Voltage overhead.

[davelectronic]

Just done some testing on the transformer with 50 watt lamps. The readings are from 1 - 6 lamps. No lamps 14.7V,  One lamp 14.5V,  Two lamps 14.4V,  Three lamps 14.2V,  Four lamps 14.1V,  Five lamps 14.0V,  Six lamps 13.9V,  I was quite impressed with that. Normally I use the the lamps for testing server modded power supplies, whilst 6 lamps is 300 watts, it comes out less on a meter. But I'm sure that at 6 lamps I was in the high twentys in number of Amps. My highest meter only goes to 20 Amps. But those reading above are one lamp added up to a maximum of 6 lamps. I could have gone past 500 watts in lamps I have, but by lamp 6 I didn't see the point. So it might be workable with an LDV regulator and a standard square package bridge rectifier ? What do you think.

[davelectronic]

Yes I might have overlooked the transistors drop across them....
I'm hopping a reasonable, maybe not to much voltage drop on AC test will be exceptable when the filtering is added. In other words as the AC didn't sag to badly, it should be ok with capacitance after rectification.

[Ian.M]

What was the current with four lamps?  If it was under 16A, then try to get a current measurement with five lamps.  Due to the burden voltage of the meter's 20A range, please read the transformer voltage while the ammeter is in circuit.

That should give enough data to determine the equivalent source impedance and run a SPICE sim to see how much ripple to expect at 20A DC load for various capacitance values so you can see how much voltage drop you can tolerate in the bridge and how that impacts the regulator headroom.

[davelectronic]

With 5 lamps it's taken my meters 20 Amps range into over range, but the voltage was 14.0V, with 4 lamps it's reading 17.67 Amps at a voltage of 14.1V steady. Not sure if this helps, but that's as much as my meter will read with 50 watt lamps.

[Ian.M]

That's great. 
No load was 14.7V, so the voltage drop @17.67A  is 14.7-14.1=0.6V.  This is consistent with the 0.3V drop for two lamps, and in the ballpark with the 0.8V drop for 6 lamps.

0.6V/17.67A is 0.034 Ohms and that's the equivalent impedance of the secondary.  (I say equivalent impedance because some of its transformed from the primary side - its *NOT* simply the DC resistance.) 

To a first approximation the transformer can be modelled as a sinusoidal voltage source 14.7V RMS with 0.034 Ohms series resistance.

I simulated it with LTspice, with a 100,000uF reservoir capacitor ans 20A load current and found that with ideal diodes you'd have 2.6V headroom for the regulator, however with the only 25A Schottky LTspice had a model for (MBRB2545CT) there was only 1V headroom.  At under 35,000uF reservoir capacitance the 25A Schottky bridge couldn't keep the ripple troughs above 12V.

A controlled MOSFET bridge with individual MOSFET Rds_on of 10 milliOhms, and  100,000uF reservoir capacitor would give you a much more respectable 1.8V headroom, considerably easing your regulator design challenge.

The last digit of the equivalent source impedance is pretty dodgy.  If you take it as 0.04 Ohms series resistance, it reduces the headroom with the above 25A Schottky to only 0.8V. I suggest re-measuring the loaded and open circuit voltages to 2 decimal places so it can be calculated more accurately.  You should also measure the mains voltage - even a 5% droop will make this supply impractical. 

[davelectronic]

Thanks for running that in simulation, I can lower my expectations to 12 Amps continuous and 15 Amps 50% duty cycle. This pertains to running an RF linear amplifier. If this transformer can run 150 Watt linear amplifier I'd be happy with that. I could drop a wire size from currently 12AWG to 14AWG, but I was hopping I could just make the grade with that many turns of cable on the secondary. I don't know what your attachment is, my tablet office app won't open it. I might be able to view it on my pc.
So my question is at 12 Amps, 12 Volts, could I get away with a standard bridge rectifier, and an LM2940-12 LDR regulator with a Volts drop of 0.5 Volts for the regulator. And 1.4 Volts for the bridge rectifier,  or do I have to go with schottky Diodes,  or the IC fets option.

[Ian.M]

The attachment is a LTspice schematic.   I'd drop a size on the wire to get more turns in. otherwise even with the MOSFET controlled rectifier option you'll have to spend far too much on caps to be worth it..

[davelectronic]

Forgot, although I'd like to stay as close to 12 Volts as possible, I can still operate the amplifier down to 11.40 Volts minimum. But the closer to 12 Volts would be better. And I can scrape by with 12 Amps continuous current.

[davelectronic]

Dam, dropping a wire size drives the cost up, I can only get this high quality silicone wire in 5 or 10 meter lengths. And the wire I unwind is £8.50 down the drain. If I go with schottky Diodes that's even more expense. Surely I can keep it above 11.40 Volts at the lower 12 Amps. Just asking what you think.

[Zero999]

Thank for the replys, now the iffy part...
It's a rewound 700 watt MOT with high temperature silicone wire secondary. The cable is rated for 45 Amps continuous current. I'm only looking for 15 - 20 Amps output, by way of series pass transistors.

Can you post a schematic of how you intend to connect the series pass transistors?

