LM317 variable power supply question.

[Doc38343]

I have a question and I know what I'm doing some will not like but it is all about the learning and is experimental project.

I have taken an ATX power supply 24pin sort and converted it to a bench supply. The steady 12v, 5v, and 3.5v are all stable and strong. I haven't loaded much current thru them but I can run normal simple things up to about 1.8 amps with no troubles. Voltages hold steady and solid every thing stays cool. OK so I wanted to learn how to make a segment of it variable. I also decided to use parts on hand so i found this LM317T and after reading a ton about it I say OK I can go low current it is only rated for 1.5 amps anyway. So I see my -12v supply rating and on the case it says 0.8 amps for -12 lead. I decided to use the -12 and +12 to get 24V @ .8 amps.... I set it all up in my breadboard even put a heatsink on the LM317 and connected a resistor and a LED for output. I used the standard setup for variable from speck sheet and also a schematic i saw online here somewhere. 5k vR and 240R as well as the caps and even did the protection diodes from the schematic. OK so i turn it on and great my voltmeter says 23.11 V at the supply input and 8.9 on the output. My LED was lit but as soon as I adjusted the pot just a little it tripped the overload. So thinking I may need to limit the LM317's current draw from the supply at the input I did some math and figured 23 volts at .8 amps (my supply limit) I got 28.xxx olms and about 16 watts! I found a couple of 5 watt 61R and put them in parallel and all was ok for a few but then when i dialed it down I got the overload trip again. I then remembered that i had setup the LED resistor for only 9V so I did the math and put a .5 watt 1k resistor there and fired it up again. Still tripped but had more range this time by a long shot. so then I took out one of the 61R power resistors to limit the current down to appx .5 amps on the LM317 input. Works full range but then as soon as I stick the probes the see what my amperage is, it trips again. I tested this a few more times and it trips as soon as any more load other than that LED is connected. Seems awful weak to me, but the LM317T is salvage so I don't know. My question is this. Is there a way to get more power out of the 24V by adding some opamps or some of the other things I've been reading about. I was going to work on current limiting as well but now I don't have enough to limit!!! LOL
Just kicken around to learn from all this and i do intend on building myself a nice bench supply after I gather all my parts but I want to understand how and why as opposed to just following a map... Thanks for any input   

Doc

[fubar.gr]

Works full range but then as soon as I stick the probes the see what my amperage is, it trips again.

I suspect that you are connecting the probes in parallel to the circuit. If you do that with the multimeter set up for measuring current, you are essentially shorting the output and that's why it trips.

You have to connect your multimeter in such a way that all current flows through the meter itself. For example, disconnect the wire that goes to one leg of the LED, then connect one multimeter probe to the wire and the other to the led. The multimeter then completes the circuit and all current passes through it.

[rdl]

You don't need any current limiting resistors for an LM317. The circuit attached to the output of the LM317 (the load) is what determines how much current is used. Something else is not right with your set up.

[Ian.M]

There is also the issue that operating it between the +12V and -12V will cause you major problems if you need to connect the circuit it is powering to to any external devices or non-battery powered test equipment. 

The PC PSU has its 0V rail grounded and *MUST* *NOT* be modified to remove this ground for safety reasons.  Many items of test equipment have one of their input or output terminals grounded (the shield on coax connectors).   If you connect the test equipment in the normal way to your circuit, you will short out the PSU -12V rail, certainly tripping its protection and shutting it down, but maybe also damaging your circuit.

[macboy]

The PC PSU has its 0V rail grounded and *MUST* *NOT* be modified to remove this ground for safety reasons.  ...

Why not? It's a flyback SMPS with transformer isolation between the primary and secondary side. Do you trust this transformer less than any other? Is there some other good reason you would like to share?

[macboy]

Works full range but then as soon as I stick the probes the see what my amperage is, it trips again.

I suspect that you are connecting the probes in parallel to the circuit. If you do that with the multimeter set up for measuring current, you are essentially shorting the output and that's why it trips.

