Current Sense Amplifiers (1/2): Why not to use an OpAmp (CMRR etc.)
ฝัง
- เผยแพร่เมื่อ 15 พ.ค. 2024
- Issues with high side sensing, common mode voltage and CMRR …
↓↓↓ Complete description, time index and links below ↓↓↓
First I have a look at a typical high side current sensing application using an OpAmp in differential amplifier configuration. There is also a short excursion why low side current sensing is easier but might not be desirable anyway.
Then I go over the limitations (common mode voltage, output swing) and a major drawback (common mode rejection ration) of general purpose operational amplifiers in this application. There’s some talk about dB and logarithm too.
00:00 Intro - there will be a second part
03:49 OpAmp - as differential amplifier for high side current sensing
10:24 Low side current sensing - the undesirable easy way out
12:14 Output swing - it’s either a negative supply or a rail-to-rail device
14:00 Common mode voltage - your positive supply needs to be as high as it gets
16:50 OpAmp differential amplifiers - why the short formula is a lie
21:35 Common mode amplification - the evil twin of differential amplification
24:26 Common mode rejection ration (CMRR) - the measure of evilness
27:41 Decibels (dB) - the engineers’ choice as far as units are concerned
30:41 Back to common mode amplification - from CMRR in dB
35:45 Wrap-Up - coming up next: examples and a live circuit
Tutorials: • Tutorials
Year Old Self-Build Analog Multimeter Teardown: • 37 Year Old Self-Build...
Current Sense Amplifiers (2/2): Examples and Circuit with LT6105: • Current Sense Amplifie... - วิทยาศาสตร์และเทคโนโลยี
Finally a video that explains so well and so practically on the subject of CMV, CMR, CMRR, how it gets injected and rejected, and why the way the metric was laid out as such on the spec sheet and how to interpret it absolutely and proportionally. My fundamentals have always been shaky until this video solidified my intuition .... the way you explain things just sticks to my head. I was here to learn current sensing but I ended up gaining a good few extra here and there
Thank you so much for the praise :-) I'm always happy when people comment that they got something out of my videos, but your feedback is really something else, very detailed. Thanks again!
Lovely and quite detailed of what numbers mean here in this case. Cannot thank you enough❤
You're very welcome 😃
This is probably the only (and absolute best) video I have found so far that explains CMMR so well, in detail, slowly, and practical, so I can listen while I'm typing and taking notes. Thank you so much. I enjoy your way of explaining a complex topic - step-by-step, drawing the schematic while explaining. I found this always the best technique to teach newcomers to complex topics and use this technique too in my classes.
You're welcome! And thank you very much for the praise!
Impressive explanation and knowledge sharing. Thanks a lot.
Thanks for the praise! And you're welcome :-)
Very good presentation.👍
Thanks for the praise 🙂
I have just found a reference to the ADS1115 here, the gentleman in question is doing something similar (ish) to me ,already some interesting stuff, especially about using multiple filters, I also decided to do some simple averaging in software and was pleased to see you suggest this, the more I dig the more interesting it gets, I think I shall need a few times through the vid to get my head around the maths but now its part two...cheers.
Yeah, you always get that +/-1 digit noise out of ADCs or - in my case - the MAX6675 thermocouple chip (which contains an ADC of course).
And by all means, enjoy part two, where it gets practical finally.
I have not seen a video this good in years. You explain everything and don’t loose us in the process. Thank you soooooo much for your efforts . Subscribed and liked
Thanks for the praise - and you're welcome!
Indeed it is good. He starts with problem. So we can see philosophy in design.
@@Spark-Hole Thank you very much!
Nice video, thank you for sharing it :)
Thanks for the praise and you're welcome 🙂
Thank you very much 🙏🏼
You're very welcome 😀
Great video. Just killed my plan to use that solution in my project. Will watch part 2 with great interest now 🤣 BTW, if I measure the resistors R1, etc.. To make sure they have the same error (let say both R1 and R2 are 0.987 for example, would that partly solve the issue. Since the video is a little old now, do you have a better suggestion than the LT6105 or this that ship still relevant? Thanks
First, yes, matching your resistors will alleviate the problem. Many current sense amplifiers at their heart do contain a differential amplifier with finely tuned (with laser etc.) resistors. Second, the LT6105 is still a production part and "recommended for new designs" by Analog Devices ;-) But there's a whole slew of current sense amplifiers out there. Don't get stuck on the LT6105 just because I used it.
Great and thank you so much.
You're welcome!
Great video 👍, have you tried measuring the hall effect sensor across the shunt ? In that case voltage does not matter neither the poles AC/DC.
Nope, I haven't. But hall effect sensors are a great option for measuring (high) currents when you need galvanic isolation (or don't want to bother with current sense amplifiers). However, hall effects sensors are a bit tricky to calibrate (unless you use a complete hall effect current sensor package) and they are not as precise as shunt resistors.
This solves a problem i have been having designing a lab psu from my project! Thank you
You're welcome!
