What is Common Mode Rejection Ratio (CMRR) and how to measure it on a high voltage differential probe or differential amplifier. Using the new Rohde & Schwarz MXO4 oscilloscope.
https://www.rohde-schwarz.com/us/products/test-and-measurement/oscilloscopes/rs-mxo-4-oscilloscope_63493-1164992.html
Extended 30min version: https://www.youtube.com/watch?v=4gi2B3LrCww
Differential Probe How It Works: https://www.youtube.com/watch?v=GOlgaEK2Hsk
High Voltage Probe reverse engineering: https://www.youtube.com/watch?v=GOk1BYKSsOI&t=2s
Oscilloscope Standard Deviation noise measurement: https://www.youtube.com/watch?v=G8Qoj3TpO9A
Use coupon code ProbeyMcprobeFace for 15% of the HVP70 differential probe:
https://www.eevblog.com/product/hvp70/
Forum: https://www.eevblog.com/forum/blog/eevblog-1521-how-to-measure-differential-probe-cmrr/
00:00 - What is CMRR?
01:53 - Why twist the wires?
03:26 - Typical CMRR figures
04:51 - How to measure CMRR - The Setup
08:54 - Remote R&S Oscilloscope desktop view & settings
11:22 - Output to input voltage ratio
11:58 - Beware of DC offset
13:36 - Does the 20MHz CMRR measaurement match?
14:02 - Input referred measurement
15:02 - Compensating for the gain
16:29 - Automated Frequency Response Analysis
Support the EEVblog on:
Patreon: http://www.patreon.com/eevblog
Web Site: http://www.eevblog.com
Other channels:
EEVblog2: http://www.youtube.com/EEVblog2
EEVdiscover: https://www.youtube.com/eevdiscover
T-Shirts: http://teespring.com/stores/eevblog
#ElectronicsCreators #CMRR #Tutorial

Hi in this video. I'm going to explain what common mode rejection ratio is and actually how to measure it in this particular case of a high voltage differential probe here. but it doesn't have to be a high voltage differential probe available on the EV blog store by the way, discount coupon code down below any differential amplifier circuit. be an Op Amp or a discrete transistor one, it will have a common mode rejection ratio.

So let's have a look at it. We're going to measure it with the brand spanking new rodent. Schwartz for Mx04 series scope here: 12-bit Joby Because why not? It's beautiful. So the common mode rejection ratio of a differential amplifier in this case, a differential probe, as the name suggests is just the ratio of the differential gain of the amplifier divided by the common mode gain of the amplifier and what is Uh differential and what is common mode.

Well, a differential amplifier measures the difference between two inputs. Here, their base essential actually is no ground reference. It is a differential signal and a differential amplifier will have a a gain of that differential signal. That's its job.

If your differential amplifier has a gain of 10 and you put one volt differential across here, it doesn't matter where it is in the circuit, it's a differential voltage. doesn't matter about the ground reference. It'll multiply that by 10 and that's its Uh differential gain. Now you divide that by the common mode gain.

Now, what is the common mode gain? Well, instead of the differential voltage across here, it is an external voltage applied to both of them at the same time. So in this particular case, okay, we've got these long leads here and we could have like external uh, either capacitive decoupling or Emi coupling into the probe like this. So they're basically getting either onto the probes in the same way. And this is why the wires are twisted like this.

And if you're measuring a differential signal, it means any external no noise or interference insured in theory apply to both wires at the same time. So it's an external reference in this particular case, referenced to the grounded output of our differential probe. So the job of a differential amplifier is to amplify the difference between its positive and negative input while rejecting all of the signal or much of the signal as it can that is applied commonly to both of these wires. So that's why it's a common mode rejection ratio.

So In theory, your differential amplifier should have an infinite common mode rejection ratio. It just measures difference here and rejects everything else. It has no gain at all of any common mode signal being picked up by both wires. But in practice, and no, that's not going to happen, just the design of the amplifier itself and most importantly, the matching of the input resistor.

