Dave explains one of the most confusing parameters in an opamp datasheet, Input Noise Voltage Density, that mysterious nV/RootHz figure.
Along with different types of opamp noise, corner noise frequency, and how to calculate output noise spectral density.
And how to use a Dynamic Signal Anayser to measure the Power Spectral Density response over frequency of various opamps.
Forum: http://www.eevblog.com/forum/blog/eevblog-528-opamp-input-noise-voltage-tutorial'>http://www.eevblog.com/forum/blog/eevblog-528-opamp-input-noise-voltage-tutorial
All EEVblog Opamp related videos are here:
https://www.youtube.com/playlist?list=PLvOlSehNtuHu2FviAaZaiyXwN41G4b1Lf
EEVblog Main Web Site:
http://www.eevblog.com
EEVblog Amazon Store:
http://astore.amazon.com/eevblogstore-20
Donations:
http://www.eevblog.com/donations/
Projects:
http://www.eevblog.com/projects/
Electronics Info Wiki:
http://www.eevblog.com/wiki/

Hi Welcome to Fundamentals Friday Today, we're going to take a look at Opamp Voltage Noise. Now this can be a real big can of worms, so I'm going to only open it just a little bit today and we're going to take a look at one of the more confusing parameters on an Opamp data sheet. And that's input noise, voltage, density, and input noise voltage. If you didn't know, well, you do now that any Opamp is going to have inherent noise in it.

just like all components and all wires and everything else has inherent Noise Within It and the Op Amp is no different and that's what we're going to take a look at. Now, we're not going to take a look at anything around the circuit, the resistor noise, and other components and stuff like that. Just what's inherit in the Opab. And to do that, we're going to start by taking a look at a a typical data sheet.

Now let's take a look at the Op07 a typical Precision Opamp. Not particularly low noise, but it is one of the Jelly Bean Precision devices now. Um, it has a parameter here called input Voltage Noise and that's the noise effectively on the input and the units are very easy. They're microv volts in Peak to Peak and it's H called E N or VN depending on the data sheet could be called other things, but they're just, uh, typical labels for it and you know that figure might be familiar to you and it's fairly easy to understand.

Okay, I've in the case of the Oppo 7 we've got. 35 microvolts peak-to Peak input noise. So if we've got a voltage follower like this with a gain of one, we're going to get an output noise or an inherent noise in our opamp. Uh, in our uh, complete amplify here of that 0.35 microv volts Peak to peak.

Real easy to understand, but there's a catch. Take a look at the conditions that that value is measured over and it's actually 0.1 Herz to 10 Hertz bandwidth. And you might be uh, familiar with this from power supply specs. for example, they might specify the output noise of your Bench Lab power supply over typically a 20 MHz bandwidth.

Well, in this case, it's a very small low frequency bandwidth and we'll find out why later. it's 0.1 HZ to 10 HZ And this is is typically how they measure it. They've uh, got the Op amp here. It may have some, may or may not have some gain, the input will be grounded, it'll all be shielded of course, and then we'll have a band pass filter of 0.1 HZ to 10 HZ.

We'll have some more gain in there because we're talking about low signal levels that'll go into a scope and they can measure that value and they'll give you a peak-to peak or a maximum peak-to Peak signal. and they'll also give you take a look at this also in the D In most data sheets, they'll also give you a typical waveform as well. Once again, that's the bandwidth limited to 0.1 to 10 Hertz Very limited frequency range, so that's well and good if you're operating down in that frequency range in your circuit. Fantastic.

You've got this real world figure here. You understand it. It's easy. it's a peak maximum voltage, and you know, uh, what your system noise is going to be at least just due to the Op Amp.
Very simple, but what happens if you want to actually operate typically over a larger frequency range? Well, we're getting into something a bit more complicated called input noise Voltage density. You'll notice it's exactly the same, but they've added this word density. And if we go back to the data sheet and take a look at some typical figures for the Opo: 7. What do we get? Well, look, you can see that the conditions there.

There's three different values and these are called The Spot frequency values. In this case, we' got 10 Htz, 100 Herz, and 1 khz and we've ' got different figures for that 10.3 10, and 9.6 respectively. And you'll notice how it's slightly higher at lower frequencies and that's important, which we'll take a look at in a minute. But it uses these bizarre units which confuses a lot of people.

And it's Nanov volts per root Hertz And here it is. Once again, it's labeled exactly the same. En VN Exactly the same. but instead of microvolts peak to Peak, we've now got a value in nanov volt per root Hertz What does that mean? In a nutshell, it's spectral density I.E the density of the noise over a specific Spectrum or frequency range.