You have to connect your multimeter in such a way that all current flows through the meter itself. For example, disconnect the wire that goes to one leg of the LED, then connect one multimeter probe to the wire and the other to the led. The multimeter then completes the circuit and all current passes through it.

This also struck me as a possible cause.
I would also ensure that there is a good size (at least 5 Watt, 10 W even better) dummy load on the +12V output. PC supplies do not like to operate with too little load.

[Ian.M]

@macboy,
If I had the PC PSU PCB in front of me so I could check clearances and see how to retain an earth return for its EMI filtering, I would be rather less pedantic about not modifying its earthing, but assuming its the usual far east cheap and nasty sort, it would be extremely irresponsible of me to encourage a relative novice to modify it.

[Doc38343]

I suspect that you are connecting the probes in parallel to the circuit. If you do that with the multimeter set up for measuring current, you are essentially shorting the output and that's why it trips.

You have to connect your multimeter in such a way that all current flows through the meter itself. For example, disconnect the wire that goes to one leg of the LED, then connect one multimeter probe to the wire and the other to the led. The multimeter then completes the circuit and all current passes through it.

I did it wrong first but i did think of this and put it in series also. I think it pushes the draw just above the limit. What I came up with is to increase the resistor on the LED since when they overload they basically are a short.. just didn't stay on long enough to burn it out ....yet 

This also struck me as a possible cause.
I would also ensure that there is a good size (at least 5 Watt, 10 W even better) dummy load on the +12V output. PC supplies do not like to operate with too little load.

Yup have a 10W 10ohm resistor already in the 5V as recommended for this model and all the full power stable rails are solid at least up to 4 amps. Haven't pressed it beyond that and don't think I will need to. This is just for learning on a personal level for actual issues I will not be using this supply at this time.

The PC PSU has its 0V rail grounded and *MUST* *NOT* be modified to remove this ground for safety reasons.  ...

Why not? It's a flyback SMPS with transformer isolation between the primary and secondary side. Do you trust this transformer less than any other? Is there some other good reason you would like to share?

This supply has a board that is completely separated. The ONLY thing that gives it a earth ground is the two screws that bolt it down. So I insulated those two holes and waala the 0v is now completely isolated from earth. I have tested it a bunch and all is well concerning this. I can touch the wires on the 12v, 5v, 3.5v and no shock... LOL just kidding I tested them with the meter. I needed this isolation so i will not burn out my new oscilloscope. I want to do some experiments on the variable supply stuff as well as some 555 timer oscillation and a few other frequency, sine and square wave tests. BTW: I took the circuit board out and examined the traces the transformer and all very carefully as well as google... some models will not work with out earth reference tied to 0v but this one will just fine. It is an older Viotek 550W it also still has the -5V lead I think because it was a universal style replacment and not an OEM model.

You don't need any current limiting resistors for an LM317. The circuit attached to the output of the LM317 (the load) is what determines how much current is used. Something else is not right with your set up.

Thanks I think you hit the nail... I had the incorrect resistor set up for the LED. I used this setup first with a 6V battery supply and was doing the variable so when i calculated the resistor for the LED it was based on 6v and not the 22.4 I get with the  power supply. I changed it with a 1K resistor and it stays on now but I think I need to increase it more so i can check the current with my meter. Later tonight possibly.

Thanks to all of you!! 

[Ian.M]

Disconnecting the earth from the primary side of the PSU is unwise as it probably has a filter capacitor network that will put a significant mains frequency leakage current onto your 0V rail.

[Doc38343]

Disconnecting the earth from the primary side of the PSU is unwise as it probably has a filter capacitor network that will put a significant mains frequency leakage current onto your 0V rail.

Sorry you misunderstood me. I did not disconnect the mains ground from the case. the case is fully grounded. It is the low voltage AC and rectified DC side of the board that is isolated from the ground. There is a very clear separation of this board directly underneath the larger step down transformer.
I fully agree with you concerning disconnecting the earth ground from the primary side of the case as well as the circuit.

