Introduction: +/- 0..15V Power Supply - Part 1

Picture of +/- 0..15V Power Supply - Part 1

This is part 1 of 2. This is the electronics part and the next installment will deal with the transforme etc. as well as the encasing and front panel.

I know some of you will see this as a complicated circuit, but, as with any circuit, if you slice it up into functional blocks, it will become pretty simple to wrap your head around - read on and I promise that it won't bite you ;)

The circuit makes for a simple, but stable, power supply regulator circuit, which only needs a single potentiometer to adjust the positive rail, the negative rail tracks the positive (but negative going of course),

Step 1: BOM and Tools

Picture of BOM and Tools

Bill of Materials
Resistors (all 0.25W):

  • R1 270 Ohm
  • R1B Depends on potentiometer value - please see text on schematic
  • R2..R5 10k
  • R6..R9 22k
  • R10 220 Ohm
  • R11 120 Ohm
  • Pot1 5k Lin potentiometer (but see text on schematic for alternatives)

Capacitors:

  • C1, C2 100n
  • C3, C5, C6, C8 220n
  • C4 10µ
  • C7 47µ

Diodes:

  • D1..D3 1N400n (4001 to 4007)
  • ZD1, ZD2 ZPD3V3 (3.3V Zener diodes, but see text for alternatives)

Integrated Circuits:

  • IC1 LM317 (Positive voltage regulator)
  • IC2 TL072 (Dual Op-amp)
  • IC3 LM337 (Negative voltage regulator)

Further:

  • K1, K2 3-way screw terminal blocks (or solder the wires)
  • J1 a wire jumper/link (a short piece of solid wire)
  • Heatsinks for the voltage regulators, but they'll be discussed in the next installment.

Tools Needed:

  • Soldering iron and related paraphernalia
  • A pair of side/diagonal/whatever cutters
  • Multimeter (DVM)

Not explicitly needed, but still nice to have:

  • A pair of flat nose or chain nose pliers
  • An oscilloscope (any kind really)
  • An ample supply of Scotch & Coke - 18+ only ;)


The reason for this
Working with op-amps, a +/- supply is nice (sometimes even needed) to have and a variable supply allows you to e.g. verify the function of a circuit at lower voltage limits, or to run at logic voltages when interfacing such, while still being able to crank it up to a total of 30V rail voltage (+/-15V). Even if you just need a single supply, this will do up to 15V and 1A, however, this is not to be considered a lab supply, as it has got no variable current limiting, but that does not detract from its usefulness, as current limiting is a "Nice to have" thing, that will allow you to have "stupid moments" at the wrong time ;)

If you would rather make e.g. a 30V/10A (or whatever) supply with bells and whistles, please go somewhere else for now, but I rarely need over 15..18V (24..28V on seldom occasions for motors, but who can live with just one supply anyway) and being an electronic engineer with decades of R&D, I am supposed to know what I need :)

Beware though, of false power claims, which seems to infest the entire web (even this site), Just like if I "sold" this supply as being capable of 55.5W, from the fact that the max. voltage/current the LM317/337 can handle is 37V at 1.5A. The 30W it _can_ handle safely (with proper heatsinking), will be enough for most circuits. And a power supply is not really a power supply until it has got power in - so, for the moment (until part 2), this is only an elaborate voltage regulator.

If this (both parts) is well received, I will later post a real lab supply and a practically noiseless (ripple+noise way below 1µV) 9V supply for e.g. guitar pedals, sensitive mike- and instrumentation circuits. Really not trying to sound like a pouting teenager here, but if interest is as low as with the first circuit I posted, it will get more feedback on my web site. Remember, views ain't feedback ;)

Step 2: Discussion of the Circuit

Picture of Discussion of the Circuit

As you can see in the picture above, I have divided the circuit into its functional blocks.

Block A - Power input
The PSU is designed for +/-18V in, but the maximum rating of the supply to the op-amp (a TL072) is +/-18V, so it has to be lowered for a couple of reasons.

First off, you should NEVER run a component at its max. ratings! These ratings are not the manufacturers guarantee, that the component will handle those numbers well, but more of a guarantee, that crossing these limits is a sure way to kill your component (if not immediately, then in a near future).

