PART 1 is HERE: http://www.youtube.com/watch?v=CIGjActDeoM
Part 2 of Dave's Open Hardware constant current linear regulated lab power supply design. An in-depth look at the LT3080 datasheet, the LM334 current source, and a few more circuit tweaks before it's ready to build in the next episode to see if it works!

Now let's take a look at the LT 380 data sheet because we might find some practical considerations in here. In fact, I guarantee we'll find some practical considerations uh in here that we have to take care of in if we want to actually build this thing as opposed to just uh, doing a conceptual uh type, uh, top level uh design schematic that we've been doing up until now. and I Really like this device because it does just exactly what we want. Now let's have a look at some of the specs: output current up to 1.1 amp I'm going to use the To220 package of this.

It's available in uh, many different Uh packages a Dfn, a Dpack, and a So 223 version which is really nice or an Mop package, but we use the To220. It's really neat. um 1% um initial ACC accuracy on the Uh Set current. That's not bad at all.

Um, and one of the neat things about this is that a single resistor programs the output voltage. although we're not going to use that today um, but we might have to take aspects of that into account as you'll see um, but it's really neat as like the unlike the Lm317 where you got to do use two resistors to actually uh, set the output voltage, this one only needs one and we'll go into why: uh, low output voltage noise it's only 40 microvolts RMS through in that uh, bandwidth and uh, it's got it supports Um input voltages up to 36 volts. Terrific. We're not going to go that high today, but hey, if you're into high voltage power supplies, it can do it.

And the Dropout voltage um, that will be at uh, full load is only 350 MTS but we can check the Uh curves on that one as you'll have to less than 1 Molt Load regulation. Line regulation's awesome. Uh, now one thing. minimum low current? There it is .5 milliamps.

We have to worry about that and it's stable With 2.2 Mic ceramic output capacitors, that's all you need uh to make this thing stable and then it's guaranteed over any load you like. Terrific. And it's course, it's got uh, fallback, current limiting, and over temperature, so it's pretty bulletproof. just like the Lm317.

And there's the typical application which we've just spent a long time talking about. You got series pass transistor, the error ramp connected, uh, straight through. Uh, but instead of having an internal voltage reference like the LM 317, it's got an internal constant current generator which is actually 10, uh, 10 microamps. and it it tells you that over here here's the Uh distribution uh graph and you'll notice you know we've been talking about Bell cures recently.

Well, there it is again. and uh, that is 10 microamps constant current down through there now. although it doesn't tell you this in the data sheet, uh uh. it basically uh, it implies that because this is only 10 microamps, this constant current generator, it's pretty darn wimpy, right? You can force a voltage into the set pin just like you can on the LM 317 and it's and it's going to be no big deal at all.

But got to remember, there's an extra 10 microamps which has to flow out through there and we'll see that that matters later. and we'll just have a quick browse through the characteristic spec table here and see if there's anything that uh takes our fancy. shall we? Now uh, set pin current as we're set as 10 microamps, we don't necessarily need to know about the Minmax values of that because we we're driving it, we're not using an external external resistance, we're just driving that pin hard. So uh, really, all we care about is that typical figure which we'll use later in a uh, simple calculation.
Now let's take a look at the output offset voltage and what this is is V out minus V set. What that means is the actual output voltage, the real output voltage minus what you've actually set, uh, what your you've set on your adjustment poort or what your software has set and what you're uh, driving into that vet pin. In this case. Now these specs are for a uh uh, a uh control voltage of um, a VIN of 1 volt with an output current of one uh 1 milliamp.

Now it could be your actual Uh value may be plusus 2 MTS Look at that so it may not be spot on to if you put exactly 1 volt on that V set pin. If you're driving that V one volt then it could be 2 molts either side of that. Not a problem in the case of this power supply. 2 MTS that's neither here nor there, but if you're designing Precision uh, really? Precision applications or Precision Power Supplies then this sort of thing can matter and you got to take it into account now.

Uh, and you notice this dot next to it? there's it's actually wider than that, over the full temperature range. It tells you up here. Bingo Little Trap The other spec. there is only for an ambient temperature of 25 C And remember, this is not the ambient temperature.

it's going to be the junction temperature of the device itself. so it's dissipating all that power and your heat sink's gone up to 60. Well, your your lab might be 25, but your heat sink and your device is at 60. So just be wear and you'll notice that there's two different specs here.