@@robertssmorgasbord i however ended up using an instrumentation amplifier, at least in the simulation. Id have to go through the math to check if that would work IRL. at least in the simulation it works...
Can't see why an instrumentation amplifier wouldn't work, as long as you stay within its input common voltages. It's a slight overkill though, you don't really need the high input impedance of an instrumentation amplifier when measuring the voltage drop across a sub-Ohm shunt resistor.
Nice video, can you do another one going into detail of low side current sensing ?
Maybe at some point. Currently I don't plan to, but I will add your suggestion to my list ;-) Usually I do this kind of videos shortly before I have a project involving the shown stuff (in this case "Small Project: Analog Ammeter with LT6105 Current Sense Amplifier" th-cam.com/video/zin56AvovXw/w-d-xo.html), and at the moment I have nothing queued up involving low side current sensing.
Hi Robert. New subscriber here. Enjoy your channel very much.
Just in case this is a test, @1:10 where you bring out a package of Op Amps, you have those marked as LM471CN. Never heard of those before. ;). Thank you for your great videos.
Ah, you're just the second person to spot this "Easter egg" ;-) These were of course LM741CNs :-) Thanks for subscribing and you're very welcome!
Good one! Very comprehensive explanation!
Little mistake i noticed: 98.02/0.04 = 2450.5 - anyhow, your dB value is correct (would be ~74 dB for 4901) ~68 dB is correct
Thanks! (also for pointing out that mistake, it's quite hard to catch those things without having another person cross-checking)
Hi, really informative video! I wanted your recommendation to design a current circuit using shunt resistors with sensitivity >500mV/A because I want to measure current in range of mA
Do you have some more information about the circuit? What's the voltage of your load circuit? What's the supply voltage of your measurement circuit. What's the maximum voltage drop you want across your shunt resistor? A sensitivity of 500mV/A would (without amplifier) just require a 500mOhm resistor. And at 1mA that would give you a voltage drop of just 500uV.
Thank you very much
You're welcome!
I really enjoyed this video but there is something I am wondering about. If you were to construct the same differential amplifier circuit but instead you powered it with a floating source for a separate ground and vcc (i.e. a battery), would this circumvent the issue of the common mode voltage at the high side of the measured circuit? It seems like the op amp would no longer 'see' this common mode voltage and would only 'see' the potential developed across Rsense but I am not sure.
Yes, in that case the op-amp output would be relative to it's own power pins. That's fine if all it's doing is to drive a panel meter for example, but not if it's meant for feedback to the circuit being sensed (a current limit circuit for example).
I really have nothing to add to @Aitch-Two-Oh comment. Of course you could do a (so called) floating measurement and thus avoid all those nasty common mode voltage problems. E.g. you measure the voltage across your shunt with a "multimeter".
Very informative, thank you.But i have a question, @12.13 why we are not seperating load and sense resistor so load will be aboce side of transistor and sense resistor will be below side of transistor?
First, you're welcome! To answer your question: The circuit you're suggesting would work of course. But keep in mind that the load would be floating between the supply voltage (transistor off) and the voltage drop across the sense resistor (transistor fully on). In many cases that wouldn't matter, e.g. driving a DC motor or a relay coil. But if you want the negative connection of your load being ground (e.g. lab power supply, the load being whatever you connect to the power supply) you can't use that circuit (or low side current sensing).
@@robertssmorgasbord Oh thank you very much again, sir can you please help me?Using this seperated circuit, i have a big electromagnet as load 1770 number of turns, 17 ohms.A diode parallel with electromagnet for protection of reverse current when power off.I controlling it with 125 hz Pwm from Arduino.Switching component is fds9953 mosfet.I placed a shunt resistor 0.5ohms and its parallel rc filter(2.2k 1uf) to smooth pwm for current sensing.Supply voltage 12 volts.I will connect Ads1115 between shunt resistor.Do you think is there a problem?I want to look current depending on pwm.
@@talhamsi Assuming I understand your circuit correctly: The cut-off frequency of your RC filter is 72Hz - that's still quite close to the 125Hz PWM. I would lower the cut-off frequency, otherwise there will be a lot of noise at your ADC input - let's say to a tenth of the PWM frequency (12.5Hz). Since the input impedance of your ADS1115 is several MOhm you could do that by simply increasing the resistor value to about 13kOhm. Also, you could use 2 or 3 of such RC filters in series to get a really smooth signal to your ADC. Of course you can always do some averaging in software. Also, such a large inductivity (the electromagnet) does cause nasty voltage spikes. Depending on how fast your free-wheeling diode parallel to it and your MOSFET in series to it switches, there might be some of those spikes still present across your shunt resistor. To be on the save side you could put two diodes in reverse polarity parallel to your shunt resistor (the voltage across it should never exceed 353mV). Just to protect the input of your ADC. Well, that's all that comes to my mind right now. Hope it helps.
