## How to analyze change in frequency?

Expand Messages
• I have a long term project which will seek to measure the degree of torsional resonance that exists at the flywheel of my car engine. I have a Hall effect
Message 1 of 23 , Mar 1, 2013
I have a long term project which will seek to measure the degree of
torsional resonance that exists at the flywheel of my car engine. I have a
Hall effect sensor mounted to sense the flywheel gear teeth going by the
sensor. As the engine goes under load the power pulses actually cause
minute changes in the frequency of those gear teeth going past the sensor.
On a V-8 engine there are four power pulses per revolution of the flywheel.
As the engine loads up and the power pulses get larger this minute change in
frequency four times per revolution will grow in amplitude. If the
mechanical drive system (engine & transmission) hit a point of rotational
resonance those minute changes in frequency will become significantly
larger.

So my question is basically how do I analyze this frequency?

I have a regular oscilloscope. I could monitor the basic waveform but how
would I detect the minute change of frequency? Would the waveform shrink
and expand if I get the time base just so?

If has been 30 years since college calc classes, am I seeking a derivative
or integral of the basic frequency? Can I achieve this with a certain mode
on the oscilloscope (I 'm a very basic scope dope)?

Ideally once I learn the terms and analysis I am trying to achieve I would
want to collect this data during operation and record it on my laptop and
sync it to the engine parameter data that I am already able to
collect/record.

What are the right terms/process I am looking for?

The sensor is already mounted in the car.

Data sheet:
http://media.digikey.com/pdf/Data%20Sheets/Hamlin%20PDFs/55505.pdf

I thought John Popelish or someone else in the group gave me the right
process a few years ago but I couldn't find that post in the archives just
now.

Thanks,

Dan Nicoson
• A couple of questions: How many teeth are there on the gear that the sensor senses? And, is there some way to tell where the 0 position of the flywheel is? I
Message 2 of 23 , Mar 1, 2013
A couple of questions: How many teeth are there on the gear that the sensor
senses? And, is there some way to tell where the "0" position of the
flywheel is?

I would suggest buying or borrowing a digital storage oscilloscope with a
large memory (say 1 million samples or so) that can store waveform data to
a USB key.

Capture a bunch of waveforms from the sensor, and analyze them at your

You'll get a frequency modulated square wave from the sensor. The higher
the RPMs, the higher the frequency and the shorter the period of the square
wave.

Write a program to measure the period of each pulse, and plot the
frequencies of the pulses (inverse of periods). You can use the Excel graph

Does the graph of the frequencies look somewhat sinusoidal? That graph
would show the frequency modulation of the sensor square wave, and the
cycles in the sinusoid should correlate with the power strokes of the
pistons. The sinusoid should have much larger amplitude when the engine is

The amplitude should also be larger at certain RPMs where there is
rotational resonance, compared with RPMs where there is no resonance.
However, my hunch is that it will be hard to see any resonance, for several
reasons: the resonant frequency may be higher than the sensor squarewave
frequency, friction in the engine and drivetrain will act to dampen any
resonance, and the resonance may be buried in random noise.

Sounds like a fun project. I'll be curious to hear how it turns out.

[Non-text portions of this message have been removed]
• as a certified master technician we use accelerometers and electronic vibration analzers for noise vibration and harshness concerns. i really am uncertain
Message 3 of 23 , Mar 1, 2013
as a certified master technician we use accelerometers and electronic
vibration analzers for noise vibration and harshness concerns.
i really am uncertain about the specifics of your investigation , the
question is what is your goal?
to lovate the frequency you have to know what component you are looking for
. then what is the magnitude of the vibration ( i.e 1 g is going to be
something heavy. is it a first second or third order vibration?

or are you simply wanting to quantify the torque impulse and it f of o?
in essence im pretty handy and trained with this by two manufacturers but i
I presume that you are looking at crankshaft acceleration to try to
calculate the impulse frequency?
are you looking for a specific vibration? is there a concern with the auto
or are you just researching dynamic phenomena ?

--
jeremy youngs

[Non-text portions of this message have been removed]
• lovate = calculate !!!! fat fingers and tired eyes :) -- jeremy youngs
Message 4 of 23 , Mar 1, 2013
lovate = calculate !!!!
fat fingers and tired eyes :)

--
jeremy youngs
• Jan, The gear in question is the starter ring on the engine flywheel, so no real way to find a zero point on a rotation. I don t actually see a reason to
Message 5 of 23 , Mar 2, 2013
Jan,

The gear in question is the starter ring on the engine flywheel, so no real
way to find a "zero" point on a rotation. I don't actually see a reason to
find a zero since we are looking for the change in frequency not a specific
position in the rotation.

The gear has 157 teeth and the operational range is 650 RPM to about 5500
RPM, that makes our base frequency range 1700hz to 14,391hz. I am hoping to
eventually analyze and display/record the level of torsional resonance in
real time. I know there are IC's that convert frequency to voltage, in this
case we need to convert the change in frequency to a voltage...

I am trying to figure out the basic sensor/processing/display/record process
now on my car so that later I can use it to analyze another combination of
engine/gearbox/load. I could end up with a toothed sensor on each segment
of the combination looking for issues as I do the full range of testing.

Thanks,

Dan

-----Original Message-----
From: Electronics_101@yahoogroups.com
[mailto:Electronics_101@yahoogroups.com] On Behalf Of Jan Kok
Sent: Saturday, March 02, 2013 12:09 AM
To: Electronics_101@yahoogroups.com
Subject: Re: [Electronics_101] How to analyze change in frequency?

A couple of questions: How many teeth are there on the gear that the sensor
senses? And, is there some way to tell where the "0" position of the
flywheel is?

I would suggest buying or borrowing a digital storage oscilloscope with a
large memory (say 1 million samples or so) that can store waveform data to a
USB key.

