Op amp based amplifier noise source clarification...

[jBam]

Hi all!  It's been a while - hope everyone is well...

I'm off on my theoretical ideas again, and would like to clarify where noise is typically generated in a basic amplifier in the hope of understanding the principles better.  Many of you have assisted me in the past with good advice on reducing noise with practical examples - but this time around I just want to make sure that I understand where it's actually coming from, and how it occurs etc...  Now I understand that there may be various micro noise sources, but I'm particularly interested in the MAIN source of noise in a basic amplifier here

With the attachment as an example, we have a very basic amplifier setup - an opamp with 3 resistors shown (only two are necessary for the amp stage, but the initial resistor will help with discussion). [EDIT: FYI - the "signal"and "signal+noise"tags are for illustration only - they're not actual sources or anything ...]

The way I understand it is:
*The signal approaches the op amp relatively unaffected - it passes through R70, but with very little current (basically none), there is little noise generated at this point;
*The signal exits the op amp, enters the feedback loop and passes through R72 and then through the drain to ground at R82 --> as we have a current to ground, we generate a relatively large amount of noise through both R72 and R82;
*This noise then enters the negative terminal of the op amp and hey-presto: appears at the output terminal inverted...

Is that right?

If so - that's good!  That means I understand the basics... If not, I'd love some one to assist me in understand this a little (or a lot) better!!

Another thing (if my understanding is correct):

I'm a little confused about what happens in the amp stage for the noise...  Does this enter the negative terminal; pop out and stabilise itself via the feedback loop at 1:1 amplification?  Or does the noise generated at R72 and R82 also actually get amplified?  I'd kind of expect it to be amplified, as noise increases with amplification; but I'm wondering if this is just perceived, and that the self generated noise here is just 1:1 and increased with increased amplification because of more current through the R72 R82 resistors (which would be interesting).  Of course, in any practical examples: you're significantly amplifying any noise in the original source too; which may add to my perceived understanding of noise increasing as amplification goes up ..


I'd really appreciate any guidance on this topic - just dreaming up more random ideas, and want to understand this topic properly before I dive into anything too deep ;P

 

[Bo Deadly]

occurs etc...  Now I understand that there may be various micro noise sources, but I'm particularly interested in the MAIN source of noise in a basic amplifier here

High gain is going to be the dominant factor. The "source" of the noise is self-noise of each part (a resistor generates noise) and EMI like hum from a ground loop, fluorescent lights and such. But under ideal operating conditions the "micro noise sources" are the only noise source. By itself they might not be much but they are if you amplify it by 60dB (1000x) like in a microphone preamp.

With the attachment as an example, we have a very basic amplifier setup - an opamp with 3 resistors shown (only two are necessary for the amp stage, but the initial resistor will help with discussion). [EDIT: FYI - the "signal"and "signal+noise"tags are for illustration only - they're not actual sources or anything ...]

The way I understand it is:
*The signal approaches the op amp relatively unaffected - it passes through R70, but with very little current (basically none), there is little noise generated at this point;

First, your circuit isn't quite right because you don't show how the + terminal is grounded so it's impedance is not defined. If we correct for this and say that the signal is going into where you labeled "Signal" and that that long vertical net is actually ground, then your circuit would work.

Then there would be a current. If say the signal was 10V at some point, that would be 10V / 10k = 1mA). But resistors are not generating more noise if they have higher currents (well slightly higher when thermal effects are considered). In fact, I'm inclined to say the circuit would probably generate less noise if it uses more current because higher currents would suggest impedances are lower and, again, noise is going to be dominated by whatever source resistances are amplified more. So lower impedance means lower noise.

 

[jBam]

Oh gosh hah... Yes sorry for the dodgy sketch!  I popped it together quickly for this post; but yes: not quite right ;P....

I'm out at the moment, but I'll tidy that pic up when I get home...

Thanks a lot for the input though squarewave ... I'll have a think about your comments and might come back to you with a few questions / clarifications!

 

[PRR]

Why "op amps"? There's noise in tube amps even when there is no gain-set resistors like you show.

> as we have a current to ground, we generate a relatively large amount of noise through both R72 and R82

Not correct.

> does the noise generated at R72 and R82 also actually get amplified?  I'd kind of expect it to be amplified, as noise increases with amplification

What is "noise"?? That is a purely human concept.

The amplifier only knows "signal".

