Continue to Site

Welcome to our site!

Electro Tech is an online community (with over 170,000 members) who enjoy talking about and building electronic circuits, projects and gadgets. To participate you need to register. Registration is free. Click here to register now.

  • Welcome to our site! Electro Tech is an online community (with over 170,000 members) who enjoy talking about and building electronic circuits, projects and gadgets. To participate you need to register. Registration is free. Click here to register now.

Low peak detector and alarm system

Status
Not open for further replies.
Wow! I wish everyone that had a question knew what they were talking about as well as you. I hope you were not offended by our simplistic aproch. Many of the posters that are new here do not know what they are working with or weather or not a simple fix is what is needed.
You do have a ligitament problem that does need a complex solution! Wish you well.
Your out of my field of expertise!
 
Last edited:
Your threshold detect should work. Perhaps your circuit is too slow. What is the rate of the pulses? If you have something like -----_---- you should be able to detect this condition with analog. Your circuit is flawed.
 
Your threshold detect should work. Perhaps your circuit is too slow. What is the rate of the pulses? If you have something like -----_---- you should be able to detect this condition with analog. Your circuit is flawed.

Exactly the situation but the circuit does not work and so we come full circle. What is wrong with my circuit, I know it shows the hated LM358 but I will use a better IC for the final build. I think I have looked at the circuit so long I just cant see past what I want to see.

Thanks all
 
It seems you need to use fast parts. LM358 is slow I think, there are many good analog designers ( I am not one) that frequent here like Ericgibbs, Rolf, and AG and many more. I am sure if you are patient, they might chime in.

The electronics world survived for a long time and produced many fine products long before the MCU came along. Fact is the MCU would be a solution, but if your not comfortable with this solution, you can do it in other ways...:)
 
Last edited:
OK i will look at the OPA series of faster amps and comparators. Also I will wait to see if Ericgibbs, Rolf or AG join the fray as you say.

Meanwhile I have changed the circuit to try and ground referance the rectifier output so removing (I Hope) the Vcc/2 component and also changed the pulse generator comparator pull up to see if that helps to improve input to the charge pump.

Looks like this now

**broken link removed**
 
Last edited by a moderator:
Wow! I wish everyone that had a question knew what they were talking about as well as you. I hope you were not offended by our simplistic aproch. Many of the posters that are new here do not know what they are working with or weather or not a simple fix is what is needed.
You do have a ligitament problem that does need a complex solution! Wish you well.
Your out of my field of expertise!

I had to think about your post for while. Appreciate your comments very much and I think forums are a place to chuck about ideas so your comments get me thinking in a new direction - that has to be good.

You cant just walk into a local pub and start chating about this kind of stuff (well not around here anyway) so all comments are welcomed and if you think of anything else - chuck it in.

Thanks
 
Doesnt having a pump impeller damaged that baddly create vibration?
I work with many types of pumping systems and I have been around enough pump systems to know when a impeller is damaged just by vibrational sound alone.

Or are these special purpose or application pumps?
A one in six sub tone can easilly be picked up and separated from the primary tone. That then can be amplified and detected all in analog.

Just curious.

Quantised said:
The vibration you both mention is correct and mainly used in larger pump installations but for my application it cannot be used. The pumps are small and mounted on large noisy machinery that would mask any signal - certainly the small changes that need to be detected to signal any impending failure.

Don't you already have the signal available? Isn't the sine-wave with the missing/attenuated cycles a signal?

What I'm saying is that the signal that you're already trying to use could be checked for frequency components outside the fundamental; so I think tcmtech's idea seems fine. I know you said you already do this for your other pumps; I'm guessing you just feed the signal through a notch filter (set to the expected signal frequency), get the average of the rectified output signal and compare this to a threshold. If the signal is not sinusoidal (i.e. has other frequency components) then the notched signal power/amplitude will increase. You may also wish to use an AGC on the input signal to give you a stable signal level.
 
Last edited:
Don't you already have the signal available? Isn't the sine-wave with the missing/attenuated cycles a signal?

What I'm saying is that the signal that you're already trying to use could be checked for frequency components outside the fundamental; so I think tcmtech's idea seems fine. I know you said you already do this for your other pumps; I'm guessing you just feed the signal through a notch filter (set to the expected signal frequency), get the average of the rectified output signal and compare this to a threshold. If the signal is not sinusoidal (i.e. has other frequency components) then the notched signal power/amplitude will increase. You may also wish to use an AGC on the input signal to give you a stable signal level.

