• State Resolved
  • Locked Locked
  • Replies 24 replies
  • Subscribers 47 subscribers
  • Views 7384 views
  • Users 0 members are here
Support feedback
Options
Options
Similar topics

This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

Amplifying an Audio Signal

ramin anushiravani
Intellectual 350 points
Other Parts Discussed in Thread: LF347, OPA2314, OPA134, LMV1012, LMV1032

Dear People! 

I'm trying to amplifying a signal from three different microphone one from Knowles (NR23158) http://www.knowles.com/search/prods_pdf/NR-23158-000.pdf it's a bidirectional mic and one from Knowles (EK23132) http://www.knowles.com/search/prods_pdf/EK-23132-000.pdf it's an omnidirectional mic and from a usual round mics ( I don't even no what kind it is!). I've been using many many different designs to amplify the signals coming from these mics. I asked a questions on TI forum before, here is a link, 

http://e2e.ti.com/support/amplifiers/precision_amplifiers/f/14/t/196007.aspx

I tried all the diagram on this post, including some others and the one in this article, 

http://www.en-genius.net/site/zones/audiovideoZONE/technical_notes/avt_030110

Here is the problem!

I've used three differen op amp, K324 , LF347  and OPA 2314 . For some reasons (unknown to me!) with the same gain on the op amps K324 has a much bigger amplitude in the output than LF347. The amplitude from the round mics are even bigger than the ones from Knowles EK23132 and that's bigger than NR23158. The design for the K324 (or C324C) http://pdf1.alldatasheet.com/datasheet-pdf/view/6751/NEC/C324C.html and for LF347 http://www.ti.com/product/lf347 is very simple. Basically, I only have a high pass filters in the input with a 1Hz cut off frequency and the rest are just some resistors and coupling capacitors to adjust the gain (very very simple design!) . For C324C I needed to apply V/2 to one of the inputs but for LF347 I just connect that other input to the ground. Anyway, It seems like the results from LF347 gives me a higher SNR ratio ( I tested all these, so take my word!) . knowles mics ( I'm using a built in preamp for the knowles mics to get higher amplitude http://www.alldatasheet.com/datasheet-pdf/pdf/353759/KNOWLES/NR-23160-000.html)  and the round mics both have an almost same SNR ratio which is weird, knowles mics are much more expensive and they're suppose to have a higher SNR ratio. 

BTW, My goal is to get a higher gain ( I should be able to hear the output easily for all the mics) and a small noise. 

I tried to change my design and solve this problem and that was the first time I posted on the TI forums. I had this idea that OPA2314 should give me a better result so I went with that. There was an incredible article http://www.en-genius.net/site/zones/audiovideoZONE/technical_notes/avt_030110 I tested the amplifier with my mics but the result is not good AT ALL! The gain is small with the same gain and the noise is higher than all the other designs. Here is the deal, I don't want to cut any frequency off, the one in the article had many filters to cut of the frequency, I don't need that, I can easily do that with the computer. What I need is a very low noise and a high amplitude, HIGH SNR! 

Anyway, so I tried to use that simple simple design in the OPA 2314 ( also OPA is 2 channels and the other ones are 4 channels, so I didn't even have the bleeding from one channel to another since I used both channels on the OPA for one mic), It doesn't work! I can see it's amplifying the signal, but it's not what it should be. First I connected one of the inputs to ground and tried to amplify my signal through the other channel, I didn't get any output, then I applied V/2, I get the output but the amplitude is too small even with a very high gain. The amplitude is a little bigger for round mics comparing to knowles ones. I modeled the circuit in TINA TI, everything is working perfectly there, but in real world not so much! 

I tested my OPA and LF347 on TINA TI and for the simple circuit they both seem to have a same noise in the output! WHY?! I thought OPA 2314 should give me a better result. Well in real world OPA is not working, so I'm kind of stuck with LF347. 

I need another design with the OPA that would actually gives me the CORRECT GAIN and small noise from the circuit. 

A little bit more detail, I'm trying to amplify the signal for 3 bidirectional and one omni at the same time. Omni always have a higher amplitude than the bidirectional ones, so I might have to go with a bigger gain with the bidirectional mics. Also, the amplitude from the knowles mics are always lower than the ones in the round mics. 

