Here is some things I learned. If you have to design a low noise circuit, you have to double check things to which you would normally not pay attention....PSRR and sometimes CMRR. Power supply disturbances and CM noise can easily exceed your noise. Make sure you have a feature on your PCB to short amplifier inputs on the PCB..don't rely on a short at input jack inches away during development.
Just something I like to throw into discussions of op-amp noise ... the easiest way to reduce op-amp noise is to not use an op-amp. There are 40+ transistors in a typical low-noise op-amp, and they all contribute to the typical EIN, which if you are wanting a high voltage gain, say 100×, can result in unusably high output noise even with allegedly "low noise" op-amps. Generally, a simple 3-transistor (or even 2-transistor) discrete amplifier circuit will be able to give you that gain with 40dB less noise. Most mic pre-amps use this kind of circuit for their initial gain for this reason, mostly using complementary doubling on the input pair to reduce non-linearity.
@@orientaldagger6920 definitely worth it. the difference is enormous. my experience is of using an NE5532AP (5nV/√Hz typ.EIN) which at 100x gain (40dB) from a mic source was borderline unusable (I think the noise was somewhere around -30/-35dB after some follow-up gain because 100x wasn't enough), whereas a balanced bjt initial mic gain stage can give the same gain and more without adding anything noticeable, and that was using cheap basic BJT's (BC556). All commercial balanced mic inputs (that I have disassembled) do this. They may also have an op-amp buffer before the next stage, but the gain is done by the transistors. The transistor circuit also has the benefits of being a true differential amplifier (no slight differences of gain from hot and cold due to resistor and gain pot tolerances) and having the possibility of balanced output to the next stage with no extra circuitry.
@@treelibrarian7618what about active rc low pass and high pass circuits? It gets really complicated to use transistors to get the similar amount of roll offs
@@Yeaitts Not sure what you mean by complicated, I have made active sallen-key RC low and high-pass with just a single-ended bjt emitter-follower as the output buffer before. They can accept the DC voltage difference from output to input due to the capacitors in the feedback path. generally, though, the signal will already be large enough and the gain low enough that a low-noise op-amp is fine from a noise perspective. My original comment is specifically about the case of Mic pre-amps where the -30dB details of the input signal is often in a similar rms voltage range to the EIN of the op-amp, so you end up with a lot getting lost in the noise floor. Due to the nature of bjt circuits there will naturally be some low end rolloff from the capacitors used for adding and removing biasing. personally I wouldn't try adding in strong (12dB/octave or more) filtering to the mic preamp directly (maybe tune up the natural low-cut fro the biasing to get more rumble filtering if needed), but rather get the signal up to line level and then think about filtering it. And use op-amps if you like, they exist for a reason, but remember there are trade-offs for that simplicity and I do recommend learning about transistors of all kinds for your future in electronic design.🙂
@@treelibrarian7618well how much transistors are you exactly talking about to get the same roll off as from an active op amp rc filter? is that effort worth it, like does it give zero noise operation considering that the input signal and the vcc is also clean? i was doing a custom bandpass adjustable filter for my subwoofer amplifier. it works fine after compensation and all but there's still this wining FM static kind of noise coming from the sub. its not that loud but it is not consistent it creates a fluctuating noise that is the real issue. if it was same our brains would just tune it out over time
We did it all the time as the signal to noise figure was very important to the customer, you had to remind them that the noise floor was also important. If we stated the signal to noise was 60dB and the noise floor was 100uV, then they would not get 60dB down on a 500mV output. But they did not like a statement like the signal to noise was 60dB or the noise floor whichever is greater..
Thank you very much Robert for bring him in your channel... it really helps a lot... Can anyone give a pdf downloadable link to A. Key's book "Operational Amplifier Noise: Techniques and Tips for Analyzing and Reducing Noise"
There's white noise, pink noise, shot noise, popcorn noise, thermal noise, Johnson noise, Nyquist noise, phase noise.... some of which are the same (or similar) phenomenon. Engineers use mathematical models to try to predict (usually estimate) the effect of 'noise' on designs. Predicting (and especially estimating) is not the same as 'explaining.' Johnson measured the noise named after him, and Nyquist developed a mathematical model to help predict it. To say thermal noise is the result of thermal agitation of electrons... IMO... isn't to say very much. You get deep into the weeds very quickly on the topic of noise.
