Simulators are great helping creating a new design but are difficult if the perfect created design is built in a imperfect real world. Capacitive wiring and pcb losses etc.
Only if you add these values in the simulator it will show these as well. If however you know that you will have these issues at forehand you don’t need a simulator in the fist place....
Only if you add these values in the simulator it will show these as well. If however you know that you will have these issues at forehand you don’t need a simulator in the fist place....
In Dutch terms: if punks come to town to trash the gardens, call the Police before the townfolk get annoyed and beat the punks.
Imagining elephants or (semi-)human punks in Holland:
If I were tracking Dutch punks (not trying to understand their talk), I would find most of their power between 100Hz and 600Hz.
Elephant is 4 times the size (linear dimension) of human. (Cube root of 8800 pounds to 180 pounds)
Logically elephants might be tracked with 25Hz-150Hz.
Some doubt because elephants have absurdly large ears for their size. Some of this is for cooling or signaling but still.... and yet an 8 foot ear-span suggests only slight directivity at 25Hz (good at 150Hz).
If elephants use much lower than 20Hz they may have to stand still to hear bass well. (Thump yourself in the side of your head. The human head-neck resonates "thump" around 20Hz. This is why F-M curves teach poor bass response at low levels: otherwise we'd go deaf walking. I don't dare thump an elephant's head-neck, but 5Hz seems reasonable, and perhaps they can't use such low frequency.)
"Elephants are capable of hearing at low frequencies and are most sensitive at 1 kHz..." Elephant - Wikipedia
Taking human peak sensitivity as 3kHz-6kHz, a 4:1 size ratio gives about 1kHz, some confirmation.
A larger microphone diaphragm gives very slight improvement of low frequency S/N at very greatly increased cost. (At the other end of the spectrum, size matters more.)
..and yet an 8 foot ear-span suggests only slight directivity at 25Hz
Wikipedia doesn't always tell you everything. Specially the things we don't know yet
A lot of things are still unknown in this field. Test where made with a portable infrasound generator (No... not a standard subwoofer) and elephants clearly recognizes sounds much lower than 20 Hz sounds. They right away turn their head towards the sound source. This Wikipedia article tells you something more about it. And the research is still on going. Instead of ear drums. Nerve fibers play a big role in perceiving infrasound. And that might be the case with elephants too.
It's believed that elephants use infrasound to communicate over distances of 4km with each other. And our AI mesh network with mics might be able to answer some of these questions in the future too. The first and main goal is still the prevention of conflict with humans.
if punks come to town to trash the gardens
Hehe. Also same difficult to scare Punks away i guess. Ohh and i am sure the police already have a Punk-Locator. With all the surveilling in cities. Punks should be a easy target for AI.
(BTW: I have nothing against punks as long they stay calm and friendly. Same with any other human)
Wikipedia doesn't always tell you everything. Specially the things we don't know yet
A lot of things are still unknown in this field. Test where made with a portable infrasound generator (No... not a standard subwoofer) and elephants clearly recognizes sounds much lower than 20 Hz sounds. They right away turn their head towards the sound source. This Wikipedia article tells you something more about it. And the research is still on going. Instead of ear drums. Nerve fibers play a big role in perceiving infrasound. And that might be the case with elephants too.
It's believed that elephants use infrasound to communicate over distances of 4km with each other. And our AI mesh network with mics might be able to answer some of these questions in the future too. The first and main goal is still the prevention of conflict with humans.
if punks come to town to trash the gardens
Hehe. Also same difficult to scare Punks away i guess. Ohh and i am sure the police already have a Punk-Locator. With all the surveilling in cities. Punks should be a easy target for AI.
(BTW: I have nothing against punks as long they stay calm and friendly. Same with any other human)
18 dB(A) improvement using a $3.16 microphone capsule, see post #30.