FYI: I was a lineman for 15 years and have mega respect for electricity. I have been hit many times in my life and did not enjoy any of them. I saw a man die in a flat 30 second incident while they were trenching for cable TV lines. He was simply leaning against the machine. My entire life I have worked in Cable TV, Installations of CATV,phone and security (low voltage stuff) and was a Class 2 engineer at Marriot Hotels (maintenance). At the hotel job I got introduced to super high Voltage 660V and the break down room where we bussed it out into the various feeds. The amperage was so high they had the switch in a room of its own enclosed in glass. When we would throw the switch to test the backup generator and arc would follow it for about 4 or 5 feet.. It was neat and scary as heck at the same time if you were in there you could definitely feel the EMF pulse in the air or it seemed like it to me.(hair would tingly). New to electornics but not to electricity, and not entirely new to electronics but i never bothered to learn the design whys I always just followed someone elses map or schematic. Just saying all this for some clarification.
Thanks for the concern also many folks just don't care.

Doc

[Ian.M]

Great. That sounds OK.

You obviously know what you are doing with AC and safety stuff, its only the fiddly electronics bits you will need to improve your understanding of.

I would however suggest making the supply into the LM317 regulator circuit switchable between 0/12V and -12/12V as the extra input voltage will double the dissipation, seriously limiting the maximum continuous output current available at low output voltages.

[macboy]

Disconnecting the earth from the primary side of the PSU is unwise as it probably has a filter capacitor network that will put a significant mains frequency leakage current onto your 0V rail.

Yes that is definitely a bad idea. The OP isolated only the secondary side from earth ground by interrupting the connection between the still-grounded case/screws and the black output wires. This retains the input EMI filter as well as the safety earth on the metal case, but floats the transformer-isolated secondary.

[Doc38343]

I would however suggest making the supply into the LM317 regulator circuit switchable between 0/12V and -12/12V as the extra input voltage will double the dissipation, seriously limiting the maximum continuous output current available at low output voltages.

Great idea.. i did not think of that for some reason but it makes perfect sense. If I understood earlier the current draw is controlled by the output drain like for example: an arduino will only pull the current it needs UNLESS something gets shorted. So at 1.25 up to 12 volts I can have lets say a little more current,.. then if I do need to analyse a circuit at above 12v I could flip the switch and have the limited less than .8A current which is still enough to activate lets say a mosfet to see if it is switching (or such as). Good idea I think... i'm gonna do it. One thought on that is the LM317 is still max rated at only 1.5 amps so::
 1. how can i protect my regulator. My full on 12v rail is capable of up to 18amps according to the chart on the supply. I know that would melt down a lot of things and burn out my single 12v wire. I can add 2 more yellows to the positive output but all the rest are tied into the stable 12v not adjustable terminal. Still a risk i would say. would adding a simple fuse be best or is there another method... like a diode or such to handle short over current jumps like if a cap is charging on the output circuit.
 2. I may be better served in this case by a new LM350T that can supply a bit more current at the lower voltage setting. (I just got in 10 of them i had ordered a while back)

#macboy you got it right. only the SECONDARY side black wires are not connected to the case ground. The PRIMARY side AND earth grounding to it and the case, is business as usual.
Thanks again guys!
A day without learning is like a day without joy... it just isn't a good day!!!
Today was a great day thanks to you all.   

[Doc38343]

the extra input voltage will double the dissipation, seriously limiting the maximum continuous output current available at low output voltages.

It just hit me. Your saying a lot of my available 0.8A current is being used up as heat dissipation on the LM317? You know a little bulb just came on over my head because i noticed today that the overload protect tripped when i turned the voltage very low. Was the opposite of my thinking!! So some current is also lost to the POT at lower voltages right? I didn't get it until just now. Today when I had it set on max 22.4v and got interrupted for a few, It never tripped out. So more resistance = more dissipation..

[Ian.M]

Not quite.  The LM317 has to drop all the excess voltage while passing the load current (+ the few mA the adjustment pot takes).  Its dissipation is the output current multiplied by the voltage drop between its input and output.  If the heatsink isn't large enough, it will overheat and go into thermal shutdown.   Flipping the switch I suggested gives you more current capability for output voltages under 10V.