Uh, depending on the different packages and the reason that they have these is because the dye used the Silicon dye is going to be different in these smaller packages as they have a larger D used in these larger uh Power type packages and the spec gets worse. Plus - 5 molts Oh man, Terrible. or plus - 6 Ms over temperature? Who cares for our case? But you know you got to be aware of that sort of stuff thing. The output offset voltage is fairly critical for is when you parallel devices is up and this is how you're going to, uh, increase the output current cuz this device is only rated to just over an amp.

What? if you want 2 amps or 3 amps or something like that? Well, you can parallel devices up like this and all you've got to do is include a Ser series balance resistor like this and you can do these. uh, do this to um, any similar type of voltage regulator as well. Now, uh, there is an alternative device, the Lt380 D1 uh I Think it's a bit rarer to get, but it actually includes a built-in balance resistor in there. and uh, the output offset voltage.
A small one like the plus - 2 Mill volts you uh see here quoted for this one, or at least nominal at uh at uh, room temp. It's if it's tight like plus - 2 mols, it means that your output um, balance resistor only has to be very small. In this case, they recommend 10 mli Ohms output resistance and it's still going to share. Uh, you know, 80 to 90% of the current uh between two devices or even better than that.

Uh, typically and uh, so you don't even have to actually buy a resistor for that. It's good enough to actually use a PCB trace for that and they actually tell you uh to. They actually recommend that a 10 10 mil width or a 10 th width uh trace on a PCB 20 th width Trace typical 1 oz or 2 oz copper. 1 oz is your normal uh weight copper.

Or if you're designing heavyduty power supplies, you might have ordered a 2 oz um copper PCB But you can get your bance resistors. you don't actually have to buy one. there's no bill, extra bill of materials item, and cost. You just include it with a PCB Trace Brilliant.

And as long as that Uh value is uh, high enough, then you can share the current between the two devices Ade adequately without one device heating up. uh, much more than the other device. but you can't make it too high cuz then you get a voltage drop. and in this case, let's say you got a 2 amp output and our Uh two balance resistors of 10 mohms.

Well, it's a total of 5 mohms cuz they're actually in parallel. It's essentially what they are. So that's 5 mohms. Uh, balance resistance or output resistance there at Uh 2 amps is going to give us a 10 mol drop on the output and that's not too bad.

At 1 volt, that's only 1 %. So the neat thing about parallel devices like this is that you can actually leave uh, one of the footprints unpopulated or multiable footprints on your Uh board. If you're designing a power supply like this and you want to save a bit of cost to begin with, OR You're designing a kit or something. Save a bit of cost.

You only have one device, or if you want two, or three, or four or more, you can actually parallel them up and you can just solder in the individual devices as you need them. Uh. Load Regulation is going to be excellent. Uh, from 1 milliamp up to it's full.

Specify: from 1 milliamp up to 1 amp. No problems. Line regulation up to 25 Vols Uh. Input: Not a problem at one, it's specified at 1 milliamp load.

Oh, n don't need to worry about that. Minimum Load: Uh Current: Now here we go: Minimum Low Current: Very, very important. We need to take that maximum figure there: 500 uh, microamps or half a milliamp as our minimum load current. If we don't do that, it doesn't tell you what's going to happen.
There's a couple of notes Here which we'll read, but it doesn't tell you. Just assume that's not going to be stable or it's going to have a larger Dropout voltage. or sorry. it's not going to allow you to go down to a as low of voltage as it could or whatever.

There's a whole bunch of different reasons not. If you don't meet that, a whole bunch of bad things can happen and ruin your day. So we have to, somehow, um, get a minimum low current over uh, our minimum low current of half a milliamp over our entire voltage range and it's only specified at a VIN range of 10 volts. If you go higher, it needs 1 milliamp minimum and there it is.

Note: three minimum low current yada yent current and it's a minimum low current required to maintain regulation. If you don't me it, it ain't going to regulate and that defeats the whole purpose of a power supply. Now the Dropout voltage of this part is interesting cuz it specifies it in two different ways. There's V control pin Dropout voltage and Vin Dropout voltage.

So if you look back back to the Circuit here, it's got VN and V control. Normally you tie these twoos together and we will in our application here today. But uh, if you want a really low Dropout voltage of this part like a low input voltage and minimum input uh Dropout voltage between the input and the output, but you happen to have in your circuit somewhere a higher control voltage then you can take advantage of that and get a lower Dropout voltage from into to out by Ty in v control up to a higher voltage up there like that. but we're going to time together and also not all parts.

Uh, have the extra V control pin some of the some of some of the packages actually. um will tie those two pins internally so you don't have it, actually have it available. So um, we need to take the worst case version of that because we're tying them together which will be there it is 1.2 Vols or at full that's at 100 Milah or at full current which which you have to take into account it could be as bad as 1.6 So our input voltage has to be at least uh, 1.6 volts above our output voltage. and that's this voltage here.

not over here. So uh, if our output voltage is uh 5 volts here, we need to have at least 6.6 volts or 1.6 volts higher here. And if we are using a 1 ohm resistor and we're drawing an amp, you're going to get an extra volt. and uh, so you need 7.6 Vols here minimum for a 5V output and the current limit here on maximum output voltage H Typical 1.4 amp, but you might.

Well, you might be able to push it that far and you probably can. but when you're designing and you want to set your maximum figure, the lowest one here is what you're going to use as opposed to say the minimum low current. you'd use the highest value. In this case, you want to be conservative and use the lowest.
So depending on the parameter, you either have to choose the maximum value or the minimum value. We'll choose 1.1 amp. That's what our circuit will be capable of. Now, as far as the output noise goes, most linear Regulators are pretty darn good and this one's no exception.

40 microvolts RMS are for the error amplifier noise. Now, if we have a look at the error amplifier here, all of the noise assuming that the input voltage is uh, perfect and there's no no noise coming in there, then all of the noise is going to be generated by the internal Uh current source and the error amplifier. So that's basically all the noise internally is going to be that 40 microvolts RMS So even if we feed force in an absolutely perfect voltage onto here with no noise at all, we're still going to get 40 microv volts or thereabouts. Worst Case: RMS Output noise.

But as you can see the Um set because this is a direct feedback loop and whatever voltage you put on here comes out here, any noise that you put on this set pin is going to come out here as well within limits of bandwidth and all sorts of uh, other things like that. So really, uh, the noise limit will depend on this. Now, if you're driving your circuit with a pot like this and you've got this fed to a uh, you know, a voltage reference or something like that, say a 2.5 volt voltage reference. You know, a really quiet low noise voltage reference, then your noise is going and you're driving that pin directly.

Then the noise is going to be pretty good. But if you're doing a Pwm signal and you're feeding that through your RC filter like that and then you're driving that with a buffer obviously, and you're driving that into the set pin like that, then um, your noise. Then any uh noise that you haven't filled it out here. Any noise on there is going to make it through to here and it's going to make it through to your output.

So filtering if you're using microcontroller control uh, filtering of your pulsewidth, modul modulated signal is important. But you can really up these values. You can, you know up them as high as you want to. Really, you know, absolutely Slaughter any noise and just kill it dead.

It's not that hard, you just need to choose High values. And if you want to care about the Uh Ripple rejection, then you have to, uh, figure out what 75 uh DB is for your various input uh Ripple which is specified at half a volt Peak to Peak. If you're powering this thing from a Uh A, Transformer and a bridge rectifier and A and a filter cap, you're going to get um 100. You know this is a full wve one cuz it's double 60 HZ so it's 120 HZ Uh Ripple There it's specified.

You know that's pretty good and you can calculate that if you're uh using um A you know a noisy input and you've got uh Ripple. But if you're using, say a battery input or something like that, then you don't have to worry about that at all. But if you actually want some real figures, you can plug those in for 75 DB there at the nominal half volt Peak to Peak and that's going to give you an output noise of less than 100 microvolts. So 100 microvolts doesn't uh, sound like much and it's not.
It's you know it's down in the noise. Although once again, if you got a really uh, low noise, very highs spec uh power supply system then it could matter. but just for a lab power supply like us, more than good enough. Order magnitude good enough but that's only at Herz that's for ripple AC Mains Ripple If you're paing this thing from an AC Mains inut what if you um, got a a switching frequency of 10 khz or 1 MHz Well where would you get that from? Well if you're powering this regulator, if this regulator is been, uh, the Lt380 is being powered from a DC Todc converter.

well that's going to have a switching frequency and that's going to have output noise and a very high efficiency uh switching regulator might be up near a megahertz or something like that. and in that case, look at the uh, look at the rejection. there at 1 MHz is down to only 20 DB That's a huge drop from 75 DB at 120 Hertz. And if we calculate what 20 DB is just like 75 DB there is.

You know the formula. you've seen it before. DB = 20 log. In this case it's going to be V out or our noise or Ripple output voltage over our V in in this case which is given as 0 5 Vols.

And if you calculate that and if you change the formula around because DB is already known and you work out what your output uh Ripple or output noise is going to be, then it's going to be 1110th of your input noise. So if your input is .5 volts, you're going to get 50 molts output noise. So it's not down in the microvolts region anymore. it's in the tens of molts and that can ruin your day if you're designing Precision apps and you've got and you haven't adequately filed your input noise like that.