THANKS FOR READING MY POST ! AND THANK YOU AGAIN FOR HELPING ME OUT! 

Regards,

Ramin

  • 0
    TI__Guru 62935 points

    Hi Ramin,

    Sorry you're still having trouble with your design.  Did you implement the last circuit I recommended?  Please include your schematic and we'll take a look.

    Regards,
    Collin Wells
    Precision Linear Applications

  • 0 in reply to Collin Wells
    Intellectual 350 points

    Hi Collin, 

    Thanks again!

    Here are the schematics. 

    2728.1.TSC

    7446.2.TSC

    3414.3.TSC

  • 0 in reply to ramin anushiravani
    Intellectual 350 points

    7356.4.TSC

    I didn't try this one yet, It seems like with positive and negative voltage the gain is positive and without it is negative?! why? and it seems like the noise is actually smaller for the opa a little bit 

  • 0 in reply to ramin anushiravani
    Intellectual 350 points

    The file number 3 is the schematics I was talking about, the simple amplifier. 

  • 0 in reply to ramin anushiravani
    Intellectual 350 points

    I tried inputting a 4.5 volts to the empty input of the + and negative volts, nothing happened. Could you please also take a look at the knowles mic datasheet, why is the amplitude coming from knowles mic so small? 

  • 0 in reply to ramin anushiravani
    Intellectual 350 points

    I tried the last circuit you recommended without the negative voltage the gain was too low. I also tried the one you had in the article ( takes me a while to troubleshoot that circuit!) the gain is not too low, but still not as high as the simple design and the noise is not too high but not as low as the simple design. I don't really need to cut the low and high freq, my noise is mixed with my signal, I need to reduced that mixed noise that's coming from the circuit. When I amplify the gain , the noise amplify with gain with the circuit in the article. 

    Thanks,

    Regards,

    Ramin

  • 0 in reply to ramin anushiravani
    TI__Guru 62935 points

    Hi Ramin,

    I'm not sure I understand your problems, could you please simply state your issue and goal in a few short sentences?.  Your circuits look fine now although I would still recommend limiting the bandwidth with filtering to lower the noise if it is a concern.  The LF347 has higher noise figures than the OPA134 so as long as you limit the bandwidth the OPA134 will have better noise performance.

    If you don't use the negative voltage with the last circuit I provided then you will lose half of your input signal because the input was referred to GND.  You can fix this by changing the bottom connection of the R3 50kOhm resistor from GND to mid-supply (4.5V in this case).  This would bias your inputs mid-supply so the full input signal will be received.

    Regards,
    Collin Wells
    Precision Linear Applications

     

  • 0 in reply to Collin Wells
    Intellectual 350 points

    Dear Collin, 

    Alright ! :)

    The goal is to have a high gain and a low noise ( cutting of low and high freq are not important for me, the noise I'm trying to cancel out is within the 20Hz to 20KHz), basically a high SNR. I need a design that gives me a high amplitude and low noise! ( higher amplitude and lower noise than what I've got before!) 

    The problem is that  the design in the article is not giving me a higher SNR than the "simple design". I thought maybe that's because of all the extra components in the circuits, and maybe if I don't have the filters in the design ( If I implement the OPA with the simple design) it will give me a higher SNR or maybe it's because the capacitors used in the LF347 are bigger than the one in the article.  I had problems with amplifying the signal using OPA too, the amplitude was too small even when the gain was set high (not too high) but then I connected negative to the ground. 

    So you think if I add that 4.5 V I will get the signal with the higher amplitude and yet smaller in noise than for one the one in LF347? 

    What do you mean by band-limiting the the bandwidth ( cutting at 20 and 20KHz like the one in the article, if yes that's not giving a better performance than LF347). 

    Thanks,

    Cheers!

    Ramin

  • 0 in reply to ramin anushiravani
    Intellectual 350 points

    Also what are your thoughts on using Ina 217 considering the output impedance of the mics. And do u have any design in mind for Ina 217 ? I tried using that too but the ic gets too hot.

  • 0 in reply to ramin anushiravani
    TI__Guru 62935 points

    Two more questions before I comment further:

    1.)  How are you determining SNR?  Are you using a network/spectral analyzer or are you listening to the audio output and deciding which sounds better?  If the noise is truly in the same band as your desired signal then I see no way to improve SNR because the noise will get gained with the signal if they are both present at the inputs.  Is there any method to reduce the noise in the 20Hz - 20kHz band.

    2.)  What is your desired supply voltage?  Is it +9VDC, +/-9VDC, +5VDC?  You used different supply voltages for your designs and you accidentally violated the common-mode input range of some of our devices when you connected them single-supply. 

    Regarding noise, there is a fundamental relationship between the bandwidth of the amplifier and the noise at the OPA output.  RMS noise = Vn * SQRT(1.57*BW) where Vn is noise spectral density in nV/rt(Hz) and BW is the bandwidth of the amplifier circuit.  So limiting the bandwidth inherently limits output noise which is why I keep suggesting limiting the bandwidth to reduce the noise.

    Regards,
    Collin Wells
    Precision Linear Applications

  • 0 in reply to Collin Wells
    Intellectual 350 points

    Dear Collin,

    For the SNR ratio, I wrote a Matlab code that take the recorded sound and give me all sorts of information, I tested the code with other data it's working fine.

    I see you're point on the bandwidth. In this case, I don't understand why the simple circuit(!) is giving a higher SNR than the one in the article. 

    There is no reason on why I used all these different power supplies, I just wanted to check if they will help some how to amplify the signal further. I'm using a voltage regulator now, and that's why I'm sticking with 5volts. 

  • 0 in reply to ramin anushiravani
    Intellectual 350 points

    Hi Collin, 

    I tired band-limiting and the only difference is the cut off freq, the noise in the output is the same in TINA TI, do you think that's because of the software?  

  • 0 in reply to ramin anushiravani
    Intellectual 350 points

    ignore my previous comment! BW is changing the noise

  • 0 in reply to ramin anushiravani
    Intellectual 350 points

    I did some testing and I think now I can give you more details on how I want the preamp to be. I need the gain to be somewhere around 50 to 60db. I tried to reduced the bandwidth and it helped reducing the noise. I'd say the circuit is a little better than the previous designs but it's still not acceptable.

    My circuit looks like the design below, the values might be a little different but close to these values. 

     3326.opa newest.TSC

    I looked at the speech signal after I recorded and it's looks a little distorted and the sparsity is not clear ( freq lines are not distinct) . My signal is mostly around 300Hz to 4KHz and the noise is also in that range. I would also like to have a more flat freq around 300Hz to 4KHz. I don't need to have the full range, something around 100 to 15KHz would work. 

    So! How can I get a 50db gain with low noise and a more flat freq in 300Hz to 4KHz range than what I've achieved in the previous design.

    Thanks!

    Regards,

    Ramin

  • 0 in reply to ramin anushiravani
    TI__Guru 62935 points

    Hi Ramin,

    Please see attached and results below.  It has a gain around 52dB and you can modify this by changing the feedback resistor values.  If you change the resistors you will have to change the feedback capacitors as well.  If you make the R bigger make the C smaller and vice versa.

    The transient results show a little less than 1Vpp with a 1mVpp input signal.

    2677.OPA376_Microphone.TSC

    Regards,
    Collin

  • 0 in reply to Collin Wells
    Intellectual 350 points

    Thanks Collin. This does work and the noise level is okay. Unfortunately, the algorithm still doesn't work with this preamp and I can't get the knowles mic (NR23158) working with these schematics. I've been using a commercial small round mics for testing all these schematics. I'll get back to you when I figured out why the algorithm is not working well. 

    Thanks,

    Regards,

    Ramin

  • 0 in reply to Collin Wells
    Intellectual 350 points

    Actually you might be able to answer this questions. The reason the circuit was oscillating before was because I used the same voltage divider to power the positive feedback of both stages or maybe two ICs. When I use an individual voltage divider for each pin, there's no oscillation anymore, but that makes the circuit bigger, so on my new design I just used the positive and negative voltage and didn't use the voltage divider. Any idea why a common voltage divider makes the op amp oscillate? Also, when I used the knowles mics, I only get the oscillation and no signals out. 