If you are concerned about over-current, you can use an extra LM317 with a low value resistor between  Out and Adj as a current limiter.  Connect its Adj terminal to the adjustable voltage regulator's input. (see Figure 43. 1A Current Regulator in its datasheet: http://www.ti.com/lit/ds/symlink/lm117.pdf

The current limiter LM317 will prevent there ever being more than 1.2V across the resistor and you can switch in different resistors for different current limits.   The voltage regulator LM317 works normally up to the current limit, at which point its input voltage drops to keep the current at the limit.  It does decrease the available headroom by about 3V, so would decrease your maximum output voltage to about 19V at full load.

[tautech]

2. I may be better served in this case by a new LM350T that can supply a bit more current at the lower voltage setting. (I just got in 10 of them i had ordered a while back)

Or LM338 that is capable of 5 Amps, 7 peak, but you still have to manage the dissipation.

[Doc38343]

#ian.m  Thanks I am actually learning a lot right here as far as actually understanding so thanks! I have seen quite a bit online about doing both current and voltage regulation and current regulation was next on my list to explore... this seems like the perfect opportunity. I just got a pack of 10 LM350T so can i use them for same? I have literally about a ton of heat sink material have been a computer shop for 20yr now and i'm one of those freaks who just can't make myself throw anything even remotely useful away. though i may sell the aluminium one day perhaps..... (I only have the one salvage LM317T on hand. have not ordered any of those) AFTER I understand this process completely i am intending on exploring the opamp option like i saw Dave do on his power supply design video. I am waiting for a bit before i build out anything i will use long term.

#tautech Yeah i almost ordered 10 of all three but my misses fussed a bit..... She says i'm just like a kid!LOL
Do you think the LM350T's will work ok? I was also thinken I could just move the larger 12v bundle (5 leads combined) and the matching ground bundle to my temporary input terminal to cover the added current draw while I'm testing things.
If i get fancy a bit couldn't i also use a mosfet to switch over to  the LN317T for the very low current over 12V variation?
I have a ON/ON DPDT switch rated at 10 amps on hand and i really like IAN's idea for switching the 0V and -12V

[Doc38343]

If you are concerned about over-current, you can use an extra LM317 with a low value resistor between  Out and Adj as a current limiter.  Connect its Adj terminal to the adjustable voltage regulator's input. (see Figure 43. 1A Current Regulator in its datasheet: http://www.ti.com/lit/ds/symlink/lm117.pdf

That is one nice data sheet  It has a LOT more example use schematics than the one i have. I was very intrigued by the #42 example for a 5A voltage and current regulator design. I just don't have those extra devices but now i'm looking... 

[Ian.M]

Yes, the LM350T will work in the current limiter circuit I described, but you can do even better.   See Figure 21. “Laboratory” Power Supply with Adjustable Current Limit and Output Voltage in https://www.onsemi.com/pub/Collateral/LM350-D.PDF

With 24V in, that would give you up to 17V out.  The -10V bias supply for the FET constant current sinks for the voltage and current adjustments only needs to supply a few tens of mA so you could simply use a 1 watt isolated DC-DC converter module (12V in, 12V out) off the PSU's old +12V rail to get a voltage more negative than the -12V rail.

However, if you get a 5A adjustable boost converter module off EBAY that can run off a minimum of 10V and output over 32V, you could run that off the high current +12V rail to power the above circuit.  For cheap modules, the 5A rating is usually the input  current ratinng and the output current is reduced in proportion to the voltage boost which is why you need a 5A rating for only 1.5A out.   The bias supply would simply come from the -12V rail .That would give you a 25V fully variable supply while still letting you use the +/-12V, +/-5V and 3.3V rails in the same circuit.  To reduce dissipation, you could even add a circuit to reduce the boost supply voltage at low output voltages.

[tautech]

#tautech Yeah i almost ordered 10 of all three but my misses fussed a bit..... She says i'm just like a kid!LOL

Like a kid in a lolly shop. 