That doesn't tell you where in the spectrum most of the improvement is made, but as low-frequency noise is mainly gate shot noise filtered by the capacitance of the electret microphone, a higher capacitance should help. I also don't know if it is the size as such or some other peculiarity in the construction of this microphone capsule that gives it such a low noise level (*). It seems worthwhile to try it, though.
(*): Actually PUI also has a smaller model that's 4 dB worse if I remember well, so it could be that 4 dB is due to diaphragm size and the other 14 dB due to something else.
Last edited:
As you note: that's not all due to size. The $1 capsules (at any price) are non-optimum, because dictation machines and telephone answerers don't need the last 10dB of S/N. CUI stays in business covering the last 2% of applications where it matters. The last 0.2% of the business goes to mikes like AKG 414 and friends (at far more than $3).
... why 9 V battery supply and not tapping off the single-board computer supply?
I use 9V only because the circuit from Sri Lanka told me so. Will definitely use 5V if possible. Will the circuit behave the same if i feed it with 5V? Or do i need to change something?
The last 0.2% of the business goes to mikes like AKG 414 and friends (at far more than $3).
I agree. Yes. But not an option. Price matters. According to the mic-specialist from YouTube the Panasonic should work okay.
"Cost wise the only mic I know that lists a frequency response lower than 20Hz is the Behringer ECM8000, which you can find at Sweetwater for about $30. It's low end it listed at 15Hz (however most measurement mics typically go down to 3Hz), it's omnidirectional and has a near ruler flat frequency response, and is used as a calibration mic. I think it even uses the WM61A as it capsule (maybe). It's possible that the frequency goes lower, but they rate microphones for human hearing. The only issue I know of is the self noise is a bit high."
But as already pointed out. The AOM-5024L-HD-R seems to be better and is going to be tested as well when i get my hands on it.
I use 9V only because the circuit from Sri Lanka told me so. Will definitely use 5V if possible. Will the circuit behave the same if i feed it with 5V? Or do i need to change something?
The last 0.2% of the business goes to mikes like AKG 414 and friends (at far more than $3).
I agree. Yes. But not an option. Price matters. According to the mic-specialist from YouTube the Panasonic should work okay.
"Cost wise the only mic I know that lists a frequency response lower than 20Hz is the Behringer ECM8000, which you can find at Sweetwater for about $30. It's low end it listed at 15Hz (however most measurement mics typically go down to 3Hz), it's omnidirectional and has a near ruler flat frequency response, and is used as a calibration mic. I think it even uses the WM61A as it capsule (maybe). It's possible that the frequency goes lower, but they rate microphones for human hearing. The only issue I know of is the self noise is a bit high."
But as already pointed out. The AOM-5024L-HD-R seems to be better and is going to be tested as well when i get my hands on it.
5 V will change what op-amps you can use, but with a 9 V battery you probably want to keep the supply current low - which means an NE5532 is not such a good choice. Sharing the supply with the computer also can lead to crosstalk issues, although it worked out fine in Hal's case.
An OPA1678 draws 4 mA typical and can work at 5 V as well as 9 V, while an NE5532 draws 8 mA typ. and needs at least 6 V. I haven't checked the TL071.
An OPA1678 draws 4 mA typical and can work at 5 V as well as 9 V, while an NE5532 draws 8 mA typ. and needs at least 6 V. I haven't checked the TL071.
Last edited:
Attached is a list with data for some of the op-amps mentioned in this thread, plus an estimate of how much signal-to-noise-ratio loss the op-amp's noise will cause when using the PUI microphone with either a 10 kohm bias resistor or the 2.2 kohm that the manufacturer recommends. These numbers would be even lower with the Panasonic microphone, because it dominates the noise even more. They apply at 1 kHz rather than 10 Hz, but as the noise density of an electret microphone capsule increases faster at low frequencies than the 1/f noise of an op-amp, the differences at 10 Hz will be even less.
I added the last line to remind you that you may not need a preamplifier at all if the noise level of the audio input of the single-board computer is low enough.
