what's wrong with this circuit?

He's right, it's not a good circuit. The current-sharing emitter resistors will degrade the regulation. The regulated point is their base junctions.

3-terminal regs have an internal feedback ararngement that regulates their output pin voltage, meaning anything after that is effectively unregulated. As a further point, look up the hFE of a 2N3055. It seriously sucks as the collector current rises. Three x 2N3055 at 20 A is gonna take a LOT more drive than a (non-moose) 7805 will provide.

max hFE of 2N3055=20

non-moose?

I agree with all that Budgie has said, in principle. However, it might be a case of horses for courses. As to whether the circuit is any good, will depend on how critical of voltage sag the circuit that you are using it to drive is. The three 0.22R current sharers in the emitters add up to a total of about 0.07R, so at 10A drain, you are looking at a voltage drop of about

0.7 volts across them, which may not be any great shakes for the driven circuit. However, as Budgie says, the transistors themselves may well be the limiting factor as to usefulness in this regard, because the voltage drop across them will increase substantially as the current drawn increases, due to their poor high current gain. Again, this is not a great problem if you are looking at a steady current requirement by the load - you just turn the sucker up a bit to maintain the output voltage that you need. Problems start when the load is dynamic.

An LM317K would be the choice for the adjustable regulator, as this would be able to supply the drive current needed. I would treat the circuit as maybe a useful primer to build if you're just getting into this sort of thing. It would be a good basis to play with, and to illustrate the principles involved, and you might even be able to modify it to employ some output feedback

Arfa

This is similar to the circuit in the Texas Instruments LM317 datasheet. Although not specified, I expect this is good for up to 5 A or more depending on the actual voltage difference between input and output and the size of the heat sink used for the power transistor, Q2.

+-------------------.C E.-------+ | Q2 _\___/_ | | 2N3055 | | | (NPN) | | | | R5 | +---------.E C.------+---/\/\---+ | Q1 _\___/_ 500 | | 2N2905 | | | (PNP) / R4 | | \ 5K | | / | | R3 | I +-------+ O | 1N4002 Vin (+) o---+-+---/\/\---+---| LM317 |---+----+--+------+-------+---o Vout (+) | 22 +-------+ | | | | | | A / _|_ | | | | \ R1 /_\ D1 | | | | / 120 | | | _|_ C1 | | | +_|_ C2 / --- 10uF +-------+---+---+ --- 47uF \ RL* | | | - | / | \ R2 +_|_ C3 | | | +->/ 5K --- 10uF | | | | \ - | | | | | | | | | Vin(-) o------+---------------+--+-----------+----------+-------+---o Vout (-)

Since the regulation point is at the emitter of the 2N3055, it will have the stability of the LM317. Multiple pass transistors can be paralleled with small emitter resistors for current equilization without sacrificing regulation. Or, MOSFETs can replace the 2N3055(s) without worrying about equilization.

--- sam | Sci.Electronics.Repair FAQ:

Repair | Main Table of Contents:

+Lasers | Sam's Laser FAQ: | Mirror Sites:

Important: Anything sent to the email address in the message header above is ignored unless my full name AND either lasers or electronics is included in the subject line. Or, you can contact me via the Feedback Form in the FAQs.

Forget hFEmax. Look up the hFE at 5 or 7A, and you need to design on the worst case (min) figure at Icmax.

There was a 5A variant of the 78xx regulators, and it was named "Moose" by IIRC National. But do the sums at 20A and see how much base current those critters want. They may be ubiquitous but that doesn't make them a good choice. There are far better bipolars for this type of application.

"Peter" hath wroth:

Argh. That's an awful design. The lack of voltage ratings on the output, xformer, and cazapitors should be a clue.

1. The output voltage isn't regulated. The feedback stops at the base of the 2N3055 xsistors. The regulation is further mangled by having 3ea 22 ohm resistors in series with the output. The voltage drop across these resistors will vary with the load current, which is not exactly my idea of voltage regulation. Ummm, doing the math: E = I * R = 5A * 7 ohms = 35 volts The 7 ohms is from 3ea 22 ohms effectively in parallel. In order for this piece of junk to output any voltage at 5A, the regulator input voltage will need to be 35 volts plus the regulators dropout voltage and Veb. It won't at all work with the 22 ohm resistors.
2. In normal operation, the 22 ohm resistors will cook. For example, if this power supply really does mangage to output 5A then the dissipation in the resistors will be: P = I^2 * R = 5^2 * 7 = 175 watts. Not even close.
3. The output isn't current limited or protected. Short the output to ground and you get 1A (output current limit of the 7805) through the base-emitter junction of the 2N3055. The usual value for such emitter follower load equalizing resistors is prehaps 0.22 ohms. Were those used in this design, the current would smoke the EB junction on of the 2n3055's.
4. The diodes listed are all plastic parts that will get rather hot near their current limits. For anything over about 8 Amps, one really should use diodes with heat sinks.
5. Using a full wave center tap xformer means that the xformer will be twice as large as a smaller xformer with a diode bridge.

Start over.

--
Jeff Liebermann     jeffl@comix.santa-cruz.ca.us
150 Felker St #D    http://www.LearnByDestroying.com
Santa Cruz CA 95060 http://802.11junk.com
Skype: JeffLiebermann     AE6KS    831-336-2558

Many thanks, guys for all the responses. They've helped clarify my thinking.

Peter

There is a decimal point on the schematic. That's zero point two two ohms. That makes the Rt 0.07R. That makes the voltage drop across them, at 5 amps,

0.35v , not 35v ... How do you arrive at a maximum output of 1A ? Whilst the 7805 can deliver a maximum output of 1 amp, this figure is multiplied by the current gain of the 2N3055's. Assuming a ( poor ) gain of 20 on them, that would result in an output current of some 20 amps with 1 amp of drive to their bases. That is the whole point of having external series pass transistors. The output current is drawn from the collector-emitter circuit, not the base-emitter circuit, so there is no reason why the base-emitter junctions should fry.

The output is current limited and protected by the fuse feeding the series pass transistors' collectors.Admittedly, this is not very elegant, but it is protection, no matter which way you look at it.

It doesn't make any difference to the size of the transformer, if you have one 20v 20A winding, or two 20v 10A windings, series'd and grounded at the junction. It's still 400vA either way

I agree that this is not a *good* design, and will suffer from poor dynamic regulation due to the current-dependant drop across the current sharing resistors, but it is at least functional, and a simple useable design to produce an adjustable, reasonably high current output. Depending on what it is needed for, it might be quite adequate, and its shortcomings, of little or no consequence.

Arfa

"Arfa Daily" hath wroth:

I'll take your word for it that there's a decimal point. It looks more like a blur or a smear to me. In the USA, the decimal point goes at the base line of the lettering. Other countries have it in different places. I guess this one is in the middle. Anyway, as I indicated in my analysis, 0.22 ohms is a more reasonable value.

The 1A was my guess as to the short circuit output current with the fuse blown. Short the output to ground with the fuse blown and the current gain doesn't even enter into the picture. 100% of everything the 7805 can deliver will go through the base-emitter junction.

With the fuse not blown, the output current will go to whatever the xformer and diodes output, minus 0.6V for emitter-base, and whatever drop is across the 0.22 ohm resistors. That should blow the fuse, followed by blowing the emitter-base junction.

Of course, the 7805 has current foldback, which will limit the amount of current that it can supply to safely reduce dissipation. It may take a while to blow up. Meanwhile, it will probably oscillate merrily.

Not if blowing the fuse also causes the emitter-base junction to blow up from excessive current. I could supply a better analysis if I knew the values of the xformer voltage and current and what value of fuse is specified.

Interestingly, the range of output voltage is rather odd. Using the

7805 example, the 270 ohm resistor and 5K pot form a divider with: 5V * 5000 / (270 * 5000) = 4.74 V Therefore, the maximum output voltage is 9.74 V (minus the Veb drop in the 2N3055. I'll call it 11 volts maximum output. The range of output voltages is 4.4V to approx 11V.

At full current (5A), the three 0.22 ohm resistors appear as a single

0.7 ohm resistor for a drop of 1.1 volts. Therefore, the output voltage will vary over a range of 0 to 1.1 volts depending on the load current. This is not what I would call good regulation. It's 4 times worse at 20Amps.

