I want to buy an accelerometer to measure the amount of out of balance of a washing machine that starts to dance. The question is, how many G's max. would I expect to try to measure when the machine is jumping around, considering that accelerometer IC's are available with various full scale sensitivities?

The accelerometer would be mounted on the external surface of the machine's internal drum enclosure(this part of the machine is supported by two heavy springs and seemingly incompetent dampers), and not on the exterior of the washing machine so well animated.

I also notice that attached to the drum enclosure (top and bottom) are two large concrete masses I would guess >25kg ea. in weight!

I am no physicist, so I don't know a G from a G-String!

I do know that the loaded water+clothes+machine weighs about 100kg when starting to dance and wander around.

The various low cost xyz axis analog accelerometers chips I am considering as candidates to buy have FS ratings of +/- 1g, 4g, 8g, 16g, 200g. Which is the right FS g for the job?

 

I want to buy an accelerometer to measure the amount of out of balance of a washing machine that starts to dance. The question is, how many G's max. would I expect to try to measure when the machine is jumping around, considering that accelerometer IC's are available with various full scale sensitivities?

The accelerometer would be mounted on the internal drum enclosure, not on the exterior of the washing machine.

I would get one that goes up to at least +/- 100g, or possibly even more. While that might seem excessive, you'd be surprised that the magnitude of the g forces that can be generated in something like that in the form of brief shocks, as parts go thumping and banging around. It's important not to overload the accelerometer, because some of them have a long recovery time (many milliseconds) after a shock overload.

I am no physicist, so I don't know a G from a G-String!

No comment.

 

Acceleration is a vector quantity. It has both magnitude and direction. It is the derivative of velocity which is also a vector quantity with both magnitude and direction. The simple physics is that acceleration can arise from both a change in the magnitude of velocity, or a change in the direction of velocity. You should understand what the instrument will measure.

A human pilot, with training, can withstand +9 g. and -6 g. I think what you see in a washing machine will be on that order of magnitude.

 

Acceleration is a vector quantity. It has both magnitude and direction. It is the derivative of velocity which is also a vector quantity with both magnitude and direction. The simple physics is that acceleration can arise from both a change in the magnitude of velocity, or a change in the direction of velocity. You should understand what the instrument will measure.

I understand that acceleration is a vector quantity, that is why I will use a X-Y-Z accelerometer and therefor I can measure the acceleration in all and any direction components of velocity. Further I can guesstimate that z-axis acceleration gives the machine "upward mobiity", while X-Y acceleration resolves to give the instantaneous direction of a dance step of the machine.

Further, I notice that the acceleration in all axes could possibly be described as being a sinusoidal function v. time.

My intention is to stop the motor when a MCU looks at the magnitudes of all of the axes and detects enough g's for the machine to not just rock and roll but roam.

 

Well you did say that you didn't know much. I apologize for taking you at your word.
At first I thought you were going to mount them on the rotating drum. That was my mistake.

 

I would get one that goes up to at least +/- 100g, or possibly even more. While that might seem excessive, you'd be surprised that the magnitude of the g forces that can be generated in something like that in the form of brief shocks, as parts go thumping and banging around. It's important not to overload the accelerometer, because some of them have a long recovery time (many milliseconds) after a shock overload.


No comment.

That magnitude does surprise me a bit, as the rotating "wet brick" of wet clothes causes the washing machine to function the same way as a vibrator used for physical therapy, but it takes several fractions of a second for large but slow oscillating amplitudes to reach maximum displacement, while fast vibrations do occur with a smaller amplitude but much higher frequency and all this is of course proportional to spin speed..and this seems to suggest to my judgement that the g forces are less??

I do know that a thump makes for big g's, but a rock and roll???

I may not know (in magnitude) a g from a G-String, but I do know a g from a G-Spot.

 

But there are thumps, big and little, amongst the rocking and the rolling, and even little thumps can involve surprisingly large g levels. I'm going to stick with my recommendation of a +/- 100g accelerometer.

 

But there are thumps, big and little, amongst the rocking and the rolling, and even little thumps can involve surprisingly large g levels. I'm going to stick with my recommendation of a +/- 100g accelerometer.

I guess we can agree to disagree.

 

But there are thumps, big and little, amongst the rocking and the rolling, and even little thumps can involve surprisingly large g levels. I'm going to stick with my recommendation of a +/- 100g accelerometer.

Papabravo(thanks for your reply!) says 6, OBWO549 says 100..who am I to believe??

 

Papabravo(thanks for your reply!) says 6, OBWO549 says 100..who am I to believe??

I would say just do as much additional research as you think necessary, and go with what seems right. I'm just giving you my off-the-cuff opinion.

 

Vibrators, g-strings, g-spots, rock and roll.

One of these things is different from the others.

 

Vibrators, g-strings, g-spots, rock and roll.

One of these things is different from the others.

g-Whiz! Joeyd999, this would be great question for an IQ test for vetting new-hire physicists at NASA!

Thanks again to everyone for their help!

Ok, I realize I could buy the +/- 200g IC with analog X-Y-Z output model and assuming g resolution is high, I should be able to scale the XYZ analog signals by feeding them into op-amp buffers that could amplify(if necessary) the signals being fed to the MCU analog to digital inputs for best use of the 10-bit (1024 levels) capability of the A2D of the MCU I am using.

I also realize that sitting on the machine during the spin cycle results in both a machine stabilizing effect and a stimulating break during a busy wash day for mommy, still I would like to automate the machine's supervision..mommy can't always keep watch of the machine.

What I must do next then is try to figure out how to interpret, to create some algorithm to relate g's v. time v. go-go dancing.

Any help with this..maybe I should post a new topic?

How to I digest all the g's readings into a simple single pin go-nogo output signal to be sent to the MCU controlling the motor?

 

Ok, I realize I could buy the +/- 200g IC with analog X-Y-Z output model and assuming g resolution is high, I should be able to scale the XYZ analog signals by feeding them into op-amp buffers that could amplify(if necessary) the signals being fed to the MCU analog to digital inputs for best use of the 10-bit (1024 levels) capability of the A2D of the MCU I am using.

I would expect even a 200g accelerometer would have enough resolution (i.e., a low enough noise level) that you could amplify the output if needed. You're looking for a vibration level, not static g's, so an AC-coupled amplifier would suffice and you wouldn't have to worry about DC accuracy, temperature drift, or any of that.

What I must do next then is try to figure out how to interpret, to create some algorithm to relate g's v. time v. go-go dancing. Any help with this..maybe I should post a new topic?

Why post a new topic? We can do this here.

How to I digest all the g's readings into a simple single pin go-nogo output signal to be sent to the MCU controlling the motor?

If I were doing it, I would amplify each of the X, Y and Z accelerometer outputs in an AC-coupled amplifier (gain TBD), run each amplifier output into an absolute-value circuit, sum the three absolute-value outputs, and present the result (perhaps with some low-pass filtering) to the ADC input on your MCU. I think this would suffice; for your purposes (having a gadget that will hit the kill switch when the washer starts doing the hokey-pokey around the laundry room floor), it seems to me that trying to vector-sum the X, Y and Z outputs would be overkill.

You'll probably have to do some testing first to determine the best amplifier gain, by monitoring the accelerometer outputs under real-load conditions with an oscilloscope, both with and without the washer dancing around the room.

That's all for tonight; I'm gonna hit the sack.

 

I would expect even a 200g accelerometer would have enough resolution (i.e., a low enough noise level) that you could amplify the output if needed. You're looking for a vibration level, not static g's, so an AC-coupled amplifier would suffice and you wouldn't have to worry about DC accuracy, temperature drift, or any of that.


Why post a new topic? We can do this here.