Gosh a new idea to assimulate, very interesting indeed. I did'nt really see how tcmtech was proposing to get the 1 in 6 low peak out from the rest but understand now.

Using the 1 low peak in 6 scenario - this waveform can theoretically be split into 2 signals that when added together give the waveform. One is 6 equal height pulse and the other is a single negative pulse at half height, one pulse every 6 cycles.

This pulse will have a frequency of 1/6th the fundamental and as it will also appear in the frequency domain as a pulse it will energy across the spectrum with most of it at the lower frequencies.

If a band pass filter is set up around this 1/6th frequency it would increase in output if a blade is broken. It is exactly the same your suggestion but this way I can add a second band pass filter at 2/6th fundamental to detect 2 broken blades - brilliant.

One of the key ideas to do this in the time domain is that it is not sensitive to RPM variations or blade count so each filter will have to be set for blade count and maybe some tracking system to follow the RPM - still it seems a very good idea and I thank you for this. For static RPM cases it looks easy.

I will have a look at the frquency response of the signal and see how much it changes for 1 low peak.
 
I wasn't alluding to using X number of band pass filters; just a single notch (band stop) filter. The notch removes the fundamental frequency that would be present when the device is rotating correctly, and the output level would be of low amplitude. When something goes wrong (blade breaks), the notch would not remove all the other frequency components created (and the 'fundamental' would likely be outside the notch bandwidth), therefore the filter output amplitude would be higher.
 
In case you wish to stay in the time domain, I drew up the attached schematic. It basically generates a reference level from the peak of the signal, and compares the input signal to some fraction of that. The output of the first comparator a pulse for every half cycle that is OK; a bad half-cycle will issue to pulse.

I tacked a charge-pump on the output of this which will give a level somewhat proportional to the frequency of the incoming pulses - if the level droops then the output is signalled as faulty.

I didn't calculate any component values for you, but I can say that the RC time constant of the peak detector output should be long (say, maybe a few seconds), the first threshold should be at halfway between the peak level of a good and bad pulse. The charge pump threshold you can get through trial & error.

I used something similar for decoding IRIG-B (see the drawings on pages 21, 30 of **broken link removed**), and it worked very well at detecting high/low amplitude pulses.
 

Attachments

  • sinthing.gif
    sinthing.gif
    29.5 KB · Views: 176
I wasn't alluding to using X number of band pass filters; just a single notch (band stop) filter. The notch removes the fundamental frequency that would be present when the device is rotating correctly, and the output level would be of low amplitude. When something goes wrong (blade breaks), the notch would not remove all the other frequency components created (and the 'fundamental' would likely be outside the notch bandwidth), therefore the filter output amplitude would be higher.

I totally understand the concept of the notch filter and thats what started me thinking and the idea is great so I was just expanding it to a band pass filter - its exactly the same idea but allows a second bandpass filter to detect 2 broken blades. Its an idea thats so easy to try I may bash one up at the w/e.
 
Thanks for time domain circuit sketch, I shall go through this see if I follow the various stages, I get them all up till the threshold of the missing pulse part but I will study it.

Many thanks for this and I like the peak detector part - it makes the pulses out easier to calibrate.

So between this and the filter method I will be busy.

I will let you know which one is best when I get it to work.
 
Sure, if you need to detect the number of blades broken then the extra filters will probably be of use. If you just want to know whether one or more blades are broken, then just the notch will suffice.

And that schematic I posted, the output of the charge pump will decrease based on the number of blades broken, so you could also potentially use that with an array of comparators to tell you how many blades are wrecked.
 
Sorry, I posted before I got your second post. All the best. I'll be interested in hearing what works.
 
Nice to see you cought what I was getting at! :)

I actualy saw this done while I worked at a local coal mine for a college credit course. (student co-op program)

They had an old portable analog machine that looked like a hybrid between a spectrum analizer and a function generator. I dont know exactly how it worked but it used a bandpass filtering setup to narrow in on a specific frequency or band of frequencies.
The high and low side noise was ignored and all you got was the range you were tuned to.
The rest was just a matter of visual interpretation of the screen waveforms you got.

They used it to listen to motors, generators and gear boxes to find if bearings were going bad or if gear teeth were chipped or wore down.

I guess it worked rather well if you knew how to use it.
Even in deafening sound and high mechanical vibration conditions it could pick out bad rollers in bearings or a bad set of gear teeth! :)
 
Status
Not open for further replies.

Latest threads

New Articles From Microcontroller Tips

Back
Top