Just like our input voltage, noise was measured from1 Herz to 10 HZ. It needs a these. This unit here actually needs a frequency range over which it's going to be valid. Otherwise, it's a meaningless figure.

Now, the confusing part about these units of nanovolts per root Hertz is that you go well. what kind of units is that? Well, it's just voltage. it's you know, even though it's called nanovolts per root Hertz The per root Hertz part just specifies that it's defined over a frequency range because it a spectral density now. So we basically it is just a voltage.

That's all there is to it. Now, the data sheet, for example, for this one at a specific frequency, has for example, 10 NTS per root Hertz Now it's very important to understand that this is not divided by root Hertz It's per root Hertz and it's actually uh referenced to 1 Hertz. So it's 10 n Volts for every 1 Hertz of bandwidth. And that's the key to understanding this thing.

So if you've only got a 1 Hertz bandwidth, then your noise is going to be square root of 1 Hertz which is the same 10 NTS. But you know, usually you're not going to be operating over a 1 Herz bandwidth. So let's look at a 1 khz bandwidth and the formula then is F Max minus F Min That's a little bit complicated, but it's basically the bandwidth you're operating under. So if you operate circuit Uh is operating from 0o Hertz up to 1 khz then you've got a bandwidth of you know 1 khz minus 0 is 1 khz 10 NTS * The < TK of 1 khz gives you a final value of 316 Nan volts Easy.

that's how much noise. RMS By the way, this is all RMS noise in your Opamp inherent in your opamp. Just like this value up here it was microv volts, but it was specified in Peak to Peak this one up here. nanov volts per root Herz specified in RMS.
So you can see that the higher frequency range you operate over, the more noise you're going to have because it's multiplied by the square root of the frequency. If we operate over 10 khz there, it's going to be bigger noise once again, or 100 khz or a megahertz. The next important thing to understand is this is what is called input referred noise or equivalent input noise. You'll see these terms are various different types of terminology, but it means that this is the noise on the in the equivalent noise on the input of the opamp.

So what that means is it gets always gets multiplied by the gain of the opamp. In this case, we just got a gain of one. So in the case of this op07 36 NTS RMS on the input, same 316 NTS RMS noise on the output. Pretty low noise, but if you suddenly whack in a gain of 1,000 in there AV equal 1,000 Bingo you've gone from 316 NTS to 316 microvolts or3 molts.

Much higher noise. Now if you remember I said this was RMS So how do you convert it to possibly a more useful uh, maximum peak-to Peak value in your system? Well, this one gets a bit fuzzy and you have to introduce probability. Now what we're talking about here is uh, white noise. or you know, purely random noise which has your typical Gausian response like this and we won't go into hugely into types of noise, but it has that ging response now.

I've drawn a voltage here I've rotated the axes like that so positive and negative voltage noise can always be equally positive and negative doesn't just go positive and basically uh, the peak value. So this is just a typical voltage Peak Like this over time. So as you can see, you know the noise is completely random and what are these Peak values here going to be? This is where you get into that probability term Sigma Now if we look at the value of uh uh, plus - 3 Sigma there. Now basically what that means is that we have a 99.9% confidence or close to it that the peak-to peak noise is going to be within that specific value.

So that 3 Sigma value What? you to get that? That's a typical figure quoted so manufacturers might uh, typically Define The convert RMS to Peak to Peak by using a multiplier of time 6 or times 6.6 6.6 will give you uh, 99.9% probability. the noise Falls within a certain range, but it doesn't guarantee it. There's a .1% chance can be outside that and well, it's up to you as the system to designer to determine what probability you need. But that's a good ballpark, so multiply that value by about 6 or 6.6 So in our example of a gain of a th000 here, what's our output noise for this Op07 with 10 Nan volts per rout Hertz specified? Well, it's going to the output is going to be 316 microvolts RMS or around about 2.1 M volts Peak to Peak with a good confidence level.
and that is going to be your output noise just solely due to your Opam not taking into account any other components or any other part of your circuit. So that's really quite easy to understand once you know. Just multiply that figure by the square root of your bandwidth and you get your output noise in RMS Very simple, but yeah, there's more to it. Let's go a little bit deeper open that can of worms.

just a little bit more. And yes, hold on to your hat. We're going into a graph of noise voltage versus frequency here on Dual log axes. So we've got our nanov volts per rout Hertz here versus frequency.

