Please Check My Simple PSU Circuit

[Mechatrommer]

the spec:
1) portable like other wallwart
2) dual rail +ve and -ve psu (tracking -ve), adjustable voltage using trimpot (Vref in schematics)
3) max voltage is up to transformer max voltage (whatever transformer i can find in my junk stock)
4) max ampere is 8A since i have a bunch of 13007 NPN (and of course to the transformer max ampere rating)
5) current limiting and the plethora of protection is currently omitted for simplicity. i will proceed to this when everything is working.

abstract:
there's example on the net (link below) but either not up to the spec, too complicated for my head to understand, or/and i dont have the parts. so i will reinvent the wheel by trying as simple as i can get with parts that i already have. my very weakness is in the transistor circuit. basically i just put them together and  tinker until they work. the best i can get so far is by using optocoupler to adjust -ve rail voltage according to Vref control adjustment (using dual comparator with +ve voltage feedback control). and i'm not sure if my transistor arrangement for -ve rail is correct, maybe this is a novel invention from me or i'm just redoing people mistake in the past, thats why i need advice, i got no luck for -ve power transistor design in the net, or i just simply dont get it.

interested at links:
0-15 Volt Tracking Power Supply
Constructing the +12 Volt Supply
0-50V 2A Bench power supply

conclusion:
from testing powering dc motor (i dont have other easy to access "higher power device" right now), so far so good (i mean dc motor did rotate fast CW and CCW resp +ve and -ve rail), but T2, R9, R10, and esp T7 can get hot very easily albeit only around 0.1A is running the motor, maybe i need to tune the value some more or change the arrangement.

obsolete. refer to latest update
problem:
the biggest problem is -ve rail is not following exactly the +ve rail magnitude. ie when i set Vref to somewhere in the middle between V+ and GND, the -ve rail seem to be pulled to GND potential, but when Vref is near GND or V+, then only then the -ve rail is closely tracking the +ve rail (not exactly same, but close). i thought the optocoupler is just like a isolated ON/OFF switch which will follow the comparator output (switching) but i'm not sure whats going on, so need big help on this. refer to schematics below, i dont have spice model for PC817 in Tina software, so i just represent it with a LED and a transistor, but the external connection is like shown. its from my try and error to get the best -ve tracking i can get. please help on there, i'm sure it will just take a blink for you to understand my circuit. thanks.


------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

update 28th Dec 2011: (portpsu1b)
the first design on -ve rail is in the Art Of Electronics 2nd Ed. entitled as "The Design of The Blunder!". so... its a retard. changed the plan, i used the N-MOSFET now, free from JunkYard Dept. re-arrange and re-testing a bit, now its better, but still, far from perfect. tracking is better than 1st design. tracking will not work if either side of the rail is heavily loaded, and the other is not. the 2nd LM393 in 1st design is also a retard, so changed to become voltage follower of the 1st LM393 output. so the latest is portpsu1b.png please review it. the 1st design (portpsu.png) is obsolete, dont talk about it anymore unless absolute necessary

ps:
1) 1x C945 and 1x LM393 dead. RIP.
2) please translate the label in the picture as:
   (a) BC394 = 13007
   (b) !NPN = C945
   (c) TLC393 = LM393 or LM193

------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

update 28th Dec 2011: (portpsu1c, few hours later)
in the cold of the night, idea popped out, changed the tracking system on the 2nd/-ve rail comparator, it turned out similar to circuit posted by Hero below, now i know what it does , now... tracking is... perfect! (portpsu1c) unless the transformer is saturated (overloaded where it cannot maintain its original sine output). and another problem is excessive heat when +ve -> -ve rail is used to power a load (only 1.5A to load). have no idea why, will figure out later, some bug sneaking i believe. for low powered device (<1A) i think this unit is now usable, thanks to me for my novel invention! and err, to Hero999 my friend the 2nd design (portpsu1b.png) is obsolete, dont talk about it anymore unless absolute necessary, but... its just hours ago? i believe no one catched that! review, C&C are highly welcomed!

ps:
1) death toll increased to 2x C945, 3x LM393 and 1x schottky. RIP.
2) translation above still apply with addition:
   (d) the Vref system is not a fixed voltage divider as shown. its actually a trimpot, go figure.

