Your electret's datasheet specifies:
It's been a while, but this is an old way of writing out sensitivity.
One bar is \$10^5\:\text{Pa}\$. So \$1\:\mu\text{bar}\$ is \$100\:\text{mPa}\$.
Today, we usually use one Pascal as the reference, so this means we need to add \$20\:\text{dB}\$ to the written specification for comparison with more common specs today.
\$60\:{\text{dB}_\text{SPL}}\$ is a loud conversational level. \$40\:{\text{dB}_\text{SPL}}\$ is the lower end of conversational levels.
Loud, the output would be \$20\:\mu\text{Pa} \times 10^{^\frac{60\:{\text{dB}_\text{SPL}}-54\:{\text{dB}_\text{SPL}}+20\:{\text{dB}_\text{SPL}}}{20}} \$ or about \$400\:{\mu\text{V}_\text{PK}}\$. It will be about 10 times less, or about \$40\:{\mu\text{V}_\text{PK}}\$ at the lower end of conversational levels.
This assumes, as stated, that \$R_\text{L}=1\:\text{k}\Omega\$. You can use a higher voltage where possible, but this also means a larger \$R_\text{L}\$. In this case, I'd just recommend a standard \$9\:\text{V}\$ alkaline battery as the voltage source. I usually anticipate that the electret will require about \$1.5\:\text{V}\$ itself. And since a standard \$9\:\text{V}\$ alkaline battery is really more like \$8.5\:\text{V}\$, I'd use \$7\:\text{V}\$ as the resistor drop. I also usually expect an electret requires \$500\:\mu\text{A}\$. So this means \$R_{_\text{L}}=\frac{8.5\:\text{V}-1.5\:\text{V}}{500\:\mu\text{A}}=14\:\text{k}\Omega\$.
I'd select \$10\:\text{k}\Omega\$, though. (Some of the electrets can use a little more current.)
Given this resistor value is 10 times the one they specified, you can expect about 10 times the AC output voltage, but obviously now with a source resistance that is also 10 times higher. So between \$400\:{\mu\text{V}_\text{PK}}\$ and \$4\:{\text{mV}_\text{PK}}\$ and with a source impedance of \$10\:\text{k}\Omega\$.
To get that up into the range of \$1\:{\text{V}_\text{RMS}}\$ would require a final voltage gain (after attentuations are accounted) of \$20\cdot\log_{10}\left(\frac{1\:{\text{V}_\text{RMS}}}{4\:{\text{mV}_\text{PK}}\,\cdot\, \frac1{\sqrt{2}}}\right)\approx 51\:\text{dB}\$ (about 350 times.)
But that's where I'd start the design process.
The first stage should be carefully designed to use low-noise transistors. (Sufficiently low-noise opamps are power-hogs, boutique, and otherwise not needed -- so BJT-only.) The voltage gain should be very modest -- perhaps a factor of 3 or 4 (\$+9.5\:\text{dB}\$ to \$+12\:\text{dB}\$.) But no more than about \$+15\:\text{dB}\$. A class-A stage should be fine, but use a current source as the collector load for it. Once that stage is done, then the 2nd stage can be designed to provide the remaining gain of around \$+39\:\text{dB}\$. (And that can be a decent audio amp.)
Regardless, this will require a minimum of two stages of amplification: the pre-amplifier (described above) and an amplifier stage that provides necessary remaining voltage gain and matching up with the ADC input for the MCU. (We've not yet discussed the MCU ADC.) But 3 stages would not be off the table.
Not doing the necessary analysis and then design and construction steps to get from A to B is likely why you may be having troubles.
Example of first stage design
There is a design process for a similar amplifier found here at the EESE site that includes the following schematic:
This is about how I would approach a first stage design using low noise BJTs (like the BC549 and BC559), except that the above design isn't just a first stage, wasn't designed for low noise, and is instead designed to directly drive a speaker.
There's a comment there, from G36, saying that the above design is similar to one by John Linsley Hood, dating back to this 1969 article (now preserved at Elliot's site).
This brings me to the point. There is an excellent pair of articles at Elliot Sound Products:
In the 2nd article see Figure 4. He chooses the same \$9\:\text{V}\$ battery and specifies the same \$10\:\text{k}\Omega\$ resistor for his electret pre-amplifier. (Once the battery is selected, the resistor falls out from that choice. We reached a similar place likely for similar reasons.)
Elliot's design uses a class-A design but one that has some exceptional properties. More on that, later.
I've redrawn it in a way I prefer. And I've replaced his output capacitors as this will be feeding into the next stage and I just wanted to test it. Click on the image to make it larger and more readable:
I've also included an electret simulation of my own in the lower left corner, for testing purposes. (And the obvious \$9\:\text{V}\$ battery source simulation, as well.)
