Measuring RPM From Video

[heyduke]

I've been meaning to get around to measuring the rpm in this video of my 40cc GZ4000 clone cutting little white ash cookies with 0.043" chain:


I tried a couple of the cuts, but this was typical:
View attachment 433957
That's 12,700rpm.

nice! wish i could get that in a top handle.

 

[Chris-PA]

nice! wish i could get that in a top handle.

I bet you could come close - the GZ3500T is a similar strato engine design, only 35cc and a clamshell. It would probably respond to mods similarly to what I did here. It has removably transfer covers too. Add a 12" 0.043" bar.

It's the only clamshell I've wanted, though I have no need for one.

 

[heyduke]

I bet you could come close - the GZ3500T is a similar strato engine design, only 35cc and a clamshell. It would probably respond to mods similarly to what I did here. It has removably transfer covers too. Add a 12" 0.043" bar.

It's the only clamshell I've wanted, though I have no need for one.

clamshell is a good way to go for top handles. they're more compact, gives the designer some options for placing the oil and fuel tanks. but regarding the 40cc redmax. that is a beautiful saw. i like the way they extended the casting so the bar is mounted in magnesium instead of plastic. it's 30cc too small for me but a thing of beauty. the construction reminds me of my old yamaha 550 seca. that was a great asphalt scratcher.

 

[LegDeLimber]

First thing is to ALWAYS have your anti virus program up to date Before downloading or running anything.

I don't know if this program will run or run well in Linux.
SpecAn_2v9b.exe
http://www.techmind.org/audio/index.html#specanaly
then choose the "Real-time Spectrum Analyser" link.
After you have it loaded and ready, go to the the controls in the top left corner.
Now click/open "Vertical Scale" and choose "A-weighted 70db" from the drop down menu.
next click/open "Horizontal Scale" and choose "Logrithmic".

Having done that, now play this video.
" john Deere G "

Whilst he's cueing up for the hitch,
that wicked lope (oh and crank yer woofers too) is really intersting to see in the 50hz to about 125~ish range (the utill looks to cut off at 50hz).
You can watch that one cylinder hitting at a rate that triggers the 50hz band
yet some of it's spuratic firings cause a VERY visible whipping of the trace
going into the mid/upper 100's and pulling sometimes the 200 band up.
The other cyl keeps that 100~ish cadence going.

Now after all of that , take a look at all of the metallic noises, turning gears etc
that fill out the range over to the (video?) sound cutoff point
at around 7khz , with just a few spuratic bits into the 10k or so region.

My point here is ti give an easy to see/locat example of the engine firing and exhaust note
VS all of the other sounds of the machine.

Just hoping to help with getting a grip on this method of looking at things.

Oh and the tractor just flat out sounds cool as h### on a real set of speakers!!

 

[heyduke]

First thing is to ALWAYS have your anti virus program up to date Before downloading or running anything.

I don't know if this program will run or run well in Linux.
SpecAn_2v9b.exe
http://www.techmind.org/audio/index.html#specanaly
then choose the "Real-time Spectrum Analyser" link.
After you have it loaded and ready, go to the the controls in the top left corner.
Now click/open "Vertical Scale" and choose "A-weighted 70db" from the drop down menu.
next click/open "Horizontal Scale" and choose "Logrithmic".

Having done that, now play this video.
" john Deere G "

Whilst he's cueing up for the hitch,
that wicked lope (oh and crank yer woofers too) is really intersting to see in the 50hz to about 125~ish range (the utill looks to cut off at 50hz).
You can watch that one cylinder hitting at a rate that triggers the 50hz band
yet some of it's spuratic firings cause a VERY visible whipping of the trace
going into the mid/upper 100's and pulling sometimes the 200 band up.
The other cyl keeps that 100~ish cadence going.

Now after all of that , take a look at all of the metallic noises, turning gears etc
that fill out the range over to the (video?) sound cutoff point
at around 7khz , with just a few spuratic bits into the 10k or so region.

My point here is ti give an easy to see/locat example of the engine firing and exhaust note
VS all of the other sounds of the machine.

