LM317 Circuit Challenge

[Roff]

Regarding your comment on Carl the curmudgeon's idea:

You might want to draw that out and compare it with the standard configuration, and then run the numbers. From your discription, if I am interpreting it correctly, the output could not come down to that range. R2's value needs to be pulled down to drop the voltage in case of a wiper fault. Its that darn common current again. R1 must go across the Adj and Out pins.

Keep in mind that only the wiper is connected to the ADJ pin, so if the wiper opens, ADJ is grounded through 10k (or whatever). Therefore the output voltage will fall to a low value, somewhat above Vref.
The problem I have with this method is that Vout is a very nonlinear function of pot rotation. Attached is a plot that I did on Excel, using Carl's example. An LTspice simulation gave a virtually identical curve.

 

[Mikebits]

I can't help but think that with a good high reliability potentiometer and an adjustable current limit that this would be a non issue. Sure the pot may cost more, but why skimp and a piece of lab gear?

 

[MrAl]

Hi again,


Well, those high quality pots start at about 10 dollars each. Seems like too much for just a simple power supply. That's probably more than half the cost of the whole project (using a home wall wart) so it seems like too much.

 

[Mikebits]

I would rather pay the insurance premium up front and have a feeling of confidence that my supply will not fail. It is like the old saying, you get what you pay for. Considering that the power distribution system is the life's blood of any project, and reliability is paramount, I would opt for the extra cost. But that is just me. Not trying to derail your thread so I will get off my preachy box and allow you to continue as you were.

 

[BrownOut]

How 'bout FET's configured as a variable resistor. If the wiper of a voltage divider fails, and the gate goes positive, that will make the channel resistance low, and effectively 'clamp' the input to a low value. Stack up a few devices to distribute vgs so they don't "pinch off" Just a thought, I havn't really thought it all the way through.

 

[mneary]

MrAl's circuit can be placed in parallel to the R2 pot. You set the LM431 to the 'safety' value, then set R2 to the precise operating value. If R2 fails, MrAl's circuit would prevent the voltage from rising above the LM431's set point.

I suggest this is a way to meet the objective as stated, but I personally wouldn't expect the R2 pot to be the most likely failure mode. The most likely failure in this circuit is that a human error would set the voltage too high. I suppose it is an improvement to make two errors necessary before damage occurs, but only if you require the settings be made by two different people.

 

[crutschow]

Regarding your comment on Carl the curmudgeon's idea: Keep in mind that only the wiper is connected to the ADJ pin, so if the wiper opens, ADJ is grounded through 10k (or whatever). Therefore the output voltage will fall to a low value, somewhat above Vref.
The problem I have with this method is that Vout is a very nonlinear function of pot rotation. Attached is a plot that I did on Excel, using Carl's example. An LTspice simulation gave a virtually identical curve.

Good observation. And I don't see any easy way to make the circuit output voltage more linear versus pot rotation.

Sometimes, simple doesn't work.

 

[MrAl]

Hi again,


Brownout:
The FET idea sounds interesting...we'd just have to find a way to make it temperature compensated so we dont loose that quality of the LM317 itself.

mneary:
Yes that's true, it can be used as a 'secondary' system as well as the primary which is what i had intended. And yes that would make the circuit require two adjustments instead of one so maybe less human error, but i tend to not worry about that too much. If someone doesnt adjust it right well i feel sorry for them
Also possible is something dropping off a shelf and hitting the control knob (yikes) and that could cause the voltage to shoot way up too, but i dont want to cover every possible human problem that could come up...just get the basic working.

More On the High Quality Pot:
I see that there are pots that are rated for 50,000 rotations for around 10 dollars USD. That would be a nice pot to use im sure, but i was hoping to keep with the garden variety. Also, just because it is a high quality pot it doesnt mean it's never going to fail. Some day it will fail and the voltage will shoot up. On the other hand, a circuit that is designed to react favorably to a 'bad' pot (arm disconnected) will never fail even after 1,000,000 rotations. This is what i call covering all the bases. A high quality pot would just put the icing on the cake


Sorry if i seemed a little critical on some of the ideas. What i am looking for i know is almost perfection and i know that is hard to get. I just feel that a little extra work now will save in problems later. I dont want to work on a circuit and accidentally blow it out, and i dont want others to have that problem either.

 

[crutschow]

Well, I don't want to argue semantics. If you don't wish to call the common current developed with Vref across R1 Iset or Iref, call it Inetwork or Idivider or what ever. It is what it is, a common current of the divider. As a common current, for a given output voltage that would necessitate a fixed ratio of the two resistors. For a 5V output the ratio would be 3, and for a 10V output the ratio would be 7, and for 15V it would be 11. If R1 were 250ohms the common current would be 5mA for all three cases. The resistance and voltage of R2 would change but the current would remain the same. It seems kinda natural to call it a reference or set current, but I guess that's just me.

