No announcement yet.

semi OT - help with some electronic circuit ideas

  • Filter
  • Time
  • Show
Clear All
new posts

  • semi OT - help with some electronic circuit ideas

    An ongoing project of mine is a soft bearing dynamic balancer. The hardware is basically done and seems to work well. This is a two plane balancer: picture a shaft to be balanced sitting on a support at each end. If you're standing in front of it, each end can move toward and away form you as any imbalance acts on it as it revolves. This motion, to and fro, can be picked up by a cheap accelerometer or perhaps even a magnet and hall sensor. The later might eliminate a lot noise.

    My original mock up used an inexpensive acceleromator, put the signal through a Butterfield filter (3 stage low pass) and I was filtering iirc for under 5Hz. After that I amplified the signal substantially then triggered a 555 which flashed a led for a short period, acting as a strobe. Commercial ones tell you where and how much mass to apply. I don't care about how much, trial and error with some clay is fine. It might take longer but its an infrequent activity...vs a balance shop using it all day long. With a strobe and mark on the shaft I should be able identify where to place the weight.

    It sort of worked, but there are some challenges.. First of all there is a lot of noise. While I might be able to filter it, I also figure I'll want variable speed so a band pass filter . I believe with the accelerometer the faster you go the more G's it'll produce, which might be a way to improve the signal to noise ratio in that at higher rpms the G force is greater, but amplitude lower. otoh if something was really off, you'd want start slowly and move up. If I used a hall sensor faster wouldn't be more sensitive, there be less amplitude so a smaller signal.

    Is there a convenient way to have some sort of adjustable band pass filter, maybe even one automated and tied to an rpm sensor? Any great ideas or other approaches on how to get and process a signal for this? My hope is to get some very precise results to let me remove a lot of vibration from some machines/motors. The mock up was just an approach I conjured, there's gotta be umpteen better ways to do obvious to someone who knows electronics.....what do you guys think? How should I come at this?

    Thanks for any ideas
    Last edited by Mcgyver; 04-22-2020, 04:17 PM.
    in Toronto Ontario - where are you?

  • #2
    That is just plain made for Digital signal processing (DSP). Generate the speed control, and vary the band pass with it. Not the nicest project to come "up to speed on" and finish in a reasonable time.

    Starting from another direction, what sort of noise do you have? I assume you are talking about "in-band" noise?

    Regular first order imbalance signal should be a once per rev signal, so you might get rid of a lot by bandpass filtering at the frequency of the RPM. Your 5 Hz is then about 300 rpm, a reasonable speed. To get the frequency up means a speedup to rather high rpms.

    Rather than trying to do a continuous sweep, you might do several discrete speeds, which your filter could be either switched to, or different filters produced for each.

    Alternately, you could do a sweepable filter and manually adjust it to whatever frequency you wanted during the test.

    There are many online and book sources for "state variable" filters which have LP, HP, and band pass outputs.

    4357 2773 5647 3671 3645 0087 1276

    CNC machines only go through the motions

    "There's no pleasing these serpents"......Lewis Carroll


    • #3
      I have to agree with J Tiers that this is begging for a digital answer. I don't know how much has been published on this kind of software, but there must be some. I would think it would be within the capability of an Arduino or similar processor. Definitely possible with even an old, cheap laptop (local pawn shop?). But the learning curve may be steep.

      One area you may look into is some of the oscilloscope software out there. I have an inexpensive PC based o-scope and the Windows based software supplied for it has many features, including even a spectrum analyzer. One idea that strikes me is that you could have a two channel o-scope display with your error signal on one channel and a tach signal on another. That would provide a nice visual presentation of both the amplitude of the error and of it's phase, relative to the tach signal. With a bit of experience and recording of the data, you may even be able to correlate the amplitude of the signal to the amount of weight or at least the torque arm that needs to be added. Even the least expensive, computer based o-scope will have sufficient bandwidth for your purposes. And if your signal level is too low, you can always add a low noise amplifier (Op Amp based) between the sensor and the o-scope.

      If you want to stick with analog, then there is a lot of information about filter design out there. I would think that Op Amp designs would be the way to go. And there is a lot published on that. A starting point may be the Op Amp Cookbook which is a classic. But the manufacturer's spec sheets and other published literature on Op Amps and designing with them are all available on-line and free. Go to suppliers of electronic components and start with a search for a 741 Op Amp. Then start clicking on the links there for the spec sheets and other literature. DigiKey, Mouser, Newark, etc. are all suppliers. Companies like Texas Instruments make the ICs and publish the literature about using them. There is enough free literature on the subject available to keep you busy for the rest of the year.