And as I said log axes, that's important. So 10 HZ 100 HZ and it's not a linear increase. same with frequency 10 100 1K And then it's your typical log axes you should be familiar with. So the black line there is is our noise voltage and you will find this.

Typically find this curve in the data sheet as well and it'll always be in this particular form. And here's where the trick with all this opamp voltage noise comes in that we've effectively got two different types of noise in our opamp and they effectively split into different parts of the frequency spectrum. The higher frequency say from around 10 HZ or 100 Herz up typically is going to be your NY and white noise that we showed before and effectively what we're using up there for our input noise voltage density. That's our White Noise up there.

But all Op amps regardless of the type are going to have this characteristic response that tails up at low frequencies and this is called one onf noise. So white noise dominates at higher frequencies, one on F noise dominates at lower frequencies. Usually you know around about 10 Herz or lower or that figure. That's why our input voltage noise here Peak to Peak was specified over that 10 HZ range because they're really looking at the one onf noise there, the low frequency stuff whereas our voltage density is looking at the higher frequency noise up here.

And yeah, they are two different things. So when we were actually uh, calculating this input noise uh, noise density over here for a 0 to 1 khz range, we were actually including this lower part down here. But because the frequency range we were uh, working over because it's log axis is so large pretty much you can ignore this tail up end and you know we can stick with the ballark figures. We got over here for our noise voltage density over that entire frequency range and we won't go into specific details of the types and noise cuz there are quite a few different types.

But suffice it to say that the white noise The High Frequency stuff is made up a combination of shot noise and thermal or jum function or Johnson noise as you may have heard it called and the one onf noise is also referred to as pink noise and that's due to, uh, what's called flicker noise but it's more typically just called one onf noise and that's the trap. With components, you can't escape this one Onf noise. It's just inherent in nature. There's absolutely nothing you can do about it.
There are things you can do in the process, uh, of manufacturing your devices to, you know, to reduce the uh flick noise, but pretty much you're going to cop it at that low frequency range. So you might think these Op Amps are less noisy at DC. Well, that's not the case. As you can see, they get much much noisier at DC they're lower noise at the higher frequencies.

It doesn't make sense, but hey, a lot of things in physics don't make sense. Next thing we know we were talking about spooky action at a distance. now. Gsy and white noise like shot and thermal noise has a uniform power density.

What that means is that it's going to be the same value regardless of the frequency. and that's why we get a flat l in there for that, but one on if noise is not a uniform Uh Power density. So that's why we get basically a flatline a straight line like that, but it has a specific slope 3db per octave. But we won't go into the details and this all comes back to why.

our input noise voltage density was uh specified in the data sheet at three particular frequencies 1 khz, 100 HZ and 10 Hertz. It's so that you can do comparisons with other opamps of how this uh noise changes and how well it performs over a frequency range like that. because if you see a large change for example, between 100 HZ and 10 Herz in for One Op amp and hardly any difference for another Op amp, then you know that that second opamp with the same figure right down to 10 htz is going to be a better opamp. And that's basically a deciding factor.

that corner frequency that we've got there that effectively determines um, how good your Op amp effectively is. The lower that corner frequency, the better your Op amp. and that's the one you're most likely going to choose. All things being equal.

And as always with data sheets, the marketers are going to fudge the numbers to give you the best possible Banner spec. So beware. You have to actually go in there and look at the graphs, look at the individual data, and compare opamps, And it can actually be pretty hard to compare opamps just from the data sheets. Not that easy.

So you got to be careful and know how to design it into your system. And you'll also notice on the data sheet that there's an identical noise spec for current as well. So it's input. uh, noise, current density, and input noise current.

And we won't go into that. That's the currents into the input to the opamp. So at the moment, as I said, we're only looking at the voltage scenario. But hey, if you got significant input currents, you have to take the input current noise in your count as well in those really critical low noise circuits.
But the same sort of fundamental Theory applies. And yes, it's all going to add up with the voltage noise as well. So you just got to be careful and by the time you actually practically build the circuit up, usually usually the external components are going to dominate your circuit more than the opamp itself. But hey, that's why they spec these things cuz a lot of critical applications.

you have to get the lowest noise off amp possible. and that's what it's all about. Frequency range. Remember how much noise density within that 1 Hertz window? And when you extrapolate these two lines here to get that uh Corner frequency crossing point where they intersect.

If you extrapolate that down, then we've got. there's 10 HZ There's 20 HZ so it's somewhere in there. Let's say about 15 HZ is our Corner frequency. For this example, we've drawn here.

Then that 15 HZ point is the point where the value of the white noise is equal to the value of the one onf noise. And of course, if you sum them together, let's say it's 10 there as shown, then you don't get 10 + 10. Of course, you get 10 times the square < TK of two. So you get about 14.1 So there you go.

That probably took a bit longer than I expected, and there's a lot more detail in here as well. But suffice it to say for your basic Op amp like that, if you're working from DC if it's all DC coupled and your full bandwidth is from DC to 1 khz, for example, you effectively do have to take into account these two different types of noise and you've got to sum them together. And when you add noises together, it's actually the root of the sum of the squares. so it's a square root of uh, this noise here squared plus this noise here, squared and You' Got to add them together and that gives you total noise.

But as we said right at the start, this is just the noise inherent in the Opam itself. It doesn't include the resistors here, which of course have that thermal Johnson noise. you might be familiar with that classic equation. the higher the resistor value, the more thermal noise you're going to get in the resistor and sorts of other stuff in your circuit.

so it can get really complicated. but I hope you found that really. It is pretty easy to understand what nanov volts per root Hertz is and how to calculate your noise. Very simple.

This is a bit more detailed of how it actually works, but let's go and see if we can actually measure exactly this graph to the bench and what tool do you use to measure the input noise voltage of something like an Op Amp? Well, you use a dynamic signal analyzer or DSA which we seen in the previous videos and this is my Hp35 60a. DSA They go from DC to about 100 khz. perfect for characterizing the Uh and seeing the one onf noise and power spec spectral density of the noise in something like an Op amp or any other circuit. It's the tool of choice.
but unfortunately this Uh 3560 isn't exactly the world's best performance. Its noise floor isn't that great in itself, so that's what we'll do. First, we'll just measure the noise floor of this unit itself with a 50 ohm Terminator on the input of course on Channel One here and we'll uh see what we get. but it's not going to be that crash hot.

but it should be good enough to at least allow us to see differences between Uh different types of Op amps. So I just run through you briefly: how to set up a dynamic signal analyzer to measure power spectral uh, density on a low voltage signal like this. Now when you first turn it on by default here we've got our frequency spectrum like this: It's displaying our frequency spectrum from 0o Herz down here to 102.4 khz and we're only uh looking at Channel 1. So there's the span.

The record length is 3.9 milliseconds for each one of those. and on our ya AIS Here we have a DB volts RMS There we go, it's doing its Auto calibration and we got a figure you know down around that 130 - 131 DB Vols RMS Mark The first thing we have to do because we're measuring low signal levels. Go input. So I've selected the input button on the front and then Channel One range.

at the moment, it's Auto ranging. We really don't want that. we want it to, um, just be fixed. And this thing I'm pressing the up down arrow keys and as you can see, there you go the channel one range up there.

the uh, highest gain range or the lowest voltage range it's got isus minus 51 DB Vols RMS and that's equivalent to I Think about 4 Ms uh, Peak or thereabouts. Next thing we want to do is turn on some averages. so I'll press the average button on the front and then we want to turn average on like that because otherwise we'll just get you know we want a smoother line. See what happens when you turn the average on There it's set for 10.

I'm going to change that to number of averages there and I'm going to enter 100 averages. So now when you press the start button and we start our acquisition, there we go. It's giving us a bit of a plot already and we can already see that we're getting a result. Here it is.

There's our pretty much flat line with the big one onf noise tailing up at the bottom. but why didn't it look like the Whiteboard? Well, because we haven't plotted uh, the frequency on a log plot yet. It's a linear plot. It's a linear axis.

Sorry. Speaking of which, we have to go uh to the input here. Set it up and just make sure we've uh, got DC couple in here. We want to go all the way down to DC.

So to change that to a log graph, we press the scale button on the front here and here it is. Xaxis. There, it is currently set to linear. We'll change that to log and Bingo.

Look at that. We're starting to get exactly the response that we wanted. Now the reason why It's um, there's not many data points down here because it has to do with the number of lines in the Uh Fft response to this thing. Now we've got a full Uh span here of 102.4 khz and this particular instrument only has 400 lines of resolution.
So if you divide um, 102.4 khz into 400, you will get. If we move our marker across here, you'll notice that, um, each step it can only measure at those frequency points there, so it's very coarse down there of course. and you'll find that the lowest step down there is going to be 1/400th of 2.4 khz. So 102.4 K ID 400.

