Voltage rating of LM317

: > >

: > >So the options are : multiple series diodes, a different : > >transformer, switched-mode regulation, or a heavy heatsink. : > >Comments please ? : >

: > I already put in my vote for a few extra diodes .... : >

: > Heavy heatsink solves a thermal problem, not a voltage differential one which : > was the thrust of your post. : : Uh-oh. I was careless with my wording there. I considered : the voltage differential issue already solved or clarified. : That is, the heavy filtering at input and the resultant : slow rise time would permit the use of an LM317, and the : problem doesn't arise anyway if I use the HV version. : : >

: > Switchmode is obviously a workable alternative, but is going to require more : > design effort for a one-off device than is usually warranted. : : This is a one-off project, true, but under rather unusual : circumstances. I live in a remote region where my clients : bought an expensive piece of equipment made by a giant : multinational company. In equivalent economic terms, it : would be something like an investment of half a million : dollars in the US. Unfortunately, it was heavily damaged : by a power surge during a storm. : : The company engineer (who flew in from a distant city) saw : some of the work I'd done earlier for the same clients, : and told them that it would be better for them in the long : run to let me redesign and build every non-mechanical : system from scratch. The power supply is only a tiny : portion of the whole thing.

I haven't read all replies so excuse me if this is redundant or irrelevant, but have you considered using two 317's to split the voltage drop between them? :

The reason for not including the zener option is simple - I forgot to, when I made the list :). Then there's the matter of availability of a high-power zener diode too.

Also consider practical design parameters. Worst-case input on the low side will be when line voltage is low and the transformer-rectifier is fully loaded. Without making detailed calculations, I estimate the DC input to be around 24V then, which allows a max drop of 9V by the resistor (for an input of 15V to the 12V regulator). This works out to 15 ohms with a dissipation of 5.4W at 0.6A (a 10W resistor will do).

Now take the other extreme case where the DC input rises to 45V at light load and abnormal line voltage. If I use a 39V zener, the resistor will briefly pump 0.4A into it at startup. I wonder if that's advisable even with, say a 5W zener diode. And I'm not sure I can get a 39V 5W zener.

OTOH, consider the case where the regulator is fully loaded to 0.6A, but the other load is off, AND line voltage is high. Estimate the regulator input at about

Another option would be a 9V 10W zener in series instead of in parallel with the regulator. There's the availability issue again and it will have to be heatsinked too.

I could lower the input with a transistor voltage divider

- say a TIP122. But that will need a heatsink too and the dissipation will simply be distributed between regulator and transistor without being reduced. The main advantage will be that a plain LM317 can be used.

I think that's a clumsy approach that can be avoided by using a 317HV and a good heatsink, or a separate transformer of lower voltage. I sometimes use a center- tapped 9V transformer which is easily available, and use it as an 18V unit at reduced current rating.

Thanks for the input. No, I hadn't considered that option. It's certainly a possibility. Since the input can drop to as low as 24V, maybe a little less, the first regulator will have to be set for an output of no more than 20V.

For stability, NatSem recommends a minimum of 1uF Tantalum or 25uF Al at the output. This again necessitates ensuring a slow input rise time or other protection scheme for the first regulator.

The voltage divider transistor I mentioned in the previous post takes a similar approach.

It is time that you do make some detailed calculations. You started out asking a question with an unreasonable change in the DC at Vin to 45 volts. First of all, that won't kill the 317 based on Vin-Vout. Nor will the concept of the uncharged cap on Vout be a concern as has been shown and particularly with the resistor in there. But a line voltage variation that would yield Vin of 45 volts would be way more than normal worst case design, and would strongly suggest a different approach. Your calculations will show that. You'd need the line voltage to rise from 120 to ~146 to get 45 volts at Vin. That's over a 21 % variation in line voltage, and if your equipment must function in that kind of environment, a basic linear regulator is not your best choice.

You misunderstand. The zener goes across the 317.

Fine. Use the HV unit.

Ed

I appreciate all the feedback and enlightening discussion that expanded from my original question. Please remember that I - and the equipment - have to work under primitive conditions. Surges of +20% do happen, though not an everyday occurence. (We use 230V mains here).

I'm including a mains window circuit and VDRs at strategic points, but I want everything to survive a momentary spike in mid-operation before the protection circuit has time to kick in. OTOH, sustained dips of more than -20% are common, and the equipment has to keep on working.