Capture a bunch of waveforms from the sensor, and analyze them at your

You'll get a frequency modulated square wave from the sensor. The higher the
RPMs, the higher the frequency and the shorter the period of the square
wave.

Write a program to measure the period of each pulse, and plot the
frequencies of the pulses (inverse of periods). You can use the Excel graph

Does the graph of the frequencies look somewhat sinusoidal? That graph would
show the frequency modulation of the sensor square wave, and the cycles in
the sinusoid should correlate with the power strokes of the pistons. The
sinusoid should have much larger amplitude when the engine is under load.

The amplitude should also be larger at certain RPMs where there is
rotational resonance, compared with RPMs where there is no resonance.
However, my hunch is that it will be hard to see any resonance, for several
reasons: the resonant frequency may be higher than the sensor squarewave
frequency, friction in the engine and drivetrain will act to dampen any
resonance, and the resonance may be buried in random noise.

Sounds like a fun project. I'll be curious to hear how it turns out.

[Non-text portions of this message have been removed]

------------------------------------

• Jeremy, I am trying set up a system that will allow me to sense when rotational resonance occurs. Just as you are measuring rigid structures with vibrations
Message 6 of 23 , Mar 2, 2013
Jeremy,

I am trying set up a system that will allow me to sense when rotational
resonance occurs. Just as you are measuring rigid structures with
vibrations looking for the frequency that causes that structure to resonate,
I am looking to find where a given engine/geabox/load combination will have
torsional resonance. Every rotational system will have at least one point
of resonance - hopefully not in the operational range.

The end goal is to figure out the basic sensor/processing/display/record
process now on my car so that later I can use it to analyze another
combination of engine/gearbox/load. I could end up with a toothed sensor on
each segment of the combination looking for issues as I do the full range of
testing.

I know this technology already exists but it is priced for manufacturers
that are developing drivelines for any number of automotive and industrial
uses. That puts it out of my hobby price range. Thus the "DIY-with-help"
approach.

Thanks,

Dan

-----Original Message-----
From: Electronics_101@yahoogroups.com
[mailto:Electronics_101@yahoogroups.com] On Behalf Of jeremy youngs
Sent: Saturday, March 02, 2013 1:31 AM
To: Electronics_101@yahoogroups.com
Subject: Re: [Electronics_101] How to analyze change in frequency?

as a certified master technician we use accelerometers and electronic
vibration analzers for noise vibration and harshness concerns.
i really am uncertain about the specifics of your investigation , the
question is what is your goal?
to lovate the frequency you have to know what component you are looking for
. then what is the magnitude of the vibration ( i.e 1 g is going to be
something heavy. is it a first second or third order vibration?

or are you simply wanting to quantify the torque impulse and it f of o?
in essence im pretty handy and trained with this by two manufacturers but i
I presume that you are looking at crankshaft acceleration to try to
calculate the impulse frequency?
are you looking for a specific vibration? is there a concern with the auto
or are you just researching dynamic phenomena ?

--
jeremy youngs

[Non-text portions of this message have been removed]

------------------------------------

• OK, Just answering Jan and Jeremy got me thinking. Without looking at any data sheets let s say we have an IC that converts frequency to voltage. The Hall
Message 7 of 23 , Mar 2, 2013
OK,

Just answering Jan and Jeremy got me thinking.

Without looking at any data sheets let's say we have an IC that converts
frequency to voltage. The Hall effect sensor puts out a square wave signal
with equal high and low durations. So each time the sensor waveform crosses
a certain voltage (say 2 volts) it allows the IC to determine frequency and
output the corresponding signal voltage.

With a completely smooth turning gear wheel at a steady RPM we will get a
very steady signal voltage. Any time the frequency starts to vary the
signal voltage will go up or down. In the case of four power pulses per
revolution the IC should see four areas of changing voltage in each rotation
of the gear wheel. If you hooked an oscilloscope to the output of this IC
you would see a steady horizontal line for a steady RPM with no power
pulses. With power pulses you would start to see those four areas of
changing voltage.

The peak amplitude of the changing voltages would signify the magnitude of
the resonance. Large resonance would be a faster change in frequency since
the mechanical amplitude of the disturbance has to happen in the same amount
of time...

I'm thinking out loud here. I need to dig into some data sheets and see how
fast these IC's really are...

Thanks,

Dan

-----Original Message-----
From: Electronics_101@yahoogroups.com
[mailto:Electronics_101@yahoogroups.com] On Behalf Of Jan Kok
Sent: Saturday, March 02, 2013 12:09 AM
To: Electronics_101@yahoogroups.com
Subject: Re: [Electronics_101] How to analyze change in frequency?

A couple of questions: How many teeth are there on the gear that the sensor
senses? And, is there some way to tell where the "0" position of the
flywheel is?

I would suggest buying or borrowing a digital storage oscilloscope with a
large memory (say 1 million samples or so) that can store waveform data to a
USB key.

Capture a bunch of waveforms from the sensor, and analyze them at your

You'll get a frequency modulated square wave from the sensor. The higher the
RPMs, the higher the frequency and the shorter the period of the square
wave.

Write a program to measure the period of each pulse, and plot the
frequencies of the pulses (inverse of periods). You can use the Excel graph

Does the graph of the frequencies look somewhat sinusoidal? That graph would
show the frequency modulation of the sensor square wave, and the cycles in
the sinusoid should correlate with the power strokes of the pistons. The
sinusoid should have much larger amplitude when the engine is under load.

The amplitude should also be larger at certain RPMs where there is
rotational resonance, compared with RPMs where there is no resonance.
However, my hunch is that it will be hard to see any resonance, for several
reasons: the resonant frequency may be higher than the sensor squarewave
frequency, friction in the engine and drivetrain will act to dampen any
resonance, and the resonance may be buried in random noise.

Sounds like a fun project. I'll be curious to hear how it turns out.

