A PC as a tachometer

[sperho]

This thread is cool for several reasons, the most important one is because it inspired me to test an idea proposed by dbabcock awhile back about using a frequency analyzer and a microphone as a 2-stroke engine tachometer. My twist on the concept is that I took steps to see if a PC with audio analysis software would suffice. The short answer is, I think so. I started by analyzing the audio tracks of a few videos posted on Arborsite, because I don't have a tach.

See this post for an explanation of my method and more importantly, any willing and able volunteers with a tachometer willing to post the video I've requested??

Post in this thread, if so. Thanks! If it works, I'll try to locate some freeware that would allow anyone with a laptop with a microphone or mic input to do this easily.

 

[sperho]

Oh, why not make it easy to follow. Here's a repost from the other thread:

----------

Playing around with some of my audio software after seeing an ingenius idea (a duh, why didn't I think of that sort of one) proffered by dbabcock in a tach thread. I was shopping for a tach, then he went off and proposed the obvious: audio frequency analysis. It turns out that one of my hobbies is phonography (recording random stuff) and I have a full version of CoolEdit that does frequency analysis. So, playing the audio track from a digital video allows me to measure with what I believe to be reasonable accuracy the frequency of a saw's power pulses and thus RPM (multiply frequency in Hz by 60) of a saw by analyzing the audio waveform.

Stihl Crazy's video showed the first saw (346 EHP) maxing out at 13,200 out of the cut right before wood entry and in the cut, it was turning around 8,700.

The second saw (260 KRS) maxed out before wood entry at around 12,000 and in the cut, ran around 9,660.

So, assuming equal sharpness and similar pressure in the wood (hard to control), the 260 HAD to win because more cuts per second were occuring. The same force into wood is key here.

In both videos, I "looked" at 1 second windows of time over the course of the cut, looking for the approximate steady-state frequency to find the in-the-cut RPMs. The out of the cut RPMs (measured just before wood entry) are almost certainly not the max-RPM of the saws, because the videos weren't taken for tuning purposes, but to record a cutting session. (i.e. Not many people let their saw spool to maximum RPM and wait a second before starting a cut.) The precison of this method is about +/- 150 RPM because of the width of the primary frequency peak in the power spectrum of the audio track. That's plenty good for practical saw tuning in my opinion, but racers would probably be better served by a precision tachometer.

I would be VERY interested in someone posting a video of a few tuning runs and have them post it, withholding what their measured RPMs were. Then, I could analyze the sound and check my technique in a "blind" sort of way and see how they compared to a tachometer's measurement. Anyone up to the challenge?? I'm going to start a new thread with a reference to this post so that a broader audience reads this buried post.

While I used a full-blown audio editing package, I'm convinced that some freeware or very cheap shareware would be out there to allow this type of RPM measurement to be easily performed by anyone with a laptop handy...

 

[dtnodya]

Great thread in my opinion!! I have been thinking about this for a while ...

Someone that knows what their saw is running (RPM's), post a video if you can. I think this could be interesting. I'll find some software...

 

[Haywire Haywood]

Good idea, If you could rig some way to have a constant pressure on the saw in the cut to reduce that variable I can see it being a great tuning tool. A programmer could come up with a GUI for a freq counter that would display in real time and hold at peak.

I'd thought about just putting a long wire on my tach and having my wife monitor it from a couple steps away. I don't know how far the signal will carry down the wire tho.

Ian

 

[scottr]

Sperho , a lot of the videos that I've watched have a muffled sound , this makes me question the frequency response of the microphone and amplifier . What bandwidth is typical for the digital video/audio recorders ? What is the capability of your system to analyze the recorded audio ?

 

[ShoerFast]

Sperho , a lot of the videos that I've watched have a muffled sound , this makes me question the frequency response of the microphone and amplifier . What bandwidth is typical for the digital video/audio recorders ? What is the capability of your system to analyze the recorded audio ?

Good question!

But the parentage of error is negligible if there getting a 13K or so out of wood speed and a 7 - 9K in the wood response, with a tach on a saw in a cut is the same.

 

[sperho]

Sperho , a lot of the videos that I've watched have a muffled sound , this makes me question the frequency response of the microphone and amplifier . What bandwidth is typical for the digital video/audio recorders ? What is the capability of your system to analyze the recorded audio ?

Good question.

16,000 RPM = 250 cylinder explosions per second, or 250 Hz. Thus, a minimum of 500 Hz bandwidth is necessary to see 250 Hz (Nyquist criterion). A telephone line has more bandwidth that required...

