Hi I got a real interesting one for you today. Check this out! Oh okay, it's a humble microcurrent Eevblog forum member in Set Man reported that he recently purchased a microcurrent and was having a problem with the offset voltage of this and was getting in the order of like millivolts. You know, four or five millivolts. Something like that offset voltage with no current on the input.

And of course the spec is much much better than that. In fact, the Maxim chip inside this thing I've done videos before is rated for a nominal naught point or a typical for naught point 1 micro volts or 100 nano volts offset voltage typical. and if you multiply that by the times 100 gain in this thing typically even though it's a two-stage let's call it point o' 1 millivolts. And even if your account for like absolute worst case datasheet values, you know the right extreme ends of the bell curve for the offset of this side chip, taken into account manufacturing process variations and temperature over the extreme range, it's still rated for a 2.5 microvolt offset voltage or what's that point? 2 5 millivolts, um, offset.

But they'll order in, at least it and they'll get in at least an order of magnitude more than that. So something was up. and then when a second user named Ki Ox actually reported and posted the findings of a similar thing I thought uh-huh let's actually investigate. So I got myself a like a new production batch of 5,000 odd serial number micro current.

Let's plug it in and see what happens. So what we'll do is, well actually our short the input. It's got a short in switch to actually do that and I'll just like put it to the middle of the range. Doesn't really matter.

But just for reference, we'll do all our tests in the 10 ohm shunt resistance range or 1 millivolt. Put my crap. So there you go. It's powered on.

Plug it in and look, no is whatsoever. There's absolutely nothing wrong with this thing. Hmm. so you know you can confirm that on any meter you happen to use.

Here you go. Point: O2, That's it. No problems whatsoever. So like, um, and you know I tested a few of these and I got basically, um, the same result.

So I like what's going on I don't like, you know, did a couple of people get bad ones? Oh oh oh oops, what's going on here? So why does the 121 GW measure different to everything else? I mean 3 millivolt may actually minus 3.7 million volts. This's our you know, the order of what they were measuring. So what's going on here? Take the exact same leads, plug them in here again to verify this is a bloody good 7 1/2 digit keysight. Me - nothing I don't understand.

Let's try the Keithley What? 10 millivolts? What the heck's going on? I mean if you have a look at the microcurrent schematic, it's the output is fairly low impedance through a protection resistor driven by the maximum chip, and there's a split rail system in there. if that moves, it shouldn't cause a problem because look at the star reference point in there. the Apple ground is taken from there and like it should work. So really, the only way that we could be getting an offset error here with a short on the input is if there was something wrong offset voltage on those maximum chips.

But this is like more than an order of magnitude out of the absolute worst case production spec over the entire temperature range. I've just never seen anything like this. So I Tried various meters in the lab here and sure enough only the Keithley DMM 75 10 the most expensive meter I've got here. highest spec meter I've got here in the lab and the 121 GW actually measure a high offset voltage and they actually measure different.

Minus four and was it - that will get in there Ten? Yep, - ten millivolts. It's a strange, but wait, you ain't seen nothing yet. Look at this. - 4 millivolts, Nothing.

and it should. This has a resolution of a hundred micro volts so we should be able to get it. Should be like reading like 40 counts there. What? Mind blowing.

And sure enough, look at this. if I put the key site and the 121 GW in parallel. Okay, - six millivolts? a match? No worries whatsoever. Switch that to Volts.

Poof. God, it's magic. So let's see what happens if we plug another set of leads in parallel here and actually plug these into the meter to confirm it. Aha.

Now we're getting somewhere. bang on and it's changed yet again. Fifteen point One millivolts. Let's go over here.

it's drop back down to nine. Aha. Now we're able to confirm it with other meters, but if we actually disconnect the Keithley meter, it causes a problem. And no, it's not the input impedance down here.

it's a standard 10 Meg. But I switch it to auto and it causes exactly the same issue. But only when we put two meters in parallel does it actually cause a problem. And sure enough, if we actually put pretty much any two meters in parallel, we're able to get something.

Once again, it's a different value every time. So you might think aha, when you put multiple 10 Meg input resistances in parallel, that might do it Well, let's check that out. there you go. And that meter in parallel with 10 Meg does absolutely nothing.