Network in here. and I've done a tear down of a high voltage differential probe down like this. I'll link it in up here and down below if you haven't seen it. And just the match in between The resistors on here is pretty much determines the common mode rejection ratio of this probe because usually like the Op Amps used inside here, they're usually pretty good.
They're going to have like a rejection ratio of like you know, over 100 DB or something whereas the resistor divider drops that down to like 40 or even less. Now a product like this Hvp 70 differential Probe, it'll typically have a common mode rejection ratio figure measured at various spot frequencies. Maybe if you're lucky you might get like a response curve of common rejection ratio because it's going to vary depending on the frequency so it's going to change. So they typically here's the values for this Hvp 70 and it gives us four spot values there.

and the ratio you can see is you know, like at 10 megahertz minus 40 DB and it is usually given in a DB figure but doesn't have to be because it's just a ratio. So you could just use the ratio figure. And the interesting thing about this is that the common mode rejection ratio is as I said, the differential gain divided by the common mode gain and that actually comes out at a positive value. But so why is the data sheet negative? Well it's kind of.

There's no like standard for this kind of thing so you just sort of have to like understand that when you're talking a negative number in this particular case, minus 60 DB common mode rejection would be better than minus 40 DB But if you had it, if you specify it as a positive one as you might get on say a Op-amp data sheet. Here's an example. Uh, you would get a the higher the value is going to be better. So that positive or negative thing just a little trap for you young players.

Just be aware of that right? So how do we measure the common mode rejection ratio and verify the common mode rejection ratio of this probe? First thing you want to do is is as I said, you want to twist the wires like this so that any external noise is equally picked up on both and then you need a signal generator. In this particular case, this new rodent Schwartz Mx04 can go up to 100 megahertz. So very nice. And then we want to feed the output of the Sig Gen into a 50 Ohm terminated load so that we don't have any transmission line issues whatsoever.

No Reflections causing problems I've done videos on that and how you can Goof that up in noise measurements and stuff like that. So I'll link in that video up here and down below if you haven't seen it. So here, I'm using an external 50 Ohm 2 watt termination. A series termination even though the scope has a built-in If you look down here, you could actually come a gutsa because this is only like got a half watt rating.

It's less than 5 volts. RMS Just you know, you don't want to blow up your scope when you do something like this because you want to use a high as high a voltage as uh, possible. But in this particular case, I'm just showing you it's better to use a high rated X external. Terminator Just so you know, blow up your really expensive, beautiful, shiny scope and then we're just tapping off right across this 50 Ohm Terminator Load here.
So here's the negative. uh, terminal. and here's the positive terminal. So what we want to do is connect both of these inputs together, short them together and connect to the positive input like this.

Why the positive input? Because it means that we're applying a voltage relative to the output here because the output is ground reference like this. so we're actually referencing it to the output. so we're effectively feeding that signal generator voltage into both of these leads. I.E A common mode signal relative to the grounded output.

Because if you remember, all the grounds on your scope are all common. So this is the input signal and the output. and they're effectively joined. They're common.

So what happens if I Just connect one of these to here? Well, you saw it. The green signals are output. The yellow is our input there, and our green signal. It gives us a nice clean output like that.

Okay, so our differential uh probe is because this one's just flapping around in the breeze right doing nothing and you'll see that just jump all over the place there. and if I touch it, look at that like we're picking up all sorts of crap and you'll get the same exactly the same thing if you connect the negative up like that, right. The exact same thing will happen because this is a differential amplifier. It it doesn't care.

Um, it's just you're just unbalancing that input. But if you connect both of them on like that to the same point, then Bingo We've got a really small signal so this probe isn't perfect. It's got a common mode rejection ratio. so there you go.

It's there's a signal being Amplified even though that input is completely shorted and you'll notice that goes away if we don't connect up to the that. Okay, there's our ground there. So we just got like the inherent noise of that amplifier and it doesn't matter what I do to the probes here, but this one, if we hook it back up, you'll notice if I start playing with those probes, things start happening. Okay, it starts like being influenced because like we've got these long leads here.

That's why a differential probe with like like really shorter leads is better. but these have them. Uh, most probes have them built in though unfortunately. So what happens if I untwist those leads Like, this? We're going to get the same signal, but we potentially have more variation.

Check it out. you actually get huge differences. Like if I like take my hand away from that, right? You can get large differences like that. So if you don't twist the leads and keep that common mode signal right, you can completely screw up and come a gutter on your measurement.
Now that I've explained what we're doing and I've shown you the setup here, there's no reason to look at this anymore. All so I'm going to actually go over to a remote desktop view and we'll do a direct screen capture of this a little bit. Just be nicer. And because I can ah isn't this? Schmick Look at this Ethernet remote control.

It's got a building web browser so we can just go to the IP address and Bam we're in so we can. Actually we can do some like configuration and file manager stuff. but let's just go to full screen here. So Channel One the yellow one that's our Sig gen there.