There is hiss *everywhere*.

Sometimes there are musicians.

The amplifier does not know hiss from musicians.

It amplifies both equally.

In a well-regulated gain chain the primary hiss is the input transducer (microphone). There is also hiss in every amplifying device and every resistor. Ideally we can find parts and gains so these do not add much noise.

 

[rackmonkey]

...But resistors are not generating more noise if they have higher currents (well slightly higher when thermal effects are considered). In fact, I'm inclined to say the circuit would probably generate less noise if it uses more current because higher currents would suggest impedances are lower and, again, noise is going to be dominated by whatever source resistances are amplified more. So lower impedance means lower noise.

Actually, flicker or 1/f noise is the source of the excess noise in carbon comp resistors (present to varying degrees in all resistive elements) and is proportional to DC current. There's a threshold of current below which thermal noise will dominate (I used to have some measurements around from a thesis or something with measured thresholds for different resistors but I can't find it), but that's a very low current. Most audio circuits will have enough DC current to make flicker noise a factor.

 

[jBam]

EDIT:  Oh - thanks Rackmonkey - that's good info... more reading for me ...  Note that I mention "thermal noise"in the below post which I was typing when you posted... But in essence, I mean all internally generated noise in this discussion]

> as we have a current to ground, we generate a relatively large amount of noise through both R72 and R82

Not correct.

Yes - good work PPR... Thanks...  You always set me straight haha ... This was the area that I'd assumed noise was mainly generated; but it seems that's not correct which is good to know...  I'd been pondering thermal noise, and came to the idea that perhaps it's most prevalent when you have actual current through a resistor, possibly due to temperature changes - again, it seems that's not correct which is good to know ...

FYI - I'm interested in op amps because that's what I'm working with here.

What is "noise"?? That is a purely human concept.

Sorry - to be clear, I mean thermal noise --> random electron movements through resistors.  For the sake of it, "hiss" is a reasonable term - although I'm just interested in understanding the self generated thermal noise of a basic op amp amplifier circuit, out of interest.  With or without an input signal...

So I've updated the picture, which now adds R161 into the mix.  I've also ran a few noise sims - sorry, it's been a while and forgot that I can do this step haha...

From these sims it seems that the most noise of this basic circuit is actually generated in R70; followed by R82 (with the values shown).  The least amount is generated in R72 (which as PRR notes - is NOT the source as I originally thought).  I've also noticed that lowering R82 will eventually take you to a point where R161 and R82 noise is the same (R82 = 100ohm), and then lower than that will result in R161 producing more noise.

So - what's this mean?  As PRR said - it's coming from everywhere.  But there are some dominant sources.  I've been dreaming up concepts for reducing thermal noise in various circuits - why? pfft - just out of interest as always haha... I don't feel any closer to any solution; well... maybe one or two things... but it's a shame that it's the initial resistor that causes the bulk of the thermal noise --> that's the only one where I can't think of a way to isolate the noise from the signal... ;P I must love conundrums...

Regardless, this has been a good learning for me

Thanks for chiming in squarewave / and PRR.

I'm open to any further comments on thermal noise generation to of course

 

[ruffrecords]

There a many noise sources of different types in your circuit.

1. All the resistors generate a Johnson noise voltage due to their temperature and value. The noise generated by the 10K input resistor for example is a little under -110dBu. The Johnson noise in the 1K reistor is 10dB lower. Both these noises appear at the inputs to the op amp and are amplified by the 20dB stage gain. So the noise at the ouput due to these alone is -90dBu.

2. Resistors exhibit excess noise due to the current flowing through them but rarely if ever is it anywhere near as large as the Johnson noise so usually you can ignore it.

3. The op amp itself generates noise both as an equivalent input voltage and also due to the input  bias current and these are the two that chip manufacturers strive to minimise.

Basic low noise rules:

1. Don't put resistors in series with the input
2. Keep feedback resistor values as low as possible
3. Choose a low noise op amp.
4. Read Doug Self's excellent book on small signal design

Cheers

Ian

 

[jBam]

Cheers

Ian - thanks so much for taking the time to clarify all of the above.  It's hugely appreciated!  It's a detailed breakdown that's really helped my understanding of noise sources ...

