hi Papa,
Any chance of having a look over your asc file.
E

 

I still don't think we have the correct circuit. the pot will now adjust the output voltage, but the output looks horrible!!
View attachment 269364
200 mV of output ripple is not a supply I would use. I still can't figure out how to get 12V of output.

The TL431 doesn't need much load capacitance to make it oscillate!

 

Try moving the capacitor from across the TL431 to from output to input. That will slow it down.

 

hi Papa,
Any chance of having a look over your asc file.
E

Sure -- no problem

 

Try moving the capacitor from across the TL531 to from output to input. That will slow it down.

The purpose of this exercise is understanding the circuit that the OP thinks he has. I have little or no interest in speculative designs for an unknown purpose.

 

I still don't think we have the correct circuit. the pot will now adjust the output voltage, but the output looks horrible!!
View attachment 269364
200 mV of output ripple is not a supply I would use. I still can't figure out how to get 12V of output.

I think people should review the datasheet, and the TP needs to check their idea of the schematic against the suggested versions in the datasheet- it could be, they are using one of the typical applications shown in the datasheet.

 

hi Papa,
The asc file you posted was not the same as your post #20.
So I edited your asc file to suit the posted version.
Used my LM317 and Pot models and a BD244A [don't have 2N5884]
This is what LTS shows.
E

 

hi Papa,
The asc file you posted was not the same as your post #20.
So I edited your asc file to suit the posted version.
Used my LM317 and Pot models and a BD244A [don't have 2N5884]
This is what LTS shows.
E
View attachment 269405

Well, that is progress. At least we are able to duplicate the TS observations. In my previous post, I added the input capacitor, C3, to see if it made a difference -- it did not. I tried another LM317 model in the circuit of post#20, and it produces something closer to the results of @ericgibbs. So it looks like not all LM317 models are created equally. I don't have the BD244A, but the transistor model does not appear to be the critical factor.

 

hi Papa,
I know it's a modification, but move C1, 100n to the Pot wiper connection to 0v.
E

 

hi Papa,
I know it's a modification, but move C1, 100n to the Pot wiper connection to 0v.
E

2 Volts P-P ripple on the 11.7 volt output. Was that supposed to be an improvement?

 

hi,
This is what I see.
E

 

Yikes! I have to say that my search for a reliable LM317 model might have to continue. I'll let you know what I find.
EDIT: I have 5 different models and all of them behave badly. I did find a note in the library for the 5 symbols that the author "appropriated" the symbol for the LT1084 and renamed it for the five versions of the LT317 that he had. I was taken in by the notion that LT had some hand in creating the impression that the models were "reliable" in some sense. Bad on me for being so trusting.

 

hi Papa,
Is this any help.?
E

 

hi Papa,
Is this any help.?
E

Thanks -- I'm sure it will be.
Edit: Except I was unable to duplicate your results with any of my TL431 models.

 

Interesting...

I will get the exact make/model of the 317 in the circuit as I would bet there are internal differences that might matter..

Im sure that is true for the TL431AC too.

So remember the actual behavior measured is a bit different then what this looks like so far. It has a pronounced 100mV swatooth oscillation with no load and around 50-100mA of load promptly snaps to 3mV pk-pk of a wideband zener like noise. This stable, but noisy, performance *seems* fairly stable.

I will poke more at it soon... I have been hit with boring work things but will get back to this.

I still dont understand those high values on the adj side of the TL431. Why would this circuit need gain on the TL431 ? If thats gain that is.

 

Interesting...

I will get the exact make/model of the 317 in the circuit as I would bet there are internal differences that might matter..

Im sure that is true for the TL431AC too.

So remember the actual behavior measured is a bit different then what this looks like so far. It has a pronounced 100mV swatooth oscillation with no load and around 50-100mA of load promptly snaps to 3mV pk-pk of a wideband zener like noise. This stable, but noisy, performance *seems* fairly stable.

I will poke more at it soon... I have been hit with boring work things but will get back to this.

I still dont understand those high values on the adj side of the TL431. Why would this circuit need gain on the TL431 ? If thats gain that is.

Here is a functional block diagram of the TL431. As you can see there is an amplifier with a 2.5V reference and no feedback to limit the gain, but when the REF pin is above the reference voltage the transistor turns on and provides a low impedance path from cathode to anode. that is why it is referred to as a shunt regulator..

 

Initially It makes more sense to me that TL431 is part current control in conjunction with a pnp transistor. It is a convenient way to use transistor to both control current and dissipate heat at low cost. maybe LM317+ TL431+ PNP T0-220/heatsink could equal a simple power supply controller.
Breaking down the circuit for the TL431 current source.
(79) TL431 Constant Current Source | All About Circuits

LM317 can provide significant current at a regulated voltage, in addition the TL431 can be used for current control.

I also think the LM317 precedes the TL431.

 

I'm not sure what to think about this situation. I can't explain the variation in the simulation results, and I can't explain the empirical observation that an implementation works. Vadj and Vin beating up and down as a function of the value of C3 is profoundly curious and suggests a circuit with manifest flaws.

 

A new piece of information has come to light. This is from the Fairchild datasheet for the 78XX and indicates that the β of the transistor and the value of the resistor are linked for using a PNP pass element with a three terminal regulator.



If this is the case, then I have to say that IMHO, a good design does not depend on the β of the transistor because a buyer of that part would typically have no control over it unless he want to screen the part or pay the supplier to do it for them.

 

That was the ticket. 10 Ω was way too big for the 2N5884. The calculated β in the simulation is about 286

What is happening is that when the current demand exceeds the ability of the LM317 and the output voltage begins to drop, Q1 turns on briefly to supply additional current which raises the output voltage and turns Q1 off. Wash, rinse, repeat. For this level of ripple there does not seem to be much point in avoiding the use of a switching regulator for its superior efficiency.