Second - when you use a mains transformer, its output has to follow the mains voltage, which can vary a bit around the nominal - e.g. the EU harmonizing of mains voltage to 230V a number of years ago, did not really change a thing, because a wide tolerance (+10%/-6%) was incorporated in the bill, as it would have been extremely expensive to actually have the entire EU change all their mains equipment, so the mains voltage in the UK can keep at 240V as always, while the rest of the EU kept to their 220V - all within regulation (wonder how much taxpayer money was poured into passing such a nonsense bill). A 230V nominal transformer will thus have a different (higher) output in theb UK, than the same transformer used in the rest of the EU.

Further, there is voltage variations up/down over a day, depending on the total load, and during weekends, where heavy industry is mostly shut down, the voltage will rise for some consumers. So, always assume the worst case variations and calculate for it.

Hence, you need to supply your op-amp a lower voltage. A quick way of doing this, is to insert zener diodes in the supply lines (the "arrows" marked V+ and V- are the supply pins of the op-amp).

This way of using the zeners does not allow for huge variations and the exact voltage depends a bit on the current drawn by the op-amp, but it saves two resistors (and if there's something I've learned from designing circuits for others to use, it is that most beginners want to shave off whatever is possible - and then some) - if demand arises, I will change it to proper zener regulators). As is, with +/-18V in, the op-amp will get around +/-14.5V to +/-14.8V, which it will handle nicely. If you stock other values of zener diodes in the range ~3V to ~6V you can use those.

Yet a note about the supply voltage: The 18V is the minimum voltage (the valleys of the rectified and filtered voltage - pic 2 above), that the regulators (LM317/LM337) need to stay in regulation at 15V. A 16-0-16V secondary transformer would do this with a sensible amount of (rectification and) filtering.

If you happen to have a transformer a bit lower, the only ill effect will be a lower maximum voltage, where the regulators keeps their output regulated fully and if you have one with a slightly higher voltage, higher voltage value zener diodes can be used, but this is really outside the scope of this instructable and will be dealt with in part two.

Block B - The positive regulator

The positive regulator, an LM317 in an almost bog standard configuration, where the only difference, from common use, is how the lower "adjust" resistor is connected.

Everyone considering electronics as either a hobby or something more serious, should stock a handful of LM317's at all time. Using the ubiquitous 78-family of regulators and having a usable stock, means that you need a lot, e.g. 7805, 7806, 7808, 7809, 7812, 7815, 7818, 7824 (and possibly a few more - and that's just the positive voltage regulators). A single LM317 (LM337 for the 79-family of negative regulators) can replace them all, as it just needs 2 "programming" resistors to give anything from 1.25V to ~35V. In short, the bread & butter of on-board power regulators - and the power supply is the foundation which all electronics circuits have to rely on (whether mains- or battery powered), so if you want to ace out in electronics, start learning the ins and outs of power supplies of _all_ kind.

Block C - The "Ground" reference for the LM317

Usually, you just connect the bottom resistor (the potentiometer here) to ground and the regulator will bottom out at 1.25V. For a lot of research, it is mandatory to be able to get lower than that, so you need a negative reference of exactly minus (-) 1.25V.

That said, the 1.25V is rarely 1.25V, but can be anything from 1.20V to 1.30V, depending on both manufacturing tolerances, adjust pin current and current draw from the output, so making a 1.25000V reference will not quite cut it. In this circuit,

I have designed a circuit that uses an op-amp to constantly monitor the instantaneous adjust voltage and then generate its negative mirror image for the reference - that way the adjust voltage can be all over its tolerance range and still maintain a rock stable lower reference voltage.

C5, by the way, is the output capacitor of the LM317 and should be placed close to the output pin of same. The reason that it is drawn a bit longer downstream is simply to make the schematic as clear as possible.

Block D - Tracking control of the negative regulator

IC2B, the other half of the dual op-amp, maintains the balanced voltage output, by comparing the midpoint of the +/- rails to ground. The reason for using R8 and R9, as opposed to a single 11k resistor (the parallel value of the two 22k's), is simply to make it easier for people stocking just the E12-row of resistor values, feel free to use a single 11k if you have it.

