I have built this audio amp circuit on a ground plane board and have taken care to limit trace lengths and provide as clean a layout as I could come up with. I'll try to attach pictures of the board layout later but I'm posting the circuit now. The amp works really well I think but I am interested in further reducing the noise (hiss). This amp has an unusually high gain because my goal was to detect very faint sounds. This project was started to give me an excuse to learn about analog circuit design. I put this together from months of casual research on amps, filters, drivers and AGC. The final design has a pretty good signal to noise ratio on the 120 ohm output with headphones but both outputs have some noticeable noise. I don't have suitable equipment to make accurate assessments regarding the actual noise level or S/N ratio. I'd like some pointers on further reducing the noise. I'm of course limited to through hole components but I am using low noise op amps and metal film resistors throughout. The input stage is a diff amp because I wanted to be able to experiment with noise cancelation. Currently I'm using only the one microphone on the board as shown on the schematic.
Please let me know what you think. My initial research included some low noise design principals but I am a novice so any suggestions are appreciated.

thanks
skip

P.S. Scematic Correction: The resistors R11, R12 and R13 are shown with incorrect values for a 300hz high pass filter. The correct values are R11 = 1k, R12 = 390 and R13 = 6.8k.

 

I don't have suitable equipment to make accurate assessments regarding the actual noise level or S/N ratio.

That might make it tough to proceed, since it's hard to optimize something you cannot measure.

Where do you think the current noise is coming from? I'm thinking of the weakest link strategy - just keep attacking the next largest source of noise. But it's hard to do that without knowing where it's coming from.

 

I have checked to see if the AGC is contributing any noticeable noise by disconnecting it from the bottom of R7. That didnt make any noticeable difference. I am wondering if the feedback resister in the 2nd stage is introducing any noise due to its somewhat high value. I have considered implementing a 3 resistor feedback pattern there to reduce the high value (249k). My main concern is with the preamp stage since that is typically where the problem is. Are there any tricks to limit or otherwise get rid of noise there? Any tricks for the use of electret mics? I know the mic itself is a noise source and i'm not sure if there are other mic options that are better. I have not tried paralleling two mics yet to improve the s/n but that is not really what I'm after here. I'm using a pretty standard electret here. I think the s/n on it was 60db? I'll post the part number tonight. I'm using a 5v regulater to power the mic and that supply looks pretty clean on my scope down to the milivolt range. My scope is of decent quality but I'm sure not the best at these low levels. I'm also wondering about parts placement on the board. I'll try to post pics tonight of the layout. I kept the layout as tight as i could get it meaning the parts density is pretty high in order to keep it all on a standard board size. I was careful to try to protect the inputs from the higher voltage outputs and such but I'm no expert. The amps are pretty low noise or as best I could find at a decent price. I think I paid $6 or $8 apiece for them. I think they are in the 4nv/hz to 6nv/hz range. The caps are film, I think polypropelene or such. I tried to research the best material for low noise but I never was quite sure I found the best.

 

Just to think outside the box here - literally - how sure are you that the noise is in the circuitry as opposed to the incoming signal? I guess that's your mic. I think it makes sense to move through the block diagram and determine where the s/n ratio is degrading, if at all.

 

Can you check the noise when the mic is disconnected?

In a well designed circuit with a low noise preamp, the biggest source of noise is likely to be the input resistors. Try reducing their value and see if that helps. Of course you also need to reduce the feedback resistor by the same percentage to keep the gain constant.

Edit: You also may want to add an RC filter to the bias voltage for the mic.

 

I checked the noise with no mic installed and it is pretty quiet, to the point that I don't consider it a problem. I don't know how much noise to expect from a mic. Is there any way to minimize the noise from the mic? I guess the problem is that the noise is in the hearing range. I suppose that leaves only the possibility of a less noisy mic. Are electrets as good as any for this application? Again my goal is to have very high sensitivity.

 

I can look at filtering the mic buss but I was thinking that the regulated 5v supply would be better. Also I have the mic bias supply being fed equally into both inputs of the differential preamp so any noise on the supply should be rejected...or at least that was the plan. I know that this design is a little bit less noisy because I remember ballancing the inputs on purpose for that reason. Plus I want to be able to attach a mic to either or both inputs.

Will this approach work as planned or do I still need to isolate the mic bias with a filter?

 

I have attached two pics of the board. I know this is not ideal but it should give you some idea. If you see anything you can suspect of being a noise problem please let me know. I guess we are focusing mainly now on the input and preamp section in the upper left corner. The board layout pretty closely matches the schematic drawing layout.

The mic I'm using is a CUI INC CMB-6544p. Sensitivity is -44db, S/N is 60db, output impedance is 1k, voltage is 4.5v.

The blue square caps are of type film (Polyester).

Thanks.
skip

 

I'd still like some feedback on the capacitor material and the microphone if anyone has any thoughts on these subjects. I'm not sure what the "self noise" is on this pcb electret mic I mentioned above but I have seen some really expensive studio mics listed as having self noise of -34db. I don't know if that is better or not because I am not able to compare it to the one I have.
I'd love to hear from someone if there is a really good quiet mic option out there or even what db level is considered really good for a mic.

