18

I'm toying with the idea of putting together a switching power supply (my first), using something like the LT1076-5 or the LM2576 controller ICs. These ICs have low external parts counts, and a relatively low switching frequency (56kHz-100kHz). Having spent some time reading datasheets for controller ICs, it's clear to me that some component placements are critical to the design. I'm wondering, then, if it's advisable or even possible to create and test the power supply on a breadboard, and later move it to a breadboard-layout protoboard.

If I'm not requiring ultra-high efficiency (any switcher's gotta be better than a linear when dropping ~35V, right?), does that make a difference? Or is it more likely to simply not work at all?

stevenvh
  • 145,145
  • 21
  • 455
  • 667
Mark
  • 4,650
  • 10
  • 48
  • 53
  • Low frequency has the advantage of radiating less, but also require bigger coils. 2~3MHz will often have enough with 3.3\$\mu\$H. May be important if you have little space. – Federico Russo Jul 19 '11 at 07:26
  • My power supply partner always builds what I call a 'Tower of Power'. He puts what he can on a breadboard and puts the high current portions on heavy wires solder up into the air. – kenny Jul 19 '11 at 12:28
  • 1
    You are getting a wide range of answers here. I suggest careful discernment of the input :-). – Russell McMahon Jul 19 '11 at 23:46

5 Answers5

15

If built carefully and sensibly with shortest minimum lead lengths, short paths to power rails and proper decoupling and filtering then a breadboard can be not too much different than a PCB based supply. Good results can be expected and noise should not be vastly worse than a typical PCB based circuit.

If built as roughly as breadboard circuits often are then bad results can be expected. However, the low frequency (50 - 100 kHz MAY even save you in those cases.

Switches have a certain amount of magic in them. In some cases/locations a few pF of stray capacitance can make things go very wrong. BUT

I have successfully built numerous switchers on breadboards (plug in style).

LT1076 Datasheet:

LM2576 datasheet

Spec sheets say these operate at 100 kHz and 52 kHz so both are relatively "breadboard friendly".

The fixed voltage LM2575 has a slight edge in lashup-proofness as it has the critical feedback divider internally, but I'd recommend going with a variable output voltage version as being more useful and flexible and being able to teach you more. The LT part looks somewhat more capable overall.

Lower than higher frequency is liable to be more successful on a breadboard, so around 100 kHz is a good starting frequency. Old tech for most ICs. Even 1 MHz may be OK but capacitive coupling increases by 10X wrt 100 kHz. A 1 pF is 10 pF equiv. A 10 pF is 100 pF equiv. A few pF seldom hurt too much at 100 kHz.

Keep leads short. Group components together that share common heavy current paths. Bypass well. Do the best breadboard job you can. Avoid long loopy wires such as usually don't matter at all. Think ahead and plan it at least a little bit. Odds are it will work.

A trap is the feedback divider network (R1 & R2 in each case on datasheet page 1 diagram, but upper/lower swapped). Here yu have a feedback input pin and a divider from output to regulate voltage. Neither datasheet shows it, but a small capacitor across top resistor of divider (feedback ping to Vout) usually helps impulse response. A small cap from the centre point = Feedback pin to anywhere else is often a disaster. Ask me how I know :-). That MAY be most sensitive spot in many circuits.

Think about current paths. Inductor /switch/diode/filter caps (in and out), Ground and power sides.

If driving an external transistor (not relevant here) keep leads short. USe reverse zener across gate-source if using a FET.

The IC's chosen make life easy at the cost of some flexibility. For "playing" look at MC34063 - I recommend them to one and all. Old. Some defects. Cheap. capable and flexible and fun and low parts count. Built in high side current limit. Can do about ANY topology (boost, buck, buck boost, CUK, SEPIC, ... .

MC34063 datasheet

  • See figs 15, 20, 21 in datasheet for step down examples.

  • Fig 15 is with internal switch. Up to 0.5A out - maybe more.

  • Fig 20 uses NPN external but I'd use an N Channel FET.

  • Fig 21 uses PNP external - I'd use a P Channel FET.

I'd prefer Fig 20, with N-Channel FET.

This will do 36V + direct (40 V rated) BUT start at say 12V to 5V to play. MUCH more energy and things to go wrong at 36V in.

Ask more questions if of interest.

ADDED: 20 July (NZT)

The example ICs which have all pins in a straight line give every prospect of good results if used following the guidelines above and data sheet guidelines.

The IC can be positioned so that power rails are fed from breadboard strips only a few tenths of an inch away and decoupled with minimal lead lengths. There are few other components and these may be placed with very short leads.

However, this is such a simple circuit that use of "vectorboard"/ veroboard / ... etc copper strip board would allow a tidy and easy implementation with slightly less to go wrong.

When using plug in bread boards some component leads are so thick that they will either not fit or will permanently "set" the breadboard springs if inserted. These can be dealt with by soldering SHORT lengths of wire to them as lead extensions and plugging these into the board. Properly done and with leds trimmed the result looks OK and is liable to be effective.

Too thin wire may also have contact problems.