We've got uh, one volt per division, 50 megahertz bandwidth because you do want to uh, bandwidth limit and this scope actually has some cool software bandwidth limited options in it um, which you might see later and uh one one big home input DC coupled or AC doesn't matter and channel two also the same 50 megahertz uh bandwidth here. leave it on uh two millivolts there, shall we? Now you can see, we've got a real fuzzy wuzzy waveform here now. Of course, this is a 12-bit scope. you don't necessarily need a 12-bit scope for you know, this particular application that we're doing right here, but 10 or 12 bits more better.

But we can actually go more than this so you can actually see up here up the top is telling this uh, just that's the basic 12-bit uh, but we can go higher because if I get my ugly mug out of here, we can see that got a HD mode down here a high definition mode and we can actually set that on. There we go. We instantly set it on and you'll notice that our 12 bits went to 16 bits up here. But before we go ahead with that, I'll just mention the signal gen here.

now. Uh, you want this to be as higher amplitude as possible because the output signal and that you're actually trying to measure that common mode signal is really low. so the higher the input signal the better. So I've gone up to the maximum amplitude here of 5 volts P to Peak here.

and we've got a frequency of 10 megahertz because that's just the uh, you know, a typical figure we've got in the data sheet which we want to try and uh, verify. So we want to clean this up a bit more. so let's do some averaging so we'll go up to the acquisition up here and we're actually in Sample mode. So we'll go down here to average mode and then them.

We can do like 40 averages something like that. We can take the time base out a bit like that. so we've got a decent number of signals. you can see that our average there.

We've got 40 averages there and uh, that's just cleaned that up a tad. But you can see how we are dealing with the wobblies down here because as I said, the test setup is everything. So if you can Shield it and keep the lead short and make sure that they're twisted and everything else, it's going to be uh, better. So what we need now is to compare the input signal to the output signal.
That'll give us our comment mode rejection ratio. in this particular case, that it's actually negative here. So at 10 megahertz, it's minus 40 DB here. So we want to flip that around here to give the output divided by the input.

Now to measure this ratio between input and output, we can even measure the uh P to Peak value or the RMS value doesn't matter RMS is you know it's better. It's more accurate, but you might think that we use this RMS value here and uh, what's that? 1.4 millivolts like that? But I've done a video on this where that that RMS value that includes any DC offset component so that's not quite what you want. So let's go into the measurement menu here: Unfortunately They don't have it in the basic category. you've got to go down to Vertical There There you go: Standard Deviation AC RMS I've done an entire video on that so we want a channel.

uh one. but I can actually uh, go in here like this: I can double click on that and I can choose specific type AC RMS Like that. So let's actually get a few waveforms on screen here so it's a bit more accurate. and once again, we can turn the statistics on there.

Come on, can't double click to get into the menu? Oops. I had the wrong bandwidth there. so we have to use the 50 megahertz bandwidth here because we're measuring uh 10 me. 20 meg is a bit close to the frequency.

You want to be a bit more than double above like that. so you know 50 is 50 is not a bad value. So we get our confu user out here and we look at the RMS value here. Don't be confused by remember how I mentioned standard deviation before you've got to watch my standard deviation video.

The standard deviation here is not referring to the AC RMS, it's referring to the standard deviation of the standard deviation Ace AC RMS signal. So it's like it's very confusing. So yeah, don't come at guts are there. So we need to, uh get our confuser out and uh, 883 microvolts so micro volts won't get any more Precision than that.

uh, divided by our input because we want a negative number so 1.75 volts. Then we want to take the log of that and then multiply that by 20 not 10 because this is a voltage. So we get minus 65, minus 66. Basically, Hmm, that doesn't sound right because our spec over here says minus 40 at 10 megahertz.

Why is it way way better? Way way better. Hmm. because this Cmrr is what's called input referred. It's referring to the input of the actual Uh amplifier in this case inside the probe here before it gets gained up by the amplifier.

Now if you noticed in the video before, we're in the Uh 10 to 1 division ratio setting, so there's a gain of 10 in there. So we have to account for that Uh gain of 10 in here in our DB figure. Now you know a good data sheet. They should actually specify that and tell you exactly what it is.
Now this is a good marketing trick because marketing can make the common mode rejection figure sound a lot better just by saying oh, that's input referred instead of like output referred. So in this particular case, uh, our times 10 uh, probe over there times 10. Of course in DB these is 20 DB and times 100 would be 40 DB times a thousand would be 60 DB It goes up 20 DB for each order of magnitude step. So we have to actually add on uh 20 DB to that minus 66 DB becomes minus 46 DB So yes, it does actually meet that specification so it beats it by 6 DB Not too shabby, so let's repeat this at one megahertz so it should get better by about 10 DB So I've got this 46 here.