Muffled sound only matters if you don't have good enough audio tools. My software runs a Fourier Transform to generate a power spectrum and things that are fairly distinct in frequency (the repetitive tone of the muffler) should, and in fact do, show up as spikes at the primary frequency and correspondly it's higher harmomics. Any computer these days has the ability to record a signal at 44.1 kHz (CD audio quality) and is tremendous overkill for picking up a useable chainsaw audio track. I do want to test the ability to accurately determine RPM with random audio quality (muffled sound included), though, so I hope someone is able to post up some videos with tached runs...

I'm not sure what you mean by "what are my capabilities". Do you have an application in mind? Chances are, CoolEdit can do it.

 

[scottr]

Good question.

16,000 RPM = 250 cylinder explosions per second, or 250 Hz. Thus, a minimum of 500 Hz bandwidth is necessary to see 250 Hz (Nyquist criterion). A telephone line has more bandwidth that required...

Muffled sound only matters if you don't have good enough audio tools. My software runs a Fourier Transform to generate a power spectrum and things that are fairly distinct in frequency (the repetitive tone of the muffler) should, and in fact do, show up as spikes at the primary frequency and correspondly it's higher harmomics. Any computer these days has the ability to record a signal at 44.1 kHz (CD audio quality) and is tremendous overkill for picking up a useable chainsaw audio track. I do want to test the ability to accurately determine RPM with random audio quality (muffled sound included), though, so I hope someone is able to post up some videos with tached runs...

I'm not sure what you mean by "what are my capabilities". Do you have an application in mind? Chances are, CoolEdit can do it.

Ok , you can find the the main sound pulse and ignore the other noise/frequencys and you can record CD audio quality on your computer . Now it would be interesting to know what the capabilities of the little digital video/audio recorder .

 

[klickitatsacket]

cool idea. I am interested in the software. It might not help with anything and then again I might find that it helps a lot. Never know until you play with it.

 

[sperho]

Ok , you can find the the main sound pulse and ignore the other noise/frequencys and you can record CD audio quality on your computer . Now it would be interesting to know what the capabilities of the little digital video/audio recorder .

I have an iRiver H340 with digital recording ability with up to 44.1 kHz data rates. Realistically, I bet nearly any digital recorder would work. Probably even the cheap little lecture recorders. I'll have to take a look at a few spec sheets. Do you have one in mind?

 

[Al Smith]

Well,if anybody could figure out the frequency involved ,Doug B would be the one. In addition,I would imagine he has already figured out how to employ band pass filters to more enhance certain rpm ranges.I wouldn't either be surprised if he has figured out how to use "junk" registers on ye olde 'puter to save certain rpm numbers for further comparisons regarding modifications to the engines.

Doug doesn't pop up too often but every so often,I sense his presence.

 

[scottr]

I have an iRiver H340 with digital recording ability with up to 44.1 kHz data rates. Realistically, I bet nearly any digital recorder would work. Probably even the cheap little lecture recorders. I'll have to take a look at a few spec sheets. Do you have one in mind?

sperho , I do not have a specific application or recorder in mind . What is the highest audio frequency that can be recorded by a 44.1 khz data rate ?

 

[sperho]

sperho , I do not have a specific application or recorder in mind . What is the highest audio frequency that can be recorded by a 44.1 khz data rate ?

22,050 Hz, which would be 1,320,000 RPM in a two stroke motor. One reason that ~40 kHz was chosen for digital audio was because that sample rate can record the upper limits of the human ear's frequency response... Any frequencies higher than that will be aliased to lower frequencies, so those are filtered out before recording.

 

[dtnodya]

Good question.

16,000 RPM = 250 cylinder explosions per second, or 250 Hz. Thus, a minimum of 500 Hz bandwidth is necessary to see 250 Hz (Nyquist criterion). A telephone line has more bandwidth that required...

Muffled sound only matters if you don't have good enough audio tools. My software runs a Fourier Transform to generate a power spectrum and things that are fairly distinct in frequency (the repetitive tone of the muffler) should, and in fact do, show up as spikes at the primary frequency and correspondly it's higher harmomics. Any computer these days has the ability to record a signal at 44.1 kHz (CD audio quality) and is tremendous overkill for picking up a useable chainsaw audio track. I do want to test the ability to accurately determine RPM with random audio quality (muffled sound included), though, so I hope someone is able to post up some videos with tached runs...

I'm not sure what you mean by "what are my capabilities". Do you have an application in mind? Chances are, CoolEdit can do it.