Yes, a linear regulator is not the best choice from a technical point of view, and I may ultimately go for a switched-mode supply. But I was exploring the feasibility of using the much simpler linear version.

I do understand that the zener goes across the 317. When the output is momentarilry shorted to ground, the only thing that limits the current is the series resistor. I = (Vin - zener V)/R. What I didn't think of was that a part of that current will be shunted by the regulator.

another method is a transistor/zener pre-regulator (its in some LM317 datasheets, or at least used to be). that also works when the output is shorted (OK the shunt zener can too, but it gets trickier).

from there you can easily make the pre-regulator drop a constant voltage across the LM317, or, say, one half of the differential voltage.

the disadvantage of the transistor is that it has no thermal protection.

using a pair of regulators is therefore a better choice. a LM317 (or

the single LM317HV solution has a lot less parts, but will have higher peak temperatures, may well be more expensive, and is probably harder to get.

regardless, I'd add in the various reverse-protection diodes, and ensure it can cope with a continuous output short-circuit.

a historic solution: you could drive the unit with a Constant-Voltage-Transformer (Also Known As (AKA) a Ferro-Resonnat Transformer)? assuming you can find somebody who still knows to wind you one....

Cheers Terry

The transistor-zener pre-regulator idea provides a solution similar to the transistor voltage divider I mentioned earlier.

The use of two regulators in tandem to distribute power dissipation is an important point. In fact, I overestimated the thermal capability of an LM317 (plain or HV). I assumed that a TO-220 package would be able to handle the 10-15W wasted power quite well with a heatsink.

But on closer inspection of the specs, I found that the j-c thermal resistance is fairly high at 4 deg C/W. Making it dissipate up to 15W with a reasonably-sized heatsink would push the junction temp closer to the max rating than I'd like.

An alternative would be to let a parallel transistor take most of the load current. Another would be to divide the load into two or more sections, and provide separate regulators for each section. And then there's always the switch-mode option.

What made me reluctant to abandon the idea of a linear regulator based on the LM317 is the elegant simplicity. (And yes, protection diodes are a must, of course). But if I'm going to have to use multiple power devices and a more elaborate circuit to distribute the voltage or current, then I might as well go for a switch-mode design. It will probably be best to S-M down to about 15V, and provide further regulation with a 7812.

The CVT option is technically interesting, and I could do the designing and winding myself, but I'd rather wind a switch-mode transformer (done it a few times) and go for the switch-mode option.

I think using something in series to drop the voltage and dissipate some power is a useful idea if sourcing a 317HV is difficult, but as you don't need regulation as such from the first device I think a voltagee regulator with heatsink would be overkill. A power resistor of 10-20 ohms would drop enough voltage at high currents - including the switch-on transient when the 317 will be 'wide open' - and at low currents the heat problem is minimal and the output cap has charged so the voltage differential is no longer a problem either. A 10-15 W power resistor would be cheaper, smaller and require less assembly than a second regulator with heatsink. You might still need a heatsink on the

Tim

yep. thermally it is better to organise the power to be evenly split, assuming equal Rthetajc.

if you use a transistor with a much lower Rthetajc, it should carry the bulk of the dissipation, which a bjt-zener circuit does nicely by returning the zener to the output (thus maintaining a constant voltage across the regulator)

thats pretty high. so I went and checked a whole bunch of regulators, and its about right. yuk. I was expecting about 1K/W....

so I went looking at some TO-220 transistors, only to discover none of my datasheets even list Rthetajc! how un-useful. so I looked at a few FETs, and found 0.5K/W for TO-220. thats a bit better!

the smps will be a lot more complex, and generate truckloads of noise (unless you are a smps expert, in which case its piss-easy).

Cheers Terry

Not an expert, but I think I can manage. It won't be the first time. Yes, noise always has to be considered. Output noise can be suppressed, but shielding EMI radiation could be a problem with the materials I have access to.

if you know how to do a good layout (and if not, read linear techs AN47, and/or M.J. Nave's "book power line filter design for SMPS"). make sure the inductance of any high dI/dt loop is as small as possible. the PCB contribution should be made negligible c.f. the magnetics, which should be minimised.

DONT use bobbin-core inductors (the shielded ones are bad; the un-shielded ones are essentially unusable if you care about EMI). gap only the centre-leg of the core, or use an appropriate koolmu/mpp toroid. and of course pay close attention to cap ESR, ripple current & lifetime.

Cheers Terry