Maybe we'll get 56, will we? So there you go. 423 microvolts divided by 1.8 volts there. and log times 20 equals minus 72. Yep, add 20 DB to that.

So it's minus 52.. So there you go. This typical spec is minus 50. we get minus 52.

yeah comes out. but if we try and measure it down at 50hz, it's here. which is supposed to be minus 80 DB So it just like it. It's 20 DB increase over 20 kilohertz.

You can see that Yeah, it's it's gone to nothing. again here. 500 microvolts per Division and there's nothing. There's nothing there.

I Mean we can take that figure and punch it into the calculator. but like there's just nothing there. We're basically measuring the RMS value of the noise at this point. Anyway, you can see the process there.

That's how we can measure the spot frequency. Now how can we get a plot over frequency? I'm glad you asked. We can do this using if we go into Apps here for Ffra or Frequency Response analyzer. So let's open this bad boy up.

Yeah, we can get a plot of this over frequency when we can also get phase as well. So we're going to put our stop frequency in here of uh, 10 megahertz and start frequency. Yeah, we can actually start down at that 50 Hertz. So we set up our input as channel one.

Our output is uh Channel 2 50 Hertz to 10 megahertz amplitude As you once again, you want the maximum amplitude, run on this now. watch down in the bottom corner down here as it's adjusted, it's set to AC and then it's adjusting the range all in real time and you can see it's slowly plotting here. It's only a small, it's got a table and a thing we could. We could zoom that later if we really wanted to.

50 hertz, 100 Hertz right? It's down in the noise starting to get out of the noise there. We can adjust the range in a minute to actually see that and boom we are done now. I Don't think there's anything in here that allows us to set the offset there. Just remember that we have to add 20 DB onto.

Uh, these figures here. So you can see that, uh, you know, around about five megahertz there. It does. Really, You know it starts to gets worse.

the higher that is. Uh, the worse it is. And you can see that the uh, the red plot here is the game we're not to. That doesn't matter for our combo rejection ratio, but if we actually extend the bandwidth on that, we should be able to actually see a phase reversal.
So at 10 megahertz here, you can see uh minus 64 which is minus uh 44.. the one megahertz figure it's so minus 71 is -51 R DB there. So yeah, that meets the Uh spec. So if I go out to the full bandwidth here 70 megahertz of this probe, let's rerun that again.

I'm not going to go low frequency this time, so it's Auto ranging. Each time it actually takes these uh samples which is really quite nice. So it's maximizing its uh dynamic range there. and it's also adjusting its bandwidth as well.

You'll see that the uh yeah, it just jumped from one to two. Mega It's three megahertz. see. so it's actually it's software adjusting that bandwidth.

This is really cool. This is a very good frequency response analyzer. so we're looking for the phase response to actually, uh, reverse here. Oh yep, there it is.

There it is that's totally expected. Not just a differential amplifier, it's a normal amplifier. Behavior But once again, phase doesn't uh, mean anything here. right up to 70 megahertz.

It's uh, it. minus 41 which is minus 21 DB So it's a fairly sharp rise. so once you get above that 10 megahertz, that's why they don't give you a figure up at 50 meg or 70. Meg They once again, marketing just does you know, stop at 10 megahertz.

There you go. common by rejection ratio. If you enjoyed it, give it a big thumbs up. as always, discuss it down below and subscribe to EV Blog 2 and my Odyssey Channel where there's exclusive videos over there if you want to see a couple of these.

I think I've got two exclusive videos of the Roden Schwartz oscilloscope. Why? it's actually been delayed? Um, and because we had to actually, uh, swap it. Yeah, this is really sweet. um.

scope. So yeah, uh. leave it in the comments. Do you want to see a tear down or do you want to see a feature review? It's got so many features, but I can show you some of the uh, really cool stuff in this.

It's going to be good. So anyway, catch you next time.

Avatar photo

By YTB

23 thoughts on “Eevblog 1521 – common mode rejection explained measured cmrr”
  1. Avataaar/Circle Created with python_avatars Dardo Sordi says:

    This is really great stuff Dave! I loved every bit of it, it's well polished.

