I have an idea to measure ultrasonic amplitudes using an interference technique so that what is actually measured is the amplitude of the "envelope" of the signal. For instance, if I have equal power 20kHz and 21kHz tones being reproduced by two different transducers, how to I measure the power or SPL of the 1kHz interference between the two? I have access to the ARTA measurement suite, for instance, but am trying to think about how to go about this... do I need to filter (bandpass?) the signal first and then analyze it, or can I measure on the interference (20+21) signal directly?
I am having a mental block on this and need to get pointed in the right direction.
-Charlie
I am having a mental block on this and need to get pointed in the right direction.
-Charlie
20 kHz is directly measureable - physically small condenser mics go well beyond that frequency with usable flatness
lots of soundcards today are usable to 40 kHz with 96 k sample rate
the 1 kHz difference frequency can only be detected when there is some nonlinearity in the system to cause IMD products
when testing for small levels of IMD is is useful to filter out the test tones for greater measurement dynamic range, and to avoid IMD conversion in the sensing electronics
lots of soundcards today are usable to 40 kHz with 96 k sample rate
the 1 kHz difference frequency can only be detected when there is some nonlinearity in the system to cause IMD products
when testing for small levels of IMD is is useful to filter out the test tones for greater measurement dynamic range, and to avoid IMD conversion in the sensing electronics
Last edited:
OK, maybe I should clarify here...
Let's say you have the typical cheap microphone that is usable up to 20kHz and then falls off rapidly. How can this be used to measure the frequency response of a transducer to, say, 50kHz or 80kHz or wherever the -20dB down point is of the transducer?
Here is my idea:
Thoughts?
-Charlie
Let's say you have the typical cheap microphone that is usable up to 20kHz and then falls off rapidly. How can this be used to measure the frequency response of a transducer to, say, 50kHz or 80kHz or wherever the -20dB down point is of the transducer?
Here is my idea:
- Assume you have a system to generate a signal and amplify it sufficiently so that there is no bandwidth limitation on the signal sent to the driver.
- Use two drivers (e.g. a pair of the same tweeter model), one as a "reference" reproducing a frequency X and the other as "DUT" reproducing X+1kHz (for instance). This will generate an interference "pattern" where the envelope of the interference will be at 1kHz.
- Somehow measure the amplitude of the 1kHz interference pattern, and from that deduce the amplitude of the DUT.
- Now we know the amplitude response at 21kHz.
- Next step is to set the reference to 21kHz and the DUT to 22kHz and again measure the amplitude at 1kHz to establish the amplitude at 22kHz.
- Continue to repeat this process and incrementally move up in frequency until there is insufficient resolution in the system to continue.
Thoughts?
-Charlie
OK, after a little bit of time spent "on the Google" I've figured out that this is called an "Envelope Detector".
https://en.wikipedia.org/wiki/Envelope_detector
Anyone want to weigh in on whether this idea can work?
-Charlie
https://en.wikipedia.org/wiki/Envelope_detector
Anyone want to weigh in on whether this idea can work?
-Charlie
I don't see how you'd separate microphone and driver FR here. Not to mention the uncertainties which would be growing quickly as you go up.
You're probably thinking along the lines of an "RF nose". That one works because there is a rectifier in front of the band-limited component. I can't think of any kind of acoustic rectifier offhand.
You're probably thinking along the lines of an "RF nose". That one works because there is a rectifier in front of the band-limited component. I can't think of any kind of acoustic rectifier offhand.
Interestingly there are sonar arrays that use the non linearity of sea water (when hit hard enough) to produce a low frequency beam from two high frequency signals (Very highly directional arrays are more managable in terms of size at high frequency).
It takes an amazing amount of power to pull it off, think well over 200dB ref 1uPa, but it does work in this case.
I would note that capacitor microphones, even good ones, are known to usually behave non linearly at high frequency, but I would not bet on a cheap mic having enough response up there to matter, you need something external to the mic to produce the non linearity (Air will do, but it would take a monster level to make it happen).
Also, don't forget that up there in the ultrasonic air starts to get VERY lossy.