I've been dreaming up design ideas that look into the question:  "If noise is introduced within a given circuit, can it be isolated?... and if it can, can it be removed??"...  Particularly in a low distortion circuit like a "perfect" amplifier as opposed to something that can dramatically modify the freq spectrum of the input source like an eq or even a compressor etc... 

I'm not sure why I want to do this - it's certainly not critical, as there's thousands of excellent designs out there with more than appropriate noise specs... but the theory has me intrigued!... Like a little puzzle

I might come back to you with another question or two at some point ... So far I can theoretically extract the self generated noise out --> I'm just not quite sure that I can then use that to cancel out noise in the circuit without adding a new noise source into the circuit again which isn't part of the extracted / isolated noise... Orr... wait t second... mayyybe... **trundles back into the studio**

 

[abbey road d enfer]

Your questions show that you need to understand how noise is generated. There are two main permanent noise components: brownian (thermal agitation) and shot noise (the fact that current is constituted of discrete particles). Any source (microphone, pick-up, synth...) has a source resistance, which creates noise. Then every component in the electroacoustical chain (including transistors and tubes) includes resistances that all contribute. Then there is also current everywhere, and this also adds noise.
You have a good starting reference there
https://en.wikipedia.org/wiki/Noise_(electronics)

 

[Newmarket]

Basic low noise rules:

1. Don't put resistors in series with the input
2. Keep feedback resistor values as low as possible
3. Choose a low noise op amp.
4. Read Doug Self's excellent book on small signal design

I'll second that and add in :

Buy and read Horowitz and Hill - 'The Art of Electronics' (Cambridge University Press)

 

[merlin]

1. Don't put resistors in series with the input
2. Keep feedback resistor values as low as possible
3. Choose a low noise op amp.
4. Read Doug Self's excellent book on small signal design

3b. So-called 'low noise' opamps are normally low voltage noise, meaning they need a low source impedance to shine. If your source impedance is high then you need a low current noise opamp.

Edit: I can't believe I wrote those back to front!  :

 

[Newmarket]

3b. So-called 'low noise' opamps are normally low current noise, meaning they need a low source impedance to shine. If your source impedance is high then you need a low voltage noise opamp.

No - if the source impedance is high then you need an op amp with low current noise (since the current noise produces a noise voltage across the source impedance ). eg for a high source impance you might opt for a FET input type opamp where the current noise is lower than a bipolar input type but the voltage noise is greater eg - compare OPA134 to NE5534.

 

[JohnRoberts]

3b. So-called 'low noise' opamps are normally low current noise, meaning they need a low source impedance to shine. If your source impedance is high then you need a low voltage noise opamp.

Total EIN (equivalent input noise), is the combination of input noise voltage and input noise current (times network resistance).

Different technology op amps have different strengths and weaknesses. Bipolar input devices will generally have lower ein noise voltage but higher ein noise current. FET input op amps will generally have higher ein noise voltage with orders of magnitude lower ein noise current.

The optimal choice depends on the impedance of the network interfacing with, and don't ignore that resistors make noise all by themselves (Johnson or thermal noise). 

The modern op amps we have today are so much better than what we had last century but you still need to match the device and technology to the application.

JR

 

[ruffrecords]

Just to put this all into context, all the effects of noise of all kinds can be reduced to a single figure for an amplifier call the noise factor (NF). Simply stated, it is the number of dB noisier that an amp is compared to a perfectly noiseless amplifier under the same conditions.

The noise is a 150 ohm resistor in a 20KHz bandwidth at 20 something degrees Celsius is about -131dBu. A noiseless amplifier with 60dB gain would raise this noise level to  -131 + 60 dBu = -71dBu. So no matter how good your mic pre is, the noise from a typical 150 ohm mic will never be less than this at this gain.

All mic pres these days are specified in terms of an equivalent input noise (EIN) with a 150 ohm resistor across the input. So if a mic pre specifies an EIN of -130dBu at 60dB gain then you know the noise at its output will be -130 +60 dBu = -70dB; in other words 1dB noisier than a perfect mic pre. So its noise factor (NF) is 1dB. No matter how much you try to improve this mic pre you will never get it more than 1dB quieter.

Most mic pres today will achieve a NF of 3dB or less. Even the old Neve 1073 achieved a NF of 5dB and that was nearly 50 years ago. My tube mic pres regularly achieve 6dB NF. So despite nearly 50 years of development, mic pres are generally only a few dB quieter.

There's more but that will do for now.