The 7805 is not an LDO regulator, so we'll need a few volts drop across it. My guess is about 5 more volts. Therefore, the xformer and full wave center tapped bridge need to supply 32V center tapped at

5Amps. That's a fairly large xformer. At 20Amps, it's a fairly huge transformer.

With full wave center tapped, you're only using one half of the xformer secondary at a time. Therefore, each *HALF* of the secondary has to supply the full current and full voltage for half a cycle. Half a cycle later, the other half of the xformer is doing the work, while the first half just sits there. To supply my calculated 16 volts of DC from the full wave center tapped system, each *HALF* of the secondary would have to supply 16VAC at 5Amps for a rating of 32VAC CT at 5Amps.

To do the same thing with a full wave center tapped arrangement, the entire secondary is used each half cycle. Therefore the xformer rating would be 16VAC at 5A or half the size. Ignoring slight efficiency differences, and a larger physical size, the xformer rating for both devices would be about 80VA, but the center tapped version would be about twice as physically large due to the doubling of the secondary windings.

One more. At 20Amps, 4700uF is inadequate filtering. I'm too lazy to do the numbers. It needs a series resistor or choke.

I beg to differ with you conclusions. The design is unsafe, has no short circuit protection, may oscillate, uses an inefficient xformer design, has improperly selected diodes, has miserable voltage regulation, and will blow up the 2n3055's if the fuse is removed or blown. Since the application has not been specified, neither you nor I can judge if the design is adequate.

--
Jeff Liebermann     jeffl@comix.santa-cruz.ca.us
150 Felker St #D    http://www.LearnByDestroying.com
Santa Cruz CA 95060 http://802.11junk.com
Skype: JeffLiebermann     AE6KS    831-336-2558

Apologies if the schematic was not clear - on the original it's .22 ohms ie 0.22 Ohms

On Sat, 6 May 2006 14:03:45 +1000, "Peter" put finger to keyboard and composed:

Remembering that this is a variable power supply, ie one where the voltage is trimmed to suit the load, and assuming that the load is reasonably constant (eg 10 +/- 1 amps), then the regulation error in this case would be +/- 70mV. This is not too bad at 13.8V, say. OTOH, a bursty load such as amateur radio equipment will be poorly regulated. Having said that, I doubt that this would matter much to most radio hams. In fact a friend had a simple design based on the same pass transistors and an LM723 regulator IC. The transformer was massive, though.

- Franc Zabkar

--
Please remove one 'i' from my address when replying by email.

Jeff Liebermann hath wroth:

Oops. I added instead of subtracted. That's what happens when I try to eat dinner in front of the computer. It should read:

Therefore, the maximum output voltage is 9.74 V (minus the Veb drop in the 2N3055. I'll call it 9 volts maximum output. The range of output voltages is 4.4V to approx 9V.

--
Jeff Liebermann     jeffl@comix.santa-cruz.ca.us
150 Felker St #D    http://www.LearnByDestroying.com
Santa Cruz CA 95060 http://802.11junk.com
Skype: JeffLiebermann     AE6KS    831-336-2558

*** That's interesting that you only see the decimal point as a blur. Are you by any chance using an LCD monitor in non-native resolution ? I say this because the point is perfectly sharp on my CRT monitor, and there has been quite a debate going on here from time to time about the relative merits of CRT over LCD when it comes to fine detail, and photographic colour rendition. Over here, the decimal point went in the centre when I was a kid, but by the time I was at senior school, it had been moved to the bottom. This is one reason that we were always taught to preceed any such value with a zero. Over here now, the decimal point tends to be omitted so as not to cause confusion with spots of fly-crap on the page. Thus, a 0.22ohm resistor, would be written as 0R22 or just R22. 22 ohms would be written as 22R. Likewise caps - 4.7uF = 4u7 6.8nF = 6n8. A better form of nomenclature, I think. *** Ah, OK ! I see where you're coming from now. Valid point - in theory. However, my data shows a max base current for the device of 7 amps. Bear in mind also, that as long as the current sharing resistors have not gone open, any short circuit current drawn, will be spread among the three 2N3055s. As the 7805 can only supply a maximum of around an amp before going into overcurrent foldback, this will be only represent around 330mA per transistor base. They should be able to handle this all day, without so much as a whimper, let alone a grunt. So yes, I agree that the circuit may well produce around an amp of short circuit current, but I do not agree that this will blow out the 2N3055s' B-E junctions. *** There is probably just about enough decoupling around to stop it oscillating under fault conditions, but had it have been my design, I agree that I would probably have put a bit more in. The regulator, properly heatsunk, should be able to cope with overcurrent foldback, which keeps the device within its SOA, all day. *** I don't believe that the B-E junctions will blow - see above