Also, I'm a little confused about these numbers anyway. My understanding is that 0db is the lowest sensitivity level of the human ear. If a mic has a sensitivity of-44db that seems extremely more sensitive than the human ear. Is that right? And if the same mic has a self noise of -34db does that mean that the lower 10db of the mics sensitivity is not usable? I have a feeling I might not be looking at this quite right...

thanks
skip

 

Also, I'm a little confused about these numbers anyway. My understanding is that 0db is the lowest sensitivity level of the human ear. If a mic has a sensitivity of-44db that seems extremely more sensitive than the human ear. Is that right? And if the same mic has a self noise of -34db does that mean that the lower 10db of the mics sensitivity is not usable? I have a feeling I might not be looking at this quite right...

thanks
skip

Take a look, especially the first few paragraphs:
https://en.wikipedia.org/wiki/Decibel

It's easy to be confused about decibels because decibels should be stated with a suffix to identify what particular reference level is being used. Trouble is the suffix is frequently omitted and it is assumed the reader just knows based on the context.

The -44db mic sensitivity should be written -44dbu which is 44db below the typical 0.775 volt 0db level used to measure audio signal levels. -44dbu is a bit more than -40dbu which is 1/100th of 0.775 volts or .00775 volts.

The 0db you refer to as the threshold of hearing refers to 0db SPL. (Sound Pressure Level)

 

The 0db you refer to as the threshold of hearing refers to 0db SPL. (Sound Pressure Level)

I may be mistaken (?) but I thought that SPL levels also require some kind of reference to be useful. That's why SPL meters you buy have scales such as "dBA" and "dBC" selectors. Those scales have built-in references and also frequency weighting that is industry standard. I don't remember what the 0dB ref is for dBA but there is some reference.

The term dB is always a ratio of two numbers, so a reference must be selected to use it for standardized measurements.

 

I may be mistaken (?) but I thought that SPL levels also require some kind of reference to be useful. That's why SPL meters you buy have scales such as "dBA" and "dBC" selectors. Those scales have built-in references and also frequency weighting that is industry standard. I don't remember what the 0dB ref is for dBA but there is some reference.

The term dB is always a ratio of two numbers, so a reference must be selected to use it for standardized measurements.

Yes they do, I should have mentioned that also .

The reason is that our ears do not have a flat frequency response. We don't hear low and high frequency's at the same level as mid frequency's. And this changes with the sound pressure level. In days past it was referred to as the Fletcher-Munson effect.

https://en.wikipedia.org/wiki/Weighting_filter

 

What bandwidth are you wanting? I would definitely add simple RC low-pass filtering to the input, to try to exclude RF and whatever else you don't absolutely need.

For the inverting input, you would probably have to split R3 and put a cap to ground from between the two new resistors, since a cap directly from the inverting input to ground would tend to cause instability.

There are probably some appropriate ideas here: http://www.analog.com/library/analogDialogue/archives/39-05/op_amp_applications_handbook.html

 

My bandwidth is currently set at 300Hz to 10KHz with the 3rd order high and low pass filters in stages 3 and 4 of the amplifier. I also have a first order low pass filter on the preamp that should roll off any RF but you may notice in the pictures of the board I have removed it because I figured it wasn't really necessary. I originally was concerned that strong RF pickup might overload the earlier stages. I'll probably put it back in place but removing it did not have any impact on the noise that I'm dealing with now. I have tried to analyze the noise with my scope but it seems to be mostly white noise covering the entire spectrum. Does FFT help analyze noise? I just had a thought and I may look into that if my scope supports it.
The board is pretty quiet without any mics attached so I'm thinking that the mic is the source of most of my problem here. I'm hoping someone can recommend a better mic.
Thanks to everyone for the feedback though. This has been interesting so far...

Skip.ele

 

I figured out how to do FFT on my scope. I have attached a few pics. These were taken with a mic attached to the inverting input in a quiet room late at night in my house.

Can someone explain what these pics show? The cursor marking the zero line for the FFT waveform is near the top in the upper left corner of the screen (purple trace and cursor). I'm assuming that means the FFT signal is showing negative db value. At 5db/div I think that would make the bottom of the purple waveform at or near -20db ???

I'm not really sure what I'm looking at here. What information about the amp or the mic can I gather from this? And what is the Sa number related to the FFT trace?

I have the bandwidth cursors (white vertical lines) set to low end (A) and near 15khz (B) which basically take up the entire screen width. This shows the practical hearing range of audio frequencies. The yellow trace is of course the amplifier output on CH1.

(Ignore the spike in the FFT trace at 10khz as this turned out to be caused by my scope. When the screen brightness was at normal this spike was present. As soon as the screen brightness/dimmer timer timed out and the screen dimmed this spike was no longer present. Took me a while to figure that one out).

Thanks for any help.

skip

 

How about using metal film resistors through out