------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

update 24th Jan 2012:
just to add the note (without bumping ), i've found similar technology has been used almost 30 years ago, and more involved than mine. the link... http://michaelgellis.tripod.com/power4.html happy chinese new year and have a good day!

------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

update 7th Feb 2012:
the latest version psu1e (skip the 1d, there's not much in it). it will become my temporary power supply to build more... -ve rail still oscillating, +ve rail now using LM317 as regulator (controlling base of pass elements), if i struggle to clean the -ve rail any longer, i'm risking a divorce! so i have to call it off for a while. the performance is as in reply #12 & #13 below. i hate to bump it, but i have to for the sake of completeness and tidyness any improvement advice is welcomed.

------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

update 19th Feb 2012:

i think this is the finale, urghh i'm tired of this. psu1f... has been tested/used for few weeks/days or so. does the job! major changes... now -ve rail wont eyeing on +ve rail anymore (heavy loading on +ve rail wont be reflected anymore on lesser loaded -ve rail). instead both +ve and -ve rail will use Vref from LM317 as a reference (the -ve rail will invert it first and then regulate the -ve rail from there). the LM317 is overkill i know, a simple zener + volt divider(trimpot) + magnifying opamp can do the same job. but, i have much more LM317 stock than zener.

Regards,
Shafri.

[Zero999]

The circuit is totally wrong and doesn't stand a chance of working. Redesign it or if you don't know how, copy a design off the Internet.

The LM393 is a comparator, not an op-amp. Using it as a linear amplifier is bound to cause intability and oscillation.

Why are you using a common emitter amplifier for the negative side? Even if the feedback loop had been designed correct (which isn't the case here, it's wrong anyway), it's prone to oscillation with capacitive loads. You should use an emitter follower for the negative side, just like the positive.

The opto-coupler is a weird idea and is totally pointless in a part of the circuit not requiring galvanic isolation.

Using two half-wave rectifiers and non-tapped transformer will work but the transformer will cook if one rail is fully loaded and the other unloaded due to DC current causing core saturation. Use a centre tapped transformer, of if that's not possible, a voltage doubler to get the negative voltage.

Here's a bipolar power supply circuit I designed awhile ago. It uses 3 pin regulators on the output which feature overcurrent and thermal protection.

http://www.silicontronics.com/index.php?action=ezportal;sa=page;p=7

[Mechatrommer]

center tap transformer is more (slight?) expensive and harder to get compared to normal 2 wire output transformer here, let me try on more economical design pleaseeee . about your circuit, its nice using OTS chip like LM217 (i used LM317), did that (for +ve rail only) but 1.5A is pretty limiting and the circuit protection tripping is really annoying for "higher pulse current during startup device". for the -ve rail chip (LM337 in your case)... currently i dont have any stock, and i dont see the point of getting it from my 1.5A limitation experience on LM317. about the optocoupler galvanitation, since the 393 output is +ve side, ie 0 to 20V, connecting directly to -ve rail control side seemed to not working (i tried that), so the opto is there to become galvanic switch? (is that the term?) or level/voltage shifter? i mean the input (from LM393 output) is 0V (off) to +ve volt (on) and the output is 0V (on) to -ve volt (off) on the T7's base. advice needed on proper setup there (opto not working as expected). i tried to control both +ve and -ve rail directly from 1 comparator, but its too much hassle (not working) i ended up using 2 comparator in the dual package. thats why 2 identical inputted LM393 there.

Why are you using a common emitter amplifier for the negative side?

as i said, its my novel invention any link/advice on proper example? and its no difference than the +ve rail, its exactly a copy of it. except the v+ to gnd vs gnd to v- potential. i'm thinking of using pnp to replace T7, but i'm not sure either whats the reason. i dont think i have PNP counterpart for the 13007. so i hope NPN can be used for this -ve rail app.

about the LM393 stability, the purpose is to shut off control (on T2, lets take +ve rail only for simplicity) immediately wherever Vpos_out > Vref. iirc, Dave gave good example on this in his PSU vblog, or maybe i'm misinterpreting... i mean immediate = saturated = high gain = comparator. it seemed to work with the buffy 1mF cap and the "simple dc motor load" on the output i may put smaller cap for hi freq filtering if needed.