Do you think the LM350T's will work ok? I was also thinken I could just move the larger 12v bundle (5 leads combined) and the matching ground bundle to my temporary input terminal to cover the added current draw while I'm testing things.

Adjustable regs are just so easy to use, who cares if they warm the room a little, they're not on ALL the time.
They're quite heavy those leads, I'd have no concerns putting 10 Amps through each conductor.

Now have a good think about how you might use this PSU and voltages required.

+ & - 15 for opamps etc won't happen with linear regs but 12 + & - is near enough.
12 + and 12-
+5 and +3.3.

Those should be your SET voltages, with outputs for each.
Then, as you have plenty of 12V current available, why not 2 identical adjustable supplies from 12V?

TBH I mostly use voltages in the lower ranges, although my home buit PSU is capable of 30V & 5A.
I used LM338 and LARGE heatsink, but a dual LM350 PSU @10.8V & 3A plus fixed outputs sounds good to me.
As you have a few  check out this datsheet with adjustable current using an additional LM350:
http://www.onsemi.com/pub_link/Collateral/LM350-D.PDF

Everybody has to build a PSU......it's a right of passage. 

EDIT
Haha, ditto datasheet. 

[Doc38343]

#Ian.M and #tautech OK I may see the light here... my thoughts are this. I don't think i need high current anyway on the variable rails of my supply. I have the permanent +12 , +5 and +3.5 with the +5 rail capable of supplying the most current according to the supply chart... I am planning on doing some Arduino projects and experiments and possibly doing some things on PIC. My steady +5v rail is bang on the 5 and very steady so i can use that for those. I also have some older 12V items like an old CB radio and I want to do some LED mods with those new strip LEDs for my bike so I can use the high output port for those things. (I got me a bag of ATTiny85's and am planning on setting up quite a mess for my bike LEDs.. LOL)
Now for the variable if I get em safe and reliable the Idea of doing two 12V and having them floating is very good up to say 2.5 amp or so. These I could then use for testing boards on the bench safely, triggering mosfets for tests and all the other things I need to be able to connect my scope to. I also like this better for not frying my new scope or my PC by powering the arduino thru USB during the test phases after I program the tinys or other shields.
Last question for tonight is that if I do build 2 12v adjustable supplies and they float can they be combined for the occasional 18V needs I have . Note the ground would still be common so they would not technically be floating from one another.. just answered my own question... so now how can i get the full 2.5 to 3amp and the 19 - 20 volt capability, I need to again isolate one of the two to fully floating right?? Jeese I'm a long winded cuss for a guy who dosen't like to type...
OK rest time for today..Thanks 

[Doc38343]

OK I need to float or get a separate +5v for my LED volt/amp meter. It cannot be grounded to the same place as the supply being measured.
can this be done without using modulation and a transformer? I need about 350mA to power a L7805 and about another 200mA for the volt/Amp meter.
I saw a 555 setup to use a transformer but I want to keep size down if i can. (so I can get the two variable supply's we were talking about to all fit in the same footprint box)

I'll attach a photo of my monster...
 

[tautech]

I struck the same problem years ago when I buit my PSU, didn't have the knowledge then and used another tiny linear transformer based PSU.....within the PSU. 
These days I'd use a buck or boost from the main transformer for the isolation required.

[Doc38343]

Hey Guys,

I liked that schematic for the LM350 #21 example that has current and voltage regulated. Those Jfets are not handy and I'm so much a noob that one is tin case 4 legs but the schematic only shows 3 pin connect...

Can they be replaced with something I may be able to salvage and also are they necessary or just for 0 setting ability?
OH they were 2N3822 the 4 lead tin and 2N5640... I looked up the spec sheet and tried to cross reference for comparable devices but I am in over on these devices... unexplored space for me.

I have every thing else available on hand on the sheet and even new parts ... thinken i may build if i can overcome the Jfet hurdle.

[tautech]

Have a look at the many LM317 datasheets for a similar setup with out a Fet, shouldn't be too hard to find, although it may not be "lab" grade then.