  • 0 in reply to ramin anushiravani
    TI__Guru 62935 points

    Hi Ramin,

    Please stop posting on two threads for the same issue, it's hard to keep track of and is likely even more difficult for the external community to follow along.

    Regarding the voltage dividers:  if you look at the circuits, when you connect the same voltage divider to both places you're effectively shorting out the first OPA because the positive input of both amplifiers get tied to the same net and the input signal likely finds some strange path around the first amp causing issues with the second.  It wouldn't be an issue if we weren't building a non-inverting system where the input signal is summed with the common-mode voltage at the non-inverting input.  But since this system is non-inverting we need two dividers.

    Regards,
    Collin Wells
    Precision Linear Applications

  • 0 in reply to ramin anushiravani
    TI__Guru 53996 points

    Hi Ramin,

    Remember, these mics have built-in FET amplifiers.The give-away is the series resistor from the output to the power supply.

    The ultimate SNR of these mics will rest on the noise and SNR performance of the internal FET's. The SNR of any system is set by the first stage (the internal FET in this case), and no amount of tricks on the post-amplifier can make up for a poor front end SNR performance (you can just minimize the damage).

    Also, noise from the supply and the series resistor can add to your noisefloor. Use a clean supply for the mic.

    Regards,

  • 0 in reply to Paul Grohe
    Intellectual 350 points

    Hi Paul, 

    Thanks! You just solved my problem with the knowles microphone! I always thought of the first stage amp as the first amplifier, not the built in fet inside the mic. What do you mean by a clean supply like a voltage regulator? 

    Regards

    Ramin

  • 0 in reply to Collin Wells
    Intellectual 350 points

    Sorry about that I will post my questions regarding the design on the other thread. 

    Regards,

    Ramin

  • 0 in reply to ramin anushiravani
    TI__Guru 53996 points

    Hi Ramin,

    Keep the supply to the microphone as clean as possible. The basic FET circuit does not have much PSRR. So pre-filter the supply line - especially if the power is derived from a switching supply.

    If you are curious, most likely the microphone uses one of these FETs:

     http://www.semicon.toshiba.co.jp/eng/product/transistor/selection/mos/fet/condence.html

    or one of our "FET" replacements:

     http://www.ti.com/product/lmv1012

     http://www.ti.com/product/lmv1032

    Regards,

  • 0 in reply to Paul Grohe
    Intellectual 350 points

    Hi Paul, 

    I'm very new to this subject. So what you're saying that preamp inside the mic can't reject the noise itself and I have to do some filtering before I connected to the amplifier. I already have many filters, but as you mentioned before those won't make up for the noise. So what kind of filter should I use? How do I know which FET is being used in my microphone ( there's nothing on the datasheet) ? There are some schematics regarding RF noise reductions on the link you sent, can I use something similar to that? I have many low and high pass filters on both stages of my preamp circuit, but that doesn't seem to help much. 

    I am using a switching supply! 

    I was actually playing around with the knowles mics, and the noise is a very big issue. Basically, before I put a small resistor through a positive voltage, connect that to microphone and that was my input to the preamp circuit ( amplitude after amplifying was too small), but then I changed that to a big (100K) resistor and the amplitude of the signal is much much bigger and also much noisier. What value resistor should I used to power the microphone with a +9v? since on the datasheet it said the microphone is functional from 1.3V to 10V and it has an output impedance of 4.4k. 

    Thanks,

    Regards,

    Ramin

  • 0 in reply to ramin anushiravani
    Intellectual 350 points

    I have another question for you. I'm amplifying signals from four different microphone, they each have their own OpAmp so there's no linkage within the opamp. I can still see that when one microphone is connected to the preamp circuit, there's linkage on all the other channels which is almost a -20db quiet high. I was using a same voltage regulator to power all the microphone so I thought maybe that was the problem but the linkage was still there. My preamp circuit are only sharing the same positive, negative and common ground, how can the signals goes back through all the amplifier? BTW the signals amplify through the linkage doesn't look like the actual input signal, but sill seems like I'm loosing my signal through other amplifier in the signal. How can I stop this linkage? and where is this linkage coming from?

    Thanks,

    Ramin