[Zero999]

center tap transformer is more (slight?) expensive and harder to get compared to normal 2 wire output transformer here, let me try on more economical design pleaseeee .

Fine cut corners if you like but you'll save more money using the correct components in the first place rather than having to replace burnt out parts later.

about your circuit, its nice using OTS chip like LM217 (i used LM317), did that (for +ve rail only) but 1.5A is pretty limiting and the circuit protection tripping is really annoying for

They also come in 3A or 5A flavours and the LM317's output current can be boosted by adding a transistor.

as i said, its my novel invention any link/advice on proper example? and its no difference than the +ve rail, its exactly a copy of it. except the v+ to gnd vs gnd to v- potential. i'm thinking of using pnp to replace T7, but i'm not sure either whats the reason. i dont think i have PNP counterpart for the 13007. so i hope NPN can be used for this -ve rail app.

No, it's totally different. A transistor with the load connected to the collector forms a common emitter amplifier which has a lot of gain and phase shift at high frequencies. A transistor with the load connected to the emitter is called an emitter follower. It has a gain of just under 1 and a low impedance output. You should use a PNP power transistor on the negative rail.

about the LM393 stability, the purpose is to shut off control (on T2, lets take +ve rail only for simplicity) immediately wherever Vpos_out > Vref. iirc, Dave gave good example on this in his PSU vblog, or maybe i'm misinterpreting... i mean immediate = saturated = high gain = comparator. it seemed to work with the buffy 1mF cap and the "simple dc motor load" on the output i may put smaller cap for hi freq filtering if needed.

The circuit as you've drawn it will probably oscillate.

I don't follow the video blog at all. Please provide a link.

[Mechatrommer]

now i see your point there. i had to move the Vneg_out to collector side so it will not stay at -Vmax potential all the time. i will study further about it thanks for highlighting. i've saved a web on boosting LM317 current using power transistor, but it requires 0.3 and 0.7 ohm resistors for current divider which i dont have. the dave's link is EEVblog #221 – Lab Power Supply Design – Part 1
edit: LM317 booster link here: http://www.eleccircuit.com/lm317t-voltage-regulator-with-pass-transistor/

[alm]

Note that you will lose overheating protection when you use an external pass transistor with the LM317. Just getting a higher current brother of the LM317 makes more sense in my opinion. It's less parts and probably not much more expensive. The only issue is that I'm not sure if these are available for negative voltages.

[Zero999]

The circuit still isn't very good I'm afraid.

The unnecessary opto-coupler is still there and now you have a common source amplifier for the negative side and you still appear to be using a comparator in the linear region. You should replace the comparator with an op-amp.

Why don't you listen?

You could use a dual ganged post with two LM338 supplies connected in series and running of a transformer with a twin secondary. You may need a trimmer on R1 of one of the regulators to adjust for the difference between the tracks of the two potentiometers.

[Mechatrommer]

Why don't you listen?

because i'm not sure in what configuration should i use an op amp to control/regulate the voltage out of the pass transistor/mosfet, i only can think of high gain op amp, in the end it will turn out to become like a comparator, so please advice. thanks for coming back. i will look into LM338. it seems i need to pause this for a few days, i hope i will be coming back soon enough.

edit: idea popped out. maybe a voltage follower op amp? gotta give it a try.

[Mechatrommer]

The circuit is totally wrong and doesn't stand a chance of working. Redesign it or...
The circuit still isn't very good I'm afraid.
maybe a voltage follower op amp? gotta give it a try.

oh my, i struggled to death to remove the noise. i thought its the simplest design i could get (comparator), where a voltage follower can do the job easier (regulation+lower noise). why someone didnt tell me earlier? (banging head on the desk).

[Zero999]

The noise is probably oscillation.

Why don't you modify my schematic?

Here's a version for 5A. If Tr1 is mounted on the same heatsink as the LM337 it will offer some overtemperature protection because it will shutdown when it gets too hot. I know you said you wanted 8A but 5A is already pushing it for a linear regulator. You could add a bypass PNP transistor to the LM338 (same principle as the LM337) and change the resistor value of Tr1's emitter resistor to get 8A if you like but you'll still have to put up with the high power dissipation.