Normally adding series pass transistors to the output incurs an extra V_BE (0.7V to 1.2V) drop on top of the regulator IC's drop-out voltage.

This would be PNP transistors, so collectors on the output,and input the the emitters. And a Base resistor, also ballast resistors.

Like this but with a low drop-out regulator rather than the LM317?

https://www.eevblog.com/forum/beginners/linear-regulated-power-supply-with-lm317t-mosfet-50n06/msg656109/#msg656109

If so then, yes you lose another 0.7V to 1.2V, across R3 in the above schematic.

[davelectronic]

Yes that configuration, I can drop to a working voltage of 11.40 Volts and a continuous current of 12 Amps. I really didn't want to rewind it with thinner wire gauge if possible. With the current wire it's 12AWG, and I couldn't get another turn on it. Kick in the teeth if I've got to drop to 14AWG to get the turns count on there.

[davelectronic]

So do you think it possible to stay above 11.40 Volts with the windings I have on there now, that using a standard rectifier, and LDR regulator of 0.5 Volts.

[Ian.M]

Dropping to 12A max eases the problem considerably.  However you'll still need a lot of capacitance if you are using an ordinary silicon bridge rectifier and you want enough headroom to regulate.  Unless either you can give us a part number for the bridge you plan to use and it has a SPICE model available, or you are prepared to measure its voltage drop at a range of currents up to 20A (or its max rating, whichever is lower) so we can fit a model to it, we cant do a lot better than say "suck it and see!".

For those playing along at home here are a couple of high current Silicon diode SPICE models:

Code: [Select]

**********************************************
*SRC=1N1183;DN1183;Diodes;Rectifier >5A;50V 35A
.model 1N1183 D (IS=23.4N RS=1.75M N=1.61 BV=66.6 IBV=4M
+ CJO=4.4N VJ=.75 M=.333 TT=144U)
* Motorola 50 Volt 20 Amp 100 us Si Diode 11-23-1990
**********************************************
*SRC=1N1200;DN1200;Diodes;Rectifier >5A;100V 12A
.model 1N1200 D (IS=29.7N RS=10.6M N=1.70 BV=133 IBV=4.5M
+ CJO=494P VJ=.75 M=.333 TT=5.76U)
* Motorola 100 Volt 12 Amp 60 us Si Diode 11-23-1990
**********************************************
*SRC=MDA2500;MDA2500;Diodes;Bridge;50V 25A, Full Wave
* single diode from MDA2500
.model 1D_MDA2500 D (IS=1.52E-08 N=1.59 BV=5.00E+01 IBV=5.00E-06
+ RS=1.79E-03 CJO=4.69E-10 VJ=.29 M=0.4 TT=4.32E-06)
* Motorola 50 Volt 25.00 Amp 3.75 us Si Diode Bridge 07-08-1990
I've revised the simulation to add them, drop the load current to 12A, and the capacitor to 27,000uF.

Personally I'd strip some dead PC PSUs for high current Schottkys. Even if you have to use two separate packages for the negative side of the bridge because you are unlikely to find common anode diodes, I reckon it will be worth it to have the extra headroom and cooler running diodes,and not need to spend so much on capacitors.

[David Hess]

30 amp dual schottky diodes in TO-220 packages are less than $1 each at Mouser and 40 amp ones can be had for less than $2.

When I built a big linear power supply like this, I used a really big surplus transformer which had a 25 volt 30+ amp center tapped secondary.  So instead of a full wave bridge, I used a full wave center tap which only requires 2 diodes and therefore has half of the voltage drop.  The diodes I used were stud mount 1N3900s on a big heat sink which are not ideal but I have a drawer full of them from the original power supplies that the transformer came out of.  In retrospect, using half of the original massive bridge rectifier would have been better.

If you could rewind your transformer for twice the voltage and a center tap, then that will halve the losses.  It would be fascinating to make a full wave center tap rectifier with a pair of n-channel MOSFETs and comparators.

[davelectronic]

I could do shottky diodes, but had the bridge rectifier KBPC3510-G data sheet below. That drops 1.1 Volts. And the LM2940-12 LDV linear regulator drops 0.5 Volts. That's what I'd hope to use. The transistors are the MJ11015 high power Darlington type transistors. Volts drop across them on diode check is 0.6 approx from Base to emitter and Base to collector.

[james_s]

All you can really do is try it, rig something up using the parts you have and then modify as needed.

[davelectronic]

All you can really do is try it, rig something up using the parts you have and then modify as needed.

I think your right James, I think, only think mind. Under load with the above components, I'm going to get 11.80 - 11.90 Volts worst outcome. I could be totally wrong. But a drop of 1.1V for the rectifier, 0.5V for the regulator, and 1.2V across the resistor transistor network. If I got 11.40 Volts hopefully a little more, at 12 Amps I'd be happy with that.