[Non-text portions of this message have been removed]

------------------------------------

• The MicroChip TC9400 will convert frequency to voltage. A bit pricey at ~\$9, but it will do the job. Kerry
Message 8 of 23 , Mar 2, 2013
The MicroChip TC9400 will convert frequency to voltage. A bit pricey at
~\$9, but it will do the job.

Kerry

A6intruder@... wrote:
> Jan,
>
> The gear in question is the starter ring on the engine flywheel, so no real
> way to find a "zero" point on a rotation. I don't actually see a reason to
> find a zero since we are looking for the change in frequency not a specific
> position in the rotation.
>
> The gear has 157 teeth and the operational range is 650 RPM to about 5500
> RPM, that makes our base frequency range 1700hz to 14,391hz. I am hoping to
> eventually analyze and display/record the level of torsional resonance in
> real time. I know there are IC's that convert frequency to voltage, in this
> case we need to convert the change in frequency to a voltage...
>
> I am trying to figure out the basic sensor/processing/display/record process
> now on my car so that later I can use it to analyze another combination of
> engine/gearbox/load. I could end up with a toothed sensor on each segment
> of the combination looking for issues as I do the full range of testing.
>
> Thanks,
>
> Dan
>
> -----Original Message-----
> From: Electronics_101@yahoogroups.com
> [mailto:Electronics_101@yahoogroups.com] On Behalf Of Jan Kok
> Sent: Saturday, March 02, 2013 12:09 AM
> To: Electronics_101@yahoogroups.com
> Subject: Re: [Electronics_101] How to analyze change in frequency?
>
> A couple of questions: How many teeth are there on the gear that the sensor
> senses? And, is there some way to tell where the "0" position of the
> flywheel is?
>
> I would suggest buying or borrowing a digital storage oscilloscope with a
> large memory (say 1 million samples or so) that can store waveform data to a
> USB key.
>
> Capture a bunch of waveforms from the sensor, and analyze them at your
>
> You'll get a frequency modulated square wave from the sensor. The higher the
> RPMs, the higher the frequency and the shorter the period of the square
> wave.
>
> Write a program to measure the period of each pulse, and plot the
> frequencies of the pulses (inverse of periods). You can use the Excel graph
> wizard to plot your data.
>
> Does the graph of the frequencies look somewhat sinusoidal? That graph would
> show the frequency modulation of the sensor square wave, and the cycles in
> the sinusoid should correlate with the power strokes of the pistons. The
> sinusoid should have much larger amplitude when the engine is under load.
>
> The amplitude should also be larger at certain RPMs where there is
> rotational resonance, compared with RPMs where there is no resonance.
> However, my hunch is that it will be hard to see any resonance, for several
> reasons: the resonant frequency may be higher than the sensor squarewave
> frequency, friction in the engine and drivetrain will act to dampen any
> resonance, and the resonance may be buried in random noise.
>
> Sounds like a fun project. I'll be curious to hear how it turns out.
>
>
> [Non-text portions of this message have been removed]
>
>
>
> ------------------------------------
>
>
>
>
>
>
>
• Solution... Use a hall sensor to detect the teeth moving nearby. Capacitor couple and amplify the signal. You will then get pulses as each of the teeth go by.
Message 9 of 23 , Mar 2, 2013
Solution...

Use a hall sensor to detect the teeth moving nearby. Capacitor couple and amplify the signal. You will then get pulses as each of the teeth go by. Convert from frequency to voltage and then monitor the voltage.

Derek Koonce
DDK Interactive Consulting Services

--- In Electronics_101@yahoogroups.com, "A6intruder@..." <A6intruder@...> wrote:
>
> Jeremy,
>
> I am trying set up a system that will allow me to sense when rotational
> resonance occurs. Just as you are measuring rigid structures with
> vibrations looking for the frequency that causes that structure to resonate,
> I am looking to find where a given engine/geabox/load combination will have
> torsional resonance. Every rotational system will have at least one point
> of resonance - hopefully not in the operational range.
>
> The end goal is to figure out the basic sensor/processing/display/record
> process now on my car so that later I can use it to analyze another
> combination of engine/gearbox/load. I could end up with a toothed sensor on
> each segment of the combination looking for issues as I do the full range of
> testing.
>
> I know this technology already exists but it is priced for manufacturers
> that are developing drivelines for any number of automotive and industrial
> uses. That puts it out of my hobby price range. Thus the "DIY-with-help"
> approach.
>
> Thanks,
>
> Dan
>
> -----Original Message-----
> From: Electronics_101@yahoogroups.com
> [mailto:Electronics_101@yahoogroups.com] On Behalf Of jeremy youngs
> Sent: Saturday, March 02, 2013 1:31 AM
> To: Electronics_101@yahoogroups.com
> Subject: Re: [Electronics_101] How to analyze change in frequency?
>
> as a certified master technician we use accelerometers and electronic
> vibration analzers for noise vibration and harshness concerns.
> i really am uncertain about the specifics of your investigation , the
> question is what is your goal?
> to lovate the frequency you have to know what component you are looking for
> . then what is the magnitude of the vibration ( i.e 1 g is going to be
> something heavy. is it a first second or third order vibration?
>
> or are you simply wanting to quantify the torque impulse and it f of o?
> in essence im pretty handy and trained with this by two manufacturers but i
> I presume that you are looking at crankshaft acceleration to try to
> calculate the impulse frequency?
> are you looking for a specific vibration? is there a concern with the auto
> or are you just researching dynamic phenomena ?
>
> --
> jeremy youngs
>
>
> [Non-text portions of this message have been removed]
>
>
>
> ------------------------------------
>
>
• I once monitored the torsional vibration at the tail end of a dynamometer using the frequency from a 60-tooth wheel mounted there. The engine torque pulses
Message 10 of 23 , Mar 2, 2013
I once monitored the torsional vibration at the tail end of a dynamometer using the frequency from a 60-tooth wheel mounted there. The engine torque pulses were clearly visible in the signal. I used a frequency to voltage converter (this was many years ago :) probably an LM2907, and just watched the output on a slow oscilloscope sweep. Most data were analysed by hand.