Now, if you were actually using this in its traditional configuration with the external resistor and you were Reliant upon the uh pin, uh Set current at 10 microamps, you'll see that that's only nominal at 25 C. When you actually go up or even down in temperature like this, it does vary a bit. You know if you go right up to 100 and go up an extra 50 uh 50 Nan volts. Woohoo! But that could be significant.

And remember: remember when you're looking at these curves, don't fall into the Trap of thinking that your product is only operating at ambient temperature at 25. it's not. This is the junction temperature of the actual Uh device itself the D temperature. And because this is a power supply and it's dissipating power, that Junction temperature could easily get up to 100 depending on how you do your thermal design.

So if you're designing really Precision uh power supplies, you need to take that sort of thing or any power system that dissipates power. You got to take these thermal graphs into consideration. And the offset voltage once again doesn't really matter for our application because it um, this is versus load current. So the output offset voltage in molts is actually going to drop based on the Uh output load current.
So the output load current 1 amp here at ambient Uh temperature or see it tells you here TG Junction the temperature of the junction, not just the ambient Uh temperature because ENT temperature makes no difference to it at all. All it cares about its temperature. Anyway, you're almost going to be at 0.5 M Vols offset there. and if your Junction is up to 125.75 M volts offset.

And if you're design in Precision applications that could matter, take it into consideration. Now looking at the minimum low current which is quite important. As we said because we have to take this into consideration, then uh at uh a Um input to Output Uh differential of 1.5 Vols then you know your your minimum low current only needs to be3 milliamps and but when it uh Rises your input to Output uh differential uh Rises to a much bigger voltage, then um, your minimum low current needs to be higher so you might put in half a milliamp. that' be that sort of tells you in the Um top level specs but you might say design it for a milliamp just to be on the safe side if you didn't care about wasting that extra half a milliamp.

Now this this lad. transient response here. We do, uh, want to consider this because this Uh tells us our typical performance when you change your load like this. Here's the Uh output load current in hundreds of milliamps.

So we're doing a 200 milliamp jump in the load current. It goes from 50 milliamps up to 250 milliamps and then back down and you can see what the output voltage how it deviates because these Regulators aren't perfect. Okay, they have a transient response. when you're out put, current suddenly changes and this is the transient response you'll get.

If you've only got 2.2 microfarad ceramic uh cap, you can expect it to change by 50 MTS Output: You can expect it to uh, droop down like that and then then recover like that. And if you use a 10 microfarad ceramic, you can see that it takes a bit longer to recover. And uh, the uh droop isn't uh Well, in this case, the Um rise isn't uh quite as much. So there you go.

but that's at a nominal 1. uh5 Vols output. But uh, the greater the um your output uh, the greater the output capacitance on your regulator, the better your load transier response can become. But just be careful, you don't want to put a massive amount of capacitance on the output of a constant current power supply like this because that capacitor can store a lot of energy.

You know, if you put in a oh, I'm going to put in a big 2200 microfarad capacitor that'll really, you know, give it lots of uh, transier, uh performance? Well, there's a downside to doing that and that's when this thing switches into constant current mode. then it can't react cuz it's still got all this energy in the cap that can get dumped High current into your load because it's not current regulating. So you want to keep in a lab power supply like this with a constant current um circuit like this, you want to keep the out output capacitance as low as possible just to ensure stability in your regulator. Although I guess you could say oh okay, you might determine Transit response is a more important thing than uh, switch it over into constant current.
but generally you're going to want to keep that thing the output capacitance low so we're probably going to want it just you know, the same value or uh, twice the recommended Uh value just to be on the safe side that it recommends for stability of this particular regulator and just for a bit of completeness there, you've also got line trans response generally not important in a power supply design like this because our input um basically this means that line Transit line means your input voltage coming into your voltage regulator as opposed to uh load Transit response which is changing Uh current on your Um output. So um, input line Transit you the the power supply that's powering your power supply is generally going to be pretty stable and it's not going to change by this sort of current like this. Uh, this sort of voltage this is in this case has got a 3vt step in your input voltage to your voltage regulator. You can see that that the output gives a droop like that of 25 molt.

So there you go. Not important, but just thought I'd Mention it. Another thing to consider is the turn on response of the Uh regulator. and that's what happens to the output voltage here when your input voltage ramps up and you don't want it to overshoot by a massive amount cuz that can damage your circuitry if it's already hooked up to the power supply.

So this one looks pretty good. It looks pretty well behaved and we'll actually when we build this thing up, we'll actually check that. Now here's an interesting graph you don't normally uh, see these. Uh, this is a bit rare.

It's the residual output volage with less than minimum load. So it's basically telling you, um, implying how this device is going to perform if you don't meet that minimum load current requirement. And you know that one of the big banner specs one of the big highlight Banner specs of this um of this voltage regulator is how low the output voltage can go. Uh, it can go all the way down to zero.

That's what it claims, but only with a minimum amount of output current. And that's what it's um, basically uh saying at this is your test resistance here. So as your test resistance gets lower and the output current increases, then let's say you've got test resistance up here of 2K Okay let's say 5 Vols There we go. 5 Vols at 2K And if you've got a 2K output resistor with an input voltage of 5 volts here, when it's trying to set, you see the set pin is uh, grounded here.
so it's trying to set Zer volts output. but you don't get Zer volts. You get there. you go.

0.55 Vols or something like that. Terrible. So you really? so that big banner spec that it shows on the front page here you know it claims. Oh, output adjustable down to Z volts.

Fantastic. Yeah, here's the devil in the detail. It only goes down to 0 volts if you've got a Z Ohm output resistor effectively. So if you've got a 1K output resistor nominal on there, you're only going to be able to go down to about you know, 0.25 Vols or something like that.

And here in lies the Trap How do we get a Uh minimum output load current here on an adjustable power supply? Because if we just still sure we could just stick a resistor here on our circuit down to ground, make that 1K No worries, we're easily going to meet our minimum uh load requirement. Let's say it's 1 volt. Um, even at Uh 1 Volt, we're going to get one milliamp that easily meets our minimum load requirement of half a milliamp and we can probably go down to 0.5 5 Vols Output: We're going to get 0.5 milliamps. Great.

Okay, that sounds like a good solution, but what if our output voltage is 10? Vols Well, we got 10 milliamps and we're effectively pissing away 10 milliamps there just on that output resistor. And aha, Here's the thing to consider. if you're trying to get milliamp accuracy over here with your uh with your constant current setting. Let's say you want to adjust it in one milliamp steps.

You know your input voltage down here 0 to 1 volt. You want to adjust it with your micro in 1 mamp steps. Well, geez, you've got 10 milliamps at 10 volts. In fact, you've got an output current which flow which is a which uh uh changes based on the output voltage.

So your microcontroller over here that's driving all this thing will have to be smart and know well it knows what the output voltage is cuz it's setting it through here so it will have to know that load's 1K and then take that into account and then compensate by driving instead of uh driving. Let's say you wanted to, you know, adjust this to 5 mli amps and that would be 5 m volts. Well if you're drawing an extra 10 milliamps out of here, you've got to actually make because you want 5 milliamps only 5 milliamps Max to go into your load. So you know that you're going to at say at 10 vol output you're going to have 10 volts coming out of 10 volts on here.

10 milliamps going down here. you have to actually compensate and set this one to 15 MTS to no and it gets really. it all gets quite ugly. So if you use a fixed resistor like that I it's not a very elegant solution I Don't necessarily like it on a variable power supply like this.

and by the way, another little uh trick. if you're um, if you really care about how much current this whole thing draws, you might want to add an LED in there like that and that could be your power LED And um, so you get two for the price of one. So instead of wasting your uh lead current just lighting up a power lead, you might get it from the output and you can use it as an output lead or something like that. but then it doesn't work down at low voltages.
If this thing goes down to you know, zero volts or 1 Vol or you know it, it, it's ugly. Anyway, we're going to I reckon we need to find a better solution than the fixed resistor. So what do we use for that output current? Well I can't think of anything better than the classic uh LM 334 constant current Source it's a single resistor. it's available in to92 package.

It's not bad, goes down to very low currents up to 10 milliamps maximum. and so I reckon we uh, set this sucker for a value of 1 milliamp. Now the Lm334 has been around since pretty much the dawn of time. It's one of those classic devices that is still incredibly useful today.

It operates from 1 Vol up to 40 volts. It's got good current regulation. You can programable current from 1 microamp to 10 milliamps two terminal operation 3% initial accuracy. If you actually care about the you know, the absolute accuracy.

We're not that fussy at all. it could be 10% We couldn't care less. and uh, and it's available in you know, a to 92 package or an So8. So very very use ible cheap as chips I Love it! One thing we do care about though is its minimum operating voltage of 0 typically 0.9 Vols um, up to 1 milliamp.