Hmm, how about 1? Meg Nope, Which a hundred? K Nope. It's not the load, so you might think okay, maybe there's something wrong with that split rail generator in there because we've just got the single coin cell battery and then it uses that Lmv three-to-one Op-amp to just split that in the middle. Well, if we put our meter on the reference output with reference ground output, which is actually the output of the virtual ground and measure our battery, then 1.33 plus and minus. one Point Three Three Perfect.

So this is just madness. What we want to do is compare that because this is the new batch number, the Five Thousand one. Let's compare this with an older batch unit I happen to have one 3690 here, so from a couple of thousand, where am I plugging this into there a couple of thousand ago? and let's try that. Nope, it's absolutely fine.

That's the offset voltage you'd expect in this worst case thing. We had plug it into both. Nah, it's hunky-dory In fact, I've done this with many old ones over a whole bunch of serial numbers. Not a problem.

and I've confirmed this with a second one as well a second recent serial number. So something very strange is going on here with the with this new batch one compared to any previous old batch version which I've measure and I've never had a single reported issue with offset voltage. Apart from user error or you know, some other such thing, it just hasn't been a problem. Some have been a little bit higher, but the this is like an order of magnitude or more higher than what you'd expect.

Absolute worst case: I mean there's not much they can go wrong on this thing. There's just the two maxima precision chopper amps on there. It never had an issue with those. I'm always buying genuine parts for those.

We've got just the offset Op amp for the split rail thing. It never had an issue with that. And there's a couple of shunt resistors and a couple of switches. And Bob's your uncle I mean what could possibly go wrong with this thing? Well, you saw that it was different depending on what sort of meter we hooked it up to.

So even though we couldn't confirm it by a resistive load, maybe something else is going on. So let's hook up the output to a scope, see what's what? All right? So let's hook up a good low noise scope. We've got the road and Schwarz one here with us. Big 10 bit converter.

I'm using a proper Ab scope probe to be in C lead using the proper probing technique. So let's switch it on. see what we get? 10 millivolts per division and that's kind of like the noise you expect. like the high bandwidth noise of course.

So I really nothing doing there and we're in high-resolution Mo. You know we can go to sample mode. That's there. There's more of the noise, which you kind of expect from a high-resolution converter.

I've done a whole video on digital scope noise in quote marks. Um, but that's fine and dandy. It's no problem. So if we hook that up to the decade resistance box in parallel, let's go to a Meg there.

Everything's hunky-dory We can go up where there we go We can go down to 10k. Everything's fine. Let's go down to 1k. Oh Doesn't like that it doesn't.

You can see that offset drop a bit and you can see it over here. Hmm, we're getting somewhere, but this is a very low load. I mean 1k. We were seeing this with 10 megohm input impedance meters.

so let's go back to our trusty 121 GW here and switch between it. Whoa. This is heavy. Look at that.

4.6 million volts all set and sure enough you can see we're at 10 millivolts per division and you can see that it is basically dropping by that. 5 millivolts there on average. I Mean you can whack every John if you want. It's not gonna be a huge amount better.

We might have to put some more averages on Beacon World. that's bad. My God. Something oscillating in this puppy have-we doesn't look great, does it? Whoa, wee lad? Wow Wow We've got a whole bunch of high frequency stuff into that high frequency oscillation in there Wow And that frequencies about 2.5 7 mega.

Hertz Well, there you go. Hey, this little baby's oscillating. but what's oscillating? It's gotta be the Op-amp I've never had never seen that maximum Op-amp oscillate before. It's crazy, and if we work in the old board, let's give that a burl.

There we go. No worries whatsoever. And on millivolt mode, look at that stable of Zulu and as you may have started to guess by now, it perhaps us something to do with the capacitive load. not just a resistive load.

or probably doesn't have anything to do a resistive load at all. really. Um, so I've got a little decade capacitance box here and I've got nothing connected in parallel at the moment. Switch it up.

I'm at what is it? Five path, ten paths. Um, I'm just putting some way hello, try and keep my fingers off it. There we go. Look at that.

That's a hundred puffs. Wow There you go. And let's put sure enough, hundred puffs. Yep, that thing oscillates.

Let's get the old one. whack it in, see if it does the same thing. Nope. No, that where we up to.

let's put in. let's go for break. put in a micro farad. Nope.