There we go gives us 256 Herz Where our marker is all the way over there. What's our marker? X There it is 256 Herz So it can only jump up in 256 Herz steps cuz that's all the Fft resolution we've got there. And of course that really shows up when you've got the log. Xaxis like that didn't really show up on the linear one because then it'll be stepping in 400 even linear increments across the screen.

Now if I press the measurement data button on the Uh front panel. Here we're in what's called uh, well, just normal uh frequency spectrum mode more correctly referred to as linear Spectrum mode and that gives us a Uh voltage response here. and as we saw before DB Volts RMS there - 1223 and if we plug that into the calculator - 123 and then we divide cuz it's in DB Remember if you want to convert it to a voltage, then we divide it by 20 and then we take the inverse log of that and we've got ourselves 78 Nan volts. But what does that mean doesn't really mean anything because that isn't our Uh Power spectral density.

So press the scale button on the front and we'll have a look at the vertical units which we've got here DB Volts RMS at the moment and as you can see, there is no option for that Uh voltage per root Hertz Because we're in the linear Spectrum mode we're not doing, we're not actually calculating the power Spectrum density. But that doesn't mean that this graph isn't correct CU it actually is. The shape of this graph is absolutely uh. Bang on to what we will get in the SP power Spectrum density.

Except our units AR up here aren't correct. where DB Vols RMS instead of that Uh voltage per uh per root Hertz So how do we do that? Uh, well, how do we convert it? Well, we can do it manually. We can do all the math ourselves to convert between the linear spectrum and the power Spectrum density. but uh, we don't need to do that.

What we can do is go into the press the measurement data on the front. This thing will do it for us. That's what these Dynamic signal analyzers are designed to do. Measure this noise specifically and there it is.

PSD Mode or Power Spectrum Density Bingo If we go into Power Spectrum density, you'll notice that the graph hasn't changed at all. and normally when you change mode, it rescales things. but it hasn't. The graph has stayed exactly the same.

But look what we've got now. it's got a little asteris next to it here and that Asis means there it is. Volts RMS per root Hertz And if we go back, that's exactly what we want. Exactly what we saw on the Whiteboard And if we go back into the measurement uh, sorry the scale here into our vertical units, we'll see because we're now in the Power Spectrum density mode that we've got root.
Hertz options here: Volts RMS Squ DB Volts RMS per Rootz Herz or volts per root Hertz That's what we want. Volts. Well, we want Nanov volts, but volts per root Herz is the same thing. It'll scale for us.

so Bingo look at what we've got now our that value there at 10 khz it's close to 10 khz is now switched over and it's calculated that it's 28 E- 9. That's Nano Of course Nanov Volts per Root Hertz Bingo We've now got our DSA to uh, check its own performance because we've uh, remember, we've got a 50 Ohm Terminator on the front and there it is. That's what it is after 100 averages down here over that. uh, well at the moment, the full span from 0 to 102 khz.

So as you can see, this instrument, um, you know, is worse than a basic uh, you know. Op07 Op Amp 28 n volts per root Hertz And as we saw in the data sheet before, just a basic Op07 is around you know, a 10 a spot frequency. Um, in this case 10 khz. It only goes up to one I I think Uh, But yeah, you know, because it's flat, it's going to be exactly the same.

It had a figure of around 10 Nan volts per root Hertz So this thing isn't good enough to measure the performance of an Op07 the way you normally do it. Although it is, you could actually use this instrument. The way you would normally do it is use an external, uh, extremely low noise purpose designed amplifier to amplify the noise before it gets into this instrument. So you use this instrument.

Um, you've already gained it up. So you bring it way above the noise floor of this instrument and then you can. you know if it's got times 100 gain then you can, just you know, change the units to compensate for that and you can actually measure the performance of an Op07. Now if we take our cursor all the way over to uh, the corner frequency down there.

Um, once again, we're very coarse cuz we're measuring the whole 102 khz bandwidth, It's it's telling us the corner frequency is about 1 khz, but I Know that's not going to be the case. What we want to do is change the span so we get more detail down on this one on F region instead of just three crappy three D points and that's easy. You just press the frequency button on the front. You can see these Dsas are specifically designed for these types of measurements.

They're optimized for it. This is what they're designed to do anyway. We can just go span like this, press the span button and then we can, uh, just type in, say well, no, let's do 1,000 Herz We'll do a Kilohertz range and then it's going to restart. You can see it's automatically restarted and it'll do the RMS averages.
It takes longer of course cuz it's lower frequency so it takes a quarter of a second per record. Uh, length like that. but there you go. This one has actually dropped off the screen.