Block E - The negative regulator

The negative regulator, the LM337, is controlled from op-amp IC2B, to mirror the positive output. The adjust pin is filtered with C7, which makes D3 needed, to protect the LM337 from fatal injuries, in case the power input is turned off.

Step 3: Component Overlay and PCB

Picture of Component Overlay and PCB

As you can see, it's very easy to build, when you make the PCB for it.

In case you do want to make a PCB, use the one on the last page of the attached .pdf-file and print it to scale (it is made in size 1:1). The text on the PCB must be readable from the solder side, so if you print it on a LASER printer, the toner side should be towards the copper -this works whether you use photo transfer or a clothes iron on a glossy magazine.

See, that was not as bad as your first look at the schematic made you think... Or what?
Need something clarified? Ned some smaller changes to the circuit? I'll do my best to help you out, but after being ill for a very long time, I just need a week of vacation to get totally into shape (I hope ;), so I'll be answering whatever questions or comments after the 16th. this month.

Next (and final) installment will be about feeding the beast and capturing it in a box with a front panel. Hope to see you back then :)

Comments

JanK209 (author)2017-07-08

Hi. Did you consider replacing TL072 with NE5532? It should give better performance overall and much lower noise on the output. It costs roughly the same and availability is good. You could also consider lowering input impedances around opamps even tenfold, further lowering the noise. NE5532 is capable of driving loads down to 600 Ohms with no problems. Putting a capacitor of high value from between R2, R3 - down to ground should improve noise considerably.

Also, it seems the second opamp controlling the LM337 has its feedback taken from a wrong place. It should take its feedback from the OUTPUT of the LM337. As shown the opamp seems to have no way of controllig the loop at all.

The outputs of the regulators are in fact controlled by the opamps, therfore C4 and C7 should be ommited, as these capacitors form a low pass filter inside the feedback loop, propably leading to oscillations on opamps output and bad performance.

I may be wrong, these are just my first impressions. Is the circuit working as is?

Omnivent (author)JanK2092017-07-20

Hi,

Thanks for commenting! Feedback makes stuff better.

"Did you consider replacing TL072 with NE5532?"
Not really. Last time I bought 5532's was like 20 years ago and back then, they were around $12 a piece, so reserved for the more exotic audio projects.

"It should give better performance overall"
I doubt that very much.

"and much lower noise on the output."
I'm getting the idea that your interest in electronics is mainly focused on DIY audio? Here, ripple suppression is more important - but it's a compromise between the two :)
(Not to forget the vain hope, to make it cheap and accessible to all sets some constraints)


"It costs roughly the same and availability is good."
That was a nice surprise to me, I found some at eBay at an even lower price then the TL072, but I keep a stock of the TL072, so that's what I went with and the low input impedance of the 5532 would interfere with the voltage setting of the LM317 anyway - I might order a batch to test out though, as they can handle higher supply rail voltages.

"You could also consider lowering input impedances around opamps even tenfold, further lowering the noise."
Not without getting in trouble in other places - and again, this is not audio, so ripple is the main adversary.

"NE5532 is capable of driving loads down to 600 Ohms with no problems."
Yes, in a low-gain closed loop, but so is most op-amps.
The "raw" output impedance is an entirely different matter though (almost 100 Ohm IIRC).

"Putting a capacitor of high value from between R2, R3 - down to ground should improve noise considerably."
Unfortunately not, it is a sure fire way of making it go ape 8-/

"Also, it seems the second opamp controlling the LM337 has its feedback taken from a wrong place. It should take its feedback from the OUTPUT of the LM337."
Yes absolutely!!

The joys of going from hand drawn paper to CAD (and in a neat way), while trying to pack for holiday - got me to connect it wrongly. This is one reason why feedback is very important - THANKS!

"As shown the opamp seems to have no way of controllig the loop at all."
"The outputs of the regulators are in fact controlled by the opamps,"
Hehe ;)

"therfore C4 and C7 should be ommited, as these capacitors form a low pass filter inside the feedback loop, propably leading to oscillations on opamps output and bad performance."
I'll admit that C7 could be left out with good results (but with a bit more ripple). C4 on the other hand, cannot be omitted!