      Note: I mentioned the 741 op amp specifically because it is my go-to op amp when I am starting a design for an op amp circuit. I work with it and if, in the end, I find that my circuit is lacking in some spec or another, then I know that I can go to the suppliers I mentioned above and find another op amp that will have an improved spec in that area while having complete pin-for-pin compatibility with the 741. I can just leave my breadboard as it is on the workbench and await the arrival of the new part. Then I unplug the 741 and plug in the better replacement and voila, the circuit works as needed. I have done this many times. I like starting with the 741. I may not wind up with it, but it is a great place to start.
      Last edited by Paul Alciatore; 04-22-2020, 07:17 PM.
      Paul A.
      SE Texas

      And if you look REAL close at an analog signal,
      You will find that it has discrete steps.


      • #4
        I always wanted to make one to balance my T&C grinder spindle and motor / components but never pursued it.
        I would be interested in seeing your progress and some pics with descriptions, parts used, etc.



        • #5
          I agree with Jerry here.

          Depending on how serious you want to get generally you need to take care of the fundamental, and usually only in one plane.

          I balanced my bench grinder through trial and error and using a simple accelerometer on a smart phone:


          • #6
            Originally posted by Mcgyver View Post
            ... Butterfield filter ...
            Butterworth (like the syrup)

            Is there a convenient way to have some sort of adjustable band pass filter, maybe even one automated and tied to an rpm sensor?
            Switched capacitor filter -- the center frequency is set by a clock, which can be derived from output of the rpm sensor.


            • #7
              Originally posted by tomato coupe View Post
              Switched capacitor filter -- the center frequency is set by a clock, which can be derived from output of the rpm sensor.
              That can be effective. It is somewhat complex.

              One issue with switched capacitor filter....... They suffer from "aliasing", so they require an analog filter anyway. Newer chips may have some of that built in, but you cannot feed them a frequency in excess of about 45% of the switching frequency or you will have problems with aliasing and "spurious output". If the switch frequency is high, this can be easily assured. (the theory says 50%, but some allowance for the "skirts" of the filter response is needed, filters are not perfect)

              Note that the "frequency" is not the filter frequency, but the highest frequency fed to it as input. Noise counts in that, so if there is considerable noise, aliasing may be an issue. With bandpass, it "may" not cause an issue.

              To do the tracking etc, you would need quite a bit of "extra stuff" that is not really worth the effort IMO, even though the solution can be "elegant". It would be easier to do it in DSP, in some ways.

              However, the regular analog state variable filter is quite effective, and should do what you need without a fuss. This is a "use sometimes" deal, not an all-day workstation every day. The solution ought to be proportioned to the need. You should do fine with an adjustable filter setting, or just several frequencies.

              The filter circuit can be built up on perfboard, which is not so true of other solutions.
              Last edited by J Tiers; 04-23-2020, 12:14 AM.
              4357 2773 5647 3671 3645 0087 1276

              CNC machines only go through the motions

              "There's no pleasing these serpents"......Lewis Carroll


              • #8
                Comes to mind that it's possible to trigger a strobe using a mechanical contact. You put a contact on an adjustable arm with a fine adjustment, then position that close to the rotating shaft. As you slowly bring it in, it begins to touch when the 'heavy' side of the rotor is towards it. Think of the electronic edge finder- pretty much the same thing except for the strobe part. You might be able to get away with driving an led string directly through the contact, instead of using any kind of strobe circuitry. At the low voltages that leds use, you should be able to get a very good 'on-off' with just small deviations of the rotating shaft. You would have to be able to make fine adjustments to the contact position without disturbing its mounting.

                As I'm thinking about this, it comes to mind that leds are not the best choice for a light source for a strobe. In this case, when you dial in the contact a tad too far, it will be in contact for a longer period, which will bring more led light into the picture, but also blur it. Probably within a range you'll be able to get the 'heavy side' indication you're looking for. I think it's worth giving it a shot.