I just stumbled upon this video and it's highly informative. I'm currently doing a project which requires current sensing. Is there any way to improve the accuracy of sensing the current flowing through the circuit using opamps (other than improving the ADC's resoulution) like having multiple stages or buffer stages.
Using additional op amps will do nothing for you but add additional errors. There are precision current sense amplifiers available that will give you +/-0.5% (e.g. www.maximintegrated.com/en/products/analog/amplifiers/MAX9922.html ). Also, if you don't need the large common voltage range of a current sense amplifier, you can use an precision instrumentation amplifier, which should give you better that +/-0.25% (e.g. www.maximintegrated.com/en/products/analog/amplifiers/MAX4208.html ).
@@robertssmorgasbord Thank you so much... Helps a lot with my project.
@@briansamuel327 You're welcome :-)
AD620 instrumentation amp is able to handle relative high voltages. But it is expensive!@@robertssmorgasbord
Hi, that was useful, thank you.
A regular, but non-trivial, requirement these days is the need to measure sleep current in the low uA or even nA, without creating a brownout condition when the processor wakes up and starts drawing mA. The current sense amplifiers I've looked at simply do not have enough gain. I wonder if you have a solution for this.
Btw, hang on to those LM471's, they're a rare find!
Ah, you noticed the LM471 thing 😅 I found them vastly superior to the LM741s 🤣 Anyway, there is a reason why multimeters usually do have a separate jacks for A/mA/µA. Covering a large dynamic range (mA to µA/nA) is quite tricky. I guess (!) you could use several shunts in parallel and switch them in and out with relays or MOSFETs). Or you could use several amplifiers with different gains and just accept that at higher currents some of them will go into saturation. Of course you need some logic (MCU) to choose which amplifier output to use (a nice ADC with an input multiplexer would come in handy here). But these are just some things that come to my mind, I never build such a thing myself.
thank you sir , i made it with lm558 and i2c oled display controlled by arduino
You're welcome! Nice project you have there. But are you sure about the LM558?
@@robertssmorgasbord sorry its lm358 i removerd it from laptop motherboard
@@user-nq2kl8gj4q Ah, that makes much more sense!
Nice video, thank you. Can I use 5V to v- and -48V to v+? Thank you in advance.
Tsi
Can you give me the time (XX:XX) in the video your question relates to or be a bit more specific?
Nice vid, btw!
Thanks!
Thanks Robert for such an informative video...Suppose, we use LM358 for high side current monitoring but take the positive terminal as the reference (ground voltage)!!!! Would CMV still be a problem. I am powering the LM358 and MCU with another power supply which is galvanically isolated from the supply under test.
I thank you for the praise! To answer your question: I guess that could work, as long as your shunt resistor (and the LM358) are before the series transistor and the load. The LM358/MCU ground would be connected just after the shunt resistor and you would use the LM358 as non-inverting amplifier. Great idea! Let me know how it works - and don't blow up your MCU ;-)
@@robertssmorgasbord It worked partly....If I measure current only, it works, but if i try to measure voltage also, it doesn't....how could I share some schematics with you - to show what I did?
@@m-kris Sorry, you're comment landed in the spam folder. Just found it there and released it. Maybe you could upload your schematic (as PDF) to a Google drive and post the link here.
Good one, Thanks! Btw i am trying to sense current through 100 ohm resistor connected across op-Amp using a DC Power supply at 11 V and current set at max. 20 mA. But the moment DC PS goes into CC mode, a ripple voltage of 20 mV is observed on DSO. Is it because DC power supply is a voltage source and i need a precision current source? I don't understand why exactly.
You're welcome! What kind/brand/type of DC power supply are you using? In most (cheaper) DC power supplies the CC mode is implemented much worse that the CV mode. E.g. my OWON P4305 overshoots enormously when I use the output on/off button in CC mode. I've never measured its ripple in CC mode, but now I guess I have to.
@@robertssmorgasbord Thanks for your response! I am using GW Instek PPH 1503D DC PS. The ripple value in CC mode is not mentioned in the datasheet. Some suggested to use Precise Current source but i couldn't understand why can't i use this DC PS in CC mode itself without voltage ripple considering 20 mA passing through my 100 ohms resistor should measure 2V across on oscilloscope.
@@vishalgdev_ Well, that's certainly not a cheap lab power supply. But if they don't specify the ripple/noise for current, they probably didn't put much effort into the CC mode and know that it's atrocious (OWON specifies 4mA RMS for my P4305). Regarding you numbers: You measured 20mV ripple across that 100 Ohm resistor, so a ripple current of 0.2mA. While that's certainly not great at a 20mA load (1% ripple), it's also not totally atrocious.
@ 26:17 You calc 98.02 / 0.04 as 4909 but if we multiply by 100 /100 it is equivalent to 9802 / 4 = approx 2500 . Somehow you got the correct result though 67.78
You're right, the 4909 is totally off. I have no idea where I got that value from. But obviously I continued working with the correct 2450.5 😅 Anyway, thanks for pointing out that mistake and kudos to you for noticing it.
