There is a missing coupling capacitor in that circuit!
I'd reduce the first feedback network to 10k/100ohms, not 1M/10k, to avoid injecting
10k's worth of Johnson noise direct into the input.
The second stage will rectify/clip as its DC bias is at ground.
MarkT:
I'd reduce the first feedback network to 10k/100ohms, not 1M/10k, to avoid injecting
10k's worth of Johnson noise direct into the input.
Okay, but now you need a much larger capacitor than 4.7uF in that feedback loop to maintain the LF response.
@aarg, @MarkT thanks for your feedback.
I've dropped the feedback R in the first stage to 1k; I dont think 0.4uV of noise in the audio band will be an issue there. Schematic amended in the post above to save confusion.
Johnson noise - interesting! It possibly WOULD be an issue in my bat detector designs with a 200kHz bandwidth and a LOT more gain.
The second stage will rectify/clip as its DC bias is at ground.
Yes, but as its a rectifier anyway ... and the input signal is too small to upset the input stage ...
Thanks for the tips everyone.
@johnerrington Re your comment about powering the LM358 from a single 5V supply, unfortunately I was already doing this and from a 3.3V supply! I was getting success but just not nearly sensitive enough. Anyway it was just an experiment and I'm re-thinking how I can achieve what I want, see comments below!
I'm trying to make the design for my multisensor as modular as possible so that others can make it. And also using as few components as possible.
So I like the idea of sticking with the MAX9184 mentioned by @MrMark (also it helps having it on a separate board as I can mount it upside-down and the mic is in exactly the right place to be near a grille I will put in the plastic casing). I have the board now but haven't had a chance to test it.
Re @DVDdoug's suggestion of a peak detector / envelope follower. Are there any reasonable ways I could get away without the need for the enevelope follower? Ideally I will be sampling the output of the above board on an analog input of the ATtiny85. Or perhaps there is a way to make an envelope follower far less complicated (I don't think it needs to be accurate)? This may be a silly suggestion but could surely just charge a capacitor as I don't care about attack / release times, which can be as long as required really. The main feature is to get a one-shot detection presumably when the signal goes above a trigger level. I appreciate comments here, in case I am wrong.
Trigger level: it is difficult to know what must be considered as a triggerable event, and my gut feeling tells me that I will need to be able to calibrate this remotely. By "remotely" I mean my circuit will be on a board inside a housing mounted to the ceiling in 20 different rooms, there will be an ESP32 running Tasmota. As yet I've no idea how to do what I want here. Tasmota does not provide an analog output that I can set remotely over wifi. I may have to build something to do this, any ideas appreciated here!
Alternatively as the MAX9184 has AGC I wonder if that feature can be used to help me determine a variable noise floor?
Re @DVDdoug's point about the 1.1V ADC reference, I am not familiar with that at all but have read briefly about it - is it available on the ATtiny85?
Max9814 looks a good option as it will give you a choice of gains.
You then need to convert the audio signal to a dc envelope, to see how big it is. I strongly suspect you will need something thike the second stage of the circuit I have shown above.
johnerrington:
I strongly suspect you will need something thike the second stage of the circuit I have shown above.
So that's the diode, which will half-wave-rectify the signal, then the resistor / capacitor tied low on the output... presumably that will do the enveloping with well chosen values?
I didn't fully understand that circuit hence asking, thanks
I think you will need some extra gain too hazymat - this whole circuit. Of course you can try with just a simple diode rectifier
johnerrington:
Max9814 looks a good option as it will give you a choice of gains.You then need to convert the audio signal to a dc envelope, to see how big it is. I strongly suspect you will need something thike the second stage of the circuit I have shown above.
The MAX9814 has a maximum gain of 60 dB, 20 dB (100x voltage) more than the circuit of post #19. The MAX9814 microphone modules will drive a passive peak detector (diode, capacitor, bleed resistor) directly should one choose to implement peak detection in hardware vs software.
Is there a reason why you are not telling us exactly what you are trying to detect ?
You simply generalized it as being in some range rather than saying "I want to detect _______"
Why are you not telling us what you want to detect ?
johnerrington:
Max9814 looks a good option as it will give you a choice of gains.You then need to convert the audio signal to a dc envelope, to see how big it is. I strongly suspect you will need something thike the second stage of the circuit I have shown above.
Its got a huge voltage noise spec of 30nV/√Hz, for a good microphone in a quiet room you need
low noise circuitry. Even the NE5532 is 15dB quieter than that.
Across the 20kHz bandwidth 30nV/√Hz is 4.2µVrms, which is about the same as the signal
levels from a microphone on a quiet room.
With a decent 'real' microphone (not electret), there is a lot you can do with the sound, however, judging from your
posts you are looking for a quick and dirty solution and not really interesting in building anything complicated.