Donald.
--
*Plain Text* email -- it's an accessibility issue
() no proprietary attachments; no html mail
/\ ascii ribbon campaign - <www.asciiribbon.org>

----- Original Message -----
> From: A6intruder@...
> To: "Electronics 101" <Electronics_101@yahoogroups.com>
> Sent: Saturday, March 2, 2013 9:48:13 AM
> Subject: RE: [Electronics_101] How to analyze change in frequency?
> Jan,
>
> The gear in question is the starter ring on the engine flywheel, so no
> real
> way to find a "zero" point on a rotation. I don't actually see a
> reason to
> find a zero since we are looking for the change in frequency not a
> specific
> position in the rotation.
>
> The gear has 157 teeth and the operational range is 650 RPM to about
> 5500
> RPM, that makes our base frequency range 1700hz to 14,391hz. I am
> hoping to
> eventually analyze and display/record the level of torsional resonance
> in
> real time. I know there are IC's that convert frequency to voltage, in
> this
> case we need to convert the change in frequency to a voltage...
>
> I am trying to figure out the basic sensor/processing/display/record
> process
> now on my car so that later I can use it to analyze another
> combination of
> engine/gearbox/load. I could end up with a toothed sensor on each
> segment
> of the combination looking for issues as I do the full range of
> testing.
>
> Thanks,
>
> Dan
>
> -----Original Message-----
> From: Electronics_101@yahoogroups.com
> [mailto:Electronics_101@yahoogroups.com] On Behalf Of Jan Kok
> Sent: Saturday, March 02, 2013 12:09 AM
> To: Electronics_101@yahoogroups.com
> Subject: Re: [Electronics_101] How to analyze change in frequency?
>
> A couple of questions: How many teeth are there on the gear that the
> sensor
> senses? And, is there some way to tell where the "0" position of the
> flywheel is?
>
> I would suggest buying or borrowing a digital storage oscilloscope
> with a
> large memory (say 1 million samples or so) that can store waveform
> data to a
> USB key.
>
> Capture a bunch of waveforms from the sensor, and analyze them at your
>
> You'll get a frequency modulated square wave from the sensor. The
> higher the
> RPMs, the higher the frequency and the shorter the period of the
> square
> wave.
>
> Write a program to measure the period of each pulse, and plot the
> frequencies of the pulses (inverse of periods). You can use the Excel
> graph
> wizard to plot your data.
>
> Does the graph of the frequencies look somewhat sinusoidal? That graph
> would
> show the frequency modulation of the sensor square wave, and the
> cycles in
> the sinusoid should correlate with the power strokes of the pistons.
> The
> sinusoid should have much larger amplitude when the engine is under
>
> The amplitude should also be larger at certain RPMs where there is
> rotational resonance, compared with RPMs where there is no resonance.
> However, my hunch is that it will be hard to see any resonance, for
> several
> reasons: the resonant frequency may be higher than the sensor
> squarewave
> frequency, friction in the engine and drivetrain will act to dampen
> any
> resonance, and the resonance may be buried in random noise.
>
> Sounds like a fun project. I'll be curious to hear how it turns out.
>
>
> [Non-text portions of this message have been removed]
>
>
>
> ------------------------------------
>
>
>
>
>
>
>
> ------------------------------------
>
>
>
>
• Modern electronics ignition / fuel injected engines have crankshaft position sensors http://en.wikipedia.org/wiki/Crankshaft_position_sensor I don t know how
Message 11 of 23 , Mar 2, 2013
Modern electronics ignition / fuel injected engines have crankshaft position sensors

http://en.wikipedia.org/wiki/Crankshaft_position_sensor

I don't know how this will help the subject but I though I'll put it out there.

Jong

On Mar 2, 2013, at 4:48 AM, "A6intruder@..." <A6intruder@...> wrote:

> Jan,
>
> The gear in question is the starter ring on the engine flywheel, so no real
> way to find a "zero" point on a rotation.
>

[Non-text portions of this message have been removed]
• ... + 1 for the hall sensor idea. A lot of motorcycles figure out their timing by a hall sensor on the flywheel/stator rotor. -- Aaron Turner
Message 12 of 23 , Mar 2, 2013
On Sat, Mar 2, 2013 at 9:05 AM, Derek <derek@...> wrote:

> **
>
>
> Solution...
>
> Use a hall sensor to detect the teeth moving nearby. Capacitor couple and
> amplify the signal. You will then get pulses as each of the teeth go by.
> Convert from frequency to voltage and then monitor the voltage.
>

+ 1 for the hall sensor idea. A lot of motorcycles figure out their timing
by a hall sensor on the flywheel/stator rotor.

--
Aaron Turner
http://tcpreplay.synfin.net/ - Pcap editing and replay tools for Unix &
Windows
Those who would give up essential Liberty, to purchase a little temporary
Safety, deserve neither Liberty nor Safety.
-- Benjamin Franklin
"carpe diem quam minimum credula postero"

[Non-text portions of this message have been removed]
• hall sensors, variable reluctance sensors, are all rotational speed tracking devices the only thing useful from these to be gained is crank position and phase
Message 13 of 23 , Mar 2, 2013
hall sensors, variable reluctance sensors, are all rotational speed
tracking devices the only thing useful from these to be gained is crank
position and phase
this will help but it is NOT all the input required an acceleometer is
absolutely essential to this application as is a transducer for measuring
the frquency of the specified noise/vibration/harhness (nvh) conern.
this is not a rogue path to follow without guidance as you have to know
WHAT you are searching for as ALL driveline noise is transmitted through
the crankshaft as they are all mechanically coupled
this pdf li nk should help
http://users.757.org/~ken/T/Toyota%20Training%20472%20-%20NVH.pdf
--
jeremy youngs

[Non-text portions of this message have been removed]
• It sounds like Jeremy has a lot of experience with this type of stuff, but when reading it I immediately thought of using an FFT function on the scope to read
Message 14 of 23 , Mar 3, 2013
It sounds like Jeremy has a lot of experience with this type of stuff, but when reading it I immediately thought of using an FFT function on the scope to read out the frequency of the sensor signal directly. On my cheap \$400 DSO, it has built in FFT functions. I would hook your sensor up to the scope, and then you could view both the fundamental and secondary frequencies directly on the scope. There may be some chips out there like certain DSP's that can do the FFT for you without the use of an expensive Digital Oscilloscope but I don't know any off hand.

--- In Electronics_101@yahoogroups.com, "A6intruder@..." <A6intruder@...> wrote:
>
> I have a long term project which will seek to measure the degree of
> torsional resonance that exists at the flywheel of my car engine. I have a
> Hall effect sensor mounted to sense the flywheel gear teeth going by the
> sensor. As the engine goes under load the power pulses actually cause
> minute changes in the frequency of those gear teeth going past the sensor.
> On a V-8 engine there are four power pulses per revolution of the flywheel.
> As the engine loads up and the power pulses get larger this minute change in
> frequency four times per revolution will grow in amplitude. If the
> mechanical drive system (engine & transmission) hit a point of rotational
> resonance those minute changes in frequency will become significantly
> larger.
>
> So my question is basically how do I analyze this frequency?
>
> I have a regular oscilloscope. I could monitor the basic waveform but how
> would I detect the minute change of frequency? Would the waveform shrink
> and expand if I get the time base just so?
>
> If has been 30 years since college calc classes, am I seeking a derivative
> or integral of the basic frequency? Can I achieve this with a certain mode
> on the oscilloscope (I 'm a very basic scope dope)?
>
> Ideally once I learn the terms and analysis I am trying to achieve I would
> want to collect this data during operation and record it on my laptop and
> sync it to the engine parameter data that I am already able to
> collect/record.
>
> What are the right terms/process I am looking for?
>
> The sensor is already mounted in the car.
>
> Data sheet:
> http://media.digikey.com/pdf/Data%20Sheets/Hamlin%20PDFs/55505.pdf
>
> I thought John Popelish or someone else in the group gave me the right
> process a few years ago but I couldn't find that post in the archives just
> now.
>
> Thanks,
>
> Dan Nicoson
>
• ... FFT will work when the engine is running at a steady speed. But FFT won t work when the engine is accelerating or there is a variable load on the engine.
Message 15 of 23 , Mar 3, 2013
On 3/3/2013 7:22 AM, Tim S wrote:
>
>
> It sounds like Jeremy has a lot of experience with this type of stuff,
> but when reading it I immediately thought of using an FFT function on
> the scope to read out the frequency of the sensor signal directly. On
> my cheap \$400 DSO, it has built in FFT functions. I would hook your
> sensor up to the scope, and then you could view both the fundamental
> and secondary frequencies directly on the scope. There may be some
> chips out there like certain DSP's that can do the FFT for you without
> the use of an expensive Digital Oscilloscope but I don't know any off
> hand.
>

FFT will work when the engine is running at a steady speed. But FFT
won't work when the engine is accelerating or there is a variable load
on the engine.

T^he other Howard

>
> --- In Electronics_101@yahoogroups.com
> <mailto:Electronics_101%40yahoogroups.com>, "A6intruder@..." wrote:
> >
> > I have a long term project which will seek to measure the degree of
> > torsional resonance that exists at the flywheel of my car engine. I
> have a
> > Hall effect sensor mounted to sense the flywheel gear teeth going by the
> > sensor. As the engine goes under load the power pulses actually cause
> > minute changes in the frequency of those gear teeth going past the
> sensor.
> > On a V-8 engine there are four power pulses per revolution of the
> flywheel.
> > As the engine loads up and the power pulses get larger this minute
> change in
> > frequency four times per revolution will grow in amplitude. If the
> > mechanical drive system (engine & transmission) hit a point of
> rotational
> > resonance those minute changes in frequency will become significantly
> > larger.
> >
> > So my question is basically how do I analyze this frequency?
> >
> > I have a regular oscilloscope. I could monitor the basic waveform
> but how
> > would I detect the minute change of frequency? Would the waveform shrink
> > and expand if I get the time base just so?
> >
> > If has been 30 years since college calc classes, am I seeking a
> derivative
> > or integral of the basic frequency? Can I achieve this with a
> certain mode
> > on the oscilloscope (I 'm a very basic scope dope)?
> >
> > Ideally once I learn the terms and analysis I am trying to achieve I
> would
> > want to collect this data during operation and record it on my
> laptop and
> > sync it to the engine parameter data that I am already able to
> > collect/record.
> >
> > What are the right terms/process I am looking for?
> >
> > The sensor is already mounted in the car.
> >
> > Data sheet:
> > http://media.digikey.com/pdf/Data%20Sheets/Hamlin%20PDFs/55505.pdf
> >
> > I thought John Popelish or someone else in the group gave me the right
> > process a few years ago but I couldn't find that post in the
> archives just
> > now.
> >
> > Thanks,
> >
> > Dan Nicoson
> >
>
>

[Non-text portions of this message have been removed]
• ... I think an FFT could be made to work, but it would take a bit of up front computation. The sample period for the FFT would have to be exactly one or an
Message 16 of 23 , Mar 3, 2013
On 03/03/2013 04:37 PM, Howard Hansen wrote:
>
> FFT will work when the engine is running at a steady
> speed. But FFT won't work when the engine is
> accelerating or there is a variable load on the engine.

