Regulated 1.4v supply.

[Boater Sam]

This weeks project, a mains power supply for battery valve radios.

The HT is done, simply a nixie tube driver board from China.
Taking 12v off a lead acid battery, fused, boost pcb adjustable from 90v to 120v as required.

The LT I want is 1.4v with protection to avoid blowing those precious valves.
I intend using a LM317T regulator with three forward biased silicon diodes across the output to prevent the voltage going over 1.8v to 2.1v powered by the same battery.

The question is will LM317T regulate down to 1.4v, if so what value resistors on the adjust pin? If not, can I reduce the output using series diodes?

Is the idea of three crowbar diodes a sound one? Should I have a series resistor in the output before the crowbar?

The 3 legged fuses are a bit out of my comfort zone.

Sam.

[Argus25]

Probably most IC regulators even LDO ones would be hard pressed to get such a low output voltage as the input voltage would need to be too low.

Probably just forward bias three silicon diodes to get about 2.1 V as the base reference for an emitter follower series regulator, the output from the emitter will be about 1.4 V , and have the diode current about 1/5 to 1/10 the load current. Then add a series fuse and those protection diodes you suggested in case the voltage jumps up due to failure of your reference diodes. The circuit will have a small negative temperature coefficient, meaning the voltage output will drop a little with heating about 2 X 2.1mV/deg C but that probably won't matter.

Otherwise you could make a super stable regulator with a voltage reference , an op amp and output transistor, but the extra complexity would hardly be worth it.

[mhennessy]

Yes, an LM317T would do this, providing it doesn't overheat. The thermal design is often the hardest part...

All an LM317 wants out of life is to ensure that it has a constant 1.25V between the output and adjust pins. We put a resistor between those pins - often 240 ohms - and that sets up a current that is close to 5mA. This current flows in the next resistor - the one between ADJ and ground. We need 0.15V across this resistor to give us 1.4V at the output, so using Ohm's law (0.15V/5.2mA), we find that we need 28.8 ohms. Call it 30 ohms.

Yes, the diodes should work. Test those separately from the LM317 though, just to save confusion. Or you could make a real crowbar, using an SCR. I think I'd install the SCR on the input to the regulator (needless to say, it'll be after a fuse), but that's just personal preference. The trigger input can be taken from the output - connect the output to the gate via a couple of diodes. Between gate and ground, put a low value resistor (100 ohms or less) and a cap to avoid false triggering (perhaps 10uF).

[G8UWM-MildMartin]

Also http://www.vintage-radio.com/project...y-set-psu.html has a handy table of resistors versus output voltage.

[mhennessy]

Note the slightly different results there because of 220R rather than 240R.

In practice, there's a fair amount of latitude with those values. The datasheet says that the minimum output current is 10mA worst-case*, so some people choose 120R as the resistor between output and adjust so that the two resistors alone are satisfying the minimum load current. In practice, if you can be sure that the LM317 will always be drawing at least 10mA, the resistors can be as high as you like. If you go too high, then the small current drawn by the adjust pin starts to affect how accurately the actual voltage matches the prediction.

* Typical operating currents are more like 3.5mA, so you could use 360 ohms and still meet the minimum load requirements - perhaps OK for a one-off, but you wouldn't risk putting that in production.

Of course, for higher output voltages, an LED is a convenient way to sink some of that current

[paulsherwin]

The LM317 does work, but it's relatively noisy and the valve filaments are very sensitive to this. The last battery eliminator I built used a 317 and I never got the supply rail completely clean, resulting in noise on LW. If I were doing it again I'd use a different regulator or even some 2*1N4001 + limiting resistor arrangement.

[mhennessy]

That's quite surprising to hear - the LM317 is a pretty quiet regulator amongst the "jellybean" types (certainly, it's quieter than the LM78xx types when you add the recommended decoupling capacitor across Radjust - though don't forget the diodes if you do!). It sounds like something was oscillating - perhaps a layout problem or a faulty component. Were you able to put a 'scope across the DC output?

[paulsherwin]

No, I didn't have a scope handy at the time. I don't think the PSU can have been oscillating though - there wasn't much there to oscillate, it was just a transformer, bridge, resistor, regulator and a few caps. It's not certain the noise was coming from the reg though as it used bog standard diodes rather than fast recovery types. I may also have used a particularly noisy 317 knowing my luck.

I would do things differently if building another one, but it's good enough for the application. It's inside a Sky Baronet. I may revisit it one day.

[mhennessy]

Yes, that sounds more likely, especially if it was a constant buzz rather than a squeal.

Any regulator has the potential to oscillate if not properly decoupled - an emitter follower is a Colpitts oscillator waiting to happen, and the high gain op-amps in there to provide regulation might well take off too. Relatively rare, happily. I'm struggling to remember the last time I saw it happen.

[paulsherwin]

It was actually more of a crackling noise rather than a constant frequency buzz. It was faint though, so it's difficult to tell, and there's always electrical interference around anyway. I have some UF4007s so will fit them when I get the chance.

[Diabolical Artificer]

I've used a LM317AHV regulated PSU to power heaters, it wasn't noisey. I used a crowbar with SCR. Putting it before the reg is a good idea Mark, didn't think of that.

There's a LM317 resistor calculator here Sam - http://www.learningaboutelectronics....calculator.php though it's easy to do on paper.

