Hardware sound evaluating solutions

[Alex_rcpilot]

Hi all.Here's my problem:

A friend of mine is trying to develope an evaluation system for automobile horn manufacturers.Owners of such factories told him that they hire experienced workers to listen to each horn,and apply proper adjustment to make the tone best resemble that of a standard horn.As you know,the horns are too loud for human to sustain for a long period of time without any harm.And human hearing varies because of lots of factors,which makes the adjustment not so close to ideal.Manufacturers want an instrument on the assembly line to do this job instead of human.Adjustment will be easy as rotating an adjustable component or two.The key is sound evaluation.We have searched the web for such chips only to find nothing but voice recognition chips.We contacted serveral domestic companies selling voice recognition chips,asked whether they have such chips we needed and they said 'no'.I think perhaps there are such things in other countries.So,here I am.

Has anyone ever heard of sound evaluating chips?I think they are less complex than voice recognition chips because the latter have to analyse a certain pattern after evaluating the waveform of input audio signal.So a sound evaluating chip is more like a semi-manufactured stuff.And I think what we need is actually parameters of the audio sample such as amplitude of the fundamental wave and its harmonic.Then a microcontroller will be able to process such data with a certain algorithm.There's a software solution in which a PC is implemented.Third-party software will aid the process of sound analysis.But they want cheaper solutions.Given this requirement,the chip we need can't be so expensive that it turns the PC into a more brilliant choice.

Better recommendations other than what I've mentioned(but not bringing back the workers ) is appreciated.Any suggestion is welcome.Thank you for the help in advance.

Regards,Alex

 

[Nigel Goodwin]

I would suggest looking at FFT (Fast Fourier Transform), using either a micro-controller or DSP (Digital Signal Processor).

Using FFT you can do spectrum analysis of the signal, and then adjust to provide the output spectrum you want.

Not a trivial task, but it sounds an interesting project - in fact, it would be a great final project for the large number of students who are always asking for suggestions on here!.

 

[Alex_rcpilot]

I would suggest looking at FFT (Fast Fourier Transform), using either a micro-controller or DSP (Digital Signal Processor).

Using FFT you can do spectrum analysis of the signal, and then adjust to provide the output spectrum you want.

Not a trivial task, but it sounds an interesting project - in fact, it would be a great final project for the large number of students who are always asking for suggestions on here!.

Thank you Nigel!I missed out the DSP solution.I wish I knew enough about DSP to implement FFT on it.Don't know if a normal PIC can handle it.But there are dsPICs,am I right?This may work!But it will be inevitable for us to reach into the very nature of spectrum analysis algorithm.That will be a lot of brain work.I would appreciate it.Actually I'm glad to give a try.Don't know if my friend who is focused on conventional MC controlling will have the patience to learn about the algorithm.

Are you suggesting that there is no such chip we need indeed?This will leave me no choice but thinking about DSP.I'm a college student too.I'm at the third grade.And my teacher promised if I want,one of my former project in the institute can work as my graduate design too.I really think I should do something better and put it in that position.This will make a fine project.

Thanks for your advice again!

 

[Nigel Goodwin]

Are you suggesting that there is no such chip we need indeed?

I wouldn't have thought so, it's MUCH too specific a requirement. You could probably do it in hardware by having a number of bandpass filters, and measuring the respective output levels of each filter - adjust the sound of the horn until it falls within the required specifications.

How many filters, what frequencies, and how narrow the filters are, would need to be determined - it's basically a simple hardware spectrum analyser?.

 

[ThermalRunaway]

What about utililzing a zero-crossing detector in order to supply time-period information to a microcontroller? If you were to write a program that measured the time-span between the zero-crossing points of the waveform, and compare this with a preset piece of data, perhaps you could get the microcontroller to adjust the frequency of the horn until it falls within specification. You can buy zero-crossing detection ICs off the shelf I believe, or you could most likely accomplish the task in software using interrupt requests.

I'm not certain that this kind of solution would fit the bill for you because I don't know enough about how the horns are setup, but it might be something to look into. Afterall, a PIC would certainly be capable of that.

Brian

 

[mstechca]

You can do basic sound comparison with your ears. Just make an oscillator, and adjust the resistor values and the capacitor values until the sound is right.

The basic formula for the tone is Frequency = 1 / (R * C) where R is resistance in ohms, and C is capacitance in farads.

If you make the resistor into a potentiometer, and connect the oscillator to the input of a PLL (phase lock loop), and connect the "horn" to the other input, the phase lock loop can give you a better indication on whether the sounds are the same or not.

 

[ThermalRunaway]

Yes but the main point of the project is to automate the process so that human intervention is not required. So using your ears is pretty much out.

Brian

 

[Alex_rcpilot]

Just dropped in,and I'm glad there're more replies now.Oh yes the hardware filter solution Nigel has referred to is interesting,and I think it's likely to work,even though it means a lot of components on a board.I think it may work well enough to make qualified adjustment.I better tell my friend to consider it.Thanks Nigel.

ThermalRunaway,I can see that you're suggesting frequency detection,and mstechca's idea is samilar.I guess the problem is not so simple.The horn is a module which characterizes comprehensive spectrum features.That's what I referred to as 'tone',instead of 'frequency'.Or perhaps 'timbre' will suit it better?I believe even the ultimate product perfectly adjusted will still differ from the standard one in some aspect.And the 'tone' detected by a simple frequency detector may not be the closest to that of the standard one.I will get some audio samples from my friend,so that I can analyze them with software on my PC before any further experiment.Thanks for the replies anyway!

Regards,Alex

 

[ThermalRunaway]

Yes I noticed you used the word "tone" and I presumed that you meant there were more than one specific sound combined to make up the total sound. I know that the horn on my car is a twin-tone type. Again I don't know exactly how the car horn works or how they're setup, but I presume you could use filters to split up the seperate frequencies and adjust them accordingly using zero-crossing detection.

Let us know how you get on with finding a solution

Brian

 

[zevon8]

I agree, this could be a very interesting project. I would think that you would need to do several tests to determine what the fundamental frequency is of the horn, and all of its dominant harmonics. These are what produce the " sound " of the horn. From there you could, as suggested above, design filters to separate these frequencies, and measure their levels. Once the levels are determined, a specification could be set.

The information could be displayed as a set of levels, showing which frequency needs adjusting.

One difficulty could be how each device under test is adjusted to either produce more or less of each required frequency. From what I recall, most car horns only have one adjustment, which basically alters the fundamental frequency, and overall loudness, peaking at its natural resonance. The mechanical properties of the horn itself will determine the harmonics. I would guess that since someone has designed the horns already, the harmonics should be taken care of, it is more a matter of adjusting the fundamental. This could leave the testing to be a go-no-go test, if the horn can be adjusted within the specification, it passes, if not, it is rejected, since I doubt you would want to start altering the mechanical properties of each horn.

 

[Alex_rcpilot]

Sorry Brain.I think I missed out the filtering concept in your previous post,which makes the idea comprehensive.Thank you for the suggestion.And I will let everybody know how it worked out when it's done.

zevon8.I agree with you.I think these horns are designed to make the adjustment 'ASAP' - as simple as possible One single component may be all to be taken care of.And if you can't get the right tone by turning the screw or something else to any position,then this horn must be rejected.Anyway let me wait for some audio samples and a chance to see the product for myself.I need to talk more with this friend.Thanks for your reply.

Regards,Alex