@@seanm8030 It is applied theory. (Just like any engineering design.) One doesn't always get lucky with a high speed design. Knowing about the types of noise, and the models for them, won't hurt anybody.
@@willthecat3861 Classifying the RF spectrum is applied theory? Seems to me like it was decided by the ITU. In addition neither I nor anybody else is arguing that knowing the mechanism for shot noise is "bad." It's simply not the subject of the video.
@@seanm8030 I dunno, if it's applied theory: that's something you're asking (or stating), not me. If it was your video, you could state the subject. Is it your video?
How about the oscillator that works at 1.4 GHz frequency and its output noise is 1.6 nV/sqrt(Hz), while total noise is 53 uV/sqrt(Hz), and phase noise is -133 dBc/Hz? The OP-AMP's bandwidth is 900 MHz at -3 dB.
As far as I knew the bigger the resistance, the bigger the voltage noise that one electron generates, just as the same current creates a bigger voltage drop over a bigger resistor. But please correct me if I am wrong.
This is an interesting fact, but if you shunt down an OpAmp power line with capacitors of all possible values (and types), this will kill a sound (if we are talking about an amplification of a sound). Some noise must be present. It’s like a mac rounded corners vs ms square ones. The rounded looks more natural
I think he said you shunt down values exceeding the frequencies you are planning to analyze, unless you mean something else! so for an audio frequency, I don't really think you need frequencies above 25 kHz, so you only remove those.
Here is some things I learned. If you have to design a low noise circuit, you have to double check things to which you would normally not pay attention....PSRR and sometimes CMRR. Power supply disturbances and CM noise can easily exceed your noise. Make sure you have a feature on your PCB to short amplifier inputs on the PCB..don't rely on a short at input jack inches away during development.
Just something I like to throw into discussions of op-amp noise ... the easiest way to reduce op-amp noise is to not use an op-amp. There are 40+ transistors in a typical low-noise op-amp, and they all contribute to the typical EIN, which if you are wanting a high voltage gain, say 100×, can result in unusably high output noise even with allegedly "low noise" op-amps. Generally, a simple 3-transistor (or even 2-transistor) discrete amplifier circuit will be able to give you that gain with 40dB less noise. Most mic pre-amps use this kind of circuit for their initial gain for this reason, mostly using complementary doubling on the input pair to reduce non-linearity.
Is it worth it? You can get opamps that have low single digit nV/rhz broadband noise.
@@orientaldagger6920 definitely worth it. the difference is enormous. my experience is of using an NE5532AP (5nV/√Hz typ.EIN) which at 100x gain (40dB) from a mic source was borderline unusable (I think the noise was somewhere around -30/-35dB after some follow-up gain because 100x wasn't enough), whereas a balanced bjt initial mic gain stage can give the same gain and more without adding anything noticeable, and that was using cheap basic BJT's (BC556). All commercial balanced mic inputs (that I have disassembled) do this. They may also have an op-amp buffer before the next stage, but the gain is done by the transistors. The transistor circuit also has the benefits of being a true differential amplifier (no slight differences of gain from hot and cold due to resistor and gain pot tolerances) and having the possibility of balanced output to the next stage with no extra circuitry.
@@treelibrarian7618what about active rc low pass and high pass circuits? It gets really complicated to use transistors to get the similar amount of roll offs
@@Yeaitts Not sure what you mean by complicated, I have made active sallen-key RC low and high-pass with just a single-ended bjt emitter-follower as the output buffer before. They can accept the DC voltage difference from output to input due to the capacitors in the feedback path. generally, though, the signal will already be large enough and the gain low enough that a low-noise op-amp is fine from a noise perspective. My original comment is specifically about the case of Mic pre-amps where the -30dB details of the input signal is often in a similar rms voltage range to the EIN of the op-amp, so you end up with a lot getting lost in the noise floor. Due to the nature of bjt circuits there will naturally be some low end rolloff from the capacitors used for adding and removing biasing. personally I wouldn't try adding in strong (12dB/octave or more) filtering to the mic preamp directly (maybe tune up the natural low-cut fro the biasing to get more rumble filtering if needed), but rather get the signal up to line level and then think about filtering it. And use op-amps if you like, they exist for a reason, but remember there are trade-offs for that simplicity and I do recommend learning about transistors of all kinds for your future in electronic design.🙂
@@treelibrarian7618well how much transistors are you exactly talking about to get the same roll off as from an active op amp rc filter? is that effort worth it, like does it give zero noise operation considering that the input signal and the vcc is also clean? i was doing a custom bandpass adjustable filter for my subwoofer amplifier. it works fine after compensation and all but there's still this wining FM static kind of noise coming from the sub. its not that loud but it is not consistent it creates a fluctuating noise that is the real issue. if it was same our brains would just tune it out over time
Great video! Noise is very important topic especially when you are dealing with sensitive circuits. Thanks Robert & Arthur.
Fascinating topic. The material was interesting from start to finish.
Great Video! I am a huge fan of Arthur. I hope there is more to come about detailed analog circuits
We did it all the time as the signal to noise figure was very important to the customer, you had to remind them that the noise floor was also important. If we stated the signal to noise was 60dB and the noise floor was 100uV, then they would not get 60dB down on a 500mV output.
But they did not like a statement like the signal to noise was 60dB or the noise floor whichever is greater..
AMAZING video. Thanks to both Robert and Arthur!
Wow ..The legend Art Kay!
Great topic Robert! Very useful. Thank you.
Nice introduction to the topic.
Where can we get a copy of the TI slides from this lecture ?
Just when I was designing active band pass filters :D
Great video ! thanks
@Robert Feranac please bring some course on analog and power electronics...
Thank you very much Robert for bring him in your channel... it really helps a lot... Can anyone give a pdf downloadable link to A. Key's book "Operational Amplifier Noise: Techniques and Tips for Analyzing and Reducing Noise"
There's white noise, pink noise, shot noise, popcorn noise, thermal noise, Johnson noise, Nyquist noise, phase noise.... some of which are the same (or similar) phenomenon. Engineers use mathematical models to try to predict (usually estimate) the effect of 'noise' on designs. Predicting (and especially estimating) is not the same as 'explaining.' Johnson measured the noise named after him, and Nyquist developed a mathematical model to help predict it. To say thermal noise is the result of thermal agitation of electrons... IMO... isn't to say very much. You get deep into the weeds very quickly on the topic of noise.
forgot to add flicker noise, and burst noise.
Sure, but this is not an electrical engineering theory discussion. This is a demonstration of techniques used to design PCAs.
@@seanm8030 It is applied theory. (Just like any engineering design.) One doesn't always get lucky with a high speed design. Knowing about the types of noise, and the models for them, won't hurt anybody.
@@willthecat3861 Classifying the RF spectrum is applied theory? Seems to me like it was decided by the ITU. In addition neither I nor anybody else is arguing that knowing the mechanism for shot noise is "bad." It's simply not the subject of the video.
@@seanm8030 I dunno, if it's applied theory: that's something you're asking (or stating), not me. If it was your video, you could state the subject. Is it your video?
How about the oscillator that works at 1.4 GHz frequency and its output noise is 1.6 nV/sqrt(Hz), while total noise is 53 uV/sqrt(Hz), and phase noise is -133 dBc/Hz? The OP-AMP's bandwidth is 900 MHz at -3 dB.
This was great! Thank you from a new subscriber!
Thank you
Interesting topic; I loved it
As far as I knew the bigger the resistance, the bigger the voltage noise that one electron generates, just as the same current creates a bigger voltage drop over a bigger resistor. But please correct me if I am wrong.
This is an interesting fact, but if you shunt down an OpAmp power line with capacitors of all possible values (and types), this will kill a sound (if we are talking about an amplification of a sound). Some noise must be present. It’s like a mac rounded corners vs ms square ones. The rounded looks more natural
I think he said you shunt down values exceeding the frequencies you are planning to analyze, unless you mean something else! so for an audio frequency, I don't really think you need frequencies above 25 kHz, so you only remove those.
Which software is used as an oscilloscope?