I added the last line to remind you that you may not need a preamplifier at all if the noise level of the audio input of the single-board computer is low enough.
Attachments
Last edited:
Okay. Hopefully the mic will be changed later also
BTW. Here is a sound example of the current design with some distant thunder. Recorded on a battery powered Raspberry Pi zero with 5USD USB soundcard.
(During the thunder sounds i stayed 1m away from the mic and talked not towards, but away from the mic. The lake is around 7m away with very small waves)
Already sounds pretty good to me. But i have nothing "real" to compare it to.
EDIT: Oh yeah. 4mA is alot. Our nodes gonna be battery/solar powered. That's gonna be a problem. Will do some more digging about OPAMPs.
But we don't have concrete plans for the power supply yet. So there might be solutions getting 10V without much losses.
BTW. Here is a sound example of the current design with some distant thunder. Recorded on a battery powered Raspberry Pi zero with 5USD USB soundcard.
(During the thunder sounds i stayed 1m away from the mic and talked not towards, but away from the mic. The lake is around 7m away with very small waves)
Already sounds pretty good to me. But i have nothing "real" to compare it to.
EDIT: Oh yeah. 4mA is alot. Our nodes gonna be battery/solar powered. That's gonna be a problem. Will do some more digging about OPAMPs.
But we don't have concrete plans for the power supply yet. So there might be solutions getting 10V without much losses.
Last edited:
For what it's worth, if I were to modify the circuit for a TLC2262 and the elephant frequency range, I'd try something like this:
MIC1: AOM-5024L-HD-R or WM-61A, whichever turns out to work best (I expect the former, but no-one specifies what happens at 5 Hz)
C3: change to 4.7 uF MKT film capacitor. I don't think you need much more than that, because the microphone is essentially a current source, so the cut-off frequency is essentially 1/(2 pi C3 (R4+R1)) rather than 1/(2 pi C3 R1).
OPAMP: TLC2262
C2: 0 (removed)
R5: replace with 220 ohm in series with 10 uF aluminum electrolytic, positive side pointing to the op-amp. The resistor is just meant for stability when someone connects a long cable, any value between 47 ohm and 1 kohm will probably do.
Connect a 10 nF MKT or NP0 ceramic multilayer capacitor in parallel with R3 if low-pass filtering is desired, otherwise 39 pF NP0 ceramic (or 100 pF if a roll-off at 16 kHz is OK).
Using the feedback capacitor as filter capacitor instead of R5-C2 has several advantages: the op-amp doesn't need to provide as much signal current as with the R5-C2 filter and you have more loop gain in the feedback loop. A disadvantage is less protection against RF signals picked up by the cable at the output.
C4: no idea, to be determined experimentally what value is needed to keep everything stable with nearly empty batteries (that is, to prevent low-frequency oscillations, a.k.a. motorboating).
MIC1: AOM-5024L-HD-R or WM-61A, whichever turns out to work best (I expect the former, but no-one specifies what happens at 5 Hz)
C3: change to 4.7 uF MKT film capacitor. I don't think you need much more than that, because the microphone is essentially a current source, so the cut-off frequency is essentially 1/(2 pi C3 (R4+R1)) rather than 1/(2 pi C3 R1).
OPAMP: TLC2262
C2: 0 (removed)
R5: replace with 220 ohm in series with 10 uF aluminum electrolytic, positive side pointing to the op-amp. The resistor is just meant for stability when someone connects a long cable, any value between 47 ohm and 1 kohm will probably do.
Connect a 10 nF MKT or NP0 ceramic multilayer capacitor in parallel with R3 if low-pass filtering is desired, otherwise 39 pF NP0 ceramic (or 100 pF if a roll-off at 16 kHz is OK).