So I think we've done something with our input scale in there. so if we press our scale button there, we can just Auto scale that and bang. That's going to bring it in line like that and look at that. Look at that we can now one that cursor we can now put it at one KZ There you go.

So it's at 1 khz there and we're getting a value of about 31 uh, nanov volts per root. Hertz that's the noise flow of this thing. As I said, not very spectacular. In fact I want to investig open this thing up? uh, have a look at the Op amps used in this and other components and see if I can actually use modern uh dropin high performance Op amster.

Actually, um, increase the performance of this thing. So uh, I'll leave that to a future video. but you can see it's essentially flat and it starts to tail up a bit There you can see it just starting to go up so you can see um because we're effectively measuring the Uh noise of the Um input noise of the the input section or the input Op amps inside this particular instrument. So we'll get exactly the same result if we were measuring an external Op amp effectively.

So the value at 1 khz here is going to be slightly lower than the value at 100 HZ which once again is going to be uh uh, lower than the value at 10 HZ here. and that's why they have those three spot values on the data sheet 1 khz, 100 Herz and then 10 HZ over here. And of course that will be a continue basically completely flat out to that 100 khz we saw uh, last time. but you can see it pretty much starting to get bad at just under 200 HZ there I've put it on 160 HZ for a reason because let's go to the data sheet for this HP DSA and here it is straight out of the user manual on the minus 51 DB Volt range I.E the highest gain range which we've got Source impedance of 50 Oh, which we got 16 RMS averages well, we've done 100.

You'll notice that it doesn't specify anything under 160 HZ it's got that 160 HZ To that 1 khz range is - 130 DB Vols per Hertz and of course you if you wanted to, you have to convert that to the power Spectrum density which we can do which we've just done with the instrument itself. So there you go. That's why they've got the figure of 160 htz in there because it its performance really starts under that 160 htz. you know really starts to be a bit.

how are you doing And one thing I want you to take note of near 50 HZ There you'll notice that we're getting no 50 HZ pickup at all. And of course this lab is just swimming in 50 HZ Mains frequency because as we saw in the tear down of this thing, it's incredibly well shielded and we've just got a 50 ohm Terminator on the front. But as I think we'll see when we try and measure a practical circuit, we're going to get at least some 50 HZ pick up. It's almost unavoidable.
Okay, so let's take note after 100 averages at our marker frequency of 1 khz, because that's a value we can get from the data sheet. For some opamps, we're getting 31.3 NTS per root Hertz So that's the basic noise flaw of our DSA here. And of course, to measure noise flaws like this, you need a Faraday You need a shielded box. One of these diecast alloy boxes.

Absolutely fantastic. sort of Industry standard way to measure these things. A little mini breadboard in there with a Um Tl072 on it, and I've got two 9V batteries. Now, if you look at the data sheet, the voltage, uh uh, the noise uh for these for all these chips is usually specified at say, plus - 15 uh, volts or sort of Maximum rail H it's going to be near enough plus - 9 Now, of course.

once you, uh, put the lid on this sucker, there's no way anything is getting in there at all. We've got our nice BNC on there. We got a shielded coax all the way to the input Bob's your uncle and of course you do want to use batteries internal to the Box. You don't want to be using an external power supply or any type of switching power supply or anything like that.

batteries the only way to do it and you'll notice no I don't need any decoupling on there. It's good enough. Uh, because we've got the uh low impedance battery directly and this thing ain't going to oscillate. So we've got our box hooked up with the T72 in it now.

I Chose the T72 cuz it's not a particularly low noise opamp. about 18 Nan volts per root Hertz at that 1 khz figure straight from the data sheet. Because it's not designed for noise, it only has the figure at 1 khz. It really, you know it's not that great.

doesn't really specify it in depth. but here we go. So that is the noise floor of our DSA. Let's Press Start and we will get using the exact same parameters we set up before.

Remember 31.3 NTS per Rootz Now of course that is below. So the noise that we're trying to measure here of this T72 is below the noise floor of this DSA But aha, remember that they sum together so we should see an increase there. Let's Press Start And away we go. And Woohoo! look at that one on F Noise has gone right off the scale there.