@@robertssmorgasbord You're welcome. A lecturer at my uni used to do incredibly complex calculations very fast by rounding everything off and calculating within 5-10% the correct answer. What I realised then was how fast I could at least get a ball park value and also cross check my answer . It has since become a habit to do back of envelope calculations. e
@@das250250 That's a very good habit. Again, kudos to you! I guess it's an extension of the "Never trust a statistic that you haven't faked yourself." principle: "Never trust a calculation you haven't messed up yourself" 😅 Most of my errors I catch when cutting the videos, but sometimes something just slips through the cracks.
Can I measure AC current through with that? Or the response of op-amp would be too slow?
Well, in principle you can measure AC current with that, if: A) The voltages at the inputs of the op-amp never exceed the allowed common mode voltages for the op-amp type. And B) the op-amp type is fast enough for your AC frequency (bandwidth, slew rate, gain bandwidth product).
Interesting. Can't we isolate the ground of the measured circuit aside and have another same ground for the differential amplifier with the 0-5 V circuit ? So we get rid of the 40V problem 14:40
But in this case, and assuming there is a common mode gain, V+ and V- will be referenced to what ?
No, unfortunately we can't. And your second comment basically contains the answer. The + and - inputs of an OpAmp are referenced to its negative power supply Vcc-. You can have a look at, for example, the LM741 datasheet ( www.ti.com/lit/gpn/LM741 ), page 7: Current into the bases of the input transistors has to flow to V(cc)-. So while in principle an OpAmp measures the voltage difference between its inputs, technically that's done by "measuring" the voltage of each, the + and - input, to V(cc)- and then "calculating" the difference between the two voltages.
So, if the data sheet says the op amp CMRR is 85dB, does that mean that the error due to the resistors mismatch in Vout of the amplifier will be 0,33% regardless the precision of the resistors?
Yes. Here's explained how that's measured for op amps: www.analog.com/media/en/training-seminars/tutorials/mt-042.pdf You will need some resistors with 0.0001% tolerance :-) Well, there's actually a second method described not relying on insanely precise resistors.
It is good continue
Thank you very much - I will!
If you want to measure high BW current measurement, this is the best configuration, no hall sensor, isolation will work any better.
Sorry, just noticed I never answered to this comment of yours ... Indeed, it's the best configuration. Isolation for high bandwidth is best implemented digitally after an ADC.
Thanks for the great video and explanation. Just a side note... I think I always hear you say "eighty nine" (89) when you actually write 98 (ninety eight). That explains the 471 error too. No big deal, it's just a fun fact. :)
Ah, somebody noticed the LM471/LM741 thing already, but you're the first to notice the 89/98 stuff - I didn't when filming/cutting the video ;-) Anyway, these are properly leftovers from my dyslexia I had as a child :-) To this day I sometimes switch digits around.
@@robertssmorgasbord Switching numbers and digits has its own one, it's called dyscalculia! :)
@@shridharambady2069 Dyscalculia - that's new to me, was never diagnosed with that :-) I'm generally OK doing mental arithmetic, was an A student in math during high school and did even the advanced math courses in senior high. But I'm still swapping around digits :-)
@@robertssmorgasbord Thats so interesting. I guess digits and numbers are letters and words as much as they are mathematical sentences. It's all inscriptions.
Went looking for someone pointing out the 471 thing xD. I thought the 89/98 swap was about second language though. I still have problems in German since they say "eight and ninety" for "98".
According to the Ramon Pallas-Areny paper, the best CCRR that can be achieved for a unity gain diff amp with 0.1% resistors will be only 54 dB.
I haven't read that paper (BWT do you have a link or the name of the paper?), but going by the him being a professor of electronic engineering and assuming you're citing him correctly I'd say that's absolutely correct. I'm just guessing here why you're posting this: Because I estimated in that video at some point a 88dB CMRR for 0.1% resistors. But my estimate was not for a unity gain (gain = 1) differential amplifier, but for one with a gain of 98. Professor Pallas-Areny's statement you cited is about unity (gain = 1) differential amplifier.
@Robert's Smorgasbord No, I posted this just to remind people that it is always beneficial to use higher k values rather than unity gain. However, as pointed out in the study, one must be careful using high k values in an attempt to maximize the CMRR as it may not be realistic due to CMRR limitations of the op-amp self, if it is to work in applications above DC. It is suggested to use lower k values and trim one of the resistors to optimize CMRR. Sorry I have the paper somewhere in my collection, but you should be able to find it in a google search.
@@dm-ot1st Sorry I didn't get the point you were making right away, but of course you're absolutely right. Things get really interesting when you start to optimize an op-amp circuit for CMRR and bandwidth (gain-bandwidth product limitations), 'cause those two are really going against each other. Anyway, thanks for your input!
Are there any significant differences between a current sense amp and an instrumentation amp?
