Please check over my LM317 power supply design!

[microbug]

Hi,
I'm making an LM317 current-limited constant-voltage power supply (I'm a semi-beginner). I am posting the schematic (attached) in case there are any glaring errors / mistakes I may have missed! I would really appreciate your help if you do spot something.
Thanks in advance!

[c4757p]

A few things. 1) How much power can your 100 ohm potentiometer for current control handle? That's a lot of current through a pot. There's a pretty easy way to divert current around it with a couple transistors, I'll see if I can dig up a schematic. 2) LM317's minimum load is 10mA. With no external load connected, your only load is R1. 1.25V/560R = 2.2mA, nowhere near 10mA. Decrease R1 or add a small dummy load. 3) In general, your capacitors are quite small. It will work, most likely, but you'll have better performance if you increase all of them by a factor of ten or so.

The only truly important error is #2 (minimum load). When not loaded enough, LM317 tends not to regulate properly.

[microbug]

1) I was thinking of using an (expensive) 10-turn 3W wirewound pot, but would welcome other ideas!
2) OK, I will add in a 100r resistor between Vout and GND. I forgot about the minimum load, but I don't want to decrease the size of R1 as it is good for my requirements as is.
3) Right - thanks!

Edit: Added new schematic.

[c4757p]

Do you have access to a negative voltage? Try this. Only 1mA goes through the pot (R2).

It's a 1k pot in series with a 200 ohm resistor, if you aren't familiar with SPICE format. You could eliminate the series resistor (it sets a maximum limit - without it, the current goes up to 1.25A).

The two transistors form a constant current source programmed to about 1mA (The V_BE threshold of Q2 divided by R4). That causes a voltage drop in the pot (1mA * pot resistance).The LM317 just regulates such that 1.25V appears between the output and adjust pins, so that leaves (1.25V - this drop) to appear on current sense resistor R1.

[microbug]

I will be running off a transformer, which is then rectified and smoothed. If there is a transformer I can use that will give me a negative voltage winding then I would use it but I don't really know what to look for - would something like 12-0-12 work?

[c4757p]

Yes, it would, but you'd be wasting half the power capacity of the transformer for this little bias voltage. Try to tell before buying it whether or not it's a split-bobbin transformer. If it's split (primary winding on one half and secondary on the other, rather than concentric), you can safely add a winding just by wrapping a few turns of wire around the existing secondary, without getting anywhere near the dangerous primary. You only need two volts or so (plus diode drop), and because you only need 1mA you can rectify it with a single, little signal diode and a small capacitor. Alternately, there are a million and one ways to generate a negative voltage from a positive one (cue Paul Price  ).

[microbug]

It might be easier to generate a negative voltage from the positive Vin, so I don't need to bother rectifying and smoothing another input. How would you go about generating a negative voltage?

[c4757p]

Apply a square wave to this circuit. You'll have to experiment with component values a bit, and I'll leave it to you to figure out where to get the square wave from (hint: 555 timer). Dave did a video on this recently.

Also, I'm playing around with that current control circuit a bit; there's no reason such a thing would need a negative rail, it's just that that particular one does. I'll see if I can get rid of that requirement.

[c4757p]

Can you afford a little bit more voltage loss? This version doesn't require as much voltage overhead, so you can get your "negative voltage" with a little trickery: put two diodes in the return path so that the output is shifted up 1.4V or so. That way, the original "ground" is 1.4V less than your new ground. For everything that comes after this part, consider the new low output to be ground and forget about the original one.

Don't let the way I drew it confuse you - R3's not part of the circuit, it's just a load to test it with. The two wires coming out to the right are the output.

In theory, Q2 and Q3 should be matched in gain, threshold and temperature; in practice, this isn't a precision circuit, just make sure they're the same part number and you'll be fine.

[Paul Price]

You can get any positive power supply to generate a negative bias voltage with a 555 timer. See attached. It generates about  minus two volts.