Using the feedback capacitor as filter capacitor instead of R5-C2 has several advantages: the op-amp doesn't need to provide as much signal current as with the R5-C2 filter and you have more loop gain in the feedback loop. A disadvantage is less protection against RF signals picked up by the cable at the output.
C4: no idea, to be determined experimentally what value is needed to keep everything stable with nearly empty batteries (that is, to prevent low-frequency oscillations, a.k.a. motorboating).
For your information, in hal4287's bird monitoring project, the loss in signal-to-noise ratio from connecting the microphone straight to his sound card without any preamplifier was at most 2.32 dB, see
https://www.diyaudio.com/forums/everything-else/352655-advice-weird-project-3.html#post6169969
This was for songbird frequencies, the difference was much less for frequencies well below 300 Hz. He used a Sound Blaster PLAY! 3, Sound Blaster PLAY! 3 - Sound Cards - Creative Labs (United States) , which is about four times as expensive as your sound card, though.
The experiment hal4287 and I did to measure the difference in signal-to-noise ratio is described right here:
https://www.diyaudio.com/forums/everything-else/352655-advice-weird-project-2.html#post6162041
He did steps A and B and I did step C.
https://www.diyaudio.com/forums/everything-else/352655-advice-weird-project-3.html#post6169969
This was for songbird frequencies, the difference was much less for frequencies well below 300 Hz. He used a Sound Blaster PLAY! 3, Sound Blaster PLAY! 3 - Sound Cards - Creative Labs (United States) , which is about four times as expensive as your sound card, though.
The experiment hal4287 and I did to measure the difference in signal-to-noise ratio is described right here:
https://www.diyaudio.com/forums/everything-else/352655-advice-weird-project-2.html#post6162041
He did steps A and B and I did step C.
Last edited:
On a system level, how are a multiple of mics input analysed (if so)?
How many mics are in a system? Distance between them?
In real time?
Is location (direction and distance) calculated by intensity? Then all mics is preferably of the same sensitivity and amps have same gain.
Location - always domestic with power available or placed in the bush, powered by batteries?
If batteries, how is signal transferred?
Maintenance (change battery etc) - how often?
//
How many mics are in a system? Distance between them?
In real time?
Is location (direction and distance) calculated by intensity? Then all mics is preferably of the same sensitivity and amps have same gain.
Location - always domestic with power available or placed in the bush, powered by batteries?
If batteries, how is signal transferred?
Maintenance (change battery etc) - how often?
//
@MArcelvdG
Very interesting information.
The circuit as well as the idea not using a preamp (i remember you mentioned it earlier already). I should definitely read more in Hal's thread.
I will do the tests the next days and let you know about the results.
How many mics are in a system? Distance between them?
IF all works out. We try to cover an area of 1000 square kilometers. Each node 1Km apart from each other (just guessing right now) But that's still many years in the future...
In real time?
Yap. Nodes will record 24/7 and the right signal (elephant sound) will trigger an algorithm which turns on the transceiver and sends data to our server.
Is location (direction and distance) calculated by intensity?
If more than one mic pics up the sound. We can calculate the location using the speed of sound (in theory). But other solutions might come in play.
Location
In bush up in the trees or so. Battery and solar powered if possible. I know. That's gonna be one of the biggest challenges.
If batteries, how is signal transferred?
Using LoRa. It's something like wifi for IoT data (can travel 20Km with line of sight and can run years with a single 9V battery). In the jungle probably just 1Km (Also still untested in the jungle).
Maintenance (change battery etc) - how often?
Preferably every three months or so. Rough calculations: We might need a 8Kg lithium for each node per month (if no solar is used)... As i said. This might become our biggest challenge.
But all these questions will be answered over time. Technology advances and we are just trying to prove the concept for now. For example IF you we are able to pick up elephant rumble from Sumatran elephants (which are smaller than in Africa) at a 1km distance. Or maybe Sumatran elephants only rumble once a week instead of many times per day? Then this system will not work at all. We don't know all that yet because nobody recorded it yet.
