers used in EKG machines or those that try and pick up muscle currents via EMG (electromyography). Ideally you want a high-gain, differential, isolated circuit because that way you can avoid inadvertently coupling in external electrical currents that may end up causing random offsets in your signal. I would steer away from an off the shelf h-bridge simply because you are using something that was never intended for a clean, spike-free output of exactly the same signals every time, there’s just no point in its normal application. Using an op-amp with a split supply set up to produce a square wave output would probably be a better bet. And finally, you might want to look at a high-resolution ADC (I.e. at least 14 bits effective/noise free, not the resolution of the ADC itself since they are always going to have a higher resolution that’s impossible to get in practice due to design limitations) to sample your signal and try and get a little more resolution from such low-amplitude signals vs your 8-bit scope. A better ADC isn’t mandatory assuming you can get enough gain from your main amplifier, though obviously this is all a matter of trying to squeeze more signal out without it vanishing into the noise.
You sir have my thumbs up for documenting your failure. It is probably the most underrated and unappreciated part of research. Hell, all failures are frowned upon. No doubt, anyone new to the field has to tread the path of same failures (may not be all, but some? definitely!) due the lack of their documentation, leading to wastage of time and funds
I've worked 10 years as a fracker and we use 3" and 4" mag meeters in some of our rate monitoring. It's neat to see the theory on how some of the stuff around you works Shame you couldn't get it to work, but interesting concept
The logic you used to eliminate the noise issue is the same logic we use when tuning controllers. You do a step change and measure the change in the process to tune a controller. Really cool video, thanks for the video.
@@beeble2003 thanks, I misinterpreted your first comment and thought you meant the inverse by including division: high signal to noise = high noise compared to signal
I work at a power plant in Sweden and i'm in awe. It's not the simplest solution but if it works with both low and high flows then i can see it being used in a number of applications. Exampel, dosing were it can be tricky due to the low flow rate to measure the flow. BOSS! COME CHECK THIS OUT!
4:34 Signal to Noise ratio here is low. The unwanted noise is almost as high as the EMF, approaching the LOWEST S/N ratio of 1:1. A much more desirable ratio is HIGHER, for example 100:1. Hate to nit-pick but the engineer in me wants to call it out. Love your work Grady and I must say you have already taught me way more than I could ever hope to teach you. That's why I'm a subscriber to your channel
1:1 is not the lowest SNR attainable. You can decrease the signal/increase the noise further (i.e. you go into fractions). This is the point where you go from 0dB to negative dB. You can do this until you approach 0:∞ ratio (i.e. an SNR of -∞dB).
Good video, glad it got posted. This is a super-tough problem and a much better showing than I would have made. I like the demonstration of how to solve the problem.
Documenting your failure is just as important as documenting your success. Absolutely. When I was a kid, I built a small scale ion thruster. I know it worked just fine. I could feel the small amount of thrust I was producing. Unfortunately, I miscalculated the amount of power I would need to get a 1 gram thrust and, as a result, never got the thing to spin on the magnetic bearing I built, and so couldn't record any data. The project got me a solid A- in freshman physics, but I still wish I could have gotten the right power.
So I thought about this some more and was thinking about ways to improve your results. 1: You may be running into reactance limitations of your coil and or the series resistance of the wire, how-much current are you driving though the coil and what is the frequency? See if you cant drive more current though it, perhaps putting some parallel capacitance on the H bridges DC rail so you can dump more current during the 33% down time of the duty cycle. 2: You may want to use a band pass filter to dump any high frequency and low frequency noise and just isolate on your driver signals range. In reality the current that flows between the electrodes will be AC with an amplitude relative to the flow rate (as you stated). If you can filter out all the noise then you will have better results. So how I would improve this experiment: 1 - increase the inductance of your coils (ferromagnetic core), measure that inductance and calculate an ideal driver frequency. 2 - increase the current (Capacitors, larger power supply, larger cross section on the coils) 3 - use a band pass filter and isolate your driver frequency 4 - ground out the fluid before it gets to the electrodes and the coils. (help kill any AC coupling and inherent charge from the pump) 5 - use a transformer on the electrodes. your amp is amplifying voltage however the setup is inducing current, I would step up the voltage to get a better reading. Also you will need to make sure that amp has a + and - supply rail since your reading and AC signal, unless you use a capacitor and an offset voltage. I cant tell exactly what amp you are using their but I would also make sure you use a linear device such as a precision op-amp, I know a lot of the amps they sell on those breakout boards are actually class D amplifiers for audio. This may get you much better results.
I enjoy your videos. I would recommend driving the magnet coils at a frequency that is not harmonically related to the line frequency or other interference sources. Amplify your sensor signal and feed it to an analog to digital convertor (ADC) at a sampling frequency that is a much higher harmonic of your driving source. When amplifying the signal, amplify it just enough to keep it close to the full scale input of your ADC without clipping. Then take a discrete Fourier transform (DFT) of your sampled signal to filter out everything except the frequency of your driving source. That will have the effect of having a very narrow bandpass filter of your sampled signal and greatly reduce the effects of noise. I've used that technique to pull signals out of the noise in other projects. It done right, you can increase the effective number of bits of your ADC by an order of magnitude or more. Main issue I could see is keeping your driving source from radiating into your sensors directly -- so layout is probably critical.
Travis Collier it measures vibration in the pipe to determine flow. like if you had your garden hose running, you can feel vibrations in the pipe as the water flows through.
There's another method of measuring flow which uses strain gages to essentially measure the torque which the flow imparts on a U-shaped bend in a pipe. This probably won't work well for this sort of setup however, as I believe it works best in higher pressure/flow scenarios.
