Interesting behavior from a diff amp - need help...

Thread Starter

skip.ele

Joined Nov 27, 2011
53
Can anyone explain why circuit A has a totally silent output while circuit B has a low amount of hiss on it's output? All resistors are metal film (1%) and so are the capacitors (5%). I was expecting the exact opposite to occur. I was expecting circuit B to be silent because the inputs are perfectly balanced (in theory) and was was expecting circuit A to have some noise (thermal or RF) on it's output because the inputs are not balanced. Please disregard the strange apearance of the circuit. It is the preamp of an audio amplifier and the mic is not shown. I stumbled across this behavior while experimenting with eliminating noise from the mic circuit. Also, this circuit is built on a 4 layer board with a ground plane and good noise immunity and the op-amp is powered from a split supply of +/- 9v.

I welcome any discussion on this. Thanks in advance.
 

Attachments

Tesla23

Joined May 10, 2009
542
Circuit B is amplifying the op amp voltage noise (16nV/rtHz) by about 51, giving about 82uV in 10kHz. Circuit A has unity gain, giving in comparison about 1.6uV.
 

jimkeith

Joined Oct 26, 2011
540
Something is amiss--check your DC levels to verify that your outputs are running close to ground potential and not stuck in saturation.
 

SgtWookie

Joined Jul 17, 2007
22,230
Tesla23 gave you the reason why.

If you want to eliminate the noise, use a better opamp.
The TL072 is a low-noise version of the TL082, and has otherwise identical characteristics.
There are much newer opamps available with much lower noise, lower THD and greater bandwidth. The MC33078 dual/MC33079 quad is one such opamp:
http://www.onsemi.com/PowerSolutions/product.do?id=MC33078

You can purchase these opamps from several authorized suppliers for under $1 USD.
 

crutschow

Joined Mar 14, 2008
34,285
Circuit A does not have unity gain. It has an AC gain of 50 from either the inverting input or the non-inverting input.

The schematic says it has a "floating input". That won't work. The input signal common must be connected to the circuit common or to the other differential input.
 

Thread Starter

skip.ele

Joined Nov 27, 2011
53
ok but remember, these circuits as shown are just a scenario I found to have unexpected results. I was curious as to why I was getting the results that I did. Now I'm looking to understand the results. I can understand the comments made above about circuit B showing noise on the output that is I guess thermal noise internal to the op-amp. I'm just not sure I understand why circuit A is quiet/no noise out. I dont understand why floating one input gets rid of the affects of the op-amp's internal noise. It still seems the me that circuit A should have the same or more noise out than circuit B.
 

Thread Starter

skip.ele

Joined Nov 27, 2011
53
Ok, i did some testing tonight and i think i am starting to understand the "unity gain" comment made by Tesla23 about circuit A. I started pulling the floating input of circuit A down to common (ground) with resistors instead of a short circuit as in circuit B. I found that as i increased the resistor value the noise on the output became less. once i increased the value of the resistor to approximately the same value as the feedback resistor (100k) the noise on the output was no longer noticeable to my ears. These resistor values approximated unity gain and thus no amplification of the noise. I am assuming that since an open circuit is equivalent to infinite resistance then for any practical purpose the gain of circuit A is unity or less. Based on this i think i understand what Tesla23 was saying. I still dont fully understand why the traditional feedback gain affects the amplification of the noise internal to the op-amp though since that noise is not coming in through the input resister(s) per circuit A as drawn. It kind of makes sense to me when i add the 100k resistor between the floating input and ground because that completes a circuit for the input current to flow.
 
Last edited:

Thread Starter

skip.ele

Joined Nov 27, 2011
53
DC biasing appears to be ok. The output is riding on only about 1mv DC. I cant imagine any shot noise coming from that.
 

thatoneguy

Joined Feb 19, 2009
6,359
Remember that components add noise, even though in tiny amounts. Try switching your transistors out for a 0.1% wirewound resistor vs 1% metal film vs 5% carbon and report the results.

Add a diode near or at its Vf point and see what the noise is then.

If you have a scope, it is very interesting to see how different components change noise.

--ETA: Add a buffer amplifier after this input amp so the noise isn't loaded down by listening to it.
 
Last edited:

Tesla23

Joined May 10, 2009
542
Thanks all.

Tesla23, can you explain why circuit A has unity gain?

Thanks.
Circuit B is easy, the gain from the non-inverting input is 1+R2/R1 = 51. This amplifies the op-amp voltage noise, which looks like the dominant noise source.

When you allow C1 to float, then the gain from the non-inverting input to the output drops to 1. It is the R2:R1 divider that gives you the gain of 51, effectively allowing R1 to float takes it out of the circuit and the feedback from R2 results in a unity gain stage. Or if you like, Av = 1+R2/R1, the floating input effectively makes R1 infinite.
 

SgtWookie

Joined Jul 17, 2007
22,230
Circuit A does not have unity gain. It has an AC gain of 50 from either the inverting input or the non-inverting input.
Were there input to the non-inverting input at C1, the gain would be 51.

The schematic says it has a "floating input". That won't work. The input signal common must be connected to the circuit common or to the other differential input.
But in this test scenario, it DOES have a floating input.

Since a floating input would have nearly infinite impedance, and the noninverting input to U1 is also nearly infinite, you essentially wind up with:
Av = 1+R2/infinity = 1+0 = 1.
 

crutschow

Joined Mar 14, 2008
34,285
Were there input to the non-inverting input at C1, the gain would be 51.

But in this test scenario, it DOES have a floating input.

Since a floating input would have nearly infinite impedance, and the noninverting input to U1 is also nearly infinite, you essentially wind up with:
Av = 1+R2/infinity = 1+0 = 1.
The circuit is a classic differential input op amp circuit, so the gain from either input has a magnitude of 50 when the other input is grounded (or has a zero impedance source). Otherwise it wouldn't reject common-mode signals.

But if the C1 input is floating then the noise gain is 1 which is why the noise is less for circuit A as compared to B. I missed that point in my first post.
 

Thread Starter

skip.ele

Joined Nov 27, 2011
53
I'm not sure what happened but this thread jumped to another one. The solution to my question was essentially this: circuit A has unity gain because the floating input's impedance is essentially infinity. Therefore the internal op-amp noise is not amplified and instead is seen on the output at only unity gain.

Thanks to all who helped me to understand this behavior.

This is the first time i have dealt with noise at this level and i guess i was assuming the noise was coming from an outside source. I never imagined that much noise coming from the op-amp itself. I'm now looking at newer op-amp chips with noise levels many times lower than the old TL082.

Thanks again.
 

SgtWookie

Joined Jul 17, 2007
22,230
I received your E-mail.
It seems that you now understand what is going on with the circuit. And yes, more a more modern opamp will definitely help.
 
Top