I think an FFT could be made to work, but it would take a
bit of up front computation.

The sample period for the FFT would have to be exactly one
or an integer multiple of rotations, and that would take
tooth counting.

The intervals tooth to tooth would have to be measured by
some high frequency counter and then reciprocled to turn
period into velocity samples, every tooth.

Then that sequence of velocities of variable periods would
have to be resampled, to make interpolated velocity samples,
uniformly spaced in time, through one (or multiple rotations).

At that point, a normal FFT could be calculated that would
produce the magnitudes and phases of the harmonics of the
rotational frequency.

There may be a few mathematical short cuts possible, that
would simply this process and still give useful results.
I'm not a good enough FFT mathematician to see them, without
a lot of digging.

Fortunately, you don't need and really cannot use the FFT of
every single rotation, so data could be taken for one
rotation and then the FFT calculated during the next few
rotations, with just tooth counting continuing, during that
process, so the next FFT could start on the same tooth, to
keep the phase information from jumping around. The running
average of several of these results could produce a usably
smooth display on some bar graph indicator.

--
Regards,

John Popelish
• ... on this project though if you dont have a magnitude of impulse (accelerometer) or the frequency of vibration (transducer pick up) one will not be able to
Message 17 of 23 , Mar 3, 2013
>
> john
>

> yes without going into the strategies and engine controller uses
> quadrature encoding in the crank sensor to monitor instantaneous crankshaft
> acceleration for misfire detection. which is exactly what you have outlined
> here.
>
on this project though if you dont have a magnitude of impulse
(accelerometer) or the frequency of vibration (transducer pick up) one
will not be able to properly calculate the data for proper fourier analysis
and will be lacking enough detail to have any substantial quantitative
results , you can measure 0 drive train harmonics by traking rotation or
acceleration ther is simply not enough information to calculate the other
polynomial variables. Simply in order to do what the o.p is aking 3 sensors
will be needed and the crank seneor would be very easy to just access at
the engine controller by simple back probing. arduino shields are available
with accelerometers and transducers availabe it is possible an electret
with enough frequency spectrum would work as well.

--
jeremy youngs

[Non-text portions of this message have been removed]
• Regarding using a Hall effect sensor to measure rotational resonance: it may work if the resonant frequency is substantially lower than the frequency of the
Message 18 of 23 , Mar 3, 2013
Regarding using a Hall effect sensor to measure rotational resonance: it
may work if the resonant frequency is substantially lower than the
frequency of the square wave that you get from the Hall effect sensor. If
the resonant frequency you're looking for is close to or higher than the
frequency of the sensor square wave (and actually, if there are strong
frequency components above the frequency of the square wave), then the Hall
effect sensor won't give you reliable info about the frequency variations,
i.e. small, high-frequency changes in the rotation rate.

It has to do with sampling theory. Suppose you use an A/D converter to
sample a 900 Hz sine wave at a 1000 samples per second rate. In 10 ms the
sine wave goes through 9 cycles, and you take 10 samples. The result is
that the samples, if plotted, look like a single cycle of a sine wave at
100 Hz. So 900 Hz looks like 100 Hz after sampling. The signal you are
sampling must not have frequency components greater than 1/2 the sampling
frequency, in order to avoid seeing spurious frequencies in the sampled
data.

With the Hall effect sensor, you can sample on the rising and falling edges
of the sensor square wave, and what you are sampling is the width of the
pulses. If you get 500 Hz from the sensor, then you can sample 1000 times
per second and the maximum resonant frequency you can see accurately is
below half that rate, i.e. less than 500 Hz.

On Sun, Mar 3, 2013 at 4:38 PM, jeremy youngs <jcyoungs76@...> wrote:

> >
> > john
> >
>
>
> > yes without going into the strategies and engine controller uses
> > quadrature encoding in the crank sensor to monitor instantaneous
> crankshaft
> > acceleration for misfire detection. which is exactly what you have
> outlined
> > here.
> >
> on this project though if you dont have a magnitude of impulse
> (accelerometer) or the frequency of vibration (transducer pick up) one
> will not be able to properly calculate the data for proper fourier analysis
> and will be lacking enough detail to have any substantial quantitative
> results , you can measure 0 drive train harmonics by traking rotation or
> acceleration ther is simply not enough information to calculate the other
> polynomial variables. Simply in order to do what the o.p is aking 3 sensors
> will be needed and the crank seneor would be very easy to just access at
> the engine controller by simple back probing. arduino shields are available
> with accelerometers and transducers availabe it is possible an electret
> with enough frequency spectrum would work as well.
>
>
>
> --
> jeremy youngs
>
>

[Non-text portions of this message have been removed]
• I believe he said the resonance he is looking for would be at 1/4 the rotational frequency. Why not measure the period between gear teeth, and compare how it
Message 19 of 23 , Mar 3, 2013
I believe he said the resonance he is looking for would be at 1/4 the rotational frequency.

Why not measure the period between gear teeth, and compare how it changes? It seems to me that what you are looking for is FM impressed on the pulses from the gear teeth.