Unlike some other suggestion offered, this circuit does not require to match some components for gain nor limit the output voltage range by a few volts thus giving you a greater range of output voltages. It is slightly more complicated but might offer better performance.

[c4757p]

Unlike some other suggestion offered, this circuit does not limit the output voltage range by a few volts giving you a greater range of output voltages.

Hey, I'm just building on the existing circuit. I assumed he was fine with the 1.25V minimum output since that's what the original circuit did. Some people (*cough*me  ) might call that a bit over-complicated. You might as well just pull out the power transistors and op amps at that point - the reason you'd use an LM317 is that while a bit limited, it's dead simple.

[Paul Price]

As I said,

It is slightly more complicated but might offer better performance, plus the supply is also a standard LM317 constant voltage supply whose output extends down to near zero volts.

As for your original design, it is simple but it simply has quite a few problems.

Firstly I like the idea of using a diode across the LM317 to protect.

I do not like the idea of C4 as it would dump a large current into a circuit when first connected, likely damaging some component.

R2 would prevent the supply working as a constant current supply, since it is shunting the output current and so your design as it is no longer a constant current supply.

R2 would also require you to continually juggle back and forth between two potentiometers to adj voltage and current, if it at all was possible to get a constant current out of your supply as it is drawn.

Thirdly you have the issue of transient response of two LM317's, they have a tendency to overshoot and undershoot the output voltage set with a quick change of load and this would result in voltage as well as current surges that could damage whatever you are working on.

Forthly, the circuit is likely to oscillate rather than regulate at some output load, current and voltage settings, because of the the very slow transient response of the two LM317 current and voltage regulated circuits cascaded and competing to control the output.

Also, the effect of two LM317's in series does seriously subtracts you from achieving the full output voltage range potential of your power supply.

Otherwise from that I congratulate you on taking the initiative to create your very own power supply design!

[microbug]

I do not like the idea of C4 as it would dump a large current into a circuit when first connected, likely damaging some component.

Should I get rid of it then? It is only there to reduce ripple / improve transient response.

R2 would prevent the supply working as a constant current supply, since it is shunting the output current and so your design as it is no longer a constant current supply.

What do you suggest? Should I get rid of it and decrease the value of R1 to meet the minimum load requirement?

Thirdly you have the issue of transient response of two LM317's, they have a tendency to overshoot and undershoot the output voltage set with a quick change of load and this would result in voltage as well as current surges that could damage whatever you are working on.

Is there any way to remedy that? I'm thinking tantalum caps...

I assumed he was fine with the 1.25V minimum output...

You assumed right!

[Paul Price]

Tantalum caps would not help.

The problem is basically the design,  you could try mine or else creating something different on your own!

Firstly your design will not be a constant current power supply at all with any load resistor on the output. This means if you eliminate R1 and R2 and have a circuit with  no R1 and no R2,  without these two components your supply will always be set to the maximum output voltage and not function at all.

If you eliminate the output load resistor you still have current surges from the output capacitor, the larger the capacitor, the slower the transient response and the greater the surge.

If you eliminate the output capacitor the circuit will likely oscillate.

[microbug]

What is yours? Do you mean the schematic you already posted or something else?

Edit: If I drop the load resistor and lower the cap to 47uf would it then be OK, if not great?

[Paul Price]

The one I posted, but there are many many others out there in electron land.

[Paul Price]

Sorry, I don't think so, for the reasons I list. Perhaps you can take a look or breadboard mine to get some ideas on how the individual components form  a constant current supply/constant voltage supply that will work.

I would add that you can completely eliminate the 555 circuit and connect your adj. pot directly to ground if you don't mind the output voltage limit that would prevent any constant current power supply to regulate current at voltages < 1.2 volts.

For a constant current power supply to work it must have "full voltage compliance."

Hint: Your circuit has implemented it's two functions backwards. The constant voltage source must come before the constant current limiting stage.

The tail must not wag the dog.

[westfw]

The constant voltage source must come before the constant current limiting stage.

That's not the way it's done in the typical 317 "app-note"...
Obviously it's a current-limiter rather than a "constant current" supply, though.

[c4757p]

I would add that you can completely eliminate the 555 circuit and connect your adj. pot directly to ground if you don't mind the output voltage limit that would prevent any constant current power supply to regulate current at voltages < 1.2 volts.

Of all the deficiencies of all the circuits presented here, I'd say that's the worst. So you lose the limit if the output is shorted? 

Hint: Your circuit has implemented it's two functions backwards. The constant voltage source must come before the constant current limiting stage.

Every LM317 CC/CV circuit I have ever seen places the current limit before the voltage regulator, with the exception of ones like yours that combine it into a single stage.

[IanB]

Hint: Your circuit has implemented it's two functions backwards. The constant voltage source must come before the constant current limiting stage.

That's contrary to every other piece of advice out there in electron land.

[IanB]

Obviously it's a current-limiter rather than a "constant current" supply, though.

In reality there is no difference between these two concepts.

Every "constant current" supply is actually a current limited supply (you can make it supply less current by putting a sufficiently high resistance across the output).

Every "constant voltage" supply is actually a voltage limited supply (you can make it supply less voltage by putting a sufficiently low resistance across the output).

There tends to be a breakdown of the English language when people talk about these things.

You can regulate the output voltage to be no more than the maximum, but sometimes less.

You can regulate the current to be no more than the maximum, but sometimes less.

A practically realizable constant voltage supply does not exist.

A practically realizable constant current supply does not exist.

[microbug]

OK, I've had a look around - would the first design here work?

[mariush]

That circuit is from the LM117/LM337 datasheets (one of them, since the same IC is made by various companies) : 

http://www.ee.buffalo.edu/courses/elab/LM117.pdf

It's on page 18.   There are other good example of circuits and the extra explanations might be helpful.

Yes, I don't see why it wouldn't work. 

Though you might want to change some components with more easy to find ones, maybe 2n5884 (http://uk.farnell.com/on-semiconductor/2n5884g/transistor-pnp-to-3/dp/9556044 ) instead of MJ4502 and some newer opamps (though the lm301 is a very generic one, probably anything else would do)

[Paul Price]

WestFw and IanB, We might just be facing a wording, semantics problem here, and your knowledge about electronics is immense, but I am very certain you are mistaken here with what you both are saying, maybe sometimes if only by your way of explanation

I hope with 10000 words or less I can clarify the situation. I don't want constant current v. current limiting to become a "What came first, the chicken or the egg?" argument!

A constant voltage source comes first and sets the nominal, non-current limited output voltage, and the current limiting or constant current circuit stage then monitors the current being drawn out of the constant  voltage source and adj's voltage output down to achieve the constant current mode of operation. We can say that the constant current stage feeds back to control the constant voltage stage to make this happen.

You might notice, in my circuit that the constant current sensing and control is  the in of the  constant voltage output stage, but that it feeds back to control the LM317 voltage output to achieve constant current operation.

IanB states also in reply,"That's contrary to every other piece of advice out there in electron land." is also mistaken in the same way as WestFw.

You might notice, for example, in my circuit (which is in fact a commonly seen implementation of many other circuits published) that the constant current sensing and control is in the  constant voltage output stage, but that it feeds back to control the LM317 voltage output to achieve constant current operation.

If you try to do it the opposite way, the constant voltage stage will never work to provide a constant current because it is always trying (at the output of your supply) to provide a constant voltage.

i repeat, The constant voltage stage most be controlled by the constant current stage but part of its circuit senses the current in the constant voltage output state, and there is usually some way of feeding back or controlling  the constant voltage stage to correct the current to the desired set value.

So, is it a just a problem of semantics, that we are both trying to say the overall method of operation is to use a part of the whole circuit that to sense the output current used by the load to and control the constant voltage output stage circuit to achieve voltage compliance to maintain the output current of a power supply?

------------------------------------
Microbug: your question, "OK, I've had a look around - would the first design here work?"

The answer is yes, it would, but it is not a very good circuit.

This circuit would have bad transient response, and this causes sudden load changes (such as happens when you first connect your power supply to a circuit and a  high current is drawn) to cause your power supply to overshoot the correct output voltage and the excess voltage could be enough to damage your test circuit you are working with.

The bottom line is that the LM317 is a very poor choice to be somehow "jerry-rigged" to create a switching power supply. The datasheet shows how it can be done, but that is misleading. The datasheet writer is just trying to sell you on the idea to use this part,  to sell you their part, so they present this schematic to try to get you to use their part in an awkward way.

I think the idea of trying my circuit is better because it is simpler, it is a linear not a switching power supply, and their exists some range of linear constant current control.

Good transient response is an important characteristic of a power supply in that it allows the power supply to act well with sudden load changes.

IanB:  We are getting into a weird battle of semantics here, but as you say it, you are wrong because you are misleading  Microbug in telling him that:
"A practically realizable constant voltage supply does not exist."

"A practically realizable constant current supply does not exist."

They do exist and are being sold and used everyday.  You are correct to say there is no such thing as a "constant voltage" or a "constant current" supply if you think such a thing can work perfect in any imaginable load condition.

Every constant voltage or constant current supply has its limitations set by the ratings of a supply parts, the maximum current that is available to be delivered safely and without overheating, there is always a limit to the maximum perfection of output regulation that can be achieved, etc.

But when you use the word "practical" here, it is misleading, "practical circuits" do exist, and can be quite tiny and work well.  Ideal or theoretical circuits often do not.

If you used the work "Ideal" or "theoretical" (instead of "practical", when you try to give the meaning that a perfect power supply, operating perfect in every possible load condition does not exist), then you could have explained the problem of "theoretical" versus "real world" practical better.
---------------------------------
IanB: in another reply you are incorrect and confusing when you say "In reality there is no difference between these two concepts." in trying to clarify if there is really any difference between constant current and constant voltage supplies. 

A constant current supply, within its range of voltage compliance, which is the limits of its constant current operation, delivers a constant current that is set by a control.

A constant current supply has many applications, for instance, it is a circuit device and tool that allows such allows smooth voltage ramp generation with a capacitor, it can also set the limit of current delivered by a power supply, and a constant current power supply is often required for battery charging methods that safely charge batteries.

A current limited supply may smoothly or abruptly switch to a current limited condition of operation, and the current limit may not be adjustable by the user, but just be set and limit the output flow of current to a safe value that would not damage the power power supply. Current limiting is a way of talking about how a power supply protects itself, rather than the circuit it is attached to. which may not be designed to be constant, just limit the current to a safe value to protect the supply or the circuit it is attached to.

An example where current limiting is used but a constant current is not delivered  is when "Foldback Current Limiting" is used. In this case the current is first limited to a higher value, then as the load is increased, the power supply voltage drops with a fast-sloped characteristic limiting curve to limit current at a lower level prevent damage to the circuit involved and to the power supply supplying the current. Any 555 timer, 7805  or LM317 regulator have current limiting, but the actual current is neither constant and varies with temperature  nor is the current limit user adjustable.
----------------------
Finally, I would point out the hype advice you find in some datasheets, they often try to get you to use as many of the devices shown as possible in a given application so as to maximize sales of the device. In many cases, the optimal circuit solution to a circuit application or problem shows only one the answer, "just use more and more of these excellent xxx devices to do it"
Not all application notes or datasheets do this, and studying them is probably one of the most important ways to learn about electronics, but remember this, the purpose of a datasheet is meant to be just about the part it is describing, it is not being intellectually honest helping you in finding better ways to solve your circuit problem.

I wish to thank everyone for their replies, and please excuse me for letting bold print get out of control somewhere.

I also am very grateful for your replies that give me food for thought (even while I am having lunch) and a giving me a chance to clarify things in electron land. Because I have only my lunch hour to reply, I might have created a poorly edited words-in-excess reply that attempts to say what I really wanted to say here.