Nope. No problems whatsoever. That's so they can do like. that's a hundred Mike hundred microfarads on the output.

The existing ones don't oscillate at all. and that's what I've always found. But there's something afoot here. Hmm.

I Smell some faint, rarefied, subtle problems here? Hmm. Alright, so let's look at both boards under the microscope here. Here's our problem: one and an older batch one. And apart from the color of the PC or the solder mask, yes, it was changed at one point.

There's like no obvious like manufacturing. You know, issues like solder joint issues or anything like that. Everything looks everything looks hunky-dory Let's go in and have a look at our Maxum tip. shall we? Sorry Sometimes it's hard to see those laser markings on their a be ia and sure enough that is correct.

As per the datasheet it's that's the manufacturing code on top for the Macs are 42:39 and I Always buy these are maximum chips from the genuine sauce because they are like it's the main thing that gives this up along with the resistors which hive comma guts are on before video linked in at the end of this that's worth a look to where I had a similar sort of our production specification problem due to the resistors and but that basically determines the full performance of this thing. So like what's going on I've I buy from like digi-key and sure enough if you have a look at the older board over here it's exactly the same a B hey I bet my bottom dollar they are genuine Maxim's So maybe let's go over here and check out the only other active our component we have in here which is our split rail Op-amp and this is an LM V three to one and the old one has the code RC one F on it. Let's have a look at this one here. Uh-huh three two one s but hey you know I mean it's a it's an LM like it's a three two one so but it.

but it does look different. it has those bars top and bottom, but technically that is different. Hmm. Now normally the offset voltage of this Lmv three two one Op-amp doesn't matter and well, it doesn't.

You can go I could go and demonstrate this, but you know, really, it's a trivial concept because the output reference point of our amplifiers and the gain of the amplifier is all determined by that split rail. So that rail can be anywhere within that three volt range doesn't have to be exactly Plus minus 1.5 volts can be plus half a volt and negative 2.5 volts for example. It doesn't matter, it's only that only becomes a problem when you've got you know, the Headroom of the amplifier to swing the output voltage. But it's actual value does not matter.

So technically, even if this thing isolated and that value changed by 5 millivolts or something, it's not really an issue. But uh-huh It also couples through via the battery in the bypass capacitor as through to the power rails of the max of the maximum Op amps. So maybe the power supply rejection ratio at that high frequency of the maximum Op amps is what's causing. It's a couple in and give that offset voltage on the output somehow something like that.

So in theory, it's possible for this LM V 3 to 1 to actually do that, but she's it's It was such a remote possibility and it's especially more surprising considering that I've already taken care of this in the design aspect of the microcurrent. I've put in a 270 AM output is this on a lot of people over the years of us with that is for and that is just for a stability of this Op-amp even though the LM V three to one is actually rated to drive a 200 Pico Farad load in a unity gain configuration without any serious resistance. Adding the series resistance on the output just increases the capacitive load that you can drive and keeps it stable. So we're seeing here oscillations of you know, like 50 to 100 powerful something was starting to go it.

I think 50 Pico Farad's so you know really something is seriously wrong here. This It's almost as if this is not an LM V 3 2 1 Is it just an LM 3 to 1 and they are two very different parts each. The difference between LM and LM V is V stands for voltage. It goes down to a lower voltage rail.

In the case of the Lmv, it's rated down. It's fully specified down at 2.7 volts, which just happens to be the dropout voltage of the battery in this particular application. which is why this little Op Amp is almost perfect for this sort of application. Now as it turns out, I did have a couple of older bags of some old leftover parts from previous runs or whatever.

this one's from 2013 and I've always used the Lmv There it is three-two-one IDB VR the Texas Instruments part and I've always bought them from digi-key or Mouser in that particular part and I've checked the number on that and Scott our C1f sure enough and and this one dates from September 2010 and it's exactly the same thing and it's an RC 1f but I did find an RC 1 K so K might be you know some other like date variant or you know something like that. but once again, they all have that RC one on them not that three to one, T or s and sure enough I did check out my Mouser orders and I did actually order. It looks like um, they must have been like out of stock or something like that perhaps of the IDB VR or whatever it is. um so I actually ordered an Lmv three two one is 5x from on semiconductor / Fairchild and if we go in and actually have a look at the data sheet of this thing, it's actually the only data sheet I can find that does not tell me what the designator is on the actual chip, the SMD code.

it just doesn't tell me. maybe I'm blind but I cannot find it in here. It's just got some crap to do with the evaluation board and things like that, but it does not have like there's the evaluation board. Great They're fantastic.