Just hoping to help with getting a grip on this method of looking at things.

Oh and the tractor just flat out sounds cool as h### on a real set of speakers!!

good to know that there's a free spectrum analyser ofr those running mr. bill's os. i have an uncle that used a two cylinder "johnny popper" until he died a few years ago. it was started by spinning a fly wheel on the side of the engine. no electric starter. everyone loves that sound. it's the polar opposite of a 372 working at 12k rpm.

 

[7sleeper]

Subscribed!

Now I just have to find out how to do it...

7

 

[Chris-PA]

Subscribed!

Now I just have to find out how to do it...

7

Somehow I missed this in the chaos of the last few days - is there anything I can do to help explain?

 

[Chris-PA]

One of the arguments thrown out by those who want to rubbish this technique has been that the various compression and conversion steps corrupt the data and introduce errors. So I decided to test this. I started with the raw HD video MTS file, and selected exactly 20s to 23s, and plotted the spectrum - the peak is at 182Hz (10,920rpm):


Then I downsampled it, converting the video to 360p .MOV format and the audio to 22,050kHz sample rate and 96kb/s bit rate. The measured frequency is unchanged:

Then I resampled that to a 360p .FLV format, and knocked the audio bit rate down to 32kb/s. The measured frequency is still the same:


Then I uploaded this to YouTube, downloaded it and looked at again - still unchanged:



By now this is a really crappy video with an audio spectrum limited to 11kHz (the lowest I could go with the program I was using), but still easily good enough to measure signals down below 300hz:

 

[heyduke]

One of the arguments thrown out by those who want to rubbish this technique has been that the various compression and conversion steps corrupt the data and introduce errors. So I decided to test this. I started with the raw HD video MTS file, and selected exactly 20s to 23s, and plotted the spectrum - the peak is at 182Hz (10,920rpm):
View attachment 493580

Then I downsampled it, converting the video to 360p .MOV format and the audio to 22,050kHz sample rate and 96kb/s bit rate. The measured frequency is unchanged:
View attachment 493581
Then I resampled that to a 360p .FLV format, and knocked the audio bit rate down to 32kb/s. The measured frequency is still the same:

View attachment 493582
Then I uploaded this to YouTube, downloaded it and looked at again - still unchanged:

View attachment 493583

By now this is a really crappy video with an audio spectrum limited to 11kHz (the lowest I could go with the program I was using), but still easily good enough to measure signals down below 300hz:

this is a good technique for measuring performance in the real world, not with a tach, saw on the bench. the software seems to be a reverse engineered free and open sorurce send-up of adobe's "audition." spectrum analysis is just one of many features. it's like having a studio and a laboratory in your lap top. i like the fact that one can measure the working rpm of any saw from a youtube vid. keep up the good work. you migfht try to compress and then normalize.

 

[Definitive Dave]

Video cameras and phones attenuate and/or clip frequencies.
I am reformed hd audio geek.
I am skeptical that a decent capture can even be made for analysis with these devices.
I am reading with interest.

My preferred mics - Schoeps mk4, mk41, mk21
My DAW runs wave lab
My DVW runs Vegas professional
My old school tape deck is an upgraded Nakamichi Dragon

 

[heyduke]

Video cameras and phones attenuate and/or clip frequencies.
I am reformed hd audio geek.
I am skeptical that a decent capture can even be made for analysis with these devices.
I am reading with interest.

My preferred mics - Schoeps mk4, mk41, mk21
My DAW runs wave lab
My DVW runs Vegas professional
My old school tape deck is an upgraded Nakamichi Dragon

i respect, applaud, your scepticism, i'm not particularly interested in your collection of hardware/software. if you want to produce empirical data to support your hypotheses, do it. my experience has been that using the spectrum analyser found in audacity is just as accurate, probably more accurate, than my wireless tachometer , which i am sure has a poorer detector and amplifier than my assorted cameras. this technique does have the advantage of being able to measure performance in the field rather than on the work bench. you have a better way, let us know. always looking for improvement.

 

[Chris-PA]

Video cameras and phones attenuate and/or clip frequencies.

