frequency analyser

[Obiw4n]

Hey!

Another work lately:

Shazam got problems with scores displayed on his counters, a lot of 14600+ but he asked me to check this with the spectro.

here are the results:



The first peaks are shazam's 14600+ runs, the one before the last is a 14100+ from the force and the last is a 14950+ from Akis.


Indeed shazam was right in his doubts, his scores are in fact below 14k. He said it must be his powerball which is too "magnetic" because he tried a lot of counters on the same powerball with the same results....he already got this problem a year ago...

Seems weird and I begin to worry on the reliability of the top 20 powerballers, don't know if one of them also got faulty powerballs....
_________________
Help Gyrotek when you buy your Powerballs, use our banner: International, France, Thanks!!

[Publicités]

[Adrena1in]

Such a cool tool that is...wish I knew how to use it properly. What settings to you put everything to to capture an entire 90s run, like the graphs you posted earlier in this thread??
_________________
250Hz: (L/R)
-> 13603/13785
<- 14026/14974
Dual - 26659
90 - 18705 - 16865/18013
60 - 12206 - 11089/12283
30 - 6383 - 6415/6796

350Hz:
-> 11688/11409
<- 12238/13053
90 - 15799 - 13873/15407
60 - 10597 - 10546/10649
30 - 5362 - 5398/5895

[mchap77]

As far as i can tell my submission is accurate....






Adrena1in, to answer your question if you are doing this from a wave file then you can only change 2 things: the points per transform should be 4096. and to be able to capture a longer wav file like a 90's run you need to have the speed(FFT's per sec) set pretty low like less than 10. You can also make the Frequency sepectrum window bigger to allow it to have more time in the window before it writes over it self. If using the mic, you can change all the options but basic ones would be to use 8 bit per sample and a sampling frequency of 11025, and same as i mentioned before for the other options. you can play with these setting to make it look nicer, but the ones i gave you seemed to work pretty good. i hope this helps
_________________
MAX SPEED:
Right: 15188 RPM (OCT/05)
Left: 14159 RPM (OCT/05)
ENDURANCE:
30 (R)6204 (L)6002
60 (L)10337 (R)11123
90 (L)15047(L+R)17509

[Adrena1in]

Thanks Mark. I've played around with the settings many times, and have never been able to get the slider to move slow enough so that it takes a minute and a half to move across the screen...the most I've managed to capture is around 40 seconds of a 2-minute WAV file.

I know Obi recommended a Points Per Transform of 4096, which makes the vertical scale of the graph so high that it goes right off my screen, and I've set the FFTs Per Sec to as low as I can, but that still doesn't help.

Perhaps my screen's just too small...I was using it on an old laptop with only a 12" screen. I'll try it again soon on my new 17" laptop and see what I can do...

Ta.
_________________
250Hz: (L/R)
-> 13603/13785
<- 14026/14974
Dual - 26659
90 - 18705 - 16865/18013
60 - 12206 - 11089/12283
30 - 6383 - 6415/6796

350Hz:
-> 11688/11409
<- 12238/13053
90 - 15799 - 13873/15407
60 - 10597 - 10546/10649
30 - 5362 - 5398/5895

[Pako]

I use a Virtual Desktop with a very high resolution (more than my monitor really supports). With the mouse I can scroll through the entire virtual screen, and can also grab its contents and save to JPG.

Pako.
_________________
REGULAR:

speed = 15136 ; 13606(L)
30s = 6662
90s = 18244 ; 17593(R) ; 15558(L)
dual = 27113
reverse = 14233(R) ; 12900(L)

SOUNDBALL:

speed = 14636

METAL:

speed = 13406
90s = 15465

[Adrena1in]

I realised something last night. With the input set to "Mike" I could set the PPTs to 4096 and the FFTs to 2, and the bar would scroll very slowly. But in the past I've always set the imput to a WAV file, and this automatically sets the FFTs to 11, and I can't make it less.

So all I need to do is get a microphone, (my laptop's built-in one isn't sensitive enough to pick up the sound from the speakers), and simply record through the Mike to get a good graph.
_________________
250Hz: (L/R)
-> 13603/13785
<- 14026/14974
Dual - 26659
90 - 18705 - 16865/18013
60 - 12206 - 11089/12283
30 - 6383 - 6415/6796

350Hz:
-> 11688/11409
<- 12238/13053
90 - 15799 - 13873/15407
60 - 10597 - 10546/10649
30 - 5362 - 5398/5895

[mroek]

Hi, all!

I'm new here, and I don't own a Powerball (but it should be on its way in the mail), but this topic is interesting. I am a software (and electronics) engineer, and a few years back I worked on something similar to this. I created a program that would allow you to measure the power of your car engine by analysing the engine sound during a full power drag. Granted, you also had to know a few parameters about the car (weight, tyres, gearbox ratios, differential ratio etc), but it was (and still is) a useable tool for the DIY-people.