  2. Avataaar/Circle Created with python_avatars GoBO207 says:

    I know this doesn't make sense, but I actually prefer looking at the scope from the cameras perspective rather than the screen capture. Wouldn't be the same for a smaller scope, but for this one…

  3. Avataaar/Circle Created with python_avatars Jim Haldeman says:

    One-of my first projects in electronics as a design tech was to build a common mode amplifier. For the life of I had no idea what there was to achieve expect build an amplifier that had a gain of 1 or less. The cathode follower ! I used an RCA low noise high frequency 6DS4.

  4. Avataaar/Circle Created with python_avatars q zorn says:

    I love these kind of videos. They make learn and relearn things I have dealt with for some time. ๐Ÿ˜Ž Thanks and happy 2023 new year. ๐Ÿฅณ

  5. Avataaar/Circle Created with python_avatars jassim javed says:

    Thank you Dave for educating us!

  6. Avataaar/Circle Created with python_avatars OvalWingNut says:

    Did I win the scope? I guessed 42 Send to Long Beach, CA. USA, 3rd house on the right. THANK YOU!

  7. Avataaar/Circle Created with python_avatars Joop Terwijn says:

    Guess you need to put the whole setup in a faraday cage to reject as much noise as possible

  8. Avataaar/Circle Created with python_avatars uLearn says:

    Dave, you should make a note that youโ€™re measuring CMRR from input to output (the smaller, the better), while the formulas you show and what you describe display CMRR as the ratio of differential gain to common mode gain (the larger, the better). I can see an astute viewer being confused. Otherwise, great video.

  9. Avataaar/Circle Created with python_avatars electronics unmessed says:

    ๐Ÿ‘Wow, pretty advanced features that you are explaining here! Many thanks.

  10. Avataaar/Circle Created with python_avatars Amrish Hirani says:

    Merry Christmas

  11. Avataaar/Circle Created with python_avatars den2k88 says:

    Too late to ask Santa this gorgeous oscilloscope

  12. Avataaar/Circle Created with python_avatars guillep2k says:

    Why Ch1 changed from 1.75V to 1.80V when changing frequencies? The attenuation in the circuit should be linear, right?

  13. Avataaar/Circle Created with python_avatars Alex R2AUK says:

    ๐Ÿ‘ ๐Ÿ”ฅ

  14. Avataaar/Circle Created with python_avatars messageobliquespe100 says:

    Iโ€™m sorry – Iโ€™m just gonna say this subject is hocus Pocus – as well as that gadget measuring it -& that fancy microwave with a funny picture on it. Hocus Pocus.

  15. Avataaar/Circle Created with python_avatars Paul Cohen says:

    Good but TOO LONG.

  16. Avataaar/Circle Created with python_avatars Ryan E says:

    I never got a satisfying answer to what "common mode" noise was. I always saw CM chokes and wondered why they were needed.

    My understanding is common mode is simply a signal that it applies to all input lines (without prejudice). Further, since lengths of wire are effectively antennas, they can pick up high frequency interference which a choke will filter out due to the magnetic flux, which differential signals will pass through due to the topology of the choke and its windings.

    This video seems to confirm my understanding. Is there any additional nuance that I'm missing?

  17. Avataaar/Circle Created with python_avatars Ghl Scitel says:

    A transformer can be used as differential probe. Did you ever measure the CMRR of e.g. an Ethernet transformer?
    Citation: "All the grounds of the inputs are common"
    That is not completely true. It may only be true if you have no ground loops.

  18. Avataaar/Circle Created with python_avatars Serg Gorod says:

    ะžั‡ะตะฝัŒ ะฒ ั‚ะตะผัƒ! Just in time!

  19. Avataaar/Circle Created with python_avatars Optron Cordian says:

    For me, a real test is to solder both inputs together on the board and then measure. Now the output can result from the difference in length and position in the twisted pair of the probe wires. However, it is not worth the effort considering the very high rejection ratio even without that.

  20. Avataaar/Circle Created with python_avatars NeverTalkToCops1 says:

    Today, Daiyve demonstrates a set up where the probe costs LESS than the oscilloscope!
    Just grab that R & S scope for $23,632.

  21. Avataaar/Circle Created with python_avatars IanScottJohnston says:

    Very nice explanation Dave, enjoyed this one very much.

  22. Avataaar/Circle Created with python_avatars PushyPawn says:

    Kamagatza ๐Ÿฅท

  23. Avataaar/Circle Created with python_avatars like dot audio says:

    Can you do balanced audio with these?

Leave a Reply

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