Yes they can, but clipping or attenuation are not functions that alter the frequency of the fundamental, which is the only information of interest here. If you clip a sine wave you produce harmonics but do not change the fundamental frequency (lots of harmonics already). Attenuation reduces amplitude but does not change the fundamental frequency.

Microphones will either respond or not to the frequencies of interest, but cannot alter the fundamental frequency - and since we're talking about 200Hz, of course just about anything will work.

I've got a nice Nakamichi CR-1A that still sounds great. These days I'm learning about tube amp designs and changing out pickups in my son's guitar. He's saving up for an Axe-Fx II, and then we'll have some signal processing fun.

 

[heyduke]

Yes they can, but clipping or attenuation are not functions that alter the frequency of the fundamental, which is the only information of interest here. If you clip a sine wave you produce harmonics but do not change the fundamental frequency (lots of harmonics already). Attenuation reduces amplitude but does not change the fundamental frequency.

Microphones will either respond or not to the frequencies of interest, but cannot alter the fundamental frequency - and since we're talking about 200Hz, of course just about anything will work.

I've got a nice Nakamichi CR-1A that still sounds great. These days I'm learning about tube amp designs and changing out pickups in my son's guitar. He's saving up for an Axe-Fx II, and then we'll have some signal processing fun.

of course microphones attenuate signals. that's their job. that's why you need a different mic for your stand-up base than for your harmonica. clipping should be controlled by the gain on the mic's preamp. that being said, if you listen to a decent recording of a saw cutting cookies from a good digital camera, you'll hear more attenuation than distortion. seems like that means the camera's audio circuit is doing it's job properly. one technique i use is to set my zoom lens fairly long. two advantages, i get less chips on the camera and sound is attenuated naturally, by distance (amplitude=inverse of the square.)

 

[LegDeLimber]

http://ptolemy.eecs.berkeley.edu/eecs20/week13/aliasing.html
I've been waiting to see if anyone would mention this.
For a few years, It pretty well drove me away from watching any wildlife shows on TV.
Recorded Wolf howls were just teeth aching because of it.

Lately, I'm a bit too tired to get into the discussion properly.
But I am a little curious of everyones thoughts about it.

 

[heyduke]

http://ptolemy.eecs.berkeley.edu/eecs20/week13/aliasing.html
I've been waiting to see if anyone would mention this.
For a few years, It pretty well drove me away from watching any wildlife shows on TV.
Recorded Wolf howls were just teeth aching because of it.

Lately, I'm a bit too tired to get into the discussion properly.
But I am a little curious of everyones thoughts about it.

ok, someone with more expertise than me should jump in on this, but the link cited refers to problems encountered with low (8 khz) sampling rates. it refers to nyquist who developed a rule for minimum sampling rates to produce desired quality. 8 khz may be adequate for telephonic applications which don't need to be sensitive to freqs. above 3khz. for high quality sampling, rates should be closer to or above 40khz. the basic rule, the nyquist rate, is that the sampling rate should be double the maximum frequency to be sampled. again, an 8khz sampling rate should be adequate for telephonic voice communication. some audio engineer on the site can slam me now. i'm just an old cowboy.

what i wonder is if you could improve the resolution of the spectrum analysis by decreasing the sampling rate to say, 800, to lose all those nasty harmonics, chris-pa?

 

[Chris-PA]

of course microphones attenuate signals. that's their job. that's why you need a different mic for your stand-up base than for your harmonica. clipping should be controlled by the gain on the mic's preamp. that being said, if you listen to a decent recording of a saw cutting cookies from a good digital camera, you'll hear more attenuation than distortion. seems like that means the camera's audio circuit is doing it's job properly. one technique i use is to set my zoom lens fairly long. two advantages, i get less chips on the camera and sound is attenuated naturally, by distance (amplitude=inverse of the square.)

Most of these cameras have some sort of automatic gain control, but I always set my camera at a distance and use the zoom too. Still, the clipping doesn't effect the frequency/rpm measurement.

http://ptolemy.eecs.berkeley.edu/eecs20/week13/aliasing.html
I've been waiting to see if anyone would mention this.
For a few years, It pretty well drove me away from watching any wildlife shows on TV.
Recorded Wolf howls were just teeth aching because of it.