*** The use of a 7805 nominally fixed regulator is, I agree, an odd choice. I would not recommend attempting to use one of these in a variable configuration. However, the alternative LM317 specified is, AIR, a genuine adjustable regulator, which goes down to its internal reference voltage of 1.2v, and up to around 35v, so by the time you had factored in drops in the series pass element, you would get down to nearly zero output, and up to somewhere near what you were putting in. *** Slight error in the math there ! 3 x 0.22 in parallel, is 0.07 ohms, not 0.7 ohms, thus at 5 amps, the drop across them is 0.35v, and about 1.4v at 20 amps *** Yep, I'll conceed that one !! Your thinking is much clearer than mine. In mitigation, I'll just say that it was in the very early hours that I was sitting here thinking about it ... !! ( but still no excuse ... ) *** That may or may not be true, depending on the application. Many loads will not mind a dirty output. I agree that if it were my design, I would probably put better filtering in, with larger caps, but there will be a degree of electronic smoothing achieved, even with this poor design, by the basic 78xx or LM317 regulator element. AIR, these devices exhibit around 70dB of ripple rejection. However, I wouldn't suggest that this circuit would achieve anything like that figure, because any good regulation or ripple rejection at the bases of the series pass elements, will be worsened by a factor of their gain. *** I don't really think that the design is fundamentally unsafe per se, and I reject your contention that the design has no short circuit protection - see above. It may oscillate under the right ( wrong ?? ) conditions, but I think that this is fairly unlikely, given that there are decouplers in the right places. I agree with your analysis of the diode and transformer specifications. I agree that the potential voltage regulation is poor compared to some other designs, but not necessarily, that it falls into the "miserable" category for low to medium demands. I dispute that it will blow the 2N3055s if the fuse fails or is removed - see above. I agree that we cannot judge the design in terms of specifics, without knowing the ultimate requirements of it, nor was I trying to, but that does not preclude judging its validity as a standalone circuit, suitable and adequate for *some* applications. Actually, if you think of it more as what it is - a variable voltage source - rather than what it's not - a properly regulated power supply, then it has many valid applications supplying non-critical loads. A few that spring immediately to mind are minidrill speed controller, model railway speed controller, sump pump speed controller, garden pond pump speed controller, low voltage lighting intensity controller, gel battery charger and so on.

Arfa

"Arfa Daily" hath wroth:

Dell P1110 19" viewable flat screen CRT. I can't afford an LCD monitor at this time.

I'm familiar with the merits of this discussion. I usually recommend that customer purchase 1600x1200 LCD displays so that running the display had 800x600 does not result in a pixel roundoff error.

Agreed. That was my problem. In the USA, the decimal point is always near the bottom of the line. I thought the spot was a smear. Please note that literally *ALL* my previous posting have the leading zero to avoid confusion. 0.22 or 0R22 would be fine. R22 is not because it is easily confused by the resitors reference designator. If the common usage has been anything other than the letter "R", I would also do it this way, but I've had far too much trouble with mixing values and reference designators on schematics than losing a decimal point. The capacitor examples might work as it's unlikely to confuse U47 (an integrated circuit) with 4U7, a cazapitor. I've done it many different ways, as demanded by various companies drafting standards.

Y'er correct. I should have looked at the data sheet first. I didn't realize that the 2n3055 can handle 7A of base-emitter current. 333mA is not going to blow up the 2N3055's. However, I find it a rather bad design that can blow the fuse, and still produce substantial output current.

I've dealt with my share of 3 terminal regulators in various designs. Depending on the construction, load impedance, lead lengths, etc, I can sometimes make them oscillate. 1uF is not enough. I usually use

0.1uF in parallel with 10uF.