Andy.

[Boater Sam]

Lots of good advice here then. Thanks guys.

I run LM317Ts in 6V & 9v wall warts that I use for transistor radios, they seem not to be noisy even without large value caps on the output. I always shunt the 220/240 ohm resistor with a 1N4148 diode. I think it reduces the ripple.

Running off a 12v lead acid battery I should not have a noise problem.
I will put a 27 ohm resistor in the input of the reg to drop some volts off.

I have diodes, no SCRs, so I'll try that way of crowbar first.
Handy calculator Andy, 'scuse the giggle. You may be too young to remember the product & advertisement.

I just want to be sure that I don't blow a set of 25mA heaters.

Sam.

[Terry_VK5TM]

There's always the Texas Instruments LP38852 0.8V - 1.8v (1.5A) adjustable LDO voltage regulator.

[Argus25]

It is interesting the noise mentioned in post #6. I first came across troublesome noise output fron TO-220 regulators such as the 7812 back in the 1970's in audiometers which were being upgraded from older + ground circuits with germanium transistors to neg ground circuits with silicon transistors. The noise on the voltage regulator output was in the audio spectrum, like a broad band hiss. The thing is, because it originates from a very low Z source impedance (the regulator output) it can be very difficult to filter it off, even with chokes and large capacitors.

I guess in most applications this regulator noise is not too problematic, but it is interesting to hear that it could be audible in this filament application.

Still the simple circuit I suggested with a few diodes and a transistor is less noisy and I always wonder why a choice sometimes to go for an IC with many semiconductor devices it contains, is preferable to a simple circuit with discretes, though I guess with the IC you get free short circuit protection and thermal overload protection built in and it saves some planning/calculations.

[mhennessy]

The diode across the 220/240 ohm resistor is only needed if you put a capacitor across the other resistor (recommended as it reduces ripple considerably). It doesn't hurt to include it either way, but it's intended to provide a path for the charge in that cap to flow safely. While you're at it, it's a good plan to put a diode between input and output - again, to provide a path for current to flow should the input be shorted (unlikely, I know).

Perhaps stating the obvious, but if using a dropper resistor ahead of the 317 - not a bad plan - then include a decoupling cap between the input to the 317 and ground.

A good source for SCRs is scrap PC power supplies. In fact, most switched-mode power supplies include a crowbar.

[kalee20]

I've used an LM317 to regulate filament supplies. It's a good regulator! Ideal application, regulates down to 1.25V - but yes it's noisy.

Anything on the filament supply is effectively applied to the grid, so a few microvolts of hash really matters. Good-quality capacitors (radial polyesters or ceramics, 0.1uF right AT the IC) will suppress this to a large extent.

Accuracy of filament voltage is quite important for long life, so the sort of precision you get from an IC regulator is what's needed.

A crowbar or whatever across the output is definitely worthwhile, it only takes a slip while experimenting to zap the filaments (I fused a DK96 in such a way, the others survived).

[Argus25]

I have used an IC (LM309k) to supply the 5V filaments in an amplifier with UX171 and UX112 valves. In that case I could not detect or hear any noise. But it might not be as sensitive a situation as in a superhet radio with 1.4V valves. I will be interested to see if any noise is audible when it is finished, or if any extra moves like added filter components were required to get rid of the noise.

[mhennessy]

I will say that I have no experience of battery valve sets - it sounds like I should get hold of one to play around with

If an IC regulator is too noisy, how about recycling the output chokes used in switched-mode power supplies? These are designed to pass lots of DC current without saturating while dropping the minimum voltage possible, and are designed to attenuate at the same sorts of frequencies that we're discussing here.

Such a choke should be placed closed to the regulator to stop any noise before it can be radiated, and good quality film caps should be used either side of it (electrolytics might be a bit lossy at those frequencies).

The trouble is, as mentioned, we need the DC precision of an IC regulator here. The LM317 is a pretty quiet IC when the adjust pin is decoupled - quieter equivalents (e.g. LT1085) are quite a lot more expensive. I suppose you could try a TL431 adjustable voltage reference (again, found in most switched mode power supplies) followed by some filtering and then an emitter follower, but it won't be as well regulated as a 317.

Just another thought - just as a bit of context, let's remember that the HT is supplied by a switched-mode boost converter

[IanNVJ35]

I never got any setup to work satisfactorily when I tried a few years ago. I ended up using a Thurly Thandar linear power supply which really worked well - but needed wires connecting so not self contained.

[kalee20]

Originally Posted by mhennessy

you could try a TL431 adjustable voltage reference (again, found in most switched mode power supplies) followed by some filtering and then an emitter follower, but it won't be as well regulated as a 317.

The TL431 is a really useful IC. With an external pass transistor, it will be as good and accurate as an LM317. However, it won't go below 2.45V.

Originally Posted by mhennessy

Just another thought - just as a bit of context, let's remember that the HT is supplied by a switched-mode boost converter

Yes! Though ripple on the HT rail is nothing like as important as ripple on the LT rail.

So - use the Nixie tube supply to provide HT and a couple of AA alkaline batteries for LT (which are at least super-quiet) to rig up a temporary trial. If the HT supply IS too noisy (radiated RFI is likely to be the issue) then you might as well forget about LT till this is tamed!