[c4757p]

A constant voltage source comes first and sets the nominal, non-current limited output voltage, and the current limiting or constant current circuit stage then monitors the current being drawn out of the constant  voltage source and adj's voltage output down to achieve the constant current mode of operation. We can say that the constant current stage feeds back to control the constant voltage stage to make this happen.

If the stages are going to be separate, the constant current must come first, because in nearly every practical power supply the usual application is CV mode and that is what must regulate most closely. You can combine the two stages into one, as you have done, in which case the two are one. If you put them in series but have them sense the same output, they are still one stage with two pass elements; they're sensing off the same point. I can't think of a single practical reason to use a two-stage solution with the current sense last and sacrifice voltage regulation accuracy.

[/b] You might notice, in my circuit that the constant current sensing and control is  the in of the  constant voltage output stage, but that it feeds back to control the LM317 voltage output to achieve constant current operation.

Your circuit does not have two stages, it modifies the LM317 to sense both voltage and current. There are two control loops running in parallel controlling the same pass element.

i repeat, The constant voltage stage most be controlled by the constant current stage but part of its circuit senses the current in the constant voltage output state, and there is usually some way of feeding back or controlling  the constant voltage stage to correct the current to the desired set value.

Restricting the maximum voltage into it is one way of accomplishing this.

The bottom line is that the LM317 is a very poor choice to be somehow "jerry-rigged" to create a switching power supply.

"Jury rig". Sorry, pet peeve. Jerry's actually a very good engineer and never rigs anything.

The datasheet shows how it can be done, but that is misleading. The datasheet writer is just trying to sell you on the idea to use this part,  to sell you their part, so they present this schematic to try to get you to use their part in an awkward way.

Just because it has happened doesn't mean this is an example of it happening.

I think the idea of trying my circuit is better because it is simpler, it is a linear not a switching power supply, and their exists some range of linear constant current control.

Simpler?

Good transient response is an important characteristic of a power supply in that it allows the power supply to act well with sudden load changes.

Every power supply has certain specifications, you don't just shoot for "as much transient response as possible". If the OP is fine with the transient response of the LM317 circuit, it's a fallacy to say the circuit is garbage because of the transient response.

IanB:  We are getting into a weird battle of semantics here, but as you say it, you are wrong because you are misleading  Microbug in telling him that:
"A practically realizable constant voltage supply does not exist."

"A practically realizable constant current supply does not exist."

They do exist and are being sold and used everyday.  You are correct to say there is no such thing as a "constant voltage" or a "constant current" supply if you think such a thing can work perfect in any imaginable load condition.

...

But when you use the work "practical" here, it is misleading, "practical circuits" do exist, and can be quite tiny and work well.  Ideal or theoretical circuits often do not.

If you used the work "Ideal" or "theoretical"

"Ideal" was implied in not otherwise qualifying "constant voltage" and "constant current", the same way that a "car" is not assumed to be a "clown car" unless otherwise stated. "Practically realizable" does not mean "practical".

An example where current limiting is used but a constant current is not delivered  is when "Foldback Current Limiting" is used. In this case the current is first limited to a higher value, then as the load is increased, the power supply voltage drops with a fast-sloped characteristic limiting curve to limit current at a lower level prevent damage to the circuit involved and to the power supply supplying the current.

Otherwise unqualified, "current limit" implied "constant current". There are variations, like a foldback limit, but none was specified. As you have defined the two, they are the same. Read your own definitions again.

Finally, I would point out the hype advice you find in some datasheets, they often try to get you to use as many of the devices shown as possible in a given application so as to maximize sales of the device. In many cases, the optimal circuit solution to a circuit application or problem shows only one the answer, "just use more and more of these excellent xxx devices to do it"[/b]

Again, and as you point out yourself, just because it has happened doesn't mean it is happening here. The two-LM317 CC/CV supply does not work nearly as poorly as you think, and no, it doesn't oscillate. In fact, it works quite well for its purpose. Nobody here is trying to make a lab-grade power supply, it's just a quick-and-simple thing with intentionally low specifications. It still works well within those.