Product detection only works when there is something causing there to be a product (in the math sense).
Regards, Dan.
It takes an amazing amount of power to pull it off, think well over 200dB ref 1uPa, but it does work in this case.
I would note that capacitor microphones, even good ones, are known to usually behave non linearly at high frequency, but I would not bet on a cheap mic having enough response up there to matter, you need something external to the mic to produce the non linearity (Air will do, but it would take a monster level to make it happen).
Also, don't forget that up there in the ultrasonic air starts to get VERY lossy.
Product detection only works when there is something causing there to be a product (in the math sense).
Regards, Dan.
You'd have to contact Knowles directly for details but they have MEMs mics that easily exceed 100 kHz. Ultrasonic Sensors :: Knowles
Even some cheap phantom powered condenser mics go ove 30 kHz.
I'd just measure directly!
Even some cheap phantom powered condenser mics go ove 30 kHz.
I'd just measure directly!
Have you heard of these neat things that we call "tweeters" over here??? All kidding aside, the passband of many modern tweeters extends to 20kHz and some up to 40kHz. For a survey, just go to Audioheuristics or Zaph Audio and look at the measurements. Sure, you can mostly measure that part directly. I want to keep going a bit, so that I can capture the roll off. It's the slope of the roll off that I am most interested in.
Anyway, if this approach could enable an el cheapo (e.g. $50) condenser mic to measure up to 50kHz, way above where the microphone can measure directly, don't you think that would be "of interest"?
-Charlie
Hmmm, now that you mention it, if the mic can not detect the "carrier" frequency can it still pick up the envelope? Here the carrier would be the 20k+21k Hz frequencies from the two transducers (that would be the lower end) and would continue up to the maximum frequency of interest, e.g. 50+ kHz (50+51). The envelope is the 1kHz tone that is generated as the waves interfere...
I thought that this was how some modern transducers operated - ultrasonic waves generated by multiple sources interfered and the result was "sound" in the audio band.
The OP wants to measure a transmitter beyond the frequency range of his receiver. Beats will be present, but the receiver can't receive them because it can't receive the signals at all. You need nonlinearity somewhere in the system to create a 1kHz tone from the beats. No nonlinearity: no 1kHz energy.
I thought about it a little more, and did an experiment, and the above seems to be the case.
Experiment: I used a pair of high quality headphones and adjusted the signal frequency up until I just passed the limit of my hearing. I did this for the left channel. Then I put the same frequency + 100Hz, 200Hz, etc. in the right channel. If I put on the headphones, I heard nothing. If I took off the headphones and placed both cups near each other and up against my ear, I heard nothing. If I did the same thing with a lower "carrier" e.g. 6kHz, I could easily hear the beating.
Offhand, what is a simple source of non-linearity that could be used? This would have to be "in the air" I guess, or somewhere in between the source (transducers) and receiver (mic).
Could a 1kHz acoustic resonator be made to "ring" by the beats, and that resonance picked up by the mic???
-Charlie
That was what I was hoping to do - computing the amplitude of frequency X+1 kHz from the known amplitude at X kHz using the beat amplitude. Once X+1 kHz has been determined, move on to X+2 kHz. This would be repeated, stepping up as far as possible by 1kHz increments, to establish the ultrasonic frequency response.
You only need to measure the beat amplitude at 1kHz each time, so I need to determine a reliable way to do that.
-Charlie
No. As I keep saying, you need nonlinearity. Without that there is no 1kHz so nothing for the resonator to ring to.CharlieLaub said:
Beats are not the same as mixing!
You need one of two things:
1. a mechanical structure which can respond to your ultrasonic frequencies and is sufficiently non-linear to generate the 1kHz difference frequency and sufficiently large to radiate enough for a microphone to pick up the 1kHz.
2. something which responds to the total ultrasonic power but can respond quick enough to cope with 1kHz variations. A tiny heat sensor might be one possibility.
Note that option 2 builds in the nonlinearity: responding to power means squaring the amplitude so you get the difference frequency generated.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.