So um, we plan to operate it at 1 milliamp or thereabouts, or half a milliamp maybe. So um, it's unfortunately it's only going to op out down to about 0.9 maybe 8. but hey8 mil. If um, we can get our power supply to operate down to say 0.8 or 0.9 uh, volts.

That's a lot better than the 1 1.25 Vols you might get on an Lm317 and then, um, you can go right down to zero if you want, but then you're Reliant upon the uh load actually providing the minimum Uh current. Then you can't use like a high impedance load actually below 0.9 Vols or or 0.8 Vols or thereout 0.8 Vols is a nice uh figure and 0.9 is nice because um, for a power supply to go down to is because the a single cell battery, a single cell you know, D Cel or able a alkaline or something like that uh will be basically pretty much dead from 0 down to 0.8 volts or 0.9 volts. So a power supply that can go down that low is pretty good and it can go lower but depend on the load. Hey, that's good enough for me.

I Like it and it's really easy to use. It's only a three pin device and you have a single set resistor like this voltage in like this I set. this will be connected to ground down here and this will be our output voltage and we'll have have a single resistor like this. and what value does it need to be? Well, you can go through all sorts of uh formulas and take into account bias, currents, and stuff like that.
But here we can cheat and uh, do the um, look at this uh graph here. R Set 68 Ohms and Bingo. Look, it settles at 1 milliamp so it looks like it's 68. Ohms will give us 1 milliamp.

68 Ohms is a nice E12 resistor value I Love it. And as you can see it, it operates down to 0.8 Vols No problems, It drops off a little bit. Um, it probably drops. Yeah, it probably dies at half at 0.8 Vols that's as probably as low as it's going to go because um, uh, we are operating uh, this thing at um, well sorry, we want a minimum, uh, load current of half a milliamp.

So really it's going to operate down to, you know there it is half a milliamp somewhere in there, around about 8 on the graph. so if we set, set it to one hey, we're going to be happy. You know we could go up to here and then get some extra voltage margin. You know if we are set was up there at you know, 5 milliamps.

We can set it higher like that but and get a bit more margin for our output voltage. but I don't think we need to do that. And if your output current doesn't happen to fall on one of these characteristic uh curves on this graph, you know if you're were 2 milliamps or something, you will have to use these formulas and you'll have to take the bias current ratio IND count which changes with your output current and that's the ratio that you can and plug into various formulas down here to calculate your resistor value. Now you remember this: 10 Micro current we've talked about quite a few times from this set.

pin on this. Uh, Lt380 doesn't just magically disappear when you drive that pin. it's got to flow out of there. It's a constant current generator, so it's going to flow out of that pin into that resistor.

and assuming that we're in constant voltage mode or l stuff is vanished, it's got to go flow out of here as well. And if we're got say, trying to set one volt on our input here, Sure, we'll get a volt here, but then it's going to be. um, there's going to be an offset error there due to that 10 microamp current. It doesn't sound like much.

but OHS law. Do the math. 10 microamps through 2K there is 20. MTS Once again, doesn't sound like much, but if you're trying to set one volt on the output there, that's a 2% error.

It's horrible. Want that? So what do we do about it? Now there's two things we can do about it. The first one, the obvious one, of course, is to lower these values until it gets to a point where you don't care anymore. And that's a perfectly reasonable design technique.

But in this case, well, we don't want to lower it too far. You know we could low have it down to four, use 470 ohms or 220 ohms or something like that. But ultimately, when you short out this Op Amp at high voltages, you're going to have a lot of excess current flowing out of there. and well, you're just wasting pissing that current away.
So I don't really like that solution at all. So the next way to do it is to put it in the feedback loop of this Op amp and compensate for it. And the way you do that is to break into here. So instead of that output going directly to the buffer there.

it's a pretty standard technique if you want to increase the output uh, impedance of your buffer or your amplifier or whatever is to Simply stick that in the feedback loop there like that. Bingo. You've instantly gotten rid of that resistor and the Opamp compensates for it. Remember, due to Op Amp action, this voltage here will be equal to this voltage.

Here, it'll do whatever it needs on the output to make that voltage the same. so you've effectively eliminated the output voltage drop across that resistor. but you still have the advantage that it's protecting the output. So when you short this, you're not going to short directly the output transistors of this opair.

Beautiful. But that still leaves this resistor up here. Well, we turns out we can actually extend that you can put all of this inside that feedback loop up here so we can connect that directly to the source pin so effectively. Uh, the voltage once again, due to Op App action that if you set one volt here, you're going to get one volt there and 1 VT on that pin.

And it's going to compensate for the drop through these resistors. So these resist can now be almost any value you like within certain limits of course. Uh, and it doesn't matter, the opamp is going to take care of it for you. Magic.

So that's what we're going to do. And uh, because you may. This voltage, uh, set here is going to come from a reference, uh, voltage. You may want to use a 2.5 volt reference voltage or something like that to match your opamps of your um, uh, microcontroller or something like that.

so you may not be using like, say, a 10vt reference. so you probably going to need I can't get rid of that? There we go. Probably going to need some gain in there as well. No problems at all.

It works exactly the same, not just in the buffer configuration, but also in the game configuration too. You put all your stuff you want to get rid of your feedback loop. Beautiful. So there you have it.

after all that design effort I'm pretty darn happy with this design I Think it's going to be the one that probably build up once again, you can add in a second device up here parallel them like that. We've mentioned stuff like that to uh increase your output current and uh, I might do that. and another thing you might want to do is uh, replace this uh, Crusty Differential amp here with a proper instrumentation amplifier like I mentioned before like an Ad620 or that's pretty expensive, an ad the lower cost ad623 or one of the high side current monitor uh chips monitor amplifier chips you can get specifically from for high side current sensing like this because it's going to be okay at 1 ohm uh like that. you know, just using a a fly? uh Jelly Bean uh sort of low-end precision.
uh Op amp that has, you know, 500 800 microvolts or something like that. you can get down to Millie aamp uh accuracy on this type of thing. but uh, if you go down if you drop that, uh, if you want to reduce your voltage, drop across there and use .1 ohms or something like that. Oh, you're generally purpose Op amps.

even your Precision ones aren't going to cut the mustard too much and you may as well go to a proper instrumentation amp or something like that. Um, now you can, uh, do lowside current sensing as well. But as as I mentioned before, you're going to get voltage drop across there. if we include the resistor on the ground return path, the current shun resistor on that ground return path.

from the output you the voltage drop there. There are ways around that, but oh, it's Dicky I Don't like it. Stick with the high side current Monitor and we're sweet. So I like that.

Let's Build It Up Said that last time, but this time I think I really mean it.

Avatar photo

By YTB

25 thoughts on “Eevblog #222 – lab power supply design – part 2”
  1. Avataaar/Circle Created with python_avatars Marcus Agrippa says:

    stop banging on shit.

  2. Avataaar/Circle Created with python_avatars newSogn says:

    Can someone help me understand why on low side current monitoring you still have a voltage drop? Let’s say the lt3080 is outputting 10v to the load, wouldn’t the voltage at the loads input still be 10v?…. As I write this I realize if a 1ohm resistor is used for current sensing then the load can only drop 9v and the current sense resistor will drop 1 volt….. so the load will actually only see 9volts….. I thought of this because for whatever reason I can’t get any op amp to work as a high side differential amplifier…..

  3. Avataaar/Circle Created with python_avatars Nachiketa Thakur says:

    did someone try making one? how was the experience?

  4. Avataaar/Circle Created with python_avatars Trent Jackson says:

    EA did a lovely lab supply 0 to 40VDC at 3A back in the mid 90s. I built and sold several.

    In more recent years I designed a supply ground up to be a bit more contemporary but using the same schema as the EA supply. The schema being the use of op amps and BJTs. The original design used needle movement meters. I used digital displays. 7 segment red LED to show voltage and a bargraph row of 10 LED to show current being drawn. The resolution of the LEDs was switchable from 10mA to 100mA. My design was good for 0 to 30VDC at 1A Dual rail tracking. I did the PCBs myself and did a low production run and flogged them off on eBay at 99-cent auctions. I lost a lot of money. It seems that Australian Made means very little these days. The money that I recieved was on par to a Chinese supply on the other side of the world. Disgusting. Australia is a first world country. Most people in Australia require at least fifty dollars or much more a day just to live comfortably. Middle-class adults would consume much much more than this. Minimum wage is close to twenty dollars a day now after tax. So that is 40Kpa.

  5. Avataaar/Circle Created with python_avatars 网汉 says:

    awesome 太牛逼了

  6. Avataaar/Circle Created with python_avatars JTB says:

    Is there a negative voltage equivalent to the lt3080?

  7. Avataaar/Circle Created with python_avatars Хлебников Денис says:

    Use 494 – it's gonna be easyer

  8. Avataaar/Circle Created with python_avatars Хлебников Денис says:

    in brif "never try to make lab power supply by your self"

  9. Avataaar/Circle Created with python_avatars ch2o2 says:

    ..amazing knowledge and presentation, that's great, Dave. But what I don't understand is – when you have so much troubles with pulling your minimum output voltage down to zero (status quo is 0,8V?) …why don't you bring the GND up to this level – which means: why not realizing a "virutal GND" which is rather easy to do with a smart op-amp configuration (especially for a skilled and smart engineer like you are)?

  10. Avataaar/Circle Created with python_avatars Felix Seib says:

    I have seen many of your videos, and I loved them because of their simplicity,
    and because you focus on the principals, and not overthinking things.

    But at this one it looks to me, like you ran into that trap.
    In the end you develop an opamp curcuit to overcome the the fact, that your are raping voltage regualtors as darlingtons.
    So… Why do you not just use a simple darlington transistor?
    The feedback loops look nearly the same then…
    The use of constant voltage regulators would just make sense, if you wanted to have a constant voltage or current,
    but if you let the opamps do the work, which is what you do, a simple darlington would get the job done.
    And as far as I can see, your regulators will not save you from building an oszilator on the current regulator by accident.
    So what should be the benefit of using this regulators, instead of a darlington?

  11. Avataaar/Circle Created with python_avatars Abdullah Almosalami says:

    Are people seriously complaining about the taps? I didn't even notice the taps until I read it in the comments because I was wholly concerned with the content Dave was trying to explain.

  12. Avataaar/Circle Created with python_avatars FalcoGer says:

    Why do you call the operational amplifier
    – a comparator
    – an error amplifier
    – a buffer
    and why?
    as far as I got it, it tries to match the voltage on the output referenced to ground to the input voltage on the + connection. Why do you sometimes draw only 3 connections? What are the other two connections for?

  13. Avataaar/Circle Created with python_avatars Rish says:

    This is a great video; I just had to laugh at the design requirement change! Even though, in this case, it makes perfect sense (for the application). Your design specs are 0V out. I wish I could tell the product spec guys that — eh; 0.9V is good enough! Instead, we end up spending more design iterations and dollars to give them the 0V out — in reality it won't be used down there!

  14. Avataaar/Circle Created with python_avatars Cpakdit No4n says:

    I havent learned so much neither at my university nor any other youtuve video, thank you so much for share your knowledge with all of us

  15. Avataaar/Circle Created with python_avatars Muhammad Mustapha Ibraheem says:

    Great job Dave, Thanks

  16. Avataaar/Circle Created with python_avatars Axel Lambert says:

    Very interesting, as usual. Thanks from France.

  17. Avataaar/Circle Created with python_avatars felixar90 says:

    Could you add some circuitry that drains the output capacitor when you hit the current limit? (And also when you decide to lower the voltage)

  18. Avataaar/Circle Created with python_avatars Phil Hooper says:

    Cam for the video content. Stayed for the content AND the accent

  19. Avataaar/Circle Created with python_avatars Sujatna Lokananta says:

    this is an advance tutorial for electronic engineer. the best electronic engineering tutorial I ever seen… increadible

  20. Avataaar/Circle Created with python_avatars M Dikdan says:

    This series of videos has taught me more than many university classes.

  21. Avataaar/Circle Created with python_avatars Hola! Jacob Corr says:

    Poor Dave, trying to make videos and you can hear Sagan crying in the background!

  22. Avataaar/Circle Created with python_avatars Power Max says:

    in LTspice, I stuck a PNP transistor where the emitter was connected to Vcc, base to the input of the LT3080, and the collector tied to the output, and I bridged the connection between the emitter and base with a 10 ohm resistor, so that the majority of the current was routed through the PNP transistor. (same config used for the classic 7805 10A linear reg) which, looking at closer, effectively made a Sziklai pair. In LTspice, this appears to work fine, although I think it does bypass device protections a bit. Can anyone give me insight if this would be problematic in real use?

  23. Avataaar/Circle Created with python_avatars BB-box says:

    Great video , might have a go at building this as i'm trying to learn more about DC power supplies LDO's bucs charge pumps etc 🙂 Think it would be a great idea if you did a video of a dual rail positive / negative dual rail tracking supply? Theres a lot of audio types that would love it !!! Or maybe dual rail is a bit over complicated for beginers like me? Anyone know if the LT3080 can be used for the negative part of a dual rail +-v supply ? have two and reverse the output of the second for the negative rail and link the amp & voltage control pins together for the tracking ?

  24. Avataaar/Circle Created with python_avatars slap_my_hand says:

    op amps are awesome.

  25. Avataaar/Circle Created with python_avatars xFuaZe says:

    Is it really the best way to up the amperes by adding more LT3080's?
    And how to increase the voltage range?

Leave a Reply

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