Sooo. First steps first: Get sound data from Sumatran elephants and after that we brake our heads about the next problems
Very interesting information.
The circuit as well as the idea not using a preamp (i remember you mentioned it earlier already). I should definitely read more in Hal's thread.
I will do the tests the next days and let you know about the results.
How many mics are in a system? Distance between them?
IF all works out. We try to cover an area of 1000 square kilometers. Each node 1Km apart from each other (just guessing right now) But that's still many years in the future...
In real time?
Yap. Nodes will record 24/7 and the right signal (elephant sound) will trigger an algorithm which turns on the transceiver and sends data to our server.
Is location (direction and distance) calculated by intensity?
If more than one mic pics up the sound. We can calculate the location using the speed of sound (in theory). But other solutions might come in play.
Location
In bush up in the trees or so. Battery and solar powered if possible. I know. That's gonna be one of the biggest challenges.
If batteries, how is signal transferred?
Using LoRa. It's something like wifi for IoT data (can travel 20Km with line of sight and can run years with a single 9V battery). In the jungle probably just 1Km (Also still untested in the jungle).
Maintenance (change battery etc) - how often?
Preferably every three months or so. Rough calculations: We might need a 8Kg lithium for each node per month (if no solar is used)... As i said. This might become our biggest challenge.
But all these questions will be answered over time. Technology advances and we are just trying to prove the concept for now. For example IF you we are able to pick up elephant rumble from Sumatran elephants (which are smaller than in Africa) at a 1km distance. Or maybe Sumatran elephants only rumble once a week instead of many times per day? Then this system will not work at all. We don't know all that yet because nobody recorded it yet.
Sooo. First steps first: Get sound data from Sumatran elephants and after that we brake our heads about the next problems
Last edited:
> maybe Sumatran elephants only rumble once a week
I bet your poachers know. Yes the pitch is very low for human hearing, but not truly inaudible. Also an old man with lost highs can be more sensitive to bass. My grandfather could hear thunder before anybody. I hear the diesel coming up the street. If I was a poacher gang, I'd give all the old men near the forest cellphones (or walkietalkies) and $5 for each valid report of an elephant.
Also review Doug Self's page on Acoustic Radar. It worked to hear bombers coming across the Channel. At long range the horn is HUGE but apparently could be effective. And you do not need vertical aiming for elephants (except Dumbo).
I bet your poachers know. Yes the pitch is very low for human hearing, but not truly inaudible. Also an old man with lost highs can be more sensitive to bass. My grandfather could hear thunder before anybody. I hear the diesel coming up the street. If I was a poacher gang, I'd give all the old men near the forest cellphones (or walkietalkies) and $5 for each valid report of an elephant.
Also review Doug Self's page on Acoustic Radar. It worked to hear bombers coming across the Channel. At long range the horn is HUGE but apparently could be effective. And you do not need vertical aiming for elephants (except Dumbo).
Last edited:
Parts are on the way and will we built soon. That's really nice of you. Thanks a lot. Will also post the results in comparison of what i have now.
I also stumbled across this paper about different microphones. But i guess everybody here has already read that
Just in case: High microphone signal-to-noise ratio enhances acoustic sampling of wildlife
By the way, for 5 V supply, R4 would have to be reduced to about 4.7 kohm (or 4.75 kohm if you prefer E96 values).
Using metal film resistors for R4 and possibly R2 and R6 could lead to a slightly lower noise floor than using carbon film (and a much lower noise floor than using carbon composition, but those are almost never used anymore). I wanted to mention that for completeness, even though I see you use metal film resistors everywhere anyway.
Using metal film resistors for R4 and possibly R2 and R6 could lead to a slightly lower noise floor than using carbon film (and a much lower noise floor than using carbon composition, but those are almost never used anymore). I wanted to mention that for completeness, even though I see you use metal film resistors everywhere anyway.