I noticed the googley eyes in the last scene(9:08), where the wave forms a mouth. And I thought: Googley eyes make everything better :D ... I wonder why I didn't notice them earlier in the video? (They were already attached at 3:35)
I'd really like to see how you manage to reduce noise. Couple of ideas: 1) since your pipe is pvc, ground the fluid and pipe to get rid of static charges, and most likely reduce the charge within the fluid. 2) Get some shielded wire for your signal pickup. Considering the voltage you are picking, even radio or nearby cell phone signals might be interfering. You did quite some work to make the testing rig. I'd rather take advantage of it. Great video by the way.
its signal to noise ratio in electrical engineering like he said actually ie SNR = 10log(Vsig/Vnoise), as Resistance cancel we can use log properties and the fact that P = V^2/R, to make it SNR = 20log(Psig/Pnoise) in dB. Look up signal process engineering or an embedded systems textbook for a better definition :)
A combination between positive displacement and induction coils is quite easy to do. I've seen industrial flowmeters with simple design as such. It is using an orifice plate, a rotating wheel with an iron core attached inside a pipe, and a coil to induce flux. The amount of voltage induced would translate to the velocity of flow. It is essentially a small hydro turbine and the orifice plate serves as a constant.
First I want to say great work on building and publishing this. It really showcases a lot of the tricks we use do measurements out in the world. Can you try the following to improve your psnr: - use half the transformer of an electric appliance (coil with core), with proven magnetic strength - put salt in the water - bigger pump to generate your data points Or better yet hack one flow meter and stick the scope probes in that. Thanks again for building this.
I'm definitely happy you made this video. I never really thought about how flow meters work, though it seems so obvious that it's not a trivial problem once you get that initial thought. I really enjoyed the video!
Trying to measure the flow of liquid as been living hell at work lately. We've tried turbine meters, ultra sonic meters, Coriolis meters, and now we're trying a torus wedge differential pressure meter.
i got to make one of these in my second year of electrical engineering. although i had machine wound electromagnets, and probably a better pump, my results weren't much better than yours. if it helps, i found distilled water gave more consistent results, and gold plated electrodes helped as well.
Yes, I have a suggestion to get this working without a total redesign. I suspect you have an issue with the magnetic circuit. Remember, a voltage is only induced when the magnetic flux changes, and with a biphasic pulse waveform it only changes at the step edges. So you need to increase the waveform frequency to the highest your ADC can cope with, and increase the energy stored in the magnetic circuit by adding a soft iron band around the sensing region, close to the outside of the coils. And try varying the delay between wave transition to triggering the ADC sample, I suspect a minimal delay might give best results. Hope this helps, and would love to see a follow up video if it does.
Sorry, in a former life I was designing electromagnetic flowmeters and, If sampling happens during the plateau phase of the biphasic pulse waveform everything is OK, just as it can work with a non-varying magnetic field (a permanent magnet). But the latter can cause polarization effect and consequently dc errors. Note that Faraday's law does not necessitate that both the conductor (in this case the liquid) and also the magnetic flux should change. The moving conductor as described in the law is represented here as an ever-renewing chain of small liquid rods, moving between the two electrodes, in a direction perpendicular to the unchanging magnetic flux.
A little bit of topic, but being practical is being doable to me, if I want to measure flow, I would not care however and whatever way (does not have to be by magnets), as long as it gives me the correct figure. Measuring flow is measuring the volume of liquid you would get in a unit of time. You have a bucket which you know its volume, you need a stop watch to time. Dividing the volume over the time, you have the flow rate. It is not the most elaborate way of measuring, however, that’s the idea you can work on to measure flow. The advantages of this approach is cheap and simple. The disadvantage are you have to disconnect your system to measure into a “bucket”, which sometimes are not convenient.
Great video! I always love your stuff. Though I will say in response to the minor dig at physicists in the beginning, the AC EM flow meter was patented by Alexander Kolin, who was a biophysicist.
KrazeeCain you can get basic scopes pretty cheaply, if you’re ok with connecting it to your computer. Let me know if you’re still interested in details.
Ya but is the black tape to set off the googly eyes or to hide the brand of the scope (which is often done in videos/films) but if the latter, its kinda pointless since Rigol was smart enough to include the brand on-screen.
Agreed. Failure videos are great and I'm happy to have watched this, but as a fellow engineer who knows all too well that the devil is in the details, I can't help but wonder what details are missing from this setup to make it work.
I am an engineer in the oil fields. We use a 2" mag meter on our production water pumps which can get flows up to 1300 barrels per day. (1706 gal/hr). you may just try to increase your fluid velocity with a smaller pipe.
5 років тому
Pal, I like your humble approach to science. Keep it up
Try using distilled water to clean your setup, then fresh distilled water in the measurement. Very slight changes in conductivity from chlorine in tap water will greatly affect your result. You may also want to try using a grounded steel bucket instead of plastic to limit static build-up. Very cool video though, I may have to build one!
My first point of call would be to reduce the electrical interference in the signal wiring. I don't think you ever showed it in the video, but I would start by using shielded twisted-pair cable to connect the electrodes to the differential amp. Make sure you connect the shield to the differential amp's ground, and connect the twisted pair to the two electrodes. Hopefully that's something you can try without too much work? If you want a cheapish source for shielded twisted-pair cable, you can buy STP Cat5 networking cables pretty cheap. Or you can buy audio cable, though that's usually on a larger spool.
Great video Practical Engineering. I always wonder about how it worked. I think you should do it right, remember not giving up is the hearth of engineering. Here is my contribution. Your piping is made of plastic, you will need a ground electrode near your other two electrodes, it will reduce noise. Another source of noise is turbulence, your pipe should have for the sake of precision a straight 20x pipe diameter upstream and 10x pipe diameter downstream the measure point. Vent to make sure you do not have air in your pipe or in the pump housing and impellers. I hope you read it and get good results.
The EMF generated is proportional to the conductivity of the fluid use a high conductivity fluid , also a larger diameter magnets 🧲 it’s going to deflect more charged particles to the electrodes. space between electrodes: a larger diameter pipe is going to increase the space between electrode. Also the EMF is proportional to the speed of the fluid, a Bernoulli design pipe it may help to increase the speed of the fluid. Great job with your videos ; you can improve all the mentioned variables to get a better signal. Good luck to your next version of an electromagnetic flow meter video. Regards from West Palm Beach , Florida , USA 🇺🇸.
If you really want to extract a weak signal from a high noise, rather than sample at the peaks and nulls of the magnet phases you can use a lock-in amplifier technique. Think of it as an ultranarrow bandpass filter centered around the frequency which your arduino generates. Even if you're using you tri-state pulses, you can choose a harmonic to lock-on to.
I was having a problem sourcing magnet wire locally due to a lack of the requisite copper mine. Then I remembered my house was just full of copper pipes. Problem solved! On a side note, do you have a line on a good plumber?
+Powder-phun If that's actually a "thing" I'm moving to Mars. Wouldn't surprise me in the slightest though, with people buying gluten free water and all
You are awesome. I watch you videos and learn many things. pls continue doing reasearch on it. I know you can do it. thanks for sharing this hard earned knowledge to all for free.
A hint from an EE: if you have to extract data from a lot of noise, and you have the excitation signal available, you can use a lock-in amplifier to do so. You can surely find all the details searching the Internet, but the principle is to extract the correlation between the stimulus and the output, so to increase the SNR; this can be done with a simple analog multiplication between the input and the output. Once you have done that, you'll be able too heavily low-pass the result of the multiplication and get a nice clean signal. The multiplication is especially easy to do with your stimulus as you simply have to invert the signals coming from the readout twice per period. One question, though: why did you use (+1, 0, -1, 0) instead of a simple (+1, -1) excitation?
Great video! I like it how you go into details of calculating the flow rate at the end. One of the characteristics for a magnetic flow meter is that it works the best when the fluid is conductive. Water on the other hand might not work as well.
@1:10 - "...and the challenges that arise when the real world doesn't quite match the theory". Brah, welcome to Electrical Engineering 101! The most precise imprecise engineering major of them all haha. In all seriousness, love the videos! EE here with a closeted passion for ME and civil engineering stuff. Keep it up!
The air core coils you're using may be a source of some of the problems. Try using a soft iron laminated core. Air cores are good when you need a homogeneous field (like in an NMR) but require huge amounts of current to get appreciable field strength (Which is why NMR machines typically have superconducting magnets). A soft iron core will help concentrate the field into the fluid. One caveat is that the iron core will have a higher inductance, which could damage your H-bridge if switched off too quickly. A capacitor in parallel with the coil will suppress that, but will slow down how fast you can switch between poles. Keep in mind you'll want to use a non-polarized capacitor or diode isolated electrolytics. I hope this helps
Thank you for the interesting and well made videos. I love how your channel shows the process that leads to a solution as well as the knowledge that is used on the way.
Ultrasonic transducers are a great way to measure flow. Phase change is relative to flow rate and changes in amplitude is related to changes in the fluid density or entrapped gases etc.
Hey, a little late, but just as a suggestion if you still have the system set up. You could spike the water with a ton of salt. That should make it far more conductive and boost the resultant signal in a more or less linear fashion according to the salt density.
Here is a data processing suggestion. Instead of a bi-phasic control signal put into the electro-magnets, you could use a pseudo-noise or PN code signal instead. The benefit here is that when you sample the output you can run the data through a matched-filter to correlate the sampled data with the known PN code, giving you a high SNR because the noise does not correlate. You can experiment with long codes and get better results of the actual flow. The down-side is that longer codes provide a slower flow reading.
Thanks for posting, even if it didn't turn out the way you hoped. This will be helpful for me if I ever get around to automating an aquaponic setup digitally. Btw, I believe the S/N ratio is low, not high. Thanks again!
The noise may come from the AC line built-in the building or the pump. To overcome the common mode noise which is far larger than the signal you want, your circuit need to add high CMRR amplifier. Instrumentation amplifier is widely used in such noisy application. There are many models of Instrumentation amplifier IC can be found. For example, TI's INA848 with CMRR>132 dB. I also found the input wires connected to the PCB can be twisted then shielded to avoid the loop magnetic induced noise. Twist the input wires of coils could be helpful also.
As a retired Electrical Engineer with 45 years of experience, I must say that I admire your procedures! You are absolutely correct to document faliures, that's how we learn!
Here are a couple of tips from Deuteron Technologies Ltd, Jerusalem Israel. First, you can get at least an order of magnitude higher magnetic field if you close the magnetic circuit with a laminated iron or ferrite core. Secondly, use a lock-in amplifier, or equivalent to look for your signal. In other words you should process the signal by continuously multiplying it by a number proportional to the applied current, and then use a low-pass filter to extract the flow signal. It's easy to see nanovolts of signal that way. Your biphasic waveform is good, but better if that was the current waveform, not the voltage waveform, so you should make a controllable current source rather than applying a voltage directly to a coil. This becomes much more important as you increase the inductance of the coils and you increase the frequency.
Very helpful info. Thanks for sharing!
Wow, that's great info.
Uhmmm what? :p
All i heard was warp core, the borg, and engage.
Magnetohydrodynamics
I absolutely love the way you put googly eyes on everything. never stop.
the issue here may not be electronic. it might be more practical. your pipe setup allows for air in the pipe. this will show up as noise. raise the discharge so the pipe stays full. long straight runs get a more laminar flow which reduces noise (10*diameter at least before meter). also make sure your bucket is full so pump does not vortex and suck air. I troubleshoot these occasionally, air in pipe is one of the most common issues.
Exactly, I would also suggest looking at some installation and operation manuals for industrial mag-flow meters for more clarification
Joshua Phillips thanks I was thinking about using one on my efi engine swap but I'm not spending big money on one. So are there pre-made ones that are under 600$ and are pretty accurate for things like water oils gas ? Or is it cheaper to make one?
Try mounting it vertically to eliminate the air, and ensure a liquid full pipe.
@@trb707 I was going to say the same thing.
I think you should try again and try to avoid some of the problems people suggested. Your failures only matter when you use them as a growing point.
Rotate the pipe into a vertical orientation. This ensures the pipe is full of liquid. This is a practice used in industry when installing mag flow meters.
Which industry? Working with mag flow meters in the water industry; they're all horizontal...
I haven't seen vertical mag in my life
Dear +Practical Engineering,
Please use the "math" function on your 'scope to multiply channels A and B, where A is the uV signal from the electrodes and the B is the current through the field-coils. Next, "measure" the average (mean) of the product-signal (i.e. the output of the 'math' channel), normalize for the excitation-amplitude and voila, you have filtered the needle out of the haystack :) The DC (average) level that you measure should be proportional to your flow. You can even measure bidirectional flow. For best results, put channels A and B on AC-coupling, HF-reject, heck you can even do triggered averaging if you trigger off of the current (channel B)! This is my favorite version of a low-budget-synchronous-detector :)
Yep, this is the reason for using the chopped Signal in the first place, so you can synchronously detect it in York noisy signal. This basically behaves like a really small bandwidth bandpass filter, thus reducing the noise voltage extremely.
same function as a Lock-in Amplifier....
A "synchronous detector" is a subset of a "lock-in amplifier" - this particular use-case doesn't warrant a (way more expensive) lock-in amplifier, i.m.o., especially since the reference excitation-signal is readily available.
hello aditya. I am looking to develop magnetic flowmeters here in india. would you be able to offer any help?
@@ATLANTECHFLOWMETERS Sure. PM: firstname.lastname (a) gmail
I really love this video, really shows how electrical engineering and engineering in general is riddled with difficulties and some times failure opposed to just the façade of perfection some people imagine.
Thanks!
The most significant advances I've made in life have often stemmed directly from initial failures. Your frank and refreshing embrace of this notion is admirable. Tellingly, your unsuccessful first attempt has led directly (through the comments as well as your own self-reflection, I should think) to the basis for a new experiment that will undoubtedly improve upon version 1.0, as well as the broader enlightenment of your audience. Kudos!
It is definitely important to document well researched failures. Extra kudos for this and all your hard work man, love the channel.
Thank you, we are currently doing an engineering school project with a homemade magnetic flow meter, and we couldn't figure out why we ended up with so much noise on our signal. But you resumed everything ! Thank you very much.
When you "spin your wheels" it helps connect the dots for me as well as it "spins" off answers to problems i cant quite grasp a way to answer the issue. This is why i am completely enjoying your channel. The parallels of how to and why are shown in a way where i can apply the therory and have found solutions through the things you explain in the method you ecplain them. I cannot thank you enough for "spinning your wheels" especially on a roof top. That one long tube from your roof top solved 4 issues that i couldnt find a way to figure out what the root cause for. Please dont stop. You keep my wheels spinning to find the answers to the issues i encounter. Thank you.
I really like the fact you still shared the results even with failed practical part. Failure IS an option, if you have data as a result.
Thanks for the videos. I tried to be an electrical engineer but i didn't realize how important trig identities are. Your videos allow me to live vicariously through an engineer.
Man, I love this channel. Some of the highest quality content on UA-cam!
Thanks. This is really encouraging to hear.
I concur! Excellent topic and treatment of the subject. Just to throw in my 2 cents: I'd love to see a followup if you do tweak the design.
I totally agree! Great video Grady!
You are exceptionally genius person. And there is a need of such teachers or mentor in our society. Respect for u sir
Having implemented many magnetic flow meters, two of the main issues to consider when installing the meter is to make sure there is no air in the line near the meter (maintain a full pipe at all times). If possible install the meter vertically, but if not, make sure it's not in the highest place in the piping or anywhere else where air might accumulate. Looking at your testing fixture I would assume that there is air going past the meter. I would have put some kind of trap in the end of the line before running the pipe back to the bucket. The other factor is to install in a place so many pipe diameters upstream and downstream to avoid turbulence and try for laminar flow. That part of the test rig looks ok to me.
You want the SNR to be as big as possible so you have to do to things:
Increase the signal: Easiest way would probably be to increase the windings on the coil and/or the voltage you're feeding them. This increases the magnetic field and therefore the induced voltage. You could also try to switch to a fluid that is less conducting than water.
Decrease the noise: Use shielding! You basically built an antenna! I would shield the set-up with a non-magnetic material like aluminium foil. You also have to use coaxial cables to the electrode in order to decrease fields feed into the cables.
Hope this helps.
If you don't have coax cables, go with twisted-pair wiring. It's easy to work with, and Cat5e cabling is available at most big-box hardware stores with 4 conductor pairs inside.
Can you help me with electromagnetic flowmeter's transmitter firmware and hardware design?
Yes man, documenting failures and errors in a neat and descriptive manner is by any mean not less important as documenting a success! Thank you!
The quality of these videos is crazy. I like how you pour your heart into these projects. I'm living with my parents still so i can't exactly do these kinds of experiments and understand hands on but these videos really help a budding engineer see what engineering is like in the real world. Keep the videos coming!
I thank you for showing the challenges we face when doing testing and research. Nothing is ever cut and dry simple testing or building.
I love the point you made about documenting failure. We don't get nearly enough of that in STEM education.
Great video and very thought provoking for someone like myself who has been working with electronics all my (Relatively long) life. I was too lazy to go through all 500 comments, particularly on a 2-year-old video, however here are a couple thoughts... First off, without an iron/ferrite core in your coils it’s going to seriously limit the amount of signal you get out of it, a core helps to localize and focus the field, which is why you’ll see so many coils with cores regardless of whether they’re generating or trying to receive fields, aside from very high frequency, low inductance coils. Second, you might want to check into isolated, high gain amplifiers used in EKG machines or those that try and pick up muscle currents via EMG (electromyography). Ideally you want a high-gain, differential, isolated circuit because that way you can avoid inadvertently coupling in external electrical currents that may end up causing random offsets in your signal. I would steer away from an off the shelf h-bridge simply because you are using something that was never intended for a clean, spike-free output of exactly the same signals every time, there’s just no point in its normal application. Using an op-amp with a split supply set up to produce a square wave output would probably be a better bet. And finally, you might want to look at a high-resolution ADC (I.e. at least 14 bits effective/noise free, not the resolution of the ADC itself since they are always going to have a higher resolution that’s impossible to get in practice due to design limitations) to sample your signal and try and get a little more resolution from such low-amplitude signals vs your 8-bit scope. A better ADC isn’t mandatory assuming you can get enough gain from your main amplifier, though obviously this is all a matter of trying to squeeze more signal out without it vanishing into the noise.
That was informative.
What a trip down memory lane, in the early 80's I used to work for a company that made mag meters. Thanks.
You sir have my thumbs up for documenting your failure. It is probably the most underrated and unappreciated part of research. Hell, all failures are frowned upon. No doubt, anyone new to the field has to tread the path of same failures (may not be all, but some? definitely!) due the lack of their documentation, leading to wastage of time and funds
Always wondered how digital flow meters work, even without refinement you effectively demonstrated both theory and practice.
Such an underrated channel, the quality of these videos is phenomenal
I've worked 10 years as a fracker and we use 3" and 4" mag meeters in some of our rate monitoring. It's neat to see the theory on how some of the stuff around you works
Shame you couldn't get it to work, but interesting concept
You engineers are absolute wizards..!
The logic you used to eliminate the noise issue is the same logic we use when tuning controllers. You do a step change and measure the change in the process to tune a controller. Really cool video, thanks for the video.
Nice Video, but a small error 4:33 the SNR is low not high, no worries hope no one got confuesd
didn't sound right as SNR is in decibels and usually never see values
@@stonerRx Doesn't sound right because "signal to noise ratio" means "signal divided by noise". High ratio = high signal.
i'm still confused, it should be noise to signal imo but naming conventions have aways been weird
@@sdaniaal It makes sense to use signal-to-noise, because it's the signal you're interested in. High signal-to-noise = high quality, etc.
@@beeble2003 thanks, I misinterpreted your first comment and thought you meant the inverse by including division: high signal to noise = high noise compared to signal
Shameful or not I am liking this video for the effort involved and explaining the concept elegantly.
@4:35 -- the signal to noise ratio would be low. The noise to signal ratio would be high.
I thought so too, I'm glad to see another comment about that discrepancy.
Yeah, that bugged me to.
Man! I'll add an annotation about this mistake. Good catch.
Correct.
Every time I watch your video, I learn something new. Thanks for being with us.
I work at a power plant in Sweden and i'm in awe. It's not the simplest solution but if it works with both low and high flows then i can see it being used in a number of applications. Exampel, dosing were it can be tricky due to the low flow rate to measure the flow. BOSS! COME CHECK THIS OUT!
4:34 Signal to Noise ratio here is low. The unwanted noise is almost as high as the EMF, approaching the LOWEST S/N ratio of 1:1. A much more desirable ratio is HIGHER, for example 100:1. Hate to nit-pick but the engineer in me wants to call it out. Love your work Grady and I must say you have already taught me way more than I could ever hope to teach you. That's why I'm a subscriber to your channel
1:1 is not the lowest SNR attainable. You can decrease the signal/increase the noise further (i.e. you go into fractions). This is the point where you go from 0dB to negative dB. You can do this until you approach 0:∞ ratio (i.e. an SNR of -∞dB).
Good video, glad it got posted. This is a super-tough problem and a much better showing than I would have made. I like the demonstration of how to solve the problem.
Do you have experience with mag meters ?
Best demonstration which i have ever seen
Great demo despite the less-than-usable result. ! I'm glad to see you used only artisan hand-wound coils from locally-sourced wire.
You "failure" saved me from spending a ton of time trying to DIY this myself, so thanks! I should just spend the money and buy an off the shelf one.
Documenting your failure is just as important as documenting your success. Absolutely. When I was a kid, I built a small scale ion thruster. I know it worked just fine. I could feel the small amount of thrust I was producing. Unfortunately, I miscalculated the amount of power I would need to get a 1 gram thrust and, as a result, never got the thing to spin on the magnetic bearing I built, and so couldn't record any data. The project got me a solid A- in freshman physics, but I still wish I could have gotten the right power.
Your videos are awesome buddy I tried engineering for 3 years before switching to business but I think I want to get back into it
So I thought about this some more and was thinking about ways to improve your results.
1: You may be running into reactance limitations of your coil and or the series resistance of the wire, how-much current are you driving though the coil and what is the frequency? See if you cant drive more current though it, perhaps putting some parallel capacitance on the H bridges DC rail so you can dump more current during the 33% down time of the duty cycle.
2: You may want to use a band pass filter to dump any high frequency and low frequency noise and just isolate on your driver signals range. In reality the current that flows between the electrodes will be AC with an amplitude relative to the flow rate (as you stated). If you can filter out all the noise then you will have better results.
So how I would improve this experiment:
1 - increase the inductance of your coils (ferromagnetic core), measure that inductance and calculate an ideal driver frequency.
2 - increase the current (Capacitors, larger power supply, larger cross section on the coils)
3 - use a band pass filter and isolate your driver frequency
4 - ground out the fluid before it gets to the electrodes and the coils. (help kill any AC coupling and inherent charge from the pump)
5 - use a transformer on the electrodes. your amp is amplifying voltage however the setup is inducing current, I would step up the voltage to get a better reading. Also you will need to make sure that amp has a + and - supply rail since your reading and AC signal, unless you use a capacitor and an offset voltage.
I cant tell exactly what amp you are using their but I would also make sure you use a linear device such as a precision op-amp, I know a lot of the amps they sell on those breakout boards are actually class D amplifiers for audio.
This may get you much better results.
Love how you are always on the edge on your comfort-zone!
I enjoy your videos. I would recommend driving the magnet coils at a frequency that is not harmonically related to the line frequency or other interference sources. Amplify your sensor signal and feed it to an analog to digital convertor (ADC) at a sampling frequency that is a much higher harmonic of your driving source. When amplifying the signal, amplify it just enough to keep it close to the full scale input of your ADC without clipping. Then take a discrete Fourier transform (DFT) of your sampled signal to filter out everything except the frequency of your driving source. That will have the effect of having a very narrow bandpass filter of your sampled signal and greatly reduce the effects of noise. I've used that technique to pull signals out of the noise in other projects. It done right, you can increase the effective number of bits of your ADC by an order of magnitude or more. Main issue I could see is keeping your driving source from radiating into your sensors directly -- so layout is probably critical.
Cool video. this reminded me of a company near where I live that makes flow meters which "listen" to the pipe in order to determine flow rate.
Travis Collier it measures vibration in the pipe to determine flow. like if you had your garden hose running, you can feel vibrations in the pipe as the water flows through.
Travis Collier you're welcome, and thanks for letting me know about acoustic anemometers.
There's another method of measuring flow which uses strain gages to essentially measure the torque which the flow imparts on a U-shaped bend in a pipe. This probably won't work well for this sort of setup however, as I believe it works best in higher pressure/flow scenarios.
Outstanding job explaining the theory and practice, along with the challenges you encountered. Very useful material.
i love how you covered the rigol and specs on the scope yet on the screen it says rigol at plain sight
he probably only covered it for the googley eyes lol
I noticed the googley eyes in the last scene(9:08), where the wave forms a mouth. And I thought: Googley eyes make everything better :D ... I wonder why I didn't notice them earlier in the video? (They were already attached at 3:35)
"Artisan electro-magnets hand wound with locally-sourced magnet wire" is what got the thumbs up. That's gold, Jerry, GOLD!
No, no, copper. ;)
top men are working on these artisan magnets
+
had he added a core to the magnets he would have had a much more powerful field :p
Who?
TOP. MEN.
But are they Antibiotic-Free²-Range Magnets?
I'd really like to see how you manage to reduce noise. Couple of ideas: 1) since your pipe is pvc, ground the fluid and pipe to get rid of static charges, and most likely reduce the charge within the fluid. 2) Get some shielded wire for your signal pickup. Considering the voltage you are picking, even radio or nearby cell phone signals might be interfering. You did quite some work to make the testing rig. I'd rather take advantage of it. Great video by the way.
I love that you put googly eyes on EVERYTHING.
Noise to signal is high, not signal to noise.
its signal to noise ratio in electrical engineering like he said actually ie SNR = 10log(Vsig/Vnoise), as Resistance cancel we can use log properties and the fact that P = V^2/R, to make it SNR = 20log(Psig/Pnoise) in dB. Look up signal process engineering or an embedded systems textbook for a better definition :)
You're right : Showing it even if its a fail is great !
A combination between positive displacement and induction coils is quite easy to do. I've seen industrial flowmeters with simple design as such. It is using an orifice plate, a rotating wheel with an iron core attached inside a pipe, and a coil to induce flux. The amount of voltage induced would translate to the velocity of flow. It is essentially a small hydro turbine and the orifice plate serves as a constant.
This channel is partially the reason I'm pursuing engineering. Starting first year in September. Any tips from anyone here?
No worries about the fails. I always wondered how these meters worked and thanks to your informative video... I now do. Thanks for a super channel!
I’m a hydraulic fitter and an electronics enthusiast, I would love to see you solve this as I would love to know how to build a cheap flow meter
"Documenting your failure is just as important as documenting your success" -true
Thanks for another great video! One suggestion I can think of to reduce the noise would be to use a water tank on your roof and eliminate the pump.
First I want to say great work on building and publishing this.
It really showcases a lot of the tricks we use do measurements out in the world.
Can you try the following to improve your psnr:
- use half the transformer of an electric appliance (coil with core), with proven magnetic strength
- put salt in the water
- bigger pump to generate your data points
Or better yet hack one flow meter and stick the scope probes in that.
Thanks again for building this.
I'm definitely happy you made this video. I never really thought about how flow meters work, though it seems so obvious that it's not a trivial problem once you get that initial thought. I really enjoyed the video!
Trying to measure the flow of liquid as been living hell at work lately. We've tried turbine meters, ultra sonic meters, Coriolis meters, and now we're trying a torus wedge differential pressure meter.
Very cool. There are so many interesting ways to do it. I tried to pick one that would be easy to demonstrate, but I was wrong!
i got to make one of these in my second year of electrical engineering. although i had machine wound electromagnets, and probably a better pump, my results weren't much better than yours. if it helps, i found distilled water gave more consistent results, and gold plated electrodes helped as well.
Practical (real) Engineering is just like that. Best luck next time! Love your videos!
Yes, I have a suggestion to get this working without a total redesign. I suspect you have an issue with the magnetic circuit. Remember, a voltage is only induced when the magnetic flux changes, and with a biphasic pulse waveform it only changes at the step edges. So you need to increase the waveform frequency to the highest your ADC can cope with, and increase the energy stored in the magnetic circuit by adding a soft iron band around the sensing region, close to the outside of the coils. And try varying the delay between wave transition to triggering the ADC sample, I suspect a minimal delay might give best results. Hope this helps, and would love to see a follow up video if it does.
Sorry, in a former life I was designing electromagnetic flowmeters and, If sampling happens during the plateau phase of the biphasic pulse waveform everything is OK, just as it can work with a non-varying magnetic field (a permanent magnet). But the latter can cause polarization effect and consequently dc errors. Note that Faraday's law does not necessitate that both the conductor (in this case the liquid) and also the magnetic flux should change. The moving conductor as described in the law is represented here as an ever-renewing chain of small liquid rods, moving between the two electrodes, in a direction perpendicular to the unchanging magnetic flux.
A little bit of topic, but being practical is being doable to me, if I want to measure flow, I would not care however and whatever way (does not have to be by magnets), as long as it gives me the correct figure. Measuring flow is measuring the volume of liquid you would get in a unit of time. You have a bucket which you know its volume, you need a stop watch to time. Dividing the volume over the time, you have the flow rate. It is not the most elaborate way of measuring, however, that’s the idea you can work on to measure flow. The advantages of this approach is cheap and simple. The disadvantage are you have to disconnect your system to measure into a “bucket”, which sometimes are not convenient.
Great video! I always love your stuff. Though I will say in response to the minor dig at physicists in the beginning, the AC EM flow meter was patented by Alexander Kolin, who was a biophysicist.
Great video. And thank you for sharing your "failures" too. I know I learn more from mine than from the successes!
You should make a video with Destin (SmarterEveryDay)!
This brings back memory’s of my MIT days. The hours spent making artisan coils for electromagnets.
If I ever get my hands on an oscilloscope, I'm totally putting googly eyes on it too.
KrazeeCain you can get basic scopes pretty cheaply, if you’re ok with connecting it to your computer. Let me know if you’re still interested in details.
@@uzaiyaro I'm interested.
Ya but is the black tape to set off the googly eyes or to hide the brand of the scope (which is often done in videos/films) but if the latter, its kinda pointless since Rigol was smart enough to include the brand on-screen.
Would love to see a redesign video of this. Success videos make failure videos all the more relevant.
Agreed. Failure videos are great and I'm happy to have watched this, but as a fellow engineer who knows all too well that the devil is in the details, I can't help but wonder what details are missing from this setup to make it work.
+Ben Burns Yes!
I am an engineer in the oil fields. We use a 2" mag meter on our production water pumps which can get flows up to 1300 barrels per day. (1706 gal/hr). you may just try to increase your fluid velocity with a smaller pipe.
Pal, I like your humble approach to science. Keep it up
As an Electrical Engineering Student, "Electrical Engineering is hard" is an understatement!
That intro sounded like a PBS or NPR show. But, it's good to see you getting the recognition from the people with money that you're worth supporting.
Try using distilled water to clean your setup, then fresh distilled water in the measurement. Very slight changes in conductivity from chlorine in tap water will greatly affect your result.
You may also want to try using a grounded steel bucket instead of plastic to limit static build-up.
Very cool video though, I may have to build one!
Awesome video! Many concepts here. Not a failure at all.
My first point of call would be to reduce the electrical interference in the signal wiring. I don't think you ever showed it in the video, but I would start by using shielded twisted-pair cable to connect the electrodes to the differential amp. Make sure you connect the shield to the differential amp's ground, and connect the twisted pair to the two electrodes. Hopefully that's something you can try without too much work? If you want a cheapish source for shielded twisted-pair cable, you can buy STP Cat5 networking cables pretty cheap. Or you can buy audio cable, though that's usually on a larger spool.
Quentin Smith buy a cheap microphone cable for $5 or so on EBay. Cut off the connectors. Voila.
Great video Practical Engineering. I always wonder about how it worked. I think you should do it right, remember not giving up is the hearth of engineering. Here is my contribution. Your piping is made of plastic, you will need a ground electrode near your other two electrodes, it will reduce noise. Another source of noise is turbulence, your pipe should have for the sake of precision a straight 20x pipe diameter upstream and 10x pipe diameter downstream the measure point. Vent to make sure you do not have air in your pipe or in the pump housing and impellers. I hope you read it and get good results.
That's pretty much verbatim from the Badger M2000 mag-flow meter manual I'm familiar with, and as such is sound advice.
The EMF generated is proportional to the conductivity of the fluid use a high conductivity fluid , also a larger diameter magnets 🧲 it’s going to deflect more charged particles to the electrodes.
space between electrodes: a larger diameter pipe is going to increase the space between electrode. Also the EMF is proportional to the speed of the fluid, a Bernoulli design pipe it may help to increase the speed of the fluid. Great job with your videos ; you can improve all the mentioned variables to get a better signal. Good luck to your next version of an electromagnetic flow meter video.
Regards from West Palm Beach , Florida , USA 🇺🇸.
If you really want to extract a weak signal from a high noise, rather than sample at the peaks and nulls of the magnet phases you can use a lock-in amplifier technique. Think of it as an ultranarrow bandpass filter centered around the frequency which your arduino generates. Even if you're using you tri-state pulses, you can choose a harmonic to lock-on to.
I was having a problem sourcing magnet wire locally due to a lack of the requisite copper mine. Then I remembered my house was just full of copper pipes. Problem solved!
On a side note, do you have a line on a good plumber?
Be sure to get a locally sourced, non-gmo, organic plumber
I suggest using gluten-free PVC piping.
+Powder-phun If that's actually a "thing" I'm moving to Mars. Wouldn't surprise me in the slightest though, with people buying gluten free water and all
LeiserGeist If it isn't a thing it should be. After all, why should your house's bowels have to suffer from gluten?
+LeiserGeist Bruh get out of my life
You are awesome. I watch you videos and learn many things. pls continue doing reasearch on it. I know you can do it. thanks for sharing this hard earned knowledge to all for free.
Very good video! Even though your demo didn't quite work, your explanation of the theory was very good!
Add an active low pass filter, if it still noisy, make a simple lock in amplifier
A hint from an EE: if you have to extract data from a lot of noise, and you have the excitation signal available, you can use a lock-in amplifier to do so.
You can surely find all the details searching the Internet, but the principle is to extract the correlation between the stimulus and the output, so to increase the SNR; this can be done with a simple analog multiplication between the input and the output.
Once you have done that, you'll be able too heavily low-pass the result of the multiplication and get a nice clean signal.
The multiplication is especially easy to do with your stimulus as you simply have to invert the signals coming from the readout twice per period.
One question, though: why did you use (+1, 0, -1, 0) instead of a simple (+1, -1) excitation?
Great video! I like it how you go into details of calculating the flow rate at the end.
One of the characteristics for a magnetic flow meter is that it works the best when the fluid is conductive. Water on the other hand might not work as well.
That would be solved by a spoon of table salt I think. Tap water in europe is around 0.05 S/m.
Amazing video!
I worked with some magnetic flow meters once and was really curious to know how they work!!!
another fantastic video don't worry about failure no one's perfect. keep it up
@1:10 - "...and the challenges that arise when the real world doesn't quite match the theory". Brah, welcome to Electrical Engineering 101! The most precise imprecise engineering major of them all haha. In all seriousness, love the videos! EE here with a closeted passion for ME and civil engineering stuff. Keep it up!
The air core coils you're using may be a source of some of the problems. Try using a soft iron laminated core. Air cores are good when you need a homogeneous field (like in an NMR) but require huge amounts of current to get appreciable field strength (Which is why NMR machines typically have superconducting magnets). A soft iron core will help concentrate the field into the fluid.
One caveat is that the iron core will have a higher inductance, which could damage your H-bridge if switched off too quickly. A capacitor in parallel with the coil will suppress that, but will slow down how fast you can switch between poles. Keep in mind you'll want to use a non-polarized capacitor or diode isolated electrolytics.
I hope this helps
Thank you for the interesting and well made videos. I love how your channel shows the process that leads to a solution as well as the knowledge that is used on the way.
a very good demo. I would suggest looking at op-amps
Another great video. Congrats on the sponsorship! I like the music you used for your into and outtro. You should keep using it!
Ultrasonic transducers are a great way to measure flow. Phase change is relative to flow rate and changes in amplitude is related to changes in the fluid density or entrapped gases etc.
Hey, a little late, but just as a suggestion if you still have the system set up.
You could spike the water with a ton of salt. That should make it far more conductive and boost the resultant signal in a more or less linear fashion according to the salt density.
Xylos144 sometimes this can damage the pump, depeneding on the type he is using
Here is a data processing suggestion. Instead of a bi-phasic control signal put into the electro-magnets, you could use a pseudo-noise or PN code signal instead. The benefit here is that when you sample the output you can run the data through a matched-filter to correlate the sampled data with the known PN code, giving you a high SNR because the noise does not correlate. You can experiment with long codes and get better results of the actual flow. The down-side is that longer codes provide a slower flow reading.
Thanks for posting, even if it didn't turn out the way you hoped. This will be helpful for me if I ever get around to automating an aquaponic setup digitally.
Btw, I believe the S/N ratio is low, not high.
Thanks again!
The noise may come from the AC line built-in the building or the pump. To overcome the common mode noise which is far larger than the signal you want, your circuit need to add high CMRR amplifier. Instrumentation amplifier is widely used in such noisy application. There are many models of Instrumentation amplifier IC can be found. For example, TI's INA848 with CMRR>132 dB. I also found the input wires connected to the PCB can be twisted then shielded to avoid the loop magnetic induced noise. Twist the input wires of coils could be helpful also.