Thanks for the info on the eve our board but then it just goes down into the physical dimensions of the packaging. gives you no identifiers whatsoever. So I don't know if this is oh I assume it is like assuming that Mouse I haven't goofed up and they haven't given haven't substituted apart. Although as I said, at the end of this video, you'll see another video where this has happened before where I think it was digi-key wasn't actually screwed up.

The parts of the manufacturer screwed up the parts sending to them and I don't know whether or not it's an LM 3-2-1 or not. But yeah, anyway, the codes not there so I I still I've looked through a lot of Lmv three-two-one data sheets cannot find that code at all. I've looked through the various SMD substitution code lists on the internet, can't find anything with three, two, one, T or s. but in any case, the on semi one or / Fairchild one basically has exactly the same specs is still rated for the 200 picofarad load and stuff like that.

It's exactly the same. So why one works robustly with like a 1000, but a hundred microfarads load on the output and the other one doesn't like not even it doesn't even meet that 200 puff. Oh, that's the crazy thing I don't understand and if you have a look at the schematic, you granted it is a bit unusual in the way that you know we don't have a direct capacitance load on the ground here for example, because the bypass caps on the max 43 9 are actually a cross V + + V - so they're directly across the battery, not actually on the output, so it's kind of hard to say what the effective capacitance is on that virtual ground relative to say the negative rail, for example. So it's hard to tell.

But like I've proven that the Texas Instrument part with exactly the same specs, it just never oscillates with any capacitive load. It's completely robust, but this one isn't, so it's almost I Starting to suspect that it's actually an incorrect or possibly even a fake part, but fake parts from the likes of Mouser Virtually unheard of. That's why you buy from Digi-key and Mouser and Finals and the other reputable catalog suppliers so that you don't get the fakes. and I have got the assembler searching to see if maybe they can find the original real for this one, but unfortunately it wasn't a manufacturer real.

it was a mouse a real. So they reeled it and put their own Mouser sticker on it. so really, it wouldn't It's probably devoid of the original manufacturers label unfortunately, so bugger cut, My guts are there. and if we have a look at the on semiconductor LM three to one then we'll If you look at the market, they do have the mark in our description here.

specific device code: I I don't know, would that be three to one? and but then it would have the assembly location, the year and the work week. So really, it ain't that and it. and it's this in one has like the bars over the top, not these little dots here. so it's definitely not an on semi LM Three Two one.

So there's only one thing left to do. who do I test my theory that it actually is that LM IV three to one is to suck it out and put on a good one. I Found another Real Listen, old digi-key Real. once again our C1 K.

So I'll whack that in, see how it goes and there we have our new chip on there. Let's plug her in and try it out. Leave that 0.01 millivolts. No worries whatsoever.

That worked a treat and we're just adding some capacitance there and have a look at the scope. Yep, no worries. Beautiful. Okay so what I've gone and done has actually got a couple of ones that I could get in stock: Lmv Three, Two one zero.

It's the Lmv three two one is 5 X which is the on semi part which I think might be the one in here, but we'll have a look at that. We've got the LM V 3, 2, 1 m5 which is ATI part which is the same as the IDB V part which I know definitely works and has been used in almost all my production units. But there's an M5 variant, so whatever, we'll use that. and there's an L M V three, two one I L ITL is that no oil T or something like that and that's an ST micro part.

But they're all LM V 3 2 1. So let's give them all a bill. And sure enough, if we take a look at the a S5x variant, it's got that three, two, one on it. but it's got three, two, one B.

So we've themed seeing what 3, 2, 1 s and 3 2, 1 T. So it looks like the culprit might be this on semi part which is the LM V 3 2 1 is 5. Sorry, But what I'll do is I'll solder this one in and give it a go see if it has the exact same problem as the S ones that are in the current board. But once again, the datasheet doesn't tell you what that s or B or T or whatever it is means and Wow There you go.

