# Single-supply op-amp help!

Discussion in 'Electronic Design' started by Hw, Aug 1, 2005.

1. ### HwGuest

Hello.

I am trying to make a single-supply op-amp circuit for an electret
microphone so I can pick-up voice 2-3 feet away (i.e. like a
speakerphone on a desk).

Circuit is split into two stages, 40 dB gain first stage with some
filtering and I'm trying to get 40 dB in second stage (no additional
filtering). I am having problems getting the 2nd stage to work
correctly.

The first stage is built following this app. note:

http://www.st.com/stonline/books/pdf/docs/8698.pdf

2nd stage is a standard non-inverting op-amp configuration with gain =
40 dB via 100k and 1k resistors.

Questions:

- Is 80 dB enough gain for a typical electret mic to pick-up voice 2-3
feet away?

- Should I AC couple between stages to avoid offset issues?

- If no to the above, do I NEED to have the second stage vcc/2
referenced, given my first stage biases the AC signal about vcc/2 ?

In my SPICE simulations, if I don't ref. the second stage to vcc/2, I
get no output, but I find that confusing since the first stage already

- Does inverting/non-inverting make a difference for second stage? I
don't think the 180 deg phase shift makes a difference since there is
only one mic but perhaps I am neglecting something.

- Anyway to keep the input impedance low for the second stage?

The ST PDF claims the input impedance is equal to R4 (18k), but
following this PDF PDF (no, I am not a Cornell student, but I couldn't
find this PDF on TI's site) the impedance would be closer to 811 ohms.

http://instruct1.cit.cornell.edu/courses/bionb440/datasheets/SingleSuppy
..pdf

Why in ST's configuration is the input impedance not equiv to R8||R7?
C4 seems like it forms a low-pass with R3...

HW.

2. ### martin griffithGuest

A single stage of 40dB should be quite OK for an electret mic capsule.
The majority of professional micamps rarely go above 66dB, and that is
usually for use with very low output ribbon mics.

Dont bother with a second stage.

Re: the input Z, the non inv input is more or less = to 18K, but this
is swamped by the low source Z of the microphone. Its only there to
set it at 1/2Vcc
but the inverting input is needs to see a low source z for noise
considerations, which is r8//r7. This is not the input z of the opamp,
but what it sees

Both inputs are Hi Z, but its the networks around the beast that
determin the operating impedances

I would replace R7 with a 5K or 2K pot in series with 470R, so you can
tweak the gain. that will give you up to 45dB gain approx.

AC couple, yes, in this type of setup

Most of this is covered in the data sheet, it just needs reading a bit
more carefully

martin

3. ### Pooh BearGuest

Excessively complicated. You can do the same probably better with any half
decent modern op-amp.

Sounds plenty to me.
Yes. Indeed essential if using a single rail supply.

Each stage needs to be biased to Vcc/2 individually.

You don't understand circuit basics. This is common with ppl who use Spice
to avoid understanding circuits.

No - phase / signal polarity is irrelevant to your application.
Low where ? Any op-amp output has quite low impedance.

Indeed - see R4.
The page cannot be found
Uh ? R8 and R7 form the feedback network. Irrelevant to input impedance.
Time to study circuit basics.

I suggest you learn some op-amp basics.

TI have an excellent app note - it's a book in effect something like 400
pages - but I forget the name and its title ( may be Op-amps for Everyone )
I may recall it later.

National semiconductor have the rather old ( refers to too many old devices
) but still useful Audio Handbook ( now reprinted ) and the Linear
Applications Handbook - may be available in hardcopy but should be available
as a pdf.

Graham

4. ### RobinGuest

If the +ve input goes, say, a little high then the output (of the
op-amp) will try to slam hard against the positive rail (because the
op-amp's gains is very high).

As soon as the output starts to move high, it is fed back to the -ve
input (reduced by the potential divider of R7,8) and this "cancels" the
above change that caused the output to move high.

This "action" makes +ve look like a "high impedance" - its the same
thing as power steering: a slight displacement allows a powerful fource
to "null it out", moving the wheels the while.

It is fun to work out; the input impedance of +ve... and then you will
never have to depend on a "PDF" again!

Here is an example of how to do it for the simpler inverting op amp
arrangement:

1) If -ve input changes by delta voltage "dV" (w.r.t. the +ve input).
2) Then the current at -ve will change by "dI".
3) And evidently dV / dI == input Z at -ve.
....
a) Meanwhile the output moves -dV * A (A == open loop gain).
b) And most of dI flows via the FB resistor (very little via -ve).
c) And most of -dV * A volts is across the FB resistor (dv << -dv * A).
d) So dI == -dV * A / FB resistor.
....
4) So from 3) above input Z at -ve == FB resistor / A

Cheers
Robin

5. ### BanGuest

There are a few mistakes in the ST pdf, 1Pa corresponds to 94dB SPL not
110dB. I also think the circuit can be improved.
You do not need 80dB of gain either. At the normal conversation we have 60dB
SPL at 1m. The speaker may talk louder (80dB) and speak closer to the mike
12.5cm (+18dB), so we have to process a max. SPL of 98dB, which corresponds
to 10mVeff or 28.25mVpp. We want to amplify this to the max. possible O/P
voltage of 2.5Vpp, so the gain needs to be 88.5 or 38.9dB. Exactly what the
amp is supposed to do.
The mike has 58dB S/N referenced to 94dB, so the noise is 7.94uVrms, which
is amplified to 715uVrms. This voltage will be -62dB below maximum, so it
will not make sense to amplify more, since the details get lost in the noise
anyway.
You are wrong with the idea of the low input impedance. C4 is for the audio
frequencies a dead short to gnd, so the input impedance of the amp is indeed
18k. This is important for the calculation of C5. The output impedance of
the mike(2.2k) is in series with it as written in the PDF. The output of the
mike is loaded with the 2k2 in parallel with the 18k, so it is 1.96k plus
100R for the EMI filter. This will reduce the output voltage by 10% or -1dB.

As to the improvement of the circuit, I can respond later, too much to do
now.

6. ### HwGuest

My confusion initially stemmed from getting one single-stage amp drive
another single-stage amp.

The first stage, 39 dB gain, works, but 2-3 feet away the output
amplitude drops and it is hard to hear the voice, hence my desire to put
a second stage.

I guess it makes sense to AC couple from stage to stage since any DC
offset would pass through the gain stages otherwise and because my
signal is AC, I can strip away (this potentially large) DC without
too much by using it in the second stage, but it is needed anyway.

The confusion I had regarding the input impedance came from page 6 of
the second PDF I was referencing (now in TinyURL form should load):

http://tinyurl.com/dutqx

In the non-inverting configuration shown, he doesn't put any bias
resistors for the AC signal coming in which seems odd since the cap
should strip away the Vcc/2 DC bias added (presumably?) in a previous
gain stage. Regardless, I suppose for his circuit the input impedance
is considered at the V+ stage since:

* V- has the signal directly coming in after the AC cap
* V+ has the feedback resistors which are referenced to Vcc/2, hence
there is a bias current flowing.

Ban of course you are right, C4's impedance for the audio frequencies is
essentially a short (~36 to ~7.2 ohms from 2kHz to 10kHz), and the
Zin_V+ for the ST op-amp circuit is 18k.

Thanks again.
HW.

7. ### Pooh BearGuest

The page cannot be found

Sorry - no luck still. Are you perhaps logged into the Cornell site ?

Graham

8. ### HwGuest

Graham,

Please try this (and no, I am not a student at Cornell or anywhere
else):

http://tinyurl.com/ck9l4

HW.