I played a bit with the sound from some Powerball videos, and it would be quite an easy task to create a new program specialized for Powerball-analysis. All that is needed is a sound recording in WAV-format (can be taken from videos). The tool I created for engine power analysis takes 8-bit mono WAV-files with a sample rate of 11025 Hz, but that could of course change if needed.

What puzzles me slightly is the actual sound frequency that the Powerball emits. I tried looking at the sound from the Akis world record video, and 16 000 RPM gives a sound of 2133 Hz. I'm wondering what creates the actual sound? The rotor itself spins at around 266 Hz, but somehow the sound frequency produced is 8 times higher. As far as I can see there are only 4 "vanes" on the rotor, so for now I am slightly puzzled.

My car analysis program can be tricked into showing Powerball RPMs by setting the number of cylinders to 32, and then multiplying the RPM readout by 4. To use, drag the WAV-file (max 30 sec long) onto the program window, change "Cylinders" to 32, and press "Analyze audio". A grayscale image will emerge (takes some time), and you should be able to see similar curves to the ones in the program references in the start of this thread. By pointing with the crosshair mouse pointer in the grayscale window, you can see a RPM readout in the bottom of the window. This readout will need to be multiplied by 4 to get the real Powerball RPM. If I created a "Powerballalyzer"-program, the readout would of course be correct.

If anyone wants to play with the program, it is freeware and can be downloaded from my homepage:

http://personal.dynator.no/

click on "RPMAnalyzer" in the left-side menu.

[Obiw4n]

Thank you a lot, I'll check this and try to see what I can do with it....
_________________
Help Gyrotek when you buy your Powerballs, use our banner: International, France, Thanks!!

[Adrena1in]

Sounds great.

I mean, when you get used to it, Frequency Analyser is superb. Enda and Rory were a little sceptical about Jaco's 15021 video, thinking it didn't quite sound or look like a 15k score. (I admit, Jaco does make it look very easy.) So I did a quick graph for them and proved it's genuine.



But something that can perhaps display an approximate figure would be excellent, (and if Rory doesn't try to include something like this in the bluetooth counter then I think he'll be denying us something special! It really wouldn't be that difficult for a clever programmer.)
_________________
250Hz: (L/R)
-> 13603/13785
<- 14026/14974
Dual - 26659
90 - 18705 - 16865/18013
60 - 12206 - 11089/12283
30 - 6383 - 6415/6796

350Hz:
-> 11688/11409
<- 12238/13053
90 - 15799 - 13873/15407
60 - 10597 - 10546/10649
30 - 5362 - 5398/5895

[mroek]

Analysis of a recorded sound with user aid is not difficult. That is how my car analysis program works. The program shows you the spectrum, and then you (the user) are responsible for telling the program where the actual curve is by setting points on it until you get a good fit.

The problem is, as outlined in the original post in this thread, that you will get harmonics, which looks like mirrored curves below and over the actual curve, so you have to choose among them. That's easy for you, not so easy to do programmatically.

If the sound is "clean", the frequency can be counted directly by timing zero-passings after removing any DC-offsets, but in practice, there will be unwanted noise.

Creating a program that has a real-time display of Powerball RPM based on the sound from a mic is feasible, but it isn't easy.

If people find the frequency analyser program sufficient, I guess there is no use in me creating a new program along the lines of my car analysis program.

[mroek]

I've looked some more into this, and I believe I was wrong about the frequency. I believe that the correct audible frequency is 4 times the spinning frequency of the rotor, which is consistent with the 4 vanes. A 250 Hz spin (15000 RPM) should therefore produce a (dominant) audible frequency of 1 kHz.

To simulate this with my program, set the number of cylinders to 16, and multiply the readings by 2. 16000 RPM will then be full scale (since the scale goes to 8000 RPM).

Sorry for the confusion!

[mroek]

It's me again. Now I am even more confused. I found a really cool real-time spectrum analyzer program, and the best part is that it is free.

Here's the program:
http://www.techmind.org/audio/#specanaly

It works very well, and can be used while you play back the sound (or the video, for that matter) on your computer. You just have to make the correct soundcard settings.

The Akis world record video shows that the dominant frequency is somewhere around 2 kHz, but the video of Jaco's 15026 score is around 1 kHz. You will also see all the harmonics, but I'm talking about the dominant frequency.

No wonder there was doubt about Jaco's score, it really sounds like a lower score.

[Adrena1in]

[mroek]

My program does not use the mic, it ONLY analyzes a WAV-file (11025 Hz, 8 bit mono).
Just convert the file to this format, and then drag-drop it onto the program window.

[Adrena1in]