That's a hundred millivolts per division. This is awful. Like we're talking a hundred and ninety millivolts offset. Now this is insane that that's like a half an order made it worse than the then the other three to one T or whatever it is which I presume is like a and on semi one as well.

So what is it with these on semi parts Wow Horrible. So that's definitely a problem. Let's actually disconnect the the meter. Oh, that's the one.

Twenty one. GW Let's switch it now. I switch to DeVault and did the did the Volt millivolt thing and it still makes no difference. Disconnect it.

No, no. look, that's just the scope. That is just the scope now. Wow, that is so this be a variant.

Whatever that is is grossly different in this particular circuit configuration. with this particular load with the X 10 Pro Bit like, let me actually disconnect the those leads from there. Wow Look at that. That's ridiculous.

And here's the three to one alt part. This is from Ste. There we go. It's K or one seven, seven.

Let's give that a try. Well go St. There we go. Point O4 and no worries whatsoever.

Clean as a whistle. There we no modify it. Let's put our capacitive load. Don't put it on the input.

Okay, let's winder up. No, that's still okay. A couple of hundred puffs? Yeah, no worries. that's whack on a hundred mic.

Whoa. Yeah, okay, it's not terrific on one hundred more. Like, is it yeah we go. Man, we've got some oscillation there, but with a hundred sorry with two micro farad's on there, hundred more a hundred Mike it's a little bit of switching noise.

don't worry about that and but otherwise it's I mean it's stable. No worries if we actually feed one milliamp into it, of course you know, No worries whatsoever. we get out. Well, that's off scale.

But anyway, there it is. There's a yellow line right on a vault. no worries. Umm, and that's with a like 33 micro farad load on the output.

so that S T1 is operating a treat. Okay, let's try the TI part. Let's try this M5. What is that? that's got a 13 on it? All right Here she goes.

Yeah, that one's not too bad, but it is showing now. slightly higher offset. we're talking not, you know. point almost point 2 millivolts there.

it's still, you know. order magnitude lower than the problem we're getting. and it's still okay. So I wouldn't quibble about that at all.

As long as it doesn't oscillate with any sort of capacitive load, then that's okay. So we'll work a load on 33 microfarad on there. So that's alright. Hundreds of puffs? Yeah, not a problem.

It gets a bit noisier up there, but still. like not in the order of millivolts. it's in the order of like hundreds of millivolts. So yeah, I would say that T that particular TI part is a pass as well.

By the way. I'm doing this on a different boards just so I did have different boards to play with and I haven't tested this one for its original offset. I probably should have, so that offset could be coming from the Maxim chips. So, but the main thing we're looking for here is that you know it doesn't oscillate like the on semi /r Fairchild part does.

So there you have it, that's a real interesting Murphy that's probably like a level 5 Murphy got you because not only do we have a a difference in between manufacturers parts when there should be no difference according to the datasheet they should both be will all of them. What do we test out? like for different chips there and only one out of those for the A S5x from our Fairchild / on semiconductor is the culprit and would you believe it that one is the one I actually originally specified in my bomb and that boring bomb comes from like 8 years ago. like way way longer than 9 years ago. but when I developed the first microcurrent I think there way before the micro current goal.

but I don't think I've ever really used that apart from obviously I'm this newer production run it looks like I still haven't confirmed. but I think the A S5x variant may have been used in there. so like just luck of the draw and a really you know I've always considered that part to be completely safe. you know because I designed in the as the aspect salt would take into account the capacity of our load and it shouldn't have been a problem and it's not on three out of the four chips we tested to a TI BRE and one St and one on semi and it doesn't even well.

I You know we could do further videos actually just testing that chip on its own. That might be isn't interesting, just putting the capacitive load on there. nothing else, just hooking it up on a bare board with nothing else in there except for that chip that might be fascinating I won't do that today, but if I do work it over just on, maybe Eevblog to channel or something like that. By the way, Eevblog to channel Linkedin at the end of this video somewhere here.

check it out. I think I'm up to like 54 55 thousand subs or something. so I was joking before to David that wouldn't be funny if we actually got a Youtube silver but and got to a hundred thousand subs for just my second channel. where I just throw random videos and stuff just you know, single take things and other stuff that doesn't really isn't published and produced for the main channel.