[Mechatrommer]

thanx alot hero. but since my original spec is from 0-(Vtransformer-Vdrop), using 317,337,338 will limit to 1.25V min. i would like to include "Dave's" spec (down to 0V) to my circuit kinda nice my current (minimum ampere) requirement is quite lenient right now, so i'm still open on that (i believe it will follow V=IR rule). i've checked LM338, nice chip i've never come across, unfortunately, there's no -ve rail complement for it afai search. + it will need to be ordered which will take sometime.

i'll be looking forward to circuit like this (boosted LM317 and LM337) if i cannot solve the 0V requirement... http://www.tcaas.btinternet.co.uk/jlhlm3x7cb.htm or similarly http://electronics-diy.com/electronic_schematic.php?id=485 but it used dual adjustment each for +ve and -ve, i'll modify it to single adjustment/tracking based on your circuit. the said booster is not using 0.3+0.7ohm divider, instead i believe most of the current will pass the top and bottom power transistor, so i can tweak and change the model there to increase/decrease power capability. i dont mind about thermal shutdown capability as alm said, as i dont think i'll be needing that, one of the purpose is to power a high current (pulse) device during a short period and maybe i'll just incorporate beep warning and thermal control cooling fan. and even when its damaged due to overcurrent, most likely it will be the power transistor which hopefully easily repairable from the simple design.

my current plan is to use NPN darlington on +ve, PNP darlington on -ve, voltage following adjustment/tracking but i havent got time to lay it out and find the appropriate parts. i'll update when i've done that. your circuit will be a good reference for me later, thanks alot for some ice breaking idea

[Zero999]

i'll be looking forward to circuit like this (boosted LM317 and LM337) if i cannot solve the 0V requirement...

For a start you should read the LM317's datasheet for a solution.

An op-amp can also be used to zero the LM317. You could use a dual op-amp IC such as the LM358 and do both jobs with one IC.
http://www.silicontronics.com/index.php?action=ezportal;sa=page;p=19

[Mechatrommer]

took me many many hours to assemble psu1e (refer to above schematics, improvement welcomed), not to mention weeks of studying it, at least i'm better at transistor now , but yet far from perfect. solder, desolder, rearrange, enough is enough, SWMBO start mumbling. but at least i got newer and better psu than my recently dead pc atx psu, its adjustable, its dual rail tracking. i need it to build better product, hence i need it quick. so its not rocket science, its not engineur or profesionelle level just a quick report for the sake of completeness theorem... might be usefull for super beginner like me, CnC welcomed.

1) first 5, how its looks like "handmade" "handcrafted", one off project
2) performance testing photo shoots
    a) while regulating at 12V...
    b) loading the +ve rail (regulation down to 7V average) 2.2A current reading
    c) loading -ve rail (+ve rail stayed at 12V, -ve rail cramped to -7V) 2.2A current reading
    d) loading both rail (sometype of push pull load), worsen the regulation to 5V, 2.9A reading
3) scope captures
    a) ripple at 0.1V/div (yellow = +ve rail, blue = -ve rail) at no load, tuning last time worked (adding caps to tracker output), not anymore now
    b) 3.7V regulation/tracking (no load)
    c) 12V tracking (no load)
    d) loading +ve rail hard (3ohm load), -ve rail just dumbly follows the loaded effect even if its not (to be improved later).
    e) loading -ve rail hard, +ve rail steady (-ve tries follow +ve, not the other way)
    f) loading both rail hard, -ve unable to follow anymore, system (volt reader) shutdown for a while during load connection.

my conclusion, the bottleneck is the transformer, cant do much about it. the pass array calculated to be able to draw 5A safely (if they are somehow load balanced by heatsink sharing, cant afford ballast resistors, space requirement not allowing), but actuality only 2.9A @ 5V from the system. so i guestimate this psu is somewhere 10W - 15W rating. cheers.

[Mechatrommer]

10 pictures per post limit...
edit 19th Feb 2012: the first 6 pictures are older pictures.

------------------------------------------------

update 19th Feb 2012:
took me one whole night to specify the psu1f (not completed, just ripple test, transient/step not yet). using (Dave's idea) constant dummy load, i can plot the chart now...
from the look of it, maybe i can specify it as 0.5V-15V, 2.5A max, 10W psu, 100mV +ve rail ripple and 500mV -ve rail ripple? so far somewhere
still poor at negative regulation, and positive as well, under varying load condition.