And look at that bump. What frequency do you reckon that is? 50 HZ Bang on! Where are we picking up our 50 HZ from? It ain't through the box, it's through the shield of the coax. That's the only place it can be getting in I Don't know. this is a you know Rg59 cable or something I don't know.

just a cheap one I had lying around. so uh yeah, you? really? Even with fully shielded coaxes and that shielded box, we're getting our 50 HZ pickup. But anyway, look, we've got almost got our 100 averages. There we go, we've gone up from 31.3 NTS per root Hertz to 38.0 28.
Uh, 38. Uh, Nan volts per root Hertz at 1 KZ So it's gone up by about 7 NTS per root Hertz And what value should have we expected? Well, 313 n per root Hertz the base noise floor we had there. We're got to square that. Remember the formula we had on the Whiteboard before and then we've got sum of the squares.

so we've got to add in the data sheet value typically 18 n volts per root Hertz at 1 khz. So uh, yeah, let's square that and then get the square root. We should get around about 36.1 and we're getting 38.1 You know 38 so you know near enough. There you go.

We were able to see a difference with that T72. Now let's get one that's even worse. 42 NTS per root Hertz it's a TL 062. It's an absolute shocker I've put it in there.

let's Press Start and there we go wo we still get our 50 HZ of course horrible one onf noise gone off the scale here. but there we go. Oh it's massive Now look at that in the order of you know, 75 nanov volts per root Hertz Awful. There we go.

after 100 averages 68.1 is that correct I don't know what is it? 31.3 squared which is the noise floor of our DSA plus the nominal 42 uh from the data sheet and then we can, uh, get that. and then we square root that we expected around about 52.3 and we're well above that so that one's not working out too great. Really is an ancient chip though. trust me, it's like 25 years old or something.

Let me check the date code. there's actually not a date code on that, but uh, this one's actually like I Had this one since I was a kid and uh, it was actually desoldered from a board. so it's ancient and shocking. Um, but anyway, it allows us just to show the difference there, what a crappy Op amp can make and how you can measure it.

And I've now put in an Analog Devices ad 712 the uh The Identical 18 NTS per root Herz of our Tl071. So let's give that one a B and see what we get. Still get our big 50 Htz. but uh, there we go.

We're getting yeah, about 40 odd, not too dissimilar to what we'll getting with the Uh Tl072. And as I said, if we really wanted to measure the performance of these Op ants properly, I would have to use an external amplifier in here I'd have to really design it properly and ironically, you need an incredibly uh, you know, low noise amplifier in there to measure low noise. Imagine trying to measure the state-ofthe-art Op amps Well, you got to be very careful in how you roll the input uh amplifiers and we would still be able to measure it easily once we got. you know, some gain in that box to get well above the noise floor there and actually be able to measure uh properly the absolute performance of the Op ANS But anyway, I hope you found that interesting.

We were able to see the differences between some opamps and if I put in a really Schmick Op amp in there, we would have actually, uh, seen it drop to pretty much the same noise floor as this particular Uh DSA So there you go. If you want to discuss it, jump on over to the Eev blog forum and uh I hope you like the video and if you did, please give it a big thumbs up. Catch you next time. Wait, hang on I Found an Ne5 34 Opamp.
Really? uh, good audio, low noise audio upamp I Think they even use a couple of them in here from what? I Uh, saw on the schematic anyway and not at the front end I don't think. but anyway, somewhere in here and uh, that has a noise figure of about four Nan volts per root. Hertz So let's give it a B and there we go. Yep, still picking up our 50.

HZ But once again, we haven't gone off scale here now. and there we go. We're not. We're almost exactly the same uh, noise floor as we got with the instrument itself.

What was it? 31.3 NTS per root? Hertz If we wait till it goes up there, we're only a couple of Nanov volts above that. So Bingo There you go. There's a good quality Op amp for you. There it is 33.7 for the record.

Beautiful. Catch you next time.

Avatar photo

By YTB

24 thoughts on “Eevblog #528 – opamp input noise voltage tutorial”
  1. Avataaar/Circle Created with python_avatars Sean M says:

    "Input referred" is actually rather important. Without saying referred you could be saying the op-amp puts that noise on the input signal, instead of the real meaning, which is noise at the output multiplied by the gain in between input and output.

  2. Avataaar/Circle Created with python_avatars Stelios Posantzis says:

    Excellent tutorial.

  3. Avataaar/Circle Created with python_avatars Adarsh Sinha says:

    Loved the video ! Thank you for this.

  4. Avataaar/Circle Created with python_avatars From Germany says:

    Uncle Bob would put in screws to lock the box. 😎

  5. Avataaar/Circle Created with python_avatars Nathaniel Kilmer says:

    Using the division sign for multiplication is the most engineer thing I've ever seen.