So anyway, so yeah, subscribe Eevblog - this tons of content over there. Let's see if we get that silver play button just for shits and giggles. Anyway, I hope you found that interesting that is absolutely fascinating that we found this problem. It's not something you'd expect in a precision instrument like this.

I've had problems with the resistors before that was like a supply manufacturing supply a distributor goof-up and I'll link that video in summer here. at the end. as I said and I you'd expect the Maxim chip to be at fault and no, it was oscillation in the virtual rail chip and against all odds like I do I don't know. So yeah, it would be interesting do a follow up video on that.

but there you go. That's how little you can get little gotchas like that. That caused an issue Wow it's like unbelievable. Anyway, just for kicks, let's hook up.

That's are we on the right range there? Let's just hook up our current Gen. Oh sorry guy. keep mixing. These knobs are but keep switching between oscilloscopes and it's like it's crazy.

There we go one volt. So let's actually do some single-shot trigger Auto normal single-shot I'm gonna switch that on and hey, there we go. Is it gonna ramp up cleanly? What I'm doing is just switching on I'm just operating the output of my programmable function. Gen here I Turned off Yeah, alright.

Go. Hey, look at that beautiful ramp thing of beauty is a joy forever anyway. I Hope you enjoyed that video. It's absolutely fascinating.

You don't see that one every day. I think that's one's a level 5 Murphy this and I'm still not sure the exact um, serial number ranges and stuff like that because I didn't keep track to that level. You know, maybe if your Apple or somebody you might track you know you might have a comprehensive a system that you know. A comprehensive production documentation system that actually documents what serial number parts came from, what supplier went into, what serial number boards, or what day and all that you know.

sort of jazz if you're a huge enterprise and you know that could be worth millions or hundreds of millions of dollars to you if you have any potential issues. But of course I'm very careful where I get the Maxum chip from of course and really, they're out and you know there's no substitutes for that chip. But of course there is the possibility of fakes for that. So I always buy those through down Mouser and digi-key but I'm pretty sure I've always bought the Lmv three to one also through mouse or digi-key or another reputable supplier.

so I'm not sure of the exact date codes and things like that are still maybe have to go through some old records I might be able to pull up something. so this is actually a problem that is not going to affect all micro carts with that particular chip on it. in all scenarios, I mean you don't, Even if you hook it up to the oscilloscope with the probe capacitance, everything, you're not going to see that. or if you hook it up to the right type of multimeter that so happens that you know that doesn't have the reactive load on it that is required to make this thing oscillate, then you're never going to see it.

So even hooked up to like a high-end multimeter and a high-end scope you don't see necessarily see the problem unless you put a a reactive enough load on there that causes instability and causes the thing to oscillate. So it's It's one of those like marginal cases that in this case I've actually got a production test. Gee, I don't have one here. It's at the assembler that actually measures the offset voltage as part of the production Go No-go test to come.

Remember the exact limit. That's like you know, half a Mil, Evol or something like that and I'd have to read my documentation on that. But like it obviously presents a small load enough that it didn't none of these units actually failed that test as far as I'm aware anyway. so I test for offset voltage I test for again on all three of the different ranges with the production test.

you know I think I may have done a video somewhere on some of those our production test cheeks Anyway, if I did I'll link them in at the end of this video. hopefully if I can find I've made too many videos I forget so I hope you found that interesting. That is like a real-world gotcha. a trap for young players, an older like on what's supposed to be an equivalent part across all manufacturers.

Guess is fortunate for us. I Guess you know like this problem is fortunate in that it allows us to see a real, a rare, quite a rare real-world problem like this. By the way, if you've got one of these new ones and it is an issue, I've only had a couple of people report it. Obviously, it's only going to be a problem on certain capacitive loads and things like that.

There may even be variations in the chip itself. The production chips I Don't know, but yeah, it looks like only several people have a report of that. But if you do have a problem with your microcurrent, it should. All previous ones should be fine.

I Think it's just this batch production run that may have had some of these on semi slash Fairchild parts fitted to it unknowingly. Anyway, yeah, if you do have one, contact me and we can arrange something so. anyway. I hope you enjoyed that.

If you did, please give it a big thumbs up. As always, discuss down below. catch you next time.