Russell McMahon
  • 147,325
  • 18
  • 210
  • 386
  • Added references to relevant circuits in application note. – Russell McMahon Jul 19 '11 at 00:41
  • I added some information about potential controller ICs, if that helps. – Mark Jul 19 '11 at 00:53
  • 1
    If "a good chance" is >50%, then *"a good chance of working, and a good chance of not working"* makes no sense. Probabilistically. – Federico Russo Jul 19 '11 at 07:22
  • 1
    1. Subsequently added chip info may explain why I wondered how I'd missed that data :-). 2. My statements about "a good chance" still happily stand. Done well and with care and understanding it has a good chance of success. Without thought care and understanding you MAY be lucky. I think the context makes that clear. Hopefully so, anyway :-). – Russell McMahon Jul 19 '11 at 20:40
  • How about on a breadboard-patterned PCB, such as [this one](http://upload.wikimedia.org/wikipedia/commons/7/75/Breadboard-144dpi.gif)? I imagine you'd still have to deal with capacitance between traces, but at least you'd have good connections... – Mark Jul 20 '11 at 14:55
  • That would be fine - especially with the single in line packages suggested - although the eg 8 pin DIP of the MC34063 would work well there too. As above - run ground and Vin on side rail and take signals to them with short links. Use good size cap on Vin-Gnd. Inductor and diode go sideways from switch out pin - can be about zero lead lengths. End result close to PCB. Note diagram in Nat Semi datasheet that shows heavy current paths. Copying their star wiring picture to some extent will help - but probably OK as is. Ensure diode is properly rated. Schottky best. – Russell McMahon Jul 20 '11 at 15:22
7

At that frequency it will probably work, yet radiate like hell, have low efficiency and bad ripple rejection. None of it will be relevant when you move it to a PCB. And due to its bad performance I wouldn't use it to power a circuit with it, but rather stick to my bench power supply. You can only use it as a proof-of-concept, if you feel you need that.
Personally I would skip the breadboard altogether and directly go for a PCB.

stevenvh
  • 145,145
  • 21
  • 455
  • 667
4

I tried running a small 5V inverter (-5V from a +5V supply) on breadboard.

This is basically a small low-power switcher in a chip with only a couple of resistors, capacitors, and a single 47µH coil (best results I found was with a toroid I wound myself).

While it worked, it really was noisy as hell. It radiated right across the board inducing a high pitched squeal into all my op-amps.

Not nice.

Majenko
  • 55,955
  • 9
  • 105
  • 187
  • 1
    Do you think that the breadboard was a major factor in the noise radiation? If the same circuit has been on a PCB and connected at the same points would the noise results have been substantially different. – Russell McMahon Jul 20 '11 at 12:18
  • On a PCB you can build in such things as guard rings, and place components so tracks are very short, etc. You can't do that on breadboard. – Majenko Jul 20 '11 at 12:33
  • 1
    ... unless you use solder and heavy wire for the power loop and breadboard the insensitive stuff. Tricky, but possible. – Mike DeSimone Jul 20 '11 at 12:55
3

Might be kind of noisy, and I wouldn't try and pull a lot of power through any breadboarded power supply, but I don't see why it shouldn't work. I'd certainly give it a go if I felt the desire.

Jay Kominek
  • 632
  • 5
  • 11
  • 1
    By "a lot of power", what are we talking here? The PS I'm designing has a maximum current of 2A at 5VDC, but in practice I'd be pulling a lot less than that through it (~100mA-500mA) – Mark Jul 19 '11 at 00:16
  • 1
    Having finally dug up a data sheet for a random breadboard, it says it's good to 36V @ 2A. My concern was initially for the current you'd be drawing, but it sounds like you've got plenty of margin there. If you're dropping 35V, your input voltage might be problematic. Some random board might be rated for even less voltage than the one I found. (Not sure how safety pedantic EE.SE is, but...) Worst case you probably melt some plastic and ruin your board. I'd try it on one of my smaller boards. – Jay Kominek Jul 19 '11 at 01:10
  • 1
    @Jay - We're not too safety pedantic here. Your comment that you wouldn't try to pull a lot of power through a breadboard is sufficient, digging up a datasheet is excellent. I'd not have guessed they were so low voltage! – Kevin Vermeer Jul 19 '11 at 02:05
1

Designing a switcher on a breadboard just makes life harder. It can be done (see the other answers), but why make work for yourself?

Basically, a breadboard adds capacitances of tens or hundreds of pF between all adjacent nodes. (Think about it: the contacts in two adjacent rows are the plates, and the plastic in between is the dieletric.) The large parallel surface area of the contacts is the killer here; on a PCB, parallel traces just have edge ("fringing") capacitance to deal with, which is much lower, and capacitance to the next plane layer down (usually ground) which is easier to predict and deal with.

I would recommend instead that you look at so-called "plug-in power modules", such as those sold by TI, that have circuit boards integrating all the good stuff in the first place, and just require input and output capacitors and some other small parts (like a resistor to set output voltage). They're a lot less of a pain to get going.

Even if there isn't an available module, you'd still be better off making a small, 2-side no-solder-mask PCB (about $100 for 10, or you could try an aggregator like DorkbotPDX) that has just the power supply, and has pins on 0.1" centers (bus wire works well here) for interfacing to your breadboard. The great thing about this is that you can reuse this power board on the real design, as well as future projects.

(On my list of "things to do when I take over the world" is making some base circuit boards for Linear Tech's µModule regulators, so you'd just buy the board with the µModule and 0.1" center pins on it, add the needed capacitors and resistors to it, and voila, power supply.)

Mike DeSimone
  • 4,705
  • 17
  • 22
  • µModule seems quite expensive: $7.78 each when buying 100 of the dinky little 500mA chips. I could build a 15A buck smps with a $3.648 BOM – Navin Jan 23 '20 at 06:37