A Flaw found in LM317 circuit, available in internet.

[ahsan_i_h]

hello everybody,

A flaw in LM317 circuit is found which is generally availabe in internet. I need varification by everyone.
There is a resistor between LM317's output and adjust. It has a maximum value. It must not go higher than 240 ohm for LM117 and must not go higher than 120 ohm for LM317. In general internet circuits, 240 ohms is shown as the maximum limit for LM317 circuits. That is the problem.

Lets calculate to prove, The minimum load current of LM117 is 5mA(datasheet), so the maximum value of that resister is 1.2v/5mA=240 ohms. But the minimum load current of LM317 is 10mA(datasheet), so the maximum value of that resistor should be 1.2v/10mA=120 ohms.

In national semiconductor datasheet all the circuits are drawn with LM117. I think, that diagrams misleaded the general internet circuits and made thinking LM317 as LM117.

About the flaw am I right? or may be I am thinking wrong or missing something?

 

[RetroTechie]

About the flaw am I right? or may be I am thinking wrong or missing something?

There is no flaw. Quoting from an LM117/LM317 datasheet (National Semiconductor):

Since the 100μA current from the adjustment terminal represents an error term, the LM117 was designed to minimize Iadj and make it very constant with line and load changes. To do this, all quiescent operating current is returned to the output establishing a minimum load current requirement. If there is insufficient load on the output, the output will rise.

That resistor (with the one from Adj pin to ground) are nominal values, picked to satisfy 2 conditions:

1. To provide reasonably good output accuracy: current through resistor divider much larger than Iadj (mA's versus 50-100 µA)
2. To make sure regulator operates normal without further load(s) on the output: current through resistor divider in the neighborhood of minimum load current

Both goals conflict with other goals like wanting as much as possible of the current as useful output current, so the values chosen are a compromise that works well for most applications. An example of what order of magnitude to put there.

Nowhere (that I know of) does it say that resistor divider is guaranteed to form a minimum load that the regulator needs. Just something that's close enough. Nor that this resistor divider should provide a minimum load - the circuit powered by the LM317 could also do that. Nor that this resistor divider should have at least X times Iadj current through it - just an arbitrary current that's large enough not to affect output accuracy too much.

That's all there is to it... not a flaw, just insufficient understanding of what that resistor divider does, and how to choose appropriate values.

 

[tpetar]

Use and read datasheet which provide manufacuter.

 

[FvM]

I agree, that the concordance of numbers suggests the idea, that the reference divider with R1 = 240 ohm is intended to work also as minimum load. As said, it does for LM117 worst case parameters. But in fact, the assumption isn't founded by the datasheet literature.

 

[ahsan_i_h]

@RetroTechie
Yes that resistor devider is guaranteed to form a minimum load. You know why? Because everyone makes 317 circuit with that two resistor and try to measure output voltage at no load. All ic manufacturer know that. Thats why when they draw a typical application circuit they draw it without load. Like in National semiconductor's LM317 datasheet. At first page shows a LM117's typical circuit without load. When we watch that circuit we believe that we will get right voltage without load. As we everyone believe that, so National semiconductor is bound to give guaranty that its resistor divider will act as a minimum load. Thats why they are showing 240 ohms in LM117 circuit.
Now If we see any LM317 circuit without load, we are sure we will get right voltage without load. So it is circuit drawer's responsibility is to give guaranty that his resistor divider will work as minimum load. So 240 ohms in LM317 circuit can not be acceptable when that circuit is showing no load.
Many internet's LM317 circuit shows 240 ohms but do not show any extra load.

---------- Post added at 22:41 ---------- Previous post was at 22:16 ----------

We do not need to think that resistor's(between output and adjust pin) value more than 240 ohms. So we do not need to consider Iadj. We will neglect Iadj for this issue.

 

[alexan_e]

I don't see the problem.
The schematics you see are suggestions to design the psu and supply a circuit.
Just because you don't see a load doesn't mean that the circuit is wrong or that the people who make it will not use a load.
Would you consider a schematic of an amplifier wrong because it doesn't have a speaker connected to it's output in the schematic?

 

[ahsan_i_h]

@alexan_e
I am talking about only a power supply circuit made by LM317. Not any other circuit related with it.
If an amplifier shows higher voltage at no load and shows lower voltage at load there is no problem. But if a power supply shows higher voltage at no load and shows lower voltage at load there is two problems.
a) How can i use such power supply at very small load when load will vary. a good linear regulator should not do that. The famous LM317 is more than that.
b) After making a power supply, we measure its voltage at no load then we connect it to the circuit. Why should I connect a LM317 regulator showing 4.0v(at no load) to a 3.3v ic? How can i say that the LM317 or psu circuit is not defective? 4.0v may not get dropped by loading.

---------- Post added at 23:44 ---------- Previous post was at 23:33 ----------

Why they show 240 ohms? why not 220 ohms or 270 ohms. Because LM117's optimum resistance(minimum loss but serving minimum load current) is that 240 ohms. They are stuck in LM117.

---------- Post added 23-05-12 at 01:20 ---------- Previous post was 22-05-12 at 23:44 ----------

To,
RetroTechie, tpetar, FvM, alexan_e

For example someone came to you and ask for a variable voltage power supply circuit using LM317 . He will use this power supply circuit to test different circuits he made. Now what R1(the resistor between output and Adj) you will suggest? most probably 220 ohms. Right? Now he will make that and he may again come to you and will say that he do not get 1.2 volt as minimum at no load. Why? LM317's worst case minimum load current is 10mA. You know 5.45mA(1.2/220) is going through the 220 ohms. So next what you will suggest? You will suggest to connect a dummy load of 220 ohms at output. Thats fine. What will happen to that dummy load when he will rise the output voltage to 25v? So for him a suggestion for dummy load can not be given, Instead we have to tell him to connect a 120 ohms resistor as R1 in place of 220 ohms.

 

[RetroTechie]

For example someone came to you and ask for a variable voltage power supply circuit using LM317 . He will use this power supply circuit to test different circuits he made. Now what R1(the resistor between output and Adj) you will suggest? most probably 220 ohms. Right? Now he will make that and he may again come to you and will say that he do not get 1.2 volt as minimum at no load.

Nope. Why? Because current through 220 Ω will be about 5.7 mA (1.25/220). That 5.7 mA >> 0.1 mA, so 1st condition is satisfied.

LM317 datasheet also says that minimum required load is typically 3.5 mA, so most LM317's will be happy. Btw that's for Vin-Vout = 40V, how is that for Vin-Vout = 5V? 10V? Now what if you have one that needs a little more? Datasheet says "If there is insufficient load on the output, the output will rise". How much? Probably not much, since 5.7 mA is already a good deal in the direction of 10 mA. So maybe it will show Vref of 1.27V in that situation. Is that a problem? Again: nope. Because your friend will have a variable resistor as 2nd one (from Adj pin to ground), and turn that till output reads (for example) 5.0V. With slightly elevated Vref, it will just be set slightly different to reach that 5.0V, than if Vref were 1.25V. Now attach some load, and say that Vref will drop to the nominal 1.25V. Then measured output might sink from 5.0 to 4.92V (5.0 / 1.27 x 1.25). Is that a problem? Again: nope, since 4.92 is fine when powering 5V circuitry.

Don't take these numbers to do the math. Like said before: that 240 Ω is just a ballpark figure. One that happens to work fine for most applications. Don't like it? Use 150, or 100 Ω. Or use 470 Ω, and measure what the LM317 that you have, does with that. Or pick 270 Ω when you want 1.25 to 8V output voltage, and a 1.5 kΩ is the only variable resistor that happens to be in your box with parts...

Besides: does it even matter what output voltage you get when NO load is attached? Nope, what matters is whether voltage is within acceptable range when the circuit is connected to its power supply. When it isn't, there's no voltage on the circuit & thus nothing will break. So you have a circuit that only uses 0.5mA? Measure your LM317 circuit with that load.

Either way I don't see a problem here. Nor a shortcoming in the datasheet info.

 

[ahsan_i_h]

You have to consider the worst case maximum limit of minimum load current and that is 10mA. You will never consider 3.5mA. Why? because National Samiconductor considered the wrost case maximum limit of minimum load current 5mA for LM117 when he put the value 240 ohm beside R1. Is there any guaranty that any one will get 3.5mA minimum load current. It can be guatanteed that no one will get more than 10mA minimum load current device.
Now you asked how much voltage will rise for insufficient load current.You guessed probably not too much. I guess probably very much. For professinal design no one should guess like that. We should act in such a way that we need not to guess. Why I guessed probably very much? Then,
Lets discuss about how LM317 works to understand minimum load current. The Lm317 has circuits inside. How that circuit gets power? For 7805 it is easy. 7805's internal circuit gets power from Vin and ground pin. And the supply current of that circuit certainly comes out from the ground pin. What about LM317? It is floating. Very small Iadj is not its supply current. It is something like opamp's input bias current. Now how its circuit gets power supply voltage? Where does go the power supply current of that circuit? That circuit gets supply voltage from dropout voltage and that circuit's supply current is the minimum load current of LM317. So below minimum load current its circuit will not function properly and will cause the output voltage to rise. I guess very much.

