117Views12Replies

Author Options:

linear regulator design, darlington, 5V dropout @ 2.5A !? Answered

The question says it all, really. I built up a linear power supply shown below. My unregulated 12V supply can fall as low as 14.7V @ ~ 2-3 A but if I end up being unable to go higher than 9.5V at 3A. Because I'm driving a motor as a load, the current drawn is very dirty. At 10V I see small dips that correlate to dips in the supply voltage when it goes just under 15V.

So what can I do to help alleviate this, as I was under the impression that I would see a absolute maximum of 3V of dropout, not a whopping 5V. (p.s. the maximum gate current into the S8050 is 10mA, but currently set to 6.77mA w/ a 10 ohm resistor)

p.p.s: MY design uses an LM324, which is really not the best, as it's slow and cannot sink much current. To alleviate the first issue I simply added current mirrors to aid the op amp in sinking current in a simalar fashion as a current source used for an open collector outputs.

What op amp should I use instead for better performance? I suspect the LM324 due to it's low slew rate and low GBP may cause poor high frequency rejection (esp. if I choose to use a switching pre-reg) and poor high frequency load regulation (causes poor transient response on the output)

Op amps I can choose from:

LM324; // slow, from my testing, this one has inputs that work to ground :D
LM224; // ditto, but in a weird package
LM386; // slow, but from my testing, this one also has inputs that work to ground.
LM741; // ancient relics of the past, not rail to rail by any means, sucks IMO.
UA1458TC // ancient relics of the past, I don't have much knowledge of them. Inputs do not work to ground.
LM318H; // its in a fancy metal case :D
TL072CP; // JFET inputs, but I think the output current is limited and the output can't go below the noninverting input.
TL2072AC;
OP37; // These things can't even give a stable output by themselves!
OP27; // ditto
LT1007; // have not messed with it much
LT1167; 
LT1191;
LT1363; CRAZY fast, 1000V/uS!!! Maybe too fast...
LT1360;
LT1112;


Discussions

0
None
iceng

Best Answer 2 years ago

5v is high, but you are using discreet NPN devices to make a darlington rather then a single unit which would have been designed to perform much better...

I would consider using a mosfet as a better regulator power component..

0
None
-max-iceng

Answer 2 years ago

Aren't MOSFETs worse to deal with to stabilize? A high-side N channel MOSFET would require voltages for the gate drive considerably higher than the output voltage, about 7V or so. Right? I would imagine this complicates the design, requiring a higher voltage (but lesser current) transformer to get at least 10v above the maximum output voltage, which I want to be 15V. a.k.a, 25V.

Also, if my supply is parasitically oscillating (it is...) then how can I stabilize it? I changed my design to use the LM358 as this is a better op amp than the LM324 and can operate properly from a single ended supply.

0
None
iceng-max-

Answer 2 years ago

Darlington is a current driven switch.

A Mosfet is a voltage driven switch.

fet switch.png
0
None
-max-iceng

Answer 2 years ago

And what sort of voodoo is that optocoupler thing? It certainly isn't a part I have just lying around. I'd like to use the parts I currently have. Maybe Ill try a TIP120 and see if I can get better performance with a few in parallel. (Hopefully negitive temp-co wont be a big issue :P)

0
None
steveastrouk-max-

Answer 2 years ago

Its a bunch of glass diodes, lit by a single LED.

0
None
-max-steveastrouk

Answer 2 years ago

Would it make more sense to a few blue LEDs, as they are already designed to emit (and focus) light onto a crystal?

0
None
-max-steveastrouk

Answer 2 years ago

Would the response time be too low? This would be a really weak gate drive... I might try it given I have a handful of 1N914's.

0
None
iceng-max-

Answer 2 years ago

The second Opto-Gen is TLP5918 photo cell that works as well in the sun or at night..

The last pic shows a simple Quad opto NPN reversing a small load just to HOODOO your voodoo.

Make friends with pack rats near you to share parts...

fetdriver.pngopto-gen TLP5918.GIFOptoH2.bmp
0
None
-max-iceng

Answer 2 years ago

Yes, I understood that for many years now. Its not a big deal to bodge on a few extra windings to a transformer for the error amps. Just a PITA.

More importantly, adding more components in series with the feedback path I feel will make things worse. I already have problems woth parasitic oscilation, so Ill need to do some research and find out how to frequrncy compansate the design effectively. The art of Electronics is a good start, do you know how to go about that?

0
None
-max-

2 years ago

https://easyeda.com/editor#id=45eefd75bd71451ba8f0...

Hopefully you can see the schematic using this link. The parts I am particularly proud of is that the output voltage can go as low as 200mV without the use of dual rail supplies, my use of diodes to act as an analog MIN() function (lowest op amp voltage wins), and my use of a constant current through a potentiometer to set generate an offset voltage for the error amp for the current (which I could change to use an adjustable current DAC for arduino control)

Things I might change include using the 7805 as a voltage reference, and using less LM334's to shave off a few pennies. (those chips are rather expensive compared to transistor current mirrors and sinks)

NOTE: My current design does not use the 7805, instead it using yet another LM334 going into a variable resistor for the voltage set and just a 33K resistor going to the 12V to set the current in the current mirrors which help the op amp and offer a minimum load for the output.

linear power supply design.png
0
None
-max--max-

Answer 2 years ago

I might add random capacitors to the potentiometers (R2 and R9) to help reduce noise and improve the voltage stability at those points. No reason not to.