Yet another linear power supply design

[crispytato]

Hi all, just a quick question, with (hopefully) a simple answer too!

I'm working on a new lab PSU with decent current capability, to be fed with a high-power 19v laptop power brick. I'm looking at 5-6A at 0-15v or thereabouts.

I settled on using the external pass element circuit shown in the LM317's data sheet, with some tweaks, the biggest of which are the replacement of the shown transistors with what I would consider to be reasonable equivalents. I'm using a 2n3906, and a tip3055 in the place of 2n2905 and a tip73.

My problem is, I can't get the output to go anywhere near the theoretical minimum voltage. I'm running it on protoboard without the negative reference at the moment, so it should go down to 1.25 volts or so at the collector of the pass transistor, but I'm seeing a minimum of 6.5V on the output. I've verified the LM317 ADJ pin is at a few millivolts off ground. Having tested this exact circuit on Multisim gives me the expected output, but in practice, this weirdness happens. As I turn the adjustment pot up, the voltage at the output increases as expected.

Included are my exact circuit, and the circuit drawn in Multisim, working properly.

Any questions about the circuit, as it's assembled, just ask. All the tracks are low impedance, relatively tightly laid out, and with no detectable oscillations going on. All components are running cool. I have tested the circuit both unloaded, and with a significant (500mA or so) load on it. No difference in the output is seen.

[Cliff Matthews]

49 ua on the adjust pin is way too low, the lm317 needs 5ma min if I recall.

[singapol]

I know nuts about circuit analysis and the use of spice but my gut feeling is your spice circuit and your actual breadboard circuit is different. Another thing is one TIP3055 may not give you 5-6A although it is rated 15A.
 In normal usage a TIP3055/2N3055 would be constructed to give 3A. You may want to look at R2. If you are using
a TO220 style LM317 then your spice model of LM317K is questionable as it is rated at 3A. Kind of makes the TIP3055 redundant.

[crispytato]

I know nuts about circuit analysis and the use of spice but my gut feeling is your spice circuit and your actual breadboard circuit is different. Another thing is one TIP3055 may not give you 5-6A although it is rated 15A.
 In normal usage a TIP3055/2N3055 would be constructed to give 3A. You may want to look at R2. If you are using
a TO220 style LM317 then your spice model of LM317K is questionable as it is rated at 3A. Kind of makes the TIP3055 redundant.

The 317K was the only package available in MS. I am using a TO220 package. As for the pass transistor, I'm using an active cooled heatsink, that according to thermal calculations should be able to dissipate around 90 watts at 25 degrees ambient (with a Tjunc max of 75 C)

you might be right that i've made an error in assembling the circuit, I'll have to take another look and make sure it's all correct.

[Kalvin]

Have you tried reducing the R2? You may also need to change the 2N3906 to something beefier like TIP42.

[crispytato]

Well, now I'm feeling rather silly. I drew out the circuit on paper, as I had it assembled, and I had messed up R4's placement. Luckily nothing blew up, I just had to swap R4 from base>collector to base>emitter. I'm feeling really dumb, because I spent at least a couple of hours troubleshooting, beeping out all the traces with my meter, everything, before I posted here, and somehow missed this one...

I shall post again once I've gotten the supply in it's beta form built up. I'm in the process of laying out a PCB, but for now it's all on proto board. Thanks for the help everyone

[Kalvin]

How do you generate the negative voltage? Do you use a simple cheap&dirty 555 as an inverting charge-pump or something more sophisticated?

[crispytato]

I have an MC34063 creating negative voltage on the board. The reason I didn't just use a charge pump was that I need to power a couple of op amps with a split supply. A bit too much for a charge pump to happily supply.

Attached is the analog portion of the schematic as it stands now. Res and cap values aren't all inputted yet, so don't trust them

U3 is set up as a summing amp to take the DAC output and offset/amplify it to get 0-16v output. I'm using external 12 bit dac and adcs for current/voltage set/read. Being I want an excuse to get some experience with using SPI in a practical solution. Haven't decided if I'll stick with the ATMEGA 2560 or a PIC24FJ (as it has an internal 12 bit ADC) yet.

[Kleinstein]

The LM317 is not really made to be used with external transistor and current limiting. The shown circuit looks like it might oscillate with some types of load. There is only limited option for compensation. Also the output voltage is controlled at the output pin of the LM317 - so voltage drop on the shunt is not compensated.

So I would consider an LM723 instead, especially if you don't need a low drift current limiting. This chip is made to drive an external power transistor. The main downside is that standard current limiting is not temperature stable.

With the typical 19 V input voltage, an 2N3055 is good for maybe 3-4 A. At 5 A one would need exceptionally good cooling and low room temperature. The TIP3055 has an slightly lower limit. So one might want to use 2 of them in parallel.

[crispytato]

The LM317 is not really made to be used with external transistor and current limiting. The shown circuit looks like it might oscillate with some types of load. There is only limited option for compensation. Also the output voltage is controlled at the output pin of the LM317 - so voltage drop on the shunt is not compensated.