I wasn't trying to argue semantics. I was just trying to point out that it was the voltage divider ratio of the two resistors, not the current that determines the output voltage. Thus you can vary either R1 or R2 to vary the output voltage. But certainly with a fixed value of R1, then the current through the divider is constant for any value of R2.

 

[Roff]

Good observation. And I don't see any easy way to make the circuit output voltage more linear versus pot rotation.

Sometimes, simple doesn't work.

Of course, you could make a feature out of the nonlinearity. You get better setting resolution at low voltages, where it's more critical.

 

[MRCecil]

I wasn't trying to argue semantics. I was just trying to point out that it was the voltage divider ratio of the two resistors, not the current that determines the output voltage. Thus you can vary either R1 or R2 to vary the output voltage. But certainly with a fixed value of R1, then the current through the divider is constant for any value of R2.

Carl,

Well, I have always proposed the idea that the chicken preceded the egg in some fashion; an acknowledgment to the physicists Ohm and Kirchhoff.

Vref begets Iref, and Iref determines the divider based on the physical laws determined by those 19th century minds and the designers of the device being discussed. I see your point clearly, but just do not feel it covers the broader implications of design criteria overall. A current analysis is necessary to consider the impacts of Iref, Iadj and other criteria if deviations from the recommended configuration and limits are made. Also, I find it much simpler in all cases since critical currents can be compared to those calculated.

As to adjusting either resistor in the divider, sure it can be done, but at the cost of predictability of the outcome and very limited range owing to changes in Iref under all load conditions to say nothing of % of regulation with varying loads.

I finally got some time today to get back to this and look up an LT device that closely matches the LM317. The only differences I noted, in a brief review, were in the necessary overhead before drop out (load dependent), and the minimum load requirement (slight). It's the LT1086. I would have preferred having the spice model for the LM317, but I cannot find it anywhere.

I ran multiple sims and the results matched very closely my numbers for the LM317 in standard configuration including Iadj. Adjusting R1 and varying Iref yielded very poor results as expected under zero/light static load conditions. I also ran sims for your suggestion of connecting only the wiper and a 10k to ground to the Adj pin. The results were of the same general sort as varying R1 as its value varies by the value of R2 north of the wiper and not being shorted by the wiper connection to the top of R2.

Vout diverges rapidly from the calculated output after R1 reaches the area of 500Ω, in normal configuration, with a ratio of 3 for Vout=5V, by stepping R1 & R2 and maintaining the ratio. With R1 at 700Ω and R2 at 2.1k, a ratio of 3, Vout=12V; proof that the ratio alone is not the controlling agent. This is due to Iref being less than the minimum load current as I suggested in an earlier post. Adding a 1k load resistor brought this regulation error down to about 1.6%, somewhat greater than the 0.1% advertised, and stayed at that error level with a 100Ω load. This lesser error can be mostly attributed to the influence of Iadj adding to the lowered Iref value through R2, which would normally swamp that impact of Iadj if Iref were in the correct range indicated by the minimum load current in the datasheet.

I could be wrong, as I just downloaded LTSpice about six weeks ago so you might want to check the results for yourself. I sure find it much simpler to use than PSpice, which I loath, in my Orcad package, except for the limited library of parts. If you decide to run a sim with the LT1086, could you let me know if your results differ?

Merv

 

[Roff]

LM317 for LTspice

MrCecil,

Here ya go.
Copy this to Notebook and save as LM317.sub in /lib/sub.

*LM317 TI voltage regulator - pin order: In, Adj, Out
*TI adjustable voltage regulator pkg:TO-3
.SUBCKT LM317 1 2 3  **Changes my be required on this line**
J1 1 3 4 JN
Q2 5 5 6 QPL .1
Q3 5 8 9 QNL  .2
Q4 8 5 7 QPL .1
Q5 81 8 3 QNL .2
Q6 3 81 10 QPL .2
Q7 12 81 13 QNL  .2
Q8 10 5 11 QPL  .2
Q9 14 12 10 QPL .2
Q10 16 5 17 QPL  .2
Q11 16 14 15 QNL .2
Q12 3 20 16 QPL .2
Q13 1 19 20 QNL .2
Q14 19 5 18 QPL .2
Q15 3 21 19 QPL .2
Q16 21 22 16 QPL .2
Q17 21 3 24 QNL   .2
Q18 22 22 16 QPL .2
Q19 22 3 241 QNL 2
Q20 3 25 16 QPL .2
Q21 25 26 3 QNL .2
Q22A 35 35 1 QPL 2
Q22B 16 35 1 QPL 2
Q23 35 16 30 QNL  2
Q24A 27 40 29 QNL .2
Q24B 27 40 28 QNL .2
Q25 1 31 41 QNL 5
Q26 1 41 32 QNL 50
D1 3 4 DZ
D2 33 1 DZ
D3 29 34 DZ
R1 1 6 310
R2 1 7 310
R3 1 11 190
R4 1 17 82
R5 1 18 5.6K
R6 4 8 100K
R7 8 81 130
R8 10 12 12.4K
R9 9 3 180
R10 13 3 4.1K
R11 14 3 5.8K
R12 15 3 72
R13 20 3 5.1K
R14 2 24 12K
R15 24 241 2.4K
R16 16 25 6.7K
R17 16 40 12K
R18 30 41 130
R19 16 31 370
R20 26 27 13K
R21 27 40 400
R22 3 41 160
R23 33 34 18K
R24 28 29 160
R25 28 32 3
R26 32 3 .1
C1 21 3 30PF
C2 21 2 30PF
C3 25 26 5PF
CBS1 5 3 2PF
CBS2 35 3 1PF
CBS3 22 3 1PF
.MODEL JN NJF(BETA=1E-4 VTO=-7)
.MODEL DZ D(BV=6.3)
.MODEL QNL NPN(EG=1.22 BF=80 RB=100 CCS=1.5PF TF=.3NS TR=6NS CJE=2PF
+ CJC=1PF VAF=100)
.MODEL QPL PNP(BF=40 RB=20 TF=.6NS TR=10NS CJE=1.5PF CJC=1PF VAF=50)
.ENDS LM317   **changes may be required on this line**