I agree. It will not blow up.

Agreed. LM317 would be a better choice. It will go down to approximately 0 volts.

Oops. I hate it when that happens. That's not as horrible regulation as I miscalculated. However, it's still not as good as it might be if the 2N3055's had been inside the feedback loop.

Well, I made more mistakes than you have. My excuse is that I was constantly getting interrupted while writing my reply and didn't have time to do much more than a spelling check.

The LM317 will certainly help with the ripple reduction. However, the ripple will appear at the collectors of the 2n3055's which does not have as much ripple rejection as the LM317. The ripple won't be huge, but it will be present.

It's a bit more important than just minimizing the ripple. There's the problem of ripple current. The input filter capacitor conducts lots of current on each half cycle. I once repaired an Astron 60A linear DC power supply. The problem was that running at almost full load, the single 250,000uF 25VDC (not sure of values) had a sufficiently high ESR that the screw terminals literally melted on the capacitor. At 5A to 20A, this is not a problem. However, even at

20A, the 4700 uF capacitor might get hot. The LM317 will clean up any ripple that's left, but the cazapitor still has to supply the power between cycle peaks.

I do. Any circuit that continues to supply power after the fuse blows, and that does not have short circuit protection, is in IMHO unsafe.

Ok. Turn the LM317 version to full output at perhaps 25VDC. Now, short the output to ground. Will the fuse blow? Maybe depending on the value selected and whether the xformer can supply the necessary power. The fuse really belongs in series with the output voltage (as well as adding a fuse on the 117VAC input). When the current goes to the limit, it is sure to blow because the 2n3055's will supply the necessary power to blow the fuse. It's not so clear whether the fuse as shown will blow. If the LM317 goes into current foldback protection mode, it won't blow. The output voltage will also drop to about Vbe plus whatever the LM317 outputs. However, put a big filter capacitor load on this thing, and it will take time for the LM317 to complain. There's probably enough power left during this time to blow the fuse. I can't really tell without values or bench testing (or modeling). I guess you could call this load dependent protection.

I can make it oscillate with an inductive load. The 1uf is strictly for improved transcient response, not filtering or stability.

Agreed. It's not as horrible as I thought. It's still quite bad and could have been done much better.

Agreed. The 7A of base-emitter current is sufficient to prevent destruction. I assumed a much lower value.

All of these will function except the gel cell battery charger. Charging a gel cell battery is not a trivial exercise and should not be done with an ordinary power supply. Go over voltage for just a small amount of time, and the battery is toast. The other applications are not particularly critical and can probably tolerate such a power source.

--
Jeff Liebermann     jeffl@comix.santa-cruz.ca.us
150 Felker St #D    http://www.LearnByDestroying.com
Santa Cruz CA 95060 http://802.11junk.com
Skype: JeffLiebermann     AE6KS    831-336-2558

What the hell are you all talking about ? Using a linear regulator is nuts, even for a mildly competent amatuer.

Get yourself a coil and design a slicer circuit. If you like American transistors get your hands on a 2N3773 and throw those 3055s out. We don't have any aluminum trees here anymore. We can't just go out and pick heatsinks in the backyard.

If nobody in here can design a chopper circuit, I will do it. I won't even use pi, in fact I think I can do it without a calculator, but don't hold my feet to the fire on that particular issue.

Chopper circuits are the bomb, they can actually put out more current than they drain. Linear regulators are a thing of the past. Of course I know that manufacturers still use 78XX series regulators in products, but there is a different reason for subregulating a supply. That is a different subject.

If you don't want to design use a upc494 and put your "fingers" in the feedbacvk loop. It can automatically give you max on times etc, and IIRC current limiting, but you could design that with any chip, once into the feedback loop. I think you don't even need a chip, unless maybe an OPAMP or comparator.

Here's cheap and dirty, let the circuit you have oscillate, put a coil in the collector of the final outputs. Let it settle on any frequency it likes. Add one coil and remove most of the capacitors is all that's needed. If it runs hot get a final with more gain. Even to do this though you wil need to modify the circuit going to the base. Just assure it turns on and off quickly and completely.

I'll get my pencil out later.

JURB