In 999 out of 1000 cases, the NF of a mic pre does not matter.

Cheers

Ian

 

[jBam]

Hi guys - thanks for the discussion; sorry - been away on business...

This:

In 999 out of 1000 cases, the NF of a mic pre does not matter.

..will help me sleep at night haha...

Still interested in the concepts I've been dwelling on; and yes - well aware that "noise" prior any idea I have is more of an issue than resolving it in the small stage of much much bigger chain of events that is the pre itself... But yes: still interesting as a concept (noise reduction / cancellation).

I'll sleep well; but keep on dreaming ... May I'll even post an idea at some point I have if I can get it stable in a simulation; or pop back with actual results if I build something odd

As always - great learning here though all, so thanks...!

 

[Newmarket]

The noise is a 150 ohm resistor in a 20KHz bandwidth at 20 something degrees Celsius is about -131dBu. A noiseless amplifier with 60dB gain would raise this noise level to  -131 + 60 dBu = -71dBu. So no matter how good your mic pre is, the noise from a typical 150 ohm mic will never be less than this at this gain...

Indeed. There are various ways to state the noise. Can be a tad confusing if not familiar.
Personally I like to look at the EIN or NF at a lower gain say +40dB. After all it's unlikely that you'll be recording with all mic preamps cranked up to +60dB which is often the max available.

 

[abbey road d enfer]

Indeed. There are various ways to state the noise. Can be a tad confusing if not familiar.
Personally I like to look at the EIN or NF at a lower gain say +40dB. After all it's unlikely that you'll be recording with all mic ampos cranked up to +60dB which is often the max available.

To my knowledge, there is no mic preamp that has as good an EIN at 40dB than at 60. That's probably why everybody publishes values at 60dB gain. I fully agree that 60dB gain is seldom used for the majority of music recording. Movie, television, theatre and nature recordists may disagree.

 

[Newmarket]

To my knowledge, there is no mic preamp that has as good an EIN at 40dB than at 60. That's probably why everybody publishes values at 60dB gain. I fully agree that 60dB gain is seldom used for the majority of music recording. Movie, television, theatre and nature recordists may disagree.

Exactly. The common topologies all have increasing resistance values where it matters as gain decreases hence EIN goes up.
Pretty much everyone serious can quote -128dBu or so  at+60dB gain which sounds great !
Obviously that is important - esp for the applications you outline + low output transduces eg ribbons - but it would be nice to see the figures at lower gains.
There's scope for keeping resistances low but it can become a bit surprising how much power you might need to dissipate and all the consequences that flow from that.

 

[JohnRoberts]

To my knowledge, there is no mic preamp that has as good an EIN at 40dB than at 60. That's probably why everybody publishes values at 60dB gain. I fully agree that 60dB gain is seldom used for the majority of music recording. Movie, television, theatre and nature recordists may disagree.

I have been looking at this for years and it seems dedicated mic preamps could be optimized for the lower gain required by the selected A/D convertor they want to feed. No sense making 30Vp-p to feed a 5V codec.  Some of the modern uber low noise op amps also drive rather high current outputs so in principle you could drop the feedback network resistance values to reduce their Johnson noise contribution.

BUT, other noise sources generally dominate in the real world.

JR

 

[ruffrecords]

To my knowledge, there is no mic preamp that has as good an EIN at 40dB than at 60. That's probably why everybody publishes values at 60dB gain. I fully agree that 60dB gain is seldom used for the majority of music recording. Movie, television, theatre and nature recordists may disagree.

I agree. At Neve back in the 70s when I knew zilch about mic pre design, we were told to consider a mic pre as having two noise sources, an input noise and an output noise. Even if the gain is reduced to zero, the output stage still contributes some noise. It is very low, probably in the -90 to -100dBu region but it is there. At high gains, the noise from the input stage is far greater than the output noise so it can be safely ignored;  at 60dB gain a perfect mic pre output noise will be around -70dBu. But drop the gain by 20dB and the input stage noise and the output stage are very similar and both contribute to EIN which is reduced as a result. As the gain is further reduced, the output noise starts to dominate and EIN drops as a consequence.

However, as the gain is reduced from 60dB to 40dB, the noise at the output of the mic pre drops by nearly the same amount so the important thing, the signal to noise ratio, does in fact increase. Further illustration of the futility of using a single figure to describe performance.

Cheers

Ian