The inputs of current sense amplifiers usually don't have a very high impedance. Which is OK in a current sense application, because you're measuring the voltage drop across a very small resistor. Instrumentation amplifiers on the other hand offer very impedance inputs. This way you can measure high impedance signals without worrying about any currents going into / coming from you amplifier inputs distorting the signal. Also, most instrumentation amplifiers not happy with input voltages over/under their positive/negative supply voltage. That's something many current sense amplifiers can handle without any problem, at least in one direction, usually input voltages far above their positive supply voltage.
@@robertssmorgasbord Thanks for the detailed reply.
@@Robert-un3cf You're welcome!
so why exactly is it a problem that the ground reference of an LED connected to an LED power supply is slightly above ground? or is that just related to cmmr?
In a LED power supply a low side shunt resistor is usually not a problem. Low side shunt resistors can become a problem when, for example, you have several power supplies in your circuit. Then the "grounds" of the different power supplies are no longer THE ground of your circuit. That is you have currents flowing between grounds of your power supplies :-(
Why do you need to have the rail voltage of the opamp as high as the load voltage?
I'm using the LM385 to sense the current on the low side and the rail voltage I have on my opamp is 12V, I'm using the LT431 to have a stable reference voltage that I use for the course/fine adjustment potentiometers that go to the non-inverting input, the output of the opamp goes to gate of the mosfet that needs the 12V to be able to turn on...
You say that the rail voltage of the opamp has to be the same as the load but I have tested my electrical DC load with an input voltage of 70V and it works perfect, how is that possible?
So either you are wrong with your statement or you're missing something but I can assure you that my input voltage is at least 70VDC coming from an old audio transformer that I've repurposed as my bench power supply so I can choose a few fixed output voltages of 25VDC, 50VDC and 70VDC and I also left the option to choose -12.5VDC/+12.5VDC, -25VDC/+25VDC, and -35VDC/+35VDC which is pretty handy to have, however the highest voltage I get out of my transformer is 70VDC and I have no issues connecting it to my DC load I made using a 500V, 19A, 235W maximum rated mosfet and the sense resistor I used is a recycled one from an old amplifier which is a x2, 0.22Ohm, 3W in one package, it's connected in parallel to the mosfets source to ground so I have a 6W, 0.11Ohm sense resistor from where I feedback to the opamp and it just work with absolutely no problems at all.
I guess you're referring to th-cam.com/video/oQrWR6Vbb6M/w-d-xo.html "Common mode voltage - your positive supply needs to be as high as it gets". I'm talking there about high side current sensing (see also the circuit diagram while I'm talking about that). When you're using low side current sensing, as you do, this does not apply. See also th-cam.com/video/oQrWR6Vbb6M/w-d-xo.html "Low side current sensing - the undesirable easy way out".
@@robertssmorgasbord aha, well I didn't know that with low side sensing it doesn't apply, I guess that I'm lucky going for low side sensing then because if I did high side sensing it would have blown up right in my face.
I can't know everything right? That's why I watched this video and commented because it didn't make much sense to me so thanks for the explanation and sharing your knowledge online, I really appreciate that.
@@RicardoPenders You're welcome! And yes, one can't know everything. I certainly don't know everything. E.g. a circuit design of mine has been shot down in flames on the EEVbolg :-) I just went there to ask if anyone sees any problems - and several, very helpful people, did. What's the saying? No such thing as a stupid question!
Would there be anything wrong (other than overkill) with using a differential amp configuration on the low side?
No, absolutely nothing wrong with that, besides that it is really overkill.
@@robertssmorgasbord I do not think it would be overkill. Without a differential amplifier you are very sensitive to any voltage drop between the sense resistor and ground. The sense resistor should be connected with separate sense connections requiring a differential amplifier.
@@larslindgren3846 Well, you are right and my statement was a wee bit too general. As soon as your precision requirements reach a certain level (and when you're using a four terminal shunt you're clearly aiming for precision) a differential amplifier is of course not an overkill but a necessity. Anyway, if you're just aiming for 1% or 2% precision with a simple shunt resistor a "good enough" OpAmp will be sufficient.
@@robertssmorgasbord I think that maybe true that you can use a singe ended amplifier when measuring small currents less than a few amps and the measurement system is on the same PCB as the power electronics and with a reasonable design. In many other reasonable cases the errors will be much larger than 2%.
Especially if the return path from the current shunt to the ground star point is a few meters and shared with other loads. I agree that the common mode rejection of a simple op-amp is often good enough if it is configured as an differential amplifier but the question was if you need to use differential connection rather than a single ended amplifier.
@@larslindgren3846 Well, it all depends on the application, doesn't it? Maybe we can agree on, that if you want to "do it right" you'll use an differential amp, and if you need to go cheap and can tolerate some larger errors you can go with an OpAmp in some cases.
Well paced and clearly explained...awesome production.
However, I was rewinding the video often to be sure the R4 were the same. The '4' was written differently: the schematic had an open four...the ideal values and equations showed both open and closed '4's.
I had issue with an instructor for over a year...grade suffered because of time wasted in clarifying these types of concerns. It was every example 9n the board, a class member would have to ask what is that...the cat scratch was indecipherable.
It is seen here that a 4 exists, but 8 was looking around for another one in case there was some alternate for an alternate design...like a trade off.
This is a channel I will recommend. Best wishes.
I amendment to my statement...what I thought was a closed four is a one...the slant is very pronounced (in the equation) and I see it as a four. Appologies.
Thanks for the praise and recommending the channel! And I'm sorry for my handwriting ... sometimes it is even worse. E.g. as a German I'm always torn between an American 1 (|) and a German 1 (1). My German 1s sometimes almost look like American 7s, and my German 7s look like, well, an American 7 but with a dash through it ( de.wikipedia.org/wiki/Sieben ).
@@jimmylightfinger1216 No apologies needed! As I've mentioned my handwriting is not the best ...
At times, I add the nose...sometimes the base. The situation will influence my usage of the nose and foot especially if I am using complex numbers or vector notation...make it contrast more clearly.
Very much appreciate the reply. 🤝
11:06 = you still need differential amplifier to eliminate the resistance of ground trace, which will add to the small resistance (milli-ohm) of shunt resistor.
You're absolutely right. I just wanted to explain why you might not want to use low side current sensing at all, not how to properly implement it ;-) BTW in some low side current sensing applications you can also work with a star ground to avoid that differential amplifier: One big trace from the low end of the shunt resistor for the high current path and a smaller one going to rest of the ground (plane) of your regulation and other circuitry. However, that might introduce some other errors (e.g. when measuring voltage to ground).
@@robertssmorgasbord Totally agree to your points. I liked your channel and lot of things to learn from them. BTW, for most motor control projects that I work on uses low side current sense for high BW measurement.
Thanks! Low side current sensing has definitely its applications. BTW I just watched your two "DC motor current measurement ..." videos.
you just remind me, sadiku, boylestad, and such more books before the pandemic :]
Wow! Being compared with Sadiku and Boylestad - who is also a Robert ;-) I'm blushing here! Thank you very much!
So the second problem is that CMRR from circuit (small variation in resistance)+ CMRR from the op amp itself make error on the output.
Indeed, I have completely omitted the CMRR of the op amp itself. I just wanted to make the point that the additional CMRR introduced by the resistors for a differential amplifier op amp circuit makes current sense amplifiers the better choice for high side current sensing.
This is an interesting topic for sure. I've been using TI's INA180x series current shunt amplifier in my BMS design, although on the low-side configuration only as it only supports up to 26V of common-mode voltage and the BMS otherwise supports battery configurations between 4 and 12 series cells (16,8V - 50,4V at full cell voltages). I've lately been interested in getting into custom DC-DC SMPS topologies via MCU control and been looking into how to implement the inductor current sensing and have wondered about this exact topic, " why not just use a OP-amp that supports high common-mode voltage for the current shunt sensing on the high-side?"
Thank you very much for that most interesting comment. At some point I plan to do some MCU controlled power electronics too (driving a 1kW/2kW single phase brushless DC motor including recuperation). But it will be a while until I come around doing that ...
I did a quick uncertainty calc just on the R values of +/-1% that SQRT(1^2 + 1^2 +1^2 +1^2) = sqrt(4) = +/-2% = Oh Look, your gain is almost -2%
Nice 😀 I should have shown that in the video 😅
you keep saying eighty nine but the actual number is 98. No problem, still a GREAT video.
First, thanks for the praise and your understanding! Second, I'm a certified dyslexic, so swapping digits in numbers comes very easy to me 😅
Great
Thanks!
LM471CN a yoke, isn´t it....
It's LM741CN of course :-) And you are the first one to notice that ;-) BTW I really did label that tube the wrong way.
I am a little bit late to the party.. I don't understand what the CMRR of the O-Amp data sheet has to do with the tolerance of the resistors. The CMRR of the Op-Amp is the output error of the INTERNAL circuit if both inputs are shortened and connected to DC different from GND (i.e 10V will bring the output to 0.1mV) .... That will add to your resistor error, so even with a great differential Op-Amp ( 120dB), the resistor error still destroys your performance. -> trimming needed.
Absolutely right. The point was that current sense amplifiers have build in, very precise (sometimes laser trimmed) resistors, and thus you don't have to fiddle with external precision resistors and/or trimming around an op amp.
BTW can you give me the time index where I refer to the op amp CMRR? I remember explaining CMRR in general and showing the CMRRs of current sense amplifiers in their datasheets, but not for normal op amps.
@@robertssmorgasbord 37:08 "As it's written in Datasheets" ... Etwas misverstaendlich.
Wie gesagt, das Datasheet it fuer den Op-amp, nicht fuer die externe Schaltung.
@@Iosens Ja, das war wohl etwas misverständlich - 'tschuldigung! Ich hoffe das Video hat Dir trotzdem etwas gebracht.