---------- Post added at 03:58 ---------- Previous post was at 03:45 ----------

To stop guessing we must take worst case minimum load current through the divider.

 

[FvM]

In a short: Professional designers need to read and understand datasheets. Usually they do.

 

[keith1200rs]

The resistors to set the output voltage are not there to provide a minimum load - you would normally have a circuit connected to a regulator which will provide that.

Keith

 

[ahsan_i_h]

Thankyou keith for your reply,

When I started to make a variable voltage power supply with LM317, then I faced a problem. As the load is unknown, so it can be a small may be far lower than minimum load current. Then I came to a conclusion that minimum load current should go through the resistor divider. Because if I take smaller current through the resistor divider,then I have to connect a dummy load to the output to maintain the minimum load current requirement. That dummy load should be able to fullfill the minimum load current requirement down to 1.2v. As a result its value will be low like 100 or 200 or 300 ohms etc. Now when i will rise the output to higher voltage like 30v, then that dummy load will get burned or unnecessarily a high power will be lost in the dummy load.

What you said is right. certainly there are good advantages if I take the minimum load current through the load,not through the divider. Because why should someone waste a power unnecessarily in the divider network which may be useful to load? And also why should someone use a bulky potentiometer(due to higher power loss in divider network) to adjust the output voltage?

But for an unknown load situation like this, should not we take the minimum load current through the divider network?

Now for any isolated power supply circuit showing 1.2-37v/1.5A ,(just showing a power supply circuit and nothing else even no note is there which tells about minimum load current requirements) does not it mean that it can power properly any unknown load in the range 1.2-37v/1.5A ?

Thats why I want to say that any isolated(means only power supply circuit and nothing else) power supply circuit of LM317, which do not take the minimum load current through its divider and which has no note about minimum load current requirement has a flaw. Such type many circuits are availabe in internet.

Ahsan

 

[keith1200rs]

You seem to be expending a lot of effort and words on something which is quite trivial. A regulator circuit shown without a load which doesn't meet the minimum load requirements of the regulator is "flawed"? If you wish to view it that way then that is your prerogative but it is up to the user of any circuit or device to ensure that it meets their requirements. Normally that would be provided by your load but if not it is your responsibility to ensure the design meets your requirements. I have certainly seen laboratory power supplies with a dummy load inside across the power terminals.

Keith

 

[ahsan_i_h]

Thankyou keith for your reply,

For any regulated power supply circuit(short circuit current limited), if the maximum current limit is not mentioned, is it not a fault of publisher or designer? because user may get unregulated voltage at any time because user dont know the current's upper limit.
In this way, For any regulated power supply circuit, if the minimum current limit is not mentioned, is it not a fault of publisher or designer? because user may get unregulated voltage at any time because user dont know the current's lower limit.

Ahsan

 

[keith1200rs]

If you don't read the datasheet then it is the fault of the designer. If you take a circuit from the internet and use it without validating it then you do so at your own risk. There are a lot of very bad circuits on the internet. Just because someone has posted a circuit on the internet doesn't mean they know what they are doing. A recent thread involved someone who was trying to drive a laser using a circuit they found on the internet. The 'design' was totally wrong and would never work reliably.

Keith

 

[Pjdd]

To sum up: The datasheet of the LM317 gives all the information needed to produce a good stable design. It is the responsibility of the user, not the manufacturer, to read, interpret and apply the information correctly. The manufacturer cannot be held responsible for some half-baked schematic published by third parties.

 

[FvM]

Besides all general consideration about quality of circuits from the internet and designers responsibility for reading datasheets exactly, there's a specific point with the LM317 regulator cicruit.

As you mentioned in your initial post, the said "R1=240 ohm" guarantees correct operation with LM117 "by design" but may fail with LM317 worst case specifications.

According to good engineering practice, you need to assure the minimum load for the actual application. Setting R1=120 ohm would be one of several ways to achieve it.

But besides follwing strict design methods, you can ask: How many exemplars of LM317 will exceed the LM117 (just a different production lot of the same chip) specification of 5 mA minimal current? And if they do, above which input voltage?

From a manufacturers viewpoint, worst case specifications aren't but a cushion to increase production yield. You may get parts with near worst case specification on a bad day. If you read data sheet fine print, you'll notice that not all specifications are necessarily production tested. This brings up a risk, that certain specifications are exceeded without notice.

Besides claiming good design practice, what do you guess? How many "240 ohm" LM317 regulator circuits will fail at no load?

On the other hand, there's a high likelihood that even profound professional designs don't take account for all possible combinations of worst case component specifications, in other words fail under certain conditions. Some of these hidden design faults are probably discovered one day during production tests, others are only triggered by a combination of conditions at the customers site.