So I would consider an LM723 instead, especially if you don't need a low drift current limiting. This chip is made to drive an external power transistor. The main downside is that standard current limiting is not temperature stable.

With the typical 19 V input voltage, an 2N3055 is good for maybe 3-4 A. At 5 A one would need exceptionally good cooling and low room temperature. The TIP3055 has an slightly lower limit. So one might want to use 2 of them in parallel.

Hey, I appreciate the input. I shall have to get a few lm723's and have a play around with them. I calculated out the thermals of the transistor jusing datasheet values for my heatsink (0.49*C/watt with active cooling), and I ended up getting a maximum of 67 watts or so allowing for 150 degrees tjunc. obviously that's a bit hopeful, I might have to move to a more powerful transistor down the line. for now, I have no need for dissipating multiple amps, at the low range of this PSU, so I can live with less current at voltages where it has a large amount of power to dissipate.

Do you have any better options for a pass element that I could consider? I have the 3055 on hand, so it's what I've used in my prototype circuit, but if there's something that can dissipate more power, in the TO247 package or similar, and doesn't cost stupid money, I'd gladly swap it out. Is it possible to move to a MOSFET, in a circuit designed for a BJT, without major re-design? To potentially lower the resistance through the pass element? Or isn't there a major advantage when it's in a partially on state?

[Kleinstein]

A MOSFET as a pass element usually needs extra voltage to drive is - so it is not attractive in this case as one would loose more voltage or needs an extra higher supply. Also many modern MOSFETs are not suitable for linear operation. So a MOSFET is not a good idea.

It is the TO218 / TO247 or similar case that sets the practical limit. So no cheap option for a single transistor to dissipate something like 100 W in real life (and not just first page DS rating). The cheap option is something like 2 of the TIP3055 in parallel. For a little more power the TIP35 might be an affordable option. The 2N3906 is also to small to drive the TIP3055 - it usually needs something like an BD135/136.

One still needs to consider the case of low voltage and high current - a short circuit is such a case. It is usually easier to have the transistors powerful enough than adding extra thermal / SOA protection. Using a kind of fold-back current limiting might be an option to get around this limit.

[Cliff Matthews]

The LM317 is not really made to be used with external transistor and current limiting. The shown circuit looks like it might oscillate with some types of load. There is only limited option for compensation. Also the output voltage is controlled at the output pin of the LM317 - so voltage drop on the shunt is not compensated.

So I would consider an LM723 instead, especially if you don't need a low drift current limiting. This chip is made to drive an external power transistor. The main downside is that standard current limiting is not temperature stable.

With the typical 19 V input voltage, an 2N3055 is good for maybe 3-4 A. At 5 A one would need exceptionally good cooling and low room temperature. The TIP3055 has an slightly lower limit. So one might want to use 2 of them in parallel.

Hey, I appreciate the input. I shall have to get a few lm723's and have a play around with them. I calculated out the thermals of the transistor jusing datasheet values for my heatsink (0.49*C/watt with active cooling), and I ended up getting a maximum of 67 watts or so allowing for 150 degrees tjunc. obviously that's a bit hopeful, I might have to move to a more powerful transistor down the line. for now, I have no need for dissipating multiple amps, at the low range of this PSU, so I can live with less current at voltages where it has a large amount of power to dissipate.

Do you have any better options for a pass element that I could consider? I have the 3055 on hand, so it's what I've used in my prototype circuit, but if there's something that can dissipate more power, in the TO247 package or similar, and doesn't cost stupid money, I'd gladly swap it out. Is it possible to move to a MOSFET, in a circuit designed for a BJT, without major re-design? To potentially lower the resistance through the pass element? Or isn't there a major advantage when it's in a partially on state?

Don't know if $4.20 is "stupid money" for a 250w 20amp pass device, but Futurelec.com has great price in this area.
http://www.futurlec.com/Transistors/MJ15003.shtml

[crispytato]

I looked at the 15003 before, but it's to-3 package only, which wouldn't work for my application.

I found this http://www.onsemi.com/pub_link/Collateral/MJL21193-D.PDF which seems to be a pretty sweet upgrade to the tip3055. Rated at 200w maximum dissiaption, and the base current seems managable at the level I'll use it at. I'll have to experiment with a few of these. I might eventually replace the heatsink/fan I have, if I can source a higher current laptop power supply. I know there are some really powerful ones available with some of the gaming laptops. 210 watts was the largest I came across if I remember right - around 11A at 19 volts output. currently the one I have is rated for 6.3A @ 19v, should be sufficient for most everything electronics related that I work on currently at least!

[Cliff Matthews]

Shows obsolete.? Mouser has the upgraded MJL21196G at $4.50
http://www.mouser.com/search/ProductDetail.aspx?R=0virtualkey0virtualkeyMJL21196G