The top graph shows the loaded microphone output, which is about \$2.5\:{\text{mV}_\text{RMS}}\$. Unloaded, this would be closer to \$2.8\:{\text{mV}_\text{RMS}}\$. But the amplifier will load it down a little.
The bottom graph shows the unloaded output of the amplifier stage. (I used a \$10\:\text{M}\Omega\$ resistor, which is essentially no load.) This is about \$10.6\:{\text{mV}_\text{RMS}}\$. So the gain is about \$+12.5\:\text{dB}\$.
addendum
I am adding this months later...
Elliot's design has some nice features:
- As it should, it uses global NFB to set the gain. In this case, the NFB factor is just \$\beta\approx \frac1{1+\frac{R_8}{R_5}}\$ (neglecting \$r_e^{\:'}\$.) The open loop gain will be in the several thousands, so the closed loop gain won't be too far afield when setting \$R_5\approx \frac{R_8}{A_v-1}\$. This is a very simple calculation to make. (It's actually \$R_5=R_8\cdot\frac{A_{vo}-A_v}{A_{vo}\left(A_v-1\right)+A_v}\$, derived from \$A_v=\frac{A_{vo}}{1+\beta\cdot A_{vo}}\$. But the simpler equation is close as \$A_{vo}\gt 5000\$ from the high impedance of the current source working against the two concerted BJTs.)
- The quiescent point is insensitive to variations in the power supply and also insensitive to variations in transistor \$\beta\$. For example, any excess base current needed by \$Q_4\$ (due to a lower value of \$\beta\$) is just passed along via \$Q_1\$ to develop an added drop across \$R_8\$, leading to an adjusted quiescent collector voltage for \$Q_4\$. This causes no problems as there is a current source feeding that node and the design includes a lot of headroom for part variations of this kind.
- The quiescent \$V_{_\text{CE}}\$ of \$Q_1\$ can be set to any comfortable value by adjusting the voltage divider of \$R_1\$ and \$R_3\$. The divider should be set for the lowest allowed power supply value so that \$Q_1\$ stays in active mode. But with that done, if the supply is increased from there the re-biasing of \$Q_1\$ is harmless as its \$V_{_\text{CE}}\$ simply picks up the slack without harm. (In this way, \$Q_1\$ and \$Q_4\$ are relative isolated from each other rather than possessing a tightly coupled design.)
- The input impedance (as seen by the electret and \$R_2\$) is quite high: roughly \$110\:\text{k}\Omega\$.
- \$R_8\$ is simultaneously a key part of the DC biasing and also the path that provides the global NFB connection into the circuit. It does both, well, simultaneously, and without sacrificing anything important. I don't see this kind of circuit parsimony as often as I'd like!
The result is very smart:
- The design can be arranged to handle supply voltages at or above \$5\:\text{V}\$ and it will work at the same quiescent point for supply voltages that are significantly higher. For higher voltages, there is only a 2nd order effect on the quiescent current due to the Early Effect operating on the current source. Other than that, the \$R_1\$ and \$R_3\$ divider voltage scales. But this is simply picked up smoothly by an increase in \$V_{_\text{CE}}\$ of \$Q_1\$ without the slightest impairment to the behavior/design goals. Very smart.
- The arrangement of \$R_8\$, \$R_9\$, and \$Q_1\$ (DC biasing) diverts only just enough current to sustain \$Q_4\$'s quiescent point and a little more for what \$R_9\$ must also sink. It self-biases very nicely and adapts to often wide BJT parameter variations of \$\beta\$ and \$I_{_\text{SAT}}\$ in a very smooth manner. Very smart.
- After all is said and done, the base recombination current of \$Q_1\$ will be very light and its emitter impedance reflected to the base almost negligible. So the input impedance (I assign \$R_2\$ to the source impedance driving the circuit) is dominated by \$R_4\$ summed with \$R_1\$. That's easy to see and adjust. But as shown, it's also a factor of 10 larger than the source impedance (\$R_2\$) which is a good design rule, as well. And the DC quiescent point for \$Q_1\$ works through \$R_8\$, which is also the NFB resistor. Very smart.
- The placement and use of \$C_3\$ and \$R_5\$ as a part of the AC closed loop gain setting group is simple and direct and again very smart.
- The whole design is well-handled parsimony, minimizing up to a point but not further. Further simplification would sacrifice something important. Further addition wouldn't provide benefits worth the added complexity. Very smart.
It's well designed and worth retaining in mind.