The problem with a cheap electret microphone is it does not detect all of the available sound. If you compare the
waveform from a good microphone with that from an electret, I believe you will find the more expensive microphone is more sensitive, consequently you can process it for specific bands using FFT and isolate particular
sounds from other sounds, or you can use a notchpass filter and detect amplitude to create a band detect , which
can be adjusted by adjusting the comparator reference voltage. Have you actually built any circuit yet ?
raschemmel:
judging from your posts you are looking for a quick and dirty solution and not really interesting in building anything complicated.
[...]
Have you actually built any circuit yet ?
Yes I have built the circuit I posted in my original post. It works fine to detect noise close-up such as a clap within 12 inches of the mic, although the output signal was not to my needs (it was not schmitt triggered)
In answer to your other point here (and your previous question on page 2 of thread), it's not a question of whether I'm interested in building anything complicated, the application in question which I described in previous posts in this thread, apologies if you missed this, is that I'm building a multisensor to be installed into each room / area within the home, and because of interest from others online, I am trying to keep the design cheap and simple. Audio sensing is not core to the project, but I do want it to be as good as possible given the space restraints within the housing as well as the cost for others to make.
You can see the project here if interested: youtube link
For the sake of context, the heart of this multisensor is PIR & microwave. Its core function will be to determine if someone is in the room. The idea of adding an audio circuit is to improve the presence sensing. I'm completely aware that noise from other rooms, street noise, thumping from above floors, etc., not to mention intentional room noise such as television / radio / music will mean it's not possible to detect presence by audio with a high degree of accuracy. So the application purposefully calls for audio sensing as a blunt tool. But combined with the other parts of the sensor (PIR/microwave as mentioned above plus additional things: ambient light level, buzzer for notification, pixel ring, temp/pressure/humidity sensor, all of which I've already included in the design), audio sensing will make a nice addition.
So yes it requires a small electret mic as it has to fit in a small case mounted on the ceiling. My ideal spec for this audio sensor would be to be able to simply trigger a digital output when someone in the room speaks or makes any noise less than "quiet room". I was also toying with the idea of having different outputs for different audio types:
any noise slightly louder than ambient room noise (including hum of fridge, etc.) - this needs remote threshold setting for each room
single clap
double clap
This would of course be even more useful, let's say if PIR/microwave movement detection AND audio detection were not triggered for eg 30 mins, lights go out, you could just clap to restore lighting again.
Likewise double clap to select next light scene, etc.
Hope the above makes sense
So in summary, I am not trying to find the best solution period, I am not trying to find an accurate or perfect solution, but I am trying to find the best solution given the constraints.
You first need to firm up this threshold sound level to an actual dB value, and look at the datasheets
for electrets - you want your signal to be well clear of the mic's noise floor for reliable detection without
false positives, then you need to figure out how many dB of amplification will put your signal comfortably
within reach of the ADC.
MarkT:
You first need to firm up this threshold sound level to an actual dB value, and look at the datasheets
for electrets - you want your signal to be well clear of the mic's noise floor for reliable detection without
false positives, then you need to figure out how many dB of amplification will put your signal comfortably
within reach of the ADC.
Thanks, I'm going to do some tests with my new sensor board this evening to see if it does a lot better than the other boards / my own sensor circuit. As long as I can get something useful out of it, i.e. trigger level at a safe distance above ambient room quietness, which triggers with speech a few metres away, then that's all I really need for now.
I built the Big Ear circuit 40 years ago and I could hear a whisper. You would have to hack it and add the detector
circuit with an arduino digital output .
raschemmel:
I built the Big Ear circuit 40 years ago and I could hear a whisper. You would have to hack it and add the detector
circuit with an arduino digital output .
Thanks for posting that link, it's a very good read and really interesting project!
FYI, I was a full time restaurant line cook teaching myself electronics at night by building circuits. I still have the 5V/12V P.S./breadboard (box with
+5V, +12V, -12V regulators inside and a solderless breadboard on top. ) It's the breadboard that literally launched my career because it's the one I
used to build a bidirectional 40-channel led sequencer to get a job as a custom VIP Party lighting controller designer that culminated in an event at Paramount Studios Sound Stage 13 (the one they filmed the first Star Trek Movie in) for the Los Angeles Museum of Fine Arts ("Tribute-82"; the l Contemporary Art Council's Awards with Leonard Nimoy and William Shatner as guest speakers. I got Shatner's autograph) The experience I got from that job got me my next job as an engineering tech prototyping op amp circuits. Consider that 30 years
later in 2010, I used it to prototype a signal conditioner circuit for a Model-S transmission test fixture at Tesla Deer Creek.
I think that's the breadboard I used to breadboard
the Big Ear circuit in 1979. Funny how that works...
I guess I got my money's worth eh ?