Steve Greenfield AE7HD

--- In Electronics_101@yahoogroups.com, Jan Kok <jan.kok.5y@...> wrote:
>
> Regarding using a Hall effect sensor to measure rotational resonance: it
> may work if the resonant frequency is substantially lower than the
> frequency of the square wave that you get from the Hall effect sensor. If
> the resonant frequency you're looking for is close to or higher than the
> frequency of the sensor square wave (and actually, if there are strong
> frequency components above the frequency of the square wave), then the Hall
> effect sensor won't give you reliable info about the frequency variations,
> i.e. small, high-frequency changes in the rotation rate.
>
> It has to do with sampling theory. Suppose you use an A/D converter to
> sample a 900 Hz sine wave at a 1000 samples per second rate. In 10 ms the
> sine wave goes through 9 cycles, and you take 10 samples. The result is
> that the samples, if plotted, look like a single cycle of a sine wave at
> 100 Hz. So 900 Hz looks like 100 Hz after sampling. The signal you are
> sampling must not have frequency components greater than 1/2 the sampling
> frequency, in order to avoid seeing spurious frequencies in the sampled
> data.
>
> With the Hall effect sensor, you can sample on the rising and falling edges
> of the sensor square wave, and what you are sampling is the width of the
> pulses. If you get 500 Hz from the sensor, then you can sample 1000 times
> per second and the maximum resonant frequency you can see accurately is
> below half that rate, i.e. less than 500 Hz.
>
>
• I believe he said the resonance he is looking for would be at 1/4 the rotational frequency. Why not measure the period between gear teeth, and compare how it
Message 20 of 23 , Mar 3, 2013
I believe he said the resonance he is looking for would be at 1/4 the
rotational frequency.

Why not measure the period between gear teeth, and compare how it changes?
It seems to me that what you are looking for is FM impressed on the pulses
from the gear teeth.

Steve Greenfield AE7HD

steve i believe accel decel curve could be extrapolated via demodulation
as you suggest. however this would assume abunch of ommitted truths that
need reconciliation. just a few i wil miss them as i dont wish to write a
thesis here :)

1 this meaure is taken at the flywheel

2 what exactly are we looking for ? just the 8 impulses in 2 revolutions?

( the op hasnt been as forward on what he is looking to do here as would be
required)

3 there will be MANY resonances as anything attached to the crankshaft will
have a natural resonant frequency

4 there will be 1st and second order harmonics dominating this analysis

some means of isolating the suspect resonance is needed ( also to note what
IS the suspect resonance? this has not been defined !!!!)

(without definition i cannot help lead to a resolution :)

5 what are the particulars of the test method i e under load ? going down
the identifying of suspect parameters and will lead to different results

in short i think an arduino with accelerometer, transducer and ignition
pulse pick up is required. i believe in essence a 100 dollar eva could be
built to experiment.

if i were limited to only looking at one characteric the most info can be
derived with a transducer monted to the engine to get precise frequency
measurements to back calculate what component is causing the resonance.
further the second most important thing is being able to quantify the
magnitude of the imulse so you can identify which resonance is THE one that
you are looking for .

last is rotational tracking this is the least useful and the first
recommended as all other parameters can be calculated using the above two
methods, however being that crankshaft pulse is readily available to
backprobe at the ecm connector it would be very handy to compare these
three in a dso type arduino box and graph them in synchrony to paint the
whole picture and make the resonance readily identifiable.

What I really want to know is what exactly is the O.P. trying to do ie why
are we measuring resonances and comparing them to other vehicles?

As an auto tech i have never fixed a vehicle in my life :)

What i do is resolve concerns and i find this concern to not currently be
defined enough to resolve :)

( oh p.s rate of resolved conerns 97 percent in 7 yrs withh chrysler :)

--
jeremy youngs

[Non-text portions of this message have been removed]
• DOH !!!! :)) simple cheap effective had to remeber bak to the first nvh class here this is the down and dirty old school tool !!! ( maybe a little ot as its
Message 21 of 23 , Mar 3, 2013
DOH !!!! :))

simple cheap effective had to remeber bak to the first nvh class here
this is the down and dirty old school tool !!!
( maybe a little ot as its not electronic !!)
http://www.treysit.com/16.html

did i just answer my own post doh enjoy

--
jeremy youngs
• I am the OP and I have tried to be completely forward as to my goal, I stated this earlier: I am trying to figure out the basic
Message 22 of 23 , Mar 4, 2013
I am the OP and I have tried to be completely forward as to my goal, I
stated this earlier:

"I am trying to figure out the basic sensor/processing/display/record
process now on my car so that later I can use it to analyze another

The point is to be able to test a given combination of engine/gearbox/load
for destructive resonant frequencies in the full range of operation, engine
RPM varying from idle (700 RPM) to max RPM (5500 RPM). I figured my turbo
charged V-8 Mustang would be a good development test bed since I can load
the engine very heavily to see "normal" power pulses with this process
which aren't resonant in this parts combination. If there were a
destructive resonant frequency in the operational range it would surface as
a much larger magnitude than the regular power pulses from the engine, at
whatever frequency it exists.

So if we can see the "standard" power pulses with this process on my car
then when I bolt a similar engine to different gearbox and load (say a
propeller) on a test stand then I can monitor to see if the new combination
has an resonant frequency within the normal range of operation.

I really doubt my car has a resonant frequency in its range of operation as
I have flogged that car on the street and at the drag strip throughout the
full range of operation and it hasn't self-destructed.

Everyone has been very helpful on this discussion, thank you. I have no
doubt the information I am seeking can be determined from the gear teeth
sensor, it is just a matter of processing that data stream. Jan has pointed
out some of the numerical limitations to that process.

Thanks,

Dan

-----Original Message-----
From: Electronics_101@yahoogroups.com
[mailto:Electronics_101@yahoogroups.com] On Behalf Of jeremy youngs
Sent: Monday, March 04, 2013 12:19 AM
To: Electronics_101@yahoogroups.com
Subject: Re: [Electronics_101] Re: How to analyze change in frequency?

I believe he said the resonance he is looking for would be at 1/4 the
rotational frequency.

Why not measure the period between gear teeth, and compare how it changes?
It seems to me that what you are looking for is FM impressed on the pulses
from the gear teeth.

Steve Greenfield AE7HD

steve i believe accel decel curve could be extrapolated via demodulation as
you suggest. however this would assume abunch of ommitted truths that need
reconciliation. just a few i wil miss them as i dont wish to write a thesis
here :)

1 this meaure is taken at the flywheel

2 what exactly are we looking for ? just the 8 impulses in 2 revolutions?