Respectively the CMRR with 1% resistors will be 34 dB at best.
Again, see my reply to your other comment: I didn't estimate the CMRR for a unity gain (gain = 1) differential amplifier, but for one with a gain of 98.
Wouldn’t a current mirror work better?
I guess you could make a current mirror work in this application ... but getting the design right (common voltage tolerance, CMRR etc.) would be quite a challenge. Would it perform better than a current sense amplifier? depends on the current sense amplifier you're comparing it to 😉 Anyway, I personally think it's not worth the hassle (again, that's my personal opinion - some current mirror gurus out there might disagree).
Is it lm741?
Indeed it is 😅
Using a isolated supply for op amp _ which you should always &
Shorting the V+ to ground of isolated supply _ will do it __
Of course you can completely circumvent the need for a differential amplifier by providing an isolated power supply for your high side current measurement circuitry. Then a simple inverting/non-inverting (op-amp) amplifier will do and all the common mode stuff goes away. I'm not sure if I agree with your "which you should always" statement though. An isolated power supply, even the cheapest isolated DC DC converter, will cost much more than a current sense amplifier chip. So if you're on a budget or have space restrictions it's not the best option.
In the era of ' Green' energy ( 1 Li-ion battery at xx Volts, DC high current motors) adding an isolated supply just for current sensing is not economically sensible
@@MrSummerbreeze01 Yeah, that's another aspect I didn't even though of. Those DC DC converters, especially the very small ones (100mW to 500mW) are not very efficient and they usually require a minimum load (10% of max. output). So you are constantly wasting energy.
Sir you didn't even explain how you got the expression of Vout
No, I did not explain that. It's the standard textbook formula for a differential OpAmp. See for example here: www.electronics-tutorials.ws/opamp/opamp_5.html . Or, if you want a deep dive that includes the errors causes by the OpAmp internals look here: en.wikipedia.org/wiki/Differential_amplifier .
What about current amplifying?
Ah, that's an interesting and kinda demanding topic. Basic circuits for that are quite easy. However, the details are a bit beyond me.
Or Application Note 92 by Jim Williams
Do you mean "Bias Voltage and Current Sense Circuits for Avalanche Photodiodes" ( www.google.com/url?sa=t&source=web&rct=j&opi=89978449&url=www.analog.com/media/en/technical-documentation/application-notes/an92f.pdf&ved=2ahUKEwjlptbpg9OFAxWp_rsIHSkMD60QFnoECBMQAQ&usg=AOvVaw05fjVkoi_OUpVE-R9fDyDU )?
/verifying Krell audioFool Electrics signal transmission theory
I like your pen ... where can I buy it ...?
It's a STABILO GREENpoint 6088. You can get at least the non-green variant (STABILO point) on Amazon: www.amazon.com/s?k=STABILO+GREENpoint&crid=274O6D2PSD3PZ&sprefix=stabilo+greenpoint%2Caps%2C147&ref=nb_sb_noss_1
The green-variant is made out of 87% recycled plastics 😉
yay
Thanks :-)
@0:55 what is that noise
I cleared my throat 😉
I don't see any problem with Low-side current sensing, It's just
Well, if you're looking for a simple, cheap, "good enough for the job" solution for current sensing, then there's really nothing wrong with low side current sensing. But as soon as your application relies on ground being really ground everywhere (not being e.g. 100mV above ground before your current sense shunt), or you have higher precision requirements (BTW 100mV is 0.02% of 500V, so in a "typical application" we're talking using your example maybe more like 100mV/12V=0.83%), high side current sensing is the way to go.
@@robertssmorgasbord I don't need ground to be ref to actual ground because I source them from solar cells. I usually work with 500 VDC most of the time, sometimes even up to 750V.
@@XiaZ Ok, I get it. I have to admit that I never had high voltage applications on my mind when I made that video. The primary application I was thinking about was precision current measurement in lab power supplies. So yes, I guess your application is an absolutely legitimate use case for low side current measurement.
I am just here to learn opamp circuit diagram for voltage measurement.
Well, I hope you've got something out of this video, it basically being about current measurement.
Good video but I would suggest neither. Lately, companies are not making op-amps for current sense with good CM parameters for high voltage or they expensive its best for HV noval designs to use a BJT current mirror pair and translate, obviously this is for applications for HV use like 100V/200V, But to your point stick to current sense op-amps for sensitive dedicated parts with good stability margins.
Kudos to you for rolling your own current mirror with BJTs! I made that series in preparation for a tiny little board for a analog ammeter in one of the first "lab" power supplies I build ( th-cam.com/video/zin56AvovXw/w-d-xo.html ). So no high voltage involved. Plus, the PCB had to fit onto the back of the analog instrument, so I was really exploring single chip (non SMD) solutions.
1.5x
Thanks - I guess.