                Putting some threads in a plastic block often ends up with the threaded rod or bolt being a bit of a tight fit. That's the zero backlash that you'll need to be able to make this fine of an adjustment possible.
                I seldom do anything within the scope of logical reason and calculated cost/benefit, etc- I'm following my passion-


                • #9
                  A few observations and suggestions:

                  1. For low level signal amplification, as well as common mode noise rejection, a good instrumentation amplifier such as the AD620 or INA118 will work better than an LM741. You can also make an instrumentation amplifier with full differential input using a quad op-amp such as LM324 or TL084.

                  2. You might consider driving the shaft of the balancer with a synchronous motor or a stepper motor, so you can precisely set the speed and even know the angular position if you use a sensor to detect a zero degree reference point.
                  Paul , P S Technology, Inc. and MrTibbs
                  USA Maryland 21030


                  • #10
                    some sort of adjustable band pass filter
                    This is exactly what an electric guitar "whah-whah" pedal is. In that case ( at least 70's era) the center frequency was altered by a foot operated variable resistor. Easy enough to change that to some other type of control.

                    I recently went through the exercise of "if my band saw has 14 inch wheels, and I want to digitally convert that to display feet per minute using stock counter chips, how many pulses per revolution do i need and should I use a single hall effect switch or photo-interrupter on the shaft with a predetermined number of slots/holes?". Of course, it ends up being much, much cheaper and simpler replace those old style counter chips with a microcontroller, keep the hardware and mechanical part dead simple, and do some modest amount of fiddling in the software. The microcontroller used in the arduino is only few bucks - after you've developed your circuit you don't need the entire arduino.
                    "A machinist's (WHAP!) best friend (WHAP! WHAP!) is his hammer. (WHAP!)" - Fred Tanner, foreman, Lunenburg Foundry and Engineering machine shop, circa 1979


                    • #11
                      I built a professional level soft bearing balancing machine and amplifier so I have some relevant experience. I used NIB magnets at each work bearing on the carriage and telephone ringer coils for the pickups - lots of turns of fine wire. The amplifier reads out an analog (regular meter with a needle) or digital signal (via a DVM) for the amplitude of unbalance and a strobe output for the phase angle. The amplifier uses 741 op amps and the filter is a tunable state-variable bandpass circuit from Don Lancaster's book Lancaster's Active Filter Cookbook, p. 162. Amplitude is the easiest parameter to measure and it has the least circuitry. Just an input amplifier, filter, and a meter circuit. All of these are available in op amp circuit cookbooks. Phase angle requires 5 or 6 stages of signal processing, I forget how many. I used a LED strobe made from a high intensity flashlight but its output is just sufficient. Newer LEDs are probably brighter. I put some effort into locating xenon flash drivers and flashtubes, thought of auto timing lights but did not build a circuit for these.

                      One of the most important parameters to control is rotor RPM. It must be constant and repeatable from one trial run to another. I run my small rotors at 30 HZ. Fast enough to get a good signal and fast enough that the filter does not have too much trouble with low frequency. The drive motor should run at a different RPM to make it easier to filter out its noise and the drive belts should not induce vibration.

                      I ended up using only the analog meter for some reason. Maybe my brain works better comparing two needle positions together with their numbers, than comparing two abstract numbers. This is completely old school simple analog, and digital will not balance a rotor any better. Where digital amplifiers become important is connected to advanced software which makes it possible to balance a rotor in fewer trial runs. Balancing software is not simple stuff. I did not go there. Oh yeah, just now thought of a detail: your input amplifier should have several ranges @ 10x per range.

                      I published a complete account of two empirical balancing methods based upon having either an amplitude readout or a phase angle strobe flash. Red the entire article, because near the end is the section on cross-effect. You cannot simply balance one end of a rotor at a time and call it good.


                      • #12
                        what device are you using? i have an adx l320 somewhere, waiting to be put to use. i bookmarked this (too complicated for me but probably not for you), no idea if its useful in any way.



                        • #13
                          Originally posted by enginuity View Post
                          I agree with Jerry here.

                          Depending on how serious you want to get generally you need to take care of the fundamental, and usually only in one plane.

                          I balanced my bench grinder through trial and error and using a simple accelerometer on a smart phone:

                          what app are you using there? the screen looks unfamiliar. i have a few of them, but am still looking for a better one. most usefull so far has been the skf app, but it is not especially consistent.


                          • #14
                            Thanks for all the ideas.....I will be processing them on the slow old analogue computer. Carl, nice work on those - thanks for the links!
                            in Toronto Ontario - where are you?


                            • #15
                              I never got around to trying it but I had always figured on using piezo electric sensors. Perhaps an old ceramic phono pickup.