Lately, I'm a bit too tired to get into the discussion properly.
But I am a little curious of everyones thoughts about it.

Aliasing is always a potential issue in any sampling system. Frequencies in the signal above half the sample rate can fold back down into the range you are interested in, and once aliased they can no longer be separated from the real signal. Here the sample rate is likely 44.1kHz, and the frequency range we're interested in is only 150-250Hz. So a signal would have to be very close to 44kHz to fold back down to 150-250Hz. Also, even if some high frequency signal folded down to this range, the real signal would still be there - just some other signal would be there too. If you look at the plots above the frequency spike are pretty narrow, making it quite unlikely to be fooled by this.

However, most such recorders have anti aliasing filters to greatly attenuate any frequencies above half the sample rate. You can see from the plots above that even when I down sampled to 22kHz there was no sign of aliasing.

what i wonder is if you could improve the resolution of the spectrum analysis by decreasing the sampling rate to say, 800hz, to lose all those nasty harmonics, chris-pa?

The higher the sample rate the better, and those harmonics don't matter. In audacity you can use the equalizer to apply a narrow bandpass filter (leaving the range of interest alone) which gets rid of most of it.

 

[heyduke]

.

The higher the sample rate the better, and those harmonics don't matter. In audacity you can use the equalizer to apply a narrow bandpass filter (leaving the range of interest alone) which gets rid of most of it.

right, my mistake the harmonics would just show up as aliased trash, folded back into the freq. band of interest. but i'd like to try putting the filtering on the front end, effectively a mic that couldn't hear anything over 300hz.

 

[Chris-PA]

This method of measuring rpm from video has generated a large amount of whining, crying and nay saying, claims that it's dumb and doesn't work, and lots of personal attacks on me for using it. So I thought I would analyze this video from one of my more vocal critics, http://www.arboristsite.com/community/members/jmssaws.135029/:



The claim made was that this saw is turning over 13k in the cut, along with the challenge to prove him a liar. Challenger accepted. So let's see:



That's 217Hz * 60 = 13,000rpm. Right on the nose and just as claimed. Therefore jmssaws is most certainly not lying (not that I ever claimed any such thing).

But that leaves us with a quandary: My method matches the claimed rpm exactly, and so it is obviously not full of **** after all. But equally obviously, it means that some other folks most certainly are full of ****, and have been all along.

Thanks for playing. Next time, try not to throw such a soft pitch.

 

[Andyshine77]

Chris sometimes BS is just easier to except than actual facts. Most have been conditioned from day one to believe complete and utter BS, so really I expect nothing else.

Sent from my HTC One_M8 using Tapatalk

 

[heyduke]

This method of measuring rpm from video has generated a large amount of whining, crying and nay saying, claims that it's dumb and doesn't work, and lots of personal attacks on me for using it. So I thought I would analyze this video from one of my more vocal critics, http://www.arboristsite.com/community/members/jmssaws.135029/:



The claim made was that this saw is turning over 13k in the cut, along with the challenge to prove him a liar. Challenger accepted. So let's see:

View attachment 510161

That's 217Hz * 60 = 13,000rpm. Right on the nose and just as claimed. Therefore jmssaws is most certainly not lying (not that I ever claimed any such thing).

But that leaves us with a quandary: My method matches the claimed rpm exactly, and so it is obviously not full of **** after all. But equally obviously, it means that some other folks most certainly are full of ****, and have been all along.

Thanks for playing. Next time, try not to throw such a soft pitch.

nice work, spectrum analysis is a valuable tool for everything from nuclear magnetic resonance to audio frequency analysis (which is what we are dealing with here.) even without doing spectrum analysis i find i can hear my saw better on a good audio recording (nikon or canon) than when i'm just listening to my saw and thinking about pussycats. i don't have any saw that runs 13k in the cut but they all make money. ya gotta wonder how long that saw will last.