When you say LM471 I think you mean LM741
Indeed I do! And I'm sorry, you're only getting a bronze medal for noticing that, because you're the third one 😉 But thanks anyway for making it to that quite dry video 😅
Just use INA254 or MAX49921
I like integrated shunt of the INA254. I also like the high precision and low CMRR of the MAX49921, but soldering TDFN packages is a wee bit beyond me.
22:10 - Yes but most engineers would use an instrumentation amplifier. Thus the internal resistors would be well matched for high CMRR. You wouldn't use discrete resistors like that. Sigh, I guess that is the problem with fly by youtube information videos.
Well, you are right about an instrumentation amplifier with its (in some cases laser trimmed) internal resistors alleviating the CMRR problem. However, most engineers wouldn't use an instrumentation amplifier most of the time in this application, because: A) Usually an instrumentation amplifier is more expensive than a current sense amplifier. B) You don't need the very high input impedance of an instrumentation amplifier in this application. C) Instrumentation amplifiers usually don't offer the wide common input voltage range (beyond the supply voltage) that current sense amplifiers usually do.
You mean LM741 not LM471….
Yes, I did 😅 Unfortunately (for me) you're not the first one to notice 😉
1st problem is not really a problem _
Sorry I have to ask, but what's the 1st problem? Been a while since I shot that video ...
You are saying Eighty Nine instead of Ninety Eight. Confused. Does your language reverse your digits when you pronounce them!! Check your video again before releasing. Just a heads up for the future.
Do you have the timestamp for when I misspoke? Anyway, 89 is of course eighty nine in English and "Neunundachzig" (nine and eighty) in German. So yes, in German you reverse the digits.
Don't follow the schematic. With Vin= 40V input, R2=1K,R4=100K, Positive input will be much higher than Vcc=5V. For sure it will damage the OPAMP.
That's what I wanted to point out there :-) An opamp will not survive that. You'll need a suitable, that is with a large enough common voltage range above its supply voltage, current sense amplifier there.
Nah, nowadays you just need to make sure the opamp can handle it. 15 years ago I would have agreed with your comment as a blanket rule, but a lot of fairly basic opamps have been available for years that can handle this like a champ without phase inversion or smoke. They can happily measure differentials that are riding on a common-mode that is 20, 40, 80, 270, even +/- 600V from a 3.3V single supply voltage. LT/Analog has opamps like the AD8479, as well as numerous "OVER-THE-TOP" devices like the LT6015 and LT3675. TI has similar opamps like the INA148 and INA149.
@@0x07AF You're absolutely right in regards to Linear Technology's "Over-The-Top" opamps (e.g. LT1615). BTW nice pieces of technology. However, INA148, INA149, LT3675 and AD8479 are not simply opamp. They all contain a resistor network around an opamp (INA148, INA149 and LT6375: "Common-Mode Voltage Difference Amplifier"; AD8479: "Very High Common-Mode Voltage Precision Difference Amplifier").
Shit - I've done exactly this: LM358 as a differential amp for sensing current in the middle of a circuit, and it just doesn't work. Now I have to learn about CMMR and shit? I fucking hate this hobby...
Nooo! Come on Frankey! You don't really hate this hobby. Remember the satisfaction when a circuit actually works ... It took me about 40 years to get where I am today (with a 25 or so year gap in between). And I'm still learning ;-)
I was about to do the same for a battery charger, but was trying to understand the implementation first. Glad I saw this video.
Now I want to know if I can compensate in software, and get away with using the LM358.
@@Subin_Roy As long as your software knows the voltage your shunt resistor is floating at, you can compensate for CMRR errors in software.
@@robertssmorgasbord Would that be a one time calibration thing, or do I have to read the floating voltage each time I need to take the current measurement?
The floating voltage will change based on the state of charge of the battery.
@@Subin_Roy The calibration would be one time. For example, you set the voltage your shunt resistor is at in 5 steps and the current going through in 5 steps. Then you measure the output of your OpAmp at all 5x5=25 combinations to derive an 5x5 (OpAmp output, shunt voltage) deviation table containing the actual current / OpAmp output at each of the 25 points. That you put into your code. In actual operation your code reads the OpAmp output and the voltage the current shunt resistor is at via the ADCs, goes to the nearest point in the table, and calculates the current as OpAmp output * table value. Of course that is a very simple approach. You could do some (linear) interpolation between the data points in the table to improve accuracy further. And the more data points your table contains, the more accurate the whole thing becomes.
Nice video, useful content.
When you wrote “add 10dB if increasing a power value by a factor of 10” (30:41), did you mean 20dB? As that would be consistent with the graph.
Thanks for the upload, it’s helpful for my project :-)
You're welcome! And no, the text is correct. But maybe it was a bad idea to put it in this context (it was just an afterthought when editing the video). The text was intended as a reminder that the dB ratio for "power quantities" like Watts is defined differently - 10 * log10( P1 / P2 ) - from that of "field quantities" like Volts and CMRR - 20 * log10( V1 / V2 ). Wikipedia has a nice article about that: en.wikipedia.org/wiki/Decibel