( the op hasnt been as forward on what he is looking to do here as would be
required)

3 there will be MANY resonances as anything attached to the crankshaft will
have a natural resonant frequency

4 there will be 1st and second order harmonics dominating this analysis

some means of isolating the suspect resonance is needed ( also to note what
IS the suspect resonance? this has not been defined !!!!)

(without definition i cannot help lead to a resolution :)

5 what are the particulars of the test method i e under load ? going down
the road? idling in the driveway? these conditions will radically affect the
identifying of suspect parameters and will lead to different results

in short i think an arduino with accelerometer, transducer and ignition
pulse pick up is required. i believe in essence a 100 dollar eva could be
built to experiment.

if i were limited to only looking at one characteric the most info can be
derived with a transducer monted to the engine to get precise frequency
measurements to back calculate what component is causing the resonance.
further the second most important thing is being able to quantify the
magnitude of the imulse so you can identify which resonance is THE one that
you are looking for .

last is rotational tracking this is the least useful and the first
recommended as all other parameters can be calculated using the above two
methods, however being that crankshaft pulse is readily available to
backprobe at the ecm connector it would be very handy to compare these three
in a dso type arduino box and graph them in synchrony to paint the whole
picture and make the resonance readily identifiable.

What I really want to know is what exactly is the O.P. trying to do ie why
are we measuring resonances and comparing them to other vehicles?

As an auto tech i have never fixed a vehicle in my life :)

What i do is resolve concerns and i find this concern to not currently be
defined enough to resolve :)

( oh p.s rate of resolved conerns 97 percent in 7 yrs withh chrysler :)

--
jeremy youngs

[Non-text portions of this message have been removed]

------------------------------------

• Hello Dan, A popular Frequency to Voltage Converter IC is the LM2907-N/LM2917-N IC. See: A recommended
Message 23 of 23 , Mar 6, 2013
Hello Dan,

A popular Frequency to Voltage Converter IC is the LM2907-N/LM2917-N
IC. See: <http://www.ti.com/lit/ds/snas555b/snas555b.pdf>
A recommended application circuit is Figure 16 on page 14. To be able
to detect variations in frequency you need a frequency to voltage
converter with a low output ripple. Hence this is why I recommended a
frequency to voltage converter with 2 pole Butterworth Filter to reduce
ripple.

To select component values for the Butterworth Filter one needs to know
the range of possible resonant frequencies. What is the range of
possible resonant frequencies? My guess is 43 to 367 Hz. This guess
comes from an engine RPM range of 650 to 5500 and 4 power pulses per
revolution.

The other Howard

On 3/2/2013 9:49 AM, A6intruder@... wrote:
>
> OK,
>
> Just answering Jan and Jeremy got me thinking.
>
> Without looking at any data sheets let's say we have an IC that converts
> frequency to voltage. The Hall effect sensor puts out a square wave signal
> with equal high and low durations. So each time the sensor waveform
> crosses
> a certain voltage (say 2 volts) it allows the IC to determine
> frequency and
> output the corresponding signal voltage.
>
> With a completely smooth turning gear wheel at a steady RPM we will get a
> very steady signal voltage. Any time the frequency starts to vary the
> signal voltage will go up or down. In the case of four power pulses per
> revolution the IC should see four areas of changing voltage in each
> rotation
> of the gear wheel. If you hooked an oscilloscope to the output of this IC
> you would see a steady horizontal line for a steady RPM with no power
> pulses. With power pulses you would start to see those four areas of
> changing voltage.
>
> The peak amplitude of the changing voltages would signify the magnitude of
> the resonance. Large resonance would be a faster change in frequency since
> the mechanical amplitude of the disturbance has to happen in the same
> amount
> of time...
>
> I'm thinking out loud here. I need to dig into some data sheets and
> see how
> fast these IC's really are...
>
> Thanks,
>
> Dan
>
> -----Original Message-----
> From: Electronics_101@yahoogroups.com
> <mailto:Electronics_101%40yahoogroups.com>
> [mailto:Electronics_101@yahoogroups.com
> <mailto:Electronics_101%40yahoogroups.com>] On Behalf Of Jan Kok
> Sent: Saturday, March 02, 2013 12:09 AM
> To: Electronics_101@yahoogroups.com
> <mailto:Electronics_101%40yahoogroups.com>
> Subject: Re: [Electronics_101] How to analyze change in frequency?
>
> A couple of questions: How many teeth are there on the gear that the
> sensor
> senses? And, is there some way to tell where the "0" position of the
> flywheel is?
>
> I would suggest buying or borrowing a digital storage oscilloscope with a
> large memory (say 1 million samples or so) that can store waveform
> data to a
> USB key.
>
> Capture a bunch of waveforms from the sensor, and analyze them at your
>
> You'll get a frequency modulated square wave from the sensor. The
> higher the
> RPMs, the higher the frequency and the shorter the period of the square
> wave.
>
> Write a program to measure the period of each pulse, and plot the
> frequencies of the pulses (inverse of periods). You can use the Excel
> graph
> wizard to plot your data.
>
> Does the graph of the frequencies look somewhat sinusoidal? That graph
> would
> show the frequency modulation of the sensor square wave, and the cycles in
> the sinusoid should correlate with the power strokes of the pistons. The
> sinusoid should have much larger amplitude when the engine is under load.
>
> The amplitude should also be larger at certain RPMs where there is
> rotational resonance, compared with RPMs where there is no resonance.
> However, my hunch is that it will be hard to see any resonance, for
> several
> reasons: the resonant frequency may be higher than the sensor squarewave
> frequency, friction in the engine and drivetrain will act to dampen any
> resonance, and the resonance may be buried in random noise.
>
> Sounds like a fun project. I'll be curious to hear how it turns out.
>
> [Non-text portions of this message have been removed]
>
> ------------------------------------
>