Help with choosing the right OpAmp

Hi Patrizia,

Welcome to e2e
I think some instrumentation amplifiers like INA163 or INA217 could fit in here.
As you need to detect a specific frequency, using a lock in amplifier seems to be a good choice. Perhaps you could add some additional band-pass filtering. I will move your post to the precision amplifiers forum, I think they can get a closer match for your application needs.

Best regards,
-Ivan Salazar
Texas Instruments

Hi Patrizia,

Normally electret microphones are much noisier than an op amp itself. The internal JFETs used in many electret microphones are often chosen for small size, low cost, and low gate to drain leakage current. All of these parameters are usually in directly conflict with low voltage noise in JFETs. I show some of the internals of a typical electret microphone in this TI design:

http://www.ti.com/lit/ug/tidu765/tidu765.pdf

I also show how the signal to noise (SNR) spec given in the datasheet of most electret mics can be used to calculated it's noise spectral density. This will also depend on the resistor you use to bias the microphone (connect from microphone to power supply). It would first be worthwhile to search for a microphone with the best signal to noise ratio that you can find. The Primo EM-173 is a very good option for example.

Choosing a low noise op amp is highly dependent on the output impedance of the source. If the bias resistor you are using with the microphone is <5k ohms, op amps with bipolar junction transistor (BJT) input types will be your best bet for very low noise such as the OPA1611 or OPA1612. If you are using a higher value resistor to bias the microphone, JFET or CMOS input amplifiers will provide the lowest noise such as the OPA1642 and OPA1652. 

Low noise design also demands that the resistor values used in the op amp feedback network be low values in order to limit their thermal noise contribution. If you haven't already, I recommend going through the TI Precision Labs videos on noise:

http://www.ti.com/lsds/ti/amplifiers-linear/precision-amplifier-precision-labs.page?DCMP=tipl&HQS=hpa-pa-opamp-tipl-vanity-tr-en

First I would like to thank both of you for your help, It feels great to be finally able to talk to experts! :)

Now, the following message is a response to John Caldwell, however I am happy for any reply/support!

Dear Mr Caldwell, I will order the microphone you have suggested for final optimizations of the sensor, it is a great choice, however, as I have already designed a housing for the sensor I am developing, I will first try it with the microphone I have to get a first idea of this amplifier. The electret microphone I use has a sensitivity of -38 dBV and a S/N ratio of > 60 dB (www.conrad.ch/.../Elektret-Mikrofonkapsel-EMY-63MP-Betriebsspannung-3-10-VDC-38-dB-Frequenzbereich-30-20000-Hz-Inhalt-1-St. The resistor used with the microphone is 2.2 kOhm, therefore I will go with OPA1611/1612 as you suggested. But what exactly is the difference of OPA1611 and OPA1612?  I noticed one has 2 channels and the other 1, but what does this mean? 

I have recalculated the theoretical voltage signal to be expected with this microphone in my application, which turns out to be some nanovolts (not microvolts as I have said earlier), is this still possible with this opamp? Or do I need a different one/a more sensitive microphone to realize this? However, in the beginning I would like to stick to the one I have if possible.

I have read the reference design PDF you have attached and I guess I can use the same design with my microphone and the OPA1611/1612 and procede in the same way to calculate the values, however, I would like to ask a few questions considering any limitations to my calculations/free design and I would be very thankful if you could help me out:

1) Can I set the maximum sound pressure expected somewhere above the expected sound pressure of the signal or must I set it much higher like somewhere between 50-75 dB due to the environmental noise (like in your case, 100dB). In the final sensor, the microphone is even shielded from environmental noise as it is placed in a thick aluminium housing. Could I just measure/estimate the sound pressure in such a shielded aluminium housing and then use this as the maximum?

2) You mention R2 and C2 form a low pass filter whereas R1 and C3 form a high pass filter. This together forms a bandpass. Now, as I am only interested in the signal at 1000 Hz, I would like to make this band smaller. Or is this not a good idea/not possible without deterioration of the system? However, as I will  use a lock-in anyway after the amplifier, the result might just be as good when I just leave it like it is, no? If you think that making the bandwith smaller is however a good idea, i have the following question, otherwise you must not read any further from this point :).

You use a response deviation of -0.1 dB at 20 Khz to calculate the location of the pole/upper frequency corner, which you then use to calculate the capacitance of C2.  Could I also use a response deviation of -0.1 dB at 1000 Hz or would you suggest a smaller value in my case? I band a small bandwith ofcourse, but still it is most important for me to  the signal too much? What response deviation would you set? I want a reasonable small bandpass but only as small so that signal attenuation is practically zero. With the equation of the pole, I could then easily determine C2 as the center/resonance frequency and R2 are basically fix, and then I could also calculate the the lower  frequency corner set by the high pass filter. With this and a high resisance R1 I can then calculate C5.When I choose an R2 to calculate C2 and choose R1 to calculate C3 respecvitely, are there any upper lower limits for these 4 variables I must consider? And by choosing a R2, which should be large, is there any limit in the Vout? I know Vout is basically limited by the power supply, but what is good enough to continue with a lock-in after preamplification such that the other parts are optimized and noise is sufficiently low?

3) For me a voltage supply of +9V is fine, also I will use the same values for R3 and R5. Should I then use the same low frequency corner as above and then calculate the value of C6 if I choose to change the bandpass?

I hope my questions are not disturbing, as I have already said, I have no experience so far and my time to realize this project is very limited (master thesis). However, with a positive result, the company will defenetively tcontinue with this project to finally develope a new photoacoustic sensor.

Thank you already in advance,

Patrizia Huggenberger

Hi Patrizia,

To answer your first question: the difference between the OPA1611 and OPA1612 is that the OPA1611 has 1 amplifier in the package, while the OPA1612 has two. The performance specifications are identical, the OPA1612 is just meant for applications which require more than one amplifier.

A quick calculation of the output noise of that microphone with a 2.2kOhm resistor gives a noise spectral density of 108.35nV/rtHz. So if your lock-in amplifier has a receiver bandwidth of 1 Hz, the noise from just the microphone will be 108.35 nVrms. I really think you should consider a lower noise microphone option.

As for your design questions:

1. Environmental noise has the potential to make the amplifier clip (reach its maximum output voltage) and prevent it from amplifying your signal. I would measure the environmental noise in the enclosure and use this to set the maximum gain since you want as much gain as possible.

2. It's a great idea to make the bandwidth as small as possible around 1kHz, this will help prevent environmental noise from clipping the amplifier. 0.1 dB is probably an acceptable deviation for the high-pass and low-pass sections of the circuit.

I would suggest that you use the highest supply voltage possible, because this allows for a noise reduction in the microphone amplifier circuit (this is explained in the document). All of the amplifiers mentioned have supply voltage ranges which extend to 36V. A 30V power supply for example would allow you to increase the microphone bias resistor to 100k ohms, which would place 5V across the mic capsule and reduce the noise gain of the amplifier. Here is an example:

Thank you! Slowly I understand the whole thing a little more! :)

But how did you calculate the noise spectral density, with which parameters? For the microphone with a sensitivity of -38 dBV with a 2.2kOhm resistor and a S/N ratio of 60 I get a microphone output current of 11.444 microA with a maximum sound pressure level of 100 dB SPL expected. For the current noise spectral density just for the microphone I then get 98.456 pA_rms/sqrt(Hz). Using the 75 kOhm resistor, this gives: 7.384 microV_rms/sqrt(Hz).

And why did you use 0.25 mA for the calculation of the microphone bias current R1? I though I should use the maximum current, which is 0.5 mA. With a power supply of 36 V and a voltage of 2 V across the mic this would be 68 kOhm, and with a 30 V powe supply and a voltage of 5 V across the mic I would get 50 kOhm.

And another question: for the calculation of the slew rate, does one always use 20 kHz?

Best wishes,

Patrizia