Here's the video that I've been using to try and replicate the tach (chaps video):

http://www.labonville.com/videos/v2.htm


I just can't figure out exactly what to do .... Anybody got any idea about Hz, like how to analyze??? I've done some half-assed attempts but don't know really know what the heck I'm doin... lol

I've been using S O N Y (it won't let you type that word for some reason .. weird) Vegas 6 for the video then Soundforge to analyze the audio ... still confused though.

 

[scottr]

22,050 Hz, which would be 1,320,000 RPM in a two stroke motor. One reason that ~40 kHz was chosen for digital audio was because that sample rate can record the upper limits of the human ear's frequency response... Any frequencies higher than that will be aliased to lower frequencies, so those are filtered out before recording.

Several guys here post video on youtube and they all sound muffled to me , they don't sound like the saw is same one that is cutting . The V8 and V Twin videos sound and look realistic . Could it be that some of the recorders are not capable of high enough sample rate to properly record the high rpm saws ?

 

[Stihl Crazy]

sperho, I posted a video in the original thread before I saw this one. Can you tell what the saw is turning in the wood?

 

[BobL]

I have analyzed the sound (frequency and volume) from a range of circular saw blades using a Mac program called SignalScope Pro. The reason for this was more for volume than rpm as I was assessing how effective the so called "low noise" circular saw blades are. It turns out they do work as claimed.

The image shows a screen capture from that program for one of my 12" standard blades.

Even though the image shows a very noisy signal, one can pick out the rated 3000 rpm (50 Hz) rotation (its the very first small peak in the spectrum) so I would imagine it could pick higher rpms provided the other noises were not too loud. The main 3000 Hz (not rpm) peak is the "fluting noise" of the teeth moving through the air and what makes your ears hurt.

The problem for a CS will also be its produce broad spectum sound which broaden any sound peaks. Nence it will probably not be possible to distinguish say 500 rpm in the 12000 rpm range, ie cannot distinguish between 11500 and 12000 rpm. If anyone is interested I can have a go at this later today and report back?

I recently bought an optical tachometer like this to use in measuring fan/motor speeds but I didn't think of using it for a CS .

It is supposed to be good up to 99,999 rpm with a resolution of 1 rpm when measuring above 1000 rpm. Of course this requires a flywheel measurements so it cannot be used in a cut.

 

[BobL]

Here's the frequency spectrum of the Labonville video in the middle of that section of video where they show the 12000 rpm reading on the tacho. It's a pity there is that melodramatic music playing at the same time.

Anyway - you can see the problem. Heaps of background noise and the main peak is at 400 Hz or 24000 rpm, although there does appear to be a peak at 200 Hz (and 800 as well)? I wouldn't bet on being able to interpret this in terms of definitive rpm but it might improve if that music wasn't there.

 

[Goat]

Now this is thinking outside the box...

I did some related work a while ago on a vaguely similar subject, except we were measuring the ion current across a spark plug immediately following a ignition event on 4 strokes.

My suggestion would be to couple the spark signal directly to soundcard. This way you don't have as much background noise to deal with.

The super cheap way would be by making wraps around your ignition wire as close to the spark plug as you can. The more wraps the higher the gain will be...you will have to tune the preamp on your recording device to make sure you have enough amplification and aren't clipping.

The more robust method is to use an inductive pickup from a timing light. Again the amount of gain, transients and noise will vary from pickup to pickup but it will give you a sharper signal trace compared to copper wire.

Lastly, a graph of rpm vs. time is a great thing to have. If you could estimate the inertial load on the saw or couple a know mass to it. You have yourself a chainsaw dyno! Even if you can't do the above it's a nice tool to take relative
measurements of your saw before & after a mod. Nice way to see if that muffler mod was worth it.

Just me 2 pence...

 

[sperho]

Thanks for pointing out that Labonville video, that is the kind of thing for which I'm looking to test the accuracy of this method.

In the Labonville video, I measure 12,000 RPM as indicated by the 200 Hz frequency of the primary audio signal emanating from the chainsaw. Not too bad of a correlation with the indicated RPM determined by the inductive tach. Below is the power spectrum with a Gaussian windowing function applied.

BobL - 400 Hz is the 2nd harmonic. I'm guessing that the window of time you are sampling is too broad and are picking up too much contamination from the soundtrack, thus the primary is more difficult for you to pick out. I sampled only about 1 second of the audio, so the main dominant audio signal is originating from the chainsaw. You can readily see that there is alot of power in the higher harmonics (even multiples of the primary) and these are the sounds that are easily heard by our ear and in some videos, are cut off, thus the saw sounding "muffled". Those signals are not required to determine the primary, but their repetitive nature helps identify the primary one to key in on. (I restricted my screenshot to the bandwith shown, but the harmonics are powerful out to 1000 Hz...)