  6. Avataaar/Circle Created with python_avatars Ross Silverwood says:

    Thank you so much for these op-amp tutorials. I have been building circuits using op amps my entire life, and I never had a true ( thorough) understanding of how they work and how to design circuits with them- until now, seeing your videos. Thanks again.

  7. Avataaar/Circle Created with python_avatars EL hub says:

    What mean by jellybean??😬

  8. Avataaar/Circle Created with python_avatars Rick West says:

    So, the elephant in the room. What happens to the noise values at DC? I want to build a low noise instrument to measure DC to 10Hz. How do I know the noise value at DC?

  9. Avataaar/Circle Created with python_avatars Jakob Furbo says:

    Why is the 50Hz noise if it is totally shielded?

  10. Avataaar/Circle Created with python_avatars K X says:

    Why don't we use 100Mohm resistors on our Opamp voltage dividers? I always see things like 47kohm.

  11. Avataaar/Circle Created with python_avatars model rogers says:

    thank you for your work, dave. you have truly made me more interesteed

  12. Avataaar/Circle Created with python_avatars N8AUM says:

    Great old video Dave, after watching one of your videos years ago about the HP DSA I was able to find one at some government auction, GLAD I did ! Wish you would do some more vids using these ol boat anchors.

  13. Avataaar/Circle Created with python_avatars Davadoff says:

    When changing the vertical axis, Dave used the non RMS option (with the RMS option being one option lower).

    On the whiteboard he said it was all RMS like the 316nV result.

    Is one of these two things a mistake? Someone else said 316nV was the p-p figure. Can anyone else confirm/explain, pls?

  14. Avataaar/Circle Created with python_avatars gary white says:

    Why is input noise divide by the gain rather than multiplied?

  15. Avataaar/Circle Created with python_avatars funkyironman69 says:

    Great video Dave.

  16. Avataaar/Circle Created with python_avatars Polonio Pi says:

    So…,I'm closed at home because of coronavirus in Spain at the order from our goverment. But, with gems to learn like your videos, i hope not to be bored much next two weeks. Thanks for this videos Dave!.

  17. Avataaar/Circle Created with python_avatars Twobob Club says:

    great vid. thanks

  18. Avataaar/Circle Created with python_avatars Sirac says:

    Actually it possible to measure the noise of an op amp with only 10nv/rtHz if your op-amp have some gain. For example, if your op-amp have a gain of 10, at the ouput, the noise will be arroud 100nv/rtHz ! Because this is the noise defined to the input of the op-amp and it will be amplified by the op-amp.

  19. Avataaar/Circle Created with python_avatars Jamie Tyson says:

    Hey Dave! I don't know if you check comments on old videos but I wanted to mention some things from my experience at work building amplifiers for signals in the 50nV range. Based on Jim Williams' work, I started using steel cans for my testing. I put my little amplifiers into Altoids mint cans with a voltage divider built-in so I can inject a signal and I have no problems with 60Hz (USA) mains frequency showing up in my measurements. A 9v battery and the amp fit nicely, and I can solder directly to the tin for a solid ground. I have found that die cast and aluminum boxes don't work as well as steel mint or and cookie tins. Ferrous material really helps with shielding! I have been able to show a 100nV signal on the DSA at work with no trouble! Also, I use an HP DSA for spot checks but final plots are done with an RME audio interface and Matlab, which easily beats the HP signal-noise specs. Thanks, and keep up the good work!

  20. Avataaar/Circle Created with python_avatars Hugo Ribeiro says:

    Very good explanation! It was a good help ;). I need to amplify the output voltage of a 1kHz notch filter, uVs… to measure a SINAD in an audio Amplifier. But I'm having troubles with the noise. Now I'll try reduce the resistors values and chose an op-amp with a reduced noise, now the ne5532… Many thanks! My Channel is in Portuguese sorry… Regards, Hugo.

  21. Avataaar/Circle Created with python_avatars Sean Wolfe says:

    What is Jelly Bean Precision?

  22. Avataaar/Circle Created with python_avatars squee222 says:

    Love your videos. Very helpful

  23. Avataaar/Circle Created with python_avatars AGS-TECH Inc. - Custom Manufacturing says:

    These videos make the subject much easier to understand by combining theory with experiments. Thank you !

  24. Avataaar/Circle Created with python_avatars Otie Brown says:

    Excellent. Clarified many questions on Op Amp noise.
    Thanks.

Leave a Reply

Your email address will not be published. Required fields are marked *