Copy this to Notebook and save as LM317.asy in /lib/sym/misc.

Version 4
SymbolType CELL
RECTANGLE Normal 64 48 -64 -32
TEXT -36 -22 Left 0 LM317
SYMATTR Value LM317
SYMATTR Prefix X
SYMATTR Value2 LM317
SYMATTR Description LM317 voltage regulator
SYMATTR ModelFile LM317.sub
PIN -64 0 LEFT 8
PINATTR PinName IN
PINATTR SpiceOrder 1
PIN 0 48 BOTTOM 8
PINATTR PinName ADJ
PINATTR SpiceOrder 2
PIN 64 0 RIGHT 8
PINATTR PinName OUT
PINATTR SpiceOrder 3

The next time you open LTspice, the LM317 should show up in the Misc directory when you open the Component window (the AND gate icon).

 

[BrianG]

Maybe I'm totally missing something here, but with the pot wired like a rheostat, the arm "lifting" will just make the pot look like a 2k ohm fixed resistor (referring to the opening post schematic). Generally, you select a pot and resistor that will give the max voltage you want at the max pot throw, so as long as the pot is selected that way, at worst, the voltage will be at that max value if the arm lifted.

 

[Roff]

Maybe I'm totally missing something here, but with the pot wired like a rheostat, the arm "lifting" will just make the pot look like a 2k ohm fixed resistor (referring to the opening post schematic). Generally, you select a pot and resistor that will give the max voltage you want at the max pot throw, so as long as the pot is selected that way, at worst, the voltage will be at that max value if the arm lifted.

This is a lab supply. If you have it set to power some 3.3V logic, and the wiper opens, you wind up with 24V across the circuit. Whatever happens next is probably bad.

 

[MRCecil]

Roff,

Thank you very much, Sir! I dislike uncertainty, and you have helped resolved some for me.

A Tip O' the Tam,

Merv

 

[MRCecil]

Roff,

I got the model in place and tested. The reaction is the same as with the LT1086, albeit starting at a different point owing to the typical lower minimum load current than the LT1086 I trust. Thanks again for supplying the files.

Merv

 

[Roff]

Roff,

I got the model in place and tested. The reaction is the same as with the LT1086, albeit starting at a different point owing to the typical lower minimum load current than the LT1086 I trust. Thanks again for supplying the files.

Merv

You're welcome.

 

[MRCecil]

Another Suggestion

MrAl,

I thought I should contribute something more than critique. Got started this morning, and here is what I came up with. It's just a shunt and appears to behave/operate without issues so far as I have tested. Some details are in the attached schematic.

Merv

 

[Roff]

MrAl,

I thought I should contribute something more than critique. Got started this morning, and here is what I came up with. It's just a shunt and appears to behave/operate without issues so far as I have tested. Some details are in the attached schematic.

Merv

Excellent, Merv! You are a genius!
I ran some transient sims. According to my sims, the output will overshoot and/or ring during wiper recovery if you don't have around 10nF or more from ADJ to GND.

 

[MRCecil]

Roff,

I did not run any transient sims owing to the fact that I was ignorant of how to switch the wiper. After reading your response, I did some searching and finally found the voltage control switch, and figured out how to manipulate it after a fashion, being very motivated to do so when seeing your nick in the Spice directive.

Yes indeed, there was potentially destructive overshoot at the 5V level I tested at with an 8Ω load or unloaded. Following your advice, I experimented and found that a 220nf integrated the leading edge of the rise nicely and helped reduce the overshoot on the lead of the falling edge to a reasonable level also. The rise at 28V takes about 4ms to rise back the programed voltage, which I think is reasonable at that level verses too much overshoot at 3.3 or 5V, which could be destructive.

Thank you for the compliment, and thank you for pointing out the shortcoming in my proposed configuration.

Merv