If I were doing it, I would amplify each of the X, Y and Z accelerometer outputs in an AC-coupled amplifier (gain TBD), run each amplifier output into an absolute-value circuit, sum the three absolute-value outputs, and present the result (perhaps with some low-pass filtering) to the ADC input on your MCU. I think this would suffice; for your purposes (having a gadget that will hit the kill switch when the washer starts doing the hokey-pokey around the laundry room floor), it seems to me that trying to vector-sum the X, Y and Z outputs would be overkill.

You'll probably have to do some testing first to determine the best amplifier gain, by monitoring the accelerometer outputs under real-load conditions with an oscilloscope, both with and without the washer dancing around the room to scale the outputs properly.

That's all for tonight; I'm gonna hit the sack.

Thanks again,

I have ordered the +/- 200g accelerometer and expect to see it at the door in about three days. Thereafter I know I will have the time this week to quickly mount the chip and on a DIY PCB with the MCU to get some readings during a live performance of my dancer. The MCU will use a simple RS232 serial link to send the raw data to a logging desktop PC program and somehow I will find a way to tag machine dance steps with data logging events even if it means I will be manually pressing a key on the keyboard to flag these dance step events to sync with the XYZ data monitoring. I can easily modify my Visual BASIC data logging program to capture data in this way. I know I can log data probably at rates as fast as a sample set /3-mS or so. Should I log data at the max MCU to PC rate available or get samples logged every .1 sec or so??

The reference design presented in the Analog Devices datasheet shows that the analog outputs all have an internal 32k resistor in series with each output pin and a capacitor placed at each output pin determines the freq. cutoff knee of the outputs. For instance, a .1uF cap results in a -3dB freq. of 50Hz.

Depending on the sample rate I should soon have a mountain of raw XYZ + dance event flags as data bytes to be saved on the PC in a data file to start to take a look at. The A2D on this MCU can yield a digital result in 32-uS (so the three A2D XYZ values can be stored in approx 100uS) and so I will program the MCU to store the results in a MCU buffer array to be then sent in data packets of 200 or so bytes to the PC each time this small MCU buffer is filled. The VB program will also save the elapsed time as integer data bytes with each packet transfer.

I don't quite see the logic of your recommendation to sum the three outputs at this point because (I am just guessing that) the Z-axis output may or may not be the most important signal to monitor?

 

I know I can log data probably at rates as fast as a sample set /3-mS or so. Should I log data at the max MCU to PC rate available or get samples logged every .1 sec or so??

Depends on how fast the dance steps are. I should think that sampling at 5-10X that rate would be enough.

I don't quite see the logic of your recommendation to sum the three outputs at this point because (I am just guessing that) the Z-axis output may or may not be the most important signal to monitor?

It could be; I was simply assuming they were all of roughly equal importance. That assumption could be wrong, of course.

 

G string: 4th string (counting from the lowest pitch) on a guitar.

ps, G forces get incredible when things knock against each other.

 

G string: 4th string (counting from the lowest pitch) on a guitar.

Yea. I usually get a shriek from the chicks when I pluck it...

 

Two words guys: Family Site.

 

I want to buy an accelerometer to measure the amount of out of balance of a washing machine that starts to dance.

It takes a pretty good sports car to stick to the road with 1G of lateral acceleration. I drove a buddy's Corvette once and enjoyed doing a tight U-turn at 30mph without swish or sway.

It's fair to say a washer will start to walk long before it receives a sustained 1G force in any direction.

I'm wondering if x-y strain gauges wouldn't be more useful for this. It's a sustained imbalance along the x or y axes that will cause the machine to walk.

Assuming you have accurate accelerometer data, what then? I'm not sure how to interpret a hammer blow - producing a 20G shock wave - versus a 10Hz wave of 0.1G impacts. Maybe this would become clear once you saw the patterns of walking versus non-walking machines.

 

Two words guys: Family Site.

Huuh?
I was talking about a lesson I received a very long time ago. I was tapping a 10 turn pot with a screwdriver to settle the wiper. The designer told me not to do that because the G forces go up in the 10's of G's.

From this, I derive the idea that anything that taps or clicks is suffering forces far in excess of the ordinary.