[Ian.M]

Simming it with a silicon bridge rectifier, it doesn't look too bad with enough good low ESR capacitance but that bridge will kill your headroom.  Due to the supply current being concentrated near the peaks of the voltage waveform, I'm seeing about 45A peak through the bridge for 12A average DC current.  That's only 20.5A RMS so well within your bridge's ratings.  However, if you look at Fig.3 of its datasheet, on the 35A curve, the instantaneous forward voltage drop PER DIODE is 1.2V.  That means you are loosing 2.4V that you can ill afford in the bridge.  For comparison, the sim says the Motorola MUR2500 50 Volt 25.00 Amp bridge I have a model for drops 0.9V per diode.  A good set of Schottkys can save you about a volt of total drop in the bridge, headroom you badly need if you want to push above 12A, or when the mains voltage drops when half the country puts the kettle on in the commercial breaks in Corrie or at the end of East Enders.

Its looking like you need 20mF minimum capacitance to keep the ripple trough above 12.5V so you've got enough headroom to regulate.  However I'm seeing 16.7A RMS ripple current in that cap for a 12A load so you are going to need a bank of them to split the ripple current up to something more reasonable so they don't puke their guts at full 12A load current.  The best deal (price per mF) on decent quality caps I could find at Mouser was http://www.mouser.co.uk/ProductDetail/United-Chemi-Con/EKYB250ELL682ML40S, (25V 6800uF, 4.22A ripple current), and you'll need at least six in parallel to meet the ripple current spec.   Eight would give you some margin, but you might as well go the whole hog and use a bank of 10 as there's a price break at qty 10.

Recalculating with 68mF capacitance, using 1/10  the worst case ESR from the above cap, and assuming 0.05R wiring resistance between the bridge and the caps, it looks like you can maintain 1V headroom at 15A load current.   The ripple current is 24.6A RMS, well within their rating.

Unless you've got some server grade 25V caps hanging around, it looks like it will cost you £13.50 for the caps.  They are going to need decent busbars as well - probably the easiest option without getting a pcb made would be to mill a break in the copper of a piece of single sided copper clad, drill it for all the caps and sweat solder two heavy gauge solid coper wires to it to improve its current carrying capability.

[davelectronic]

Thanks for the info Ian, I've never used simulation, so no idea how that looks. But having watched a few videos on it, so have some idea. I can see how close it is, and yes i got the bridge rectifier wrong, it says 1.1Fv per element, so that makes sense. I might just rewind it, as much as it pains me to waste cable. The cable on there now is 12AWG with an outer diameter of 4.5mm, a drop in cable size to 14AWG is 3.6mm outer diameter. That would allow the more turns count, but i would definitely have to drop my expectation to 12 - 15 Amps then, not much of a big deal.

As its just a matter of removing bolts, I'm tempted to build it, try it, then rewind the transformer if its not executable with the 12 AWG cable. Might as well see.

[David Hess]

Unless you've got some server grade 25V caps hanging around, it looks like it will cost you £13.50 for the caps.  They are going to need decent busbars as well - probably the easiest option without getting a pcb made would be to mill a break in the copper of a piece of single sided copper clad, drill it for all the caps and sweat solder two heavy gauge solid copper wires to it to improve its current carrying capability.

I built a couple of power supplies like I described above and the stranded copper wire between the transformer and rectifiers was so large that it supported the flat heat sink that the stud package rectifiers were mounted on; that is not a good construction practice but it worked out.

The screw terminal capacitors were so large that the regulator board including the heat sinks mounted to the input capacitors instead of the reverse or having the input capacitors separate.

The regulator circuit was the one Hero999 shows but with an LM337 negative regulator instead of an LM317 positive regulator so I could use 2N3055 NPN transistors; it worked fine but if I did it again, I would use the configuration shown below from National Semiconductor's Voltage Regulator Handbook.

I added a fan directly above the pass transistor heat sink and made a simple temperature control circuit using an operational amplifier and the -2mV/C temperature coefficient of a diode epoxied to the heat sink.  This lowered the thermal resistance of the heat sink by an order of magnitude which is no surprise.  Since the error amplifier was an integrator, the voltage driving the fan was roughly proportional to power dissipated (the temperature sensor operated at a constant temperature) although I never took advantage of that; at nominal output current, the fan was completely silent.

The only issue I ever had was with the emitter ballast resistors which were not rated for enough power and not cooled by the fan.  At an unrealistically high output current for an extended time, they would glow dull red which was more alarming than a problem.

If I did a high power regulator like this again, I would add external controlled foldback current limiting to the protection already built into the integrated regulator.  I might make a delta Vbe based temperature control for the fan so no adjustments are needed for the offset which varies between diodes.  An SCR crowbar on the output would be a good idea to protect whatever is being powered from a pass transistor short; this requires a fuse which is probably located between the input capacitor and the regulator.

[davelectronic]

My circuit is pretty much the same, with out that diode on the regulators input. I've built a few now, all with multiple pnp pass transistors. The last one was a 13.20 Volt psu which can do 15 Amps, by way of a 7812K and 4 x MJ4502 transistors, its in use now on a cb radio linear amplifier, and power a couple of radio scanners. That one has a 300VA toroidal transformer in it with a secondary of 15 Volts. But reads AC 17.6V no load. Its 22 volts DC rectified and filtered, