1
\$\begingroup\$

I am looking for solutions to detect the rotational speed and direction of a permanent magnet rotor. I thought this would be easy, but every concept I considered falls short somehow...

Does an off-the-shelf solution exist (there should be?!) that solves this?

I need something with a consumption in the uW range, the ability to count up to 200 Hz and <2 cm scale.

I imagined that this would be common enough that there should exist an IC with only plus, minus, dir and freq pins, or similar, but haven't found anything.

If you don't have tips for something ready made, what concepts would you recommend? I looked at MRT or hall sensors first, but could only find some that are low power and some that are fast enough, none that are both.
I have also tried using analog signals from sensor coils, but I experience some (possibly solvable) signal processing issues due to placement limits.
I know that I looked at reed switches too, but can't recall why I dropped them. Perhaps lifetime concerns (I expect 10^11 cycles in some cases).

update: The helicopter perspective of the more specific problem I am actually trying to solve is that I have a 3 cm diam hydro turbine in a much bigger pipe of clean water with flows of 0.2-0.3 m/s and I want to use the turbine to charge a battery and generate a signal that measures the rotation, which is fed to an MCU, powered by the battery. I generate about a 1mW and need every last bit of power I can get for the other tasks on the MCU. The turbine is supposed to sit there for a very long time, sometimes with 2m gravel on top of it. The turbine has two coils that form a single phase, no other arrangement is possible due to space limits. I can fit additional, small, coils or sensors at limited positions around the rotor.

\$\endgroup\$
15
  • 1
    \$\begingroup\$ If you were to generate power from the rotor would this ease the uW power budget? \$\endgroup\$ Commented Apr 22 at 10:08
  • \$\begingroup\$ Two Hall sensors and an MCU would do it, but not within uW budget. \$\endgroup\$ Commented Apr 22 at 11:31
  • \$\begingroup\$ @Frog it would not. I already do and the power level is sub mW and I need it elsewhere. I was hoping to integrate with the rotor magnetically instead of with the stator electrically. \$\endgroup\$ Commented Apr 28 at 6:35
  • 1
    \$\begingroup\$ Now redundant (probably). The question as is is wasting your time and ours :-(. |||| If you only have one phase, is it a brush motor or ...? || I suspect you could tell us quite a lot more that may help us t o help you. So far there is a lot of magic / interpolation required to understand what you are doing. || uW range is too broad. What is your upper and desirable uW range? I am operating an LED usefully on 400 uW and you could certainly operate one well below that with enough power for an optical sensor. || If brushed, are the brushes at 180 degrees electrical or ...? \$\endgroup\$ Commented Apr 28 at 12:49
  • 1
    \$\begingroup\$ Why not use an off-the-shelf ultrasonic flowmeter? Are you looking to make a self-contained totalizer? For the turbine solution - what is the maximum pressure drop you are allowed? A small turbine in a free-flowing pipe is going to be fairly inefficient. Have you ducted the turbine? How long is the duct? Try to speed up the flow through the turbine. It doesn't have to be the same as the flow speed in the pipe. You're dealing with a fluid dynamics problem, so solve that first and you should have easily one order of magnitude more power at your disposal. \$\endgroup\$ Commented Apr 30 at 10:12

4 Answers 4

Reset to default
1
\$\begingroup\$

This MAY do what you want.
You haven't provided enough information on how the rotor is powered and controlled.

I "devised" this circuit 20+ years ago to measure the speed of a 4 phase person powered exercise bike alternator. About the simplest longtailed pair imaginable.
Inputs were from two alternator phase windings.
The alternator was resistively loaded with ~ 25 kHz variable duty cycle PWM applied to bridge rectified DC - so the input signal was an utter mess.
Despite that this circuit cleanly separated the speed signal.
Filter capacitors at th bases may help.

If I saw this circuit I wouldn't believe it . You don't have to - try it and see :-)

Resistor values are semi arbitrary. You could experiment with higher values. MOSFETS may work - but note that the bipolar transistors self clamp the base voltages whereas FETs will not. Input was 12VAC - 200 VAC range (usually say 12-50VAC or so)

enter image description here

\$\endgroup\$
2
  • \$\begingroup\$ Thank you! I was hoping to couple with the existing device magnetically, which is why I didn't provide more details. I only have one phase, so I will never be able to get direction from the stator without adding something. \$\endgroup\$ Commented Apr 28 at 6:26
  • \$\begingroup\$ Even if I were to add more to the stator I don't expect to induce more than some 10mV in any additional coil that I could physically fit. Perhaps I could use amplifiers, but that eats into the power budget and it doesn't seem like the correct solution. Hmm.... Thank you for your efforts so far though! \$\endgroup\$ Commented Apr 29 at 8:20
1
\$\begingroup\$

I just looked at a datasheet for a Hall effect sensor, it says it turns on in 32 µs for a supply current of 12 mA. So you could turn it on at 400 Hz giving an average current of 153 µA. Maybe some of the pure "analog" sensors may have a faster turnon time, and it may be possible to supply them from a lower pulsed supply voltage to reduce power.

Even the reed switch solution may turn out to draw similar current if you want to track 200 Hz - current flow into debounce capacitors through pullup resistors may require lowish pullup resistor values. Plus those capacitors are going to gently burn the reed switches as they are shorted into the switch every cycle.

\$\endgroup\$
1
  • \$\begingroup\$ Thank you for the suggestion. As i stated in a comment above the generator generates sub mW power and I need all I can get to charge my battery, so 150uA per sensor is way too much. I would also need power and components to generate the enable signal, but perhaps that's not too much. Your comments about similar consumption for a reed solution are probably correct and appreciated. \$\endgroup\$ Commented Apr 30 at 9:36
0
\$\begingroup\$

A simple solution would be to use two reed switches and a flip-flop circuit. All quite simple really and power consumption will be a few uA. Max freq will be limited by the speed of the switches and that seems to be sub ms for most models.

I still haven't found any reason reed switches would not work and looking around a bit has somewhat settled my unease with the lifetime cycles.

The switches are spaced close together such that they are both affected by one magnet at a time. The output of the first switch is simply a pulse signal whose frequency represents the magnetic frequency. Plain as is.

As for direction... Each switch 1 activation also triggers a flip-flop reading, and if switch 2 is active or not, determines which switch activated first and thereby the rotational direction.

If the receiver of the signals works like most directional pulse readers I've seen, a freq flank triggers an interrupt that samples the dir pin and increases or decreases a counter. In that case, even the flip-flop could be omitted as the dir signal is only relevant at the time of a flank on the freq pin and does not need to be persisted between flanks.

With this solution the pulse width will vary with freq. If the pulse shape of the freq signal is important, a monostable multivibrator could be added before outputting the signal to a receiver.

flip-flop

Schematic

\$\endgroup\$
5
  • \$\begingroup\$ Reeds were always a trivially obvious solution - apart from lifetime. You specifically excluded them in your question. You state up to E11 cycles in some applications and up to 200 Hz operation. So 15+ years at 200 Hz and longer at slower. Running a reed for 100 billion operations is pushing their limits compared to figures found almost anywhere. "Billions" is mentioned in some locations. Low loading may save you - bur lack of wetting may 'unsave' you. || A long tailed pair as per my answer + coils and magnets as per yours would solve lifetime issues. \$\endgroup\$ Commented Apr 29 at 7:47
  • \$\begingroup\$ I started from a very complex solution and it took getting frustrated enough before I truly went into carte blanch mode to even think of this trivially simple solution. But looking at it further I think you are right and that lifetime will be an issue indeed... :( \$\endgroup\$ Commented Apr 29 at 8:13
  • \$\begingroup\$ Thank you. Whats stack exchange policy now that I changed my mind? Should I edit the answer? Delete it? Or simply not mark it as accepted? \$\endgroup\$ Commented Apr 29 at 8:21
  • \$\begingroup\$ It would be a shame to delete it. It solves your problem in SOME ways and is not a terrible answer. I made the point about E11 operations being 15 years at maximum rotation rate because I suspected you may not have thought it through AND/OR that the combined figure is quite likely to be above yoir real target. I don't know the application but is ir really meant to live in a swirling water environment for 15+ years? Maybe yes, but a fuller description of the task would really help. (As I just perhaps maybe have already said :-) ). \$\endgroup\$ Commented Apr 29 at 10:34
  • \$\begingroup\$ THe very best reeds at very light loads are said to reach "billions of operations". You'd want to be designing such with immensely good understanding. || Making ANYTHING last 15 +years reliably takes a mix of real engineering, luck and over doing things. Knowing the whole task (or more of it) helps us help. || I could do this iptically (unless there are more unknown fatal constraints). \$\endgroup\$ Commented Apr 29 at 10:37
0
\$\begingroup\$

Update: After beating myself bloody, I made it work with a variation of this solution. I moved the coils around a bit and changed the resistor values. Thank you all who engaged with this question, Russell McMahon in particular.

I'll explain my current solution that sort of solves the problem, but not quite. Perhaps someone can suggest a modification that would make it better. I can possibly add more details as the discussion goes along.

In addition to the main power-generating phase, I've added two small auxiliary coils at a small angular offset. Each coil is connected to an attenuation net with ac coupling and a dc offset. That signal is fed to a low-voltage, nanopower comparator, which in turn feeds a flip-flop circuit that generates the output signal. This kind of works.

attenuation net

The issue with this solution is that I can't fit the aux coils perpendicular to the radial direction, which creates a very wobbly waveform. The waveform issues make the comparator fire more than it should, which destroys the signal.

I also considered using a fixed reference voltage fed to the comparator(would only require 1 aux coil), but found that the DC offset varies with frequency, which makes the comparator not fire at all.

This is the best image I can find from my testing, where you see the mentioned waveform in the wiggly light blue and dark blue probes. The induced voltage is significantly lower at 3 or 4 Hz where I need to solution to work. There, I get about 10- 15 mv amplitude. waveform

\$\endgroup\$
4
  • \$\begingroup\$ Can you show FD simulation results for the fluid flow through the turbine? It looks to me like your turbine is unnecessarily wimpy. You can extract a lot of energy from the flow you mention. The idea is to minimize or eliminate turbulence so that the energy is not wasted. I'd start there first. What is the allowable pressure delta across your sensor? That will give you the upper limit for available power. Go from there. \$\endgroup\$ Commented Apr 30 at 10:16
  • \$\begingroup\$ @Kubahasn'tforgottenMonica Again, thanks but no thanks. For now at least. The wavy lines are wavy because of the orientation of the coils, which is due to space limitations. If you didn't look at the wavy lines but rather the pink one and (correctly) guessed that it comes from the power-generating phase, you are correct. The non-sine shape comes from the rectifier diodes. I have an M.Sc. in fluid dynamics and I assure you that this problem is not solved easily on the fluid side of things. Pressure drop is the wrong way to think about it, since it sits in a much bigger pipe. \$\endgroup\$ Commented Apr 30 at 11:05
  • \$\begingroup\$ R14 & R15 can be combined (or R14 probably just eliminated). Did you try my two transistor comparator which can be made as micropower as you wish. || For a single input + reference comparator you can derive mean DC level with a simple RC filter feeding the reference input. This is a very common data recovery method. Making RC time constant longer makes reference steadier but less responsive to dc level changes with time. A 3 pole low pass Bessel filter gives you excellent filtering but good response to DC changes and takes ONE opamp section. I am using OPA504 at present with 500 nA quiescent. \$\endgroup\$ Commented May 2 at 14:05
  • \$\begingroup\$ Gain bandwidth is VERY low but may suiffice. There are many others wityh performance/ Iquiescent tradeoffs. || Similar with comparators. || Look at LCSC.COM for good pricing and a substantial selector guide. Also digikey.com and mouser.com with higher prices plus good selector guide. \$\endgroup\$ Commented May 2 at 14:07

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.