"I may be wrong, these are just my first impressions. Is the circuit working as is?"
Yes, in several iterations, but I have spotted a few things that I think will improve the overall response, so I'll redraw the circuit (+ PCB etc.) and post it as soon as everything is verified (have an appointment on the hospital tomorrow, so might be somewhat delayed).

For now, thanks a lot for taking the time to "proof read"!

JanK209 (author)Omnivent2017-07-25

Hi,

I suggested NE5532 over TL072 mainly because of its higher bandwidth, which is useful in a voltage regulator scenario. Also commercial low noise power supplies are almost impossible to get, and others are impossibly cheap. I see it as a gap that only DIY can fill.

From what I see voltage "reference" is cleverly stolen from LM317s relatively clean output and formed over R3. It seems to me it doesn't have to follow any deviations in the output, for the opamp to be regulated - we want it stable and ripple-free, right?

I ran a quick simulation in falstad, and bypassing R3 and R4 improved load regulation a lot (with C4 removed). It will have very similar beneficial effect on ripple rejection as well. You can also see what C4 does to the opamps output - it goes into saturation trying to discharge/charge C4 through its open loop output impedance, as the load changes. These spikes show themselves in the output, obviously.

LM317 itself has PSSR around 60dB at reasonable bandwidth (with ADJ. pin bypassed) and I wouldn't use it as voltage reference without proper bypassing, when we have an easy option not to.

Its a simple model so I may be terribly wrong here. You can have a look yourself: http://tinyurl.com/y9jwqtqz The schematic is a bit messy - sorry!

Omnivent made it! (author)JanK2092017-08-01

Hi,

"[...] commercial low noise power supplies are almost impossible to get, and others are impossibly cheap."
I guess you meant "impossibly expensive" here? You can get very fine supplies if you are willing to sell your soul and your first born, but they would be of even less interrest to this sites users than this - if a circuit has got 5 components or more and is not glued up from Chinese eBay modules, the whole thing is more or less ignored "learn something, yeech".

"we want it stable and ripple-free, right?"
Exactly!

"I ran a quick simulation in falstad [...] "
Well ;)

"LM317 itself has PSSR around 60dB at reasonable bandwidth (with ADJ. pin bypassed) and I wouldn't use it as voltage reference without proper bypassing,"
Which is the purpose of C4, which adds a further ~40dB

"Its a simple model so I may be terribly wrong here."
Yeah and it's falstad ;) Almost impossible to decipher(!) and simming just a bit of a circuit seldom tell the overall response.

To be open to your suggestions and fair in the comparison, I ran a couple of sims. Attached is the output from both circuits -
same settings on both runs, each of ~5s (which is a very long time in real time, for a full circuit), to make sure they had time to stabilize properly.
The load consists of 13 Ohm resistors (one on each output), so current is a shade over 1A.
Conclusion:
Your suggestions made the outputs around 80% worse! ;-P
The circuit is so well balanced that it almost hurts 13.484V both plus and minus i.e. any unbalance will be less than 0.004% worst case and that's pretty fly for a fairly simple circuit IMO.

I'll get around to uploading the revised .pdf, schematic etc very soon (I hope ;)

Thanks anyway, it's healthy to get second opinions and constantly quetion what you do to stay sharp.

Regards

JanK209 (author)Omnivent2017-08-09

Hi,

80% worse means just shy of 6dB less regulation. Seems to me it can be fixed just by using same value capacitors (47uF) in both sims, at least theoretically. What my solution was supposed to cure was step response to fast load changes. Why? Because LM317 is a c.a. 1960 chip built around 741-like opamp and its way slower than excellent TL072. I suggested the change to let TL072 do the job of load regulation, because as is - LM317 is doing it for the most part. Since you have built the sim already could You try and run it with a lets say 10kHz 500mA p-p square wave current sink loading the positive output (as well as some 130 Ohm resistor)? It would be nice to see if there even is anything to fix there. If You would be doing that I would be interested to also see the raw output of TL072 in both sims, as with C4 it was saturating even in falstad.

As for circuit balance It is very optimistic. In the real world You'll be using 5% tolerance resistors. Not that it matters a lot in a power supply.

Thanks for taking the time.

About This Instructable

1,238views

34favorites

License:

More by Omnivent:+/- 0..15V Power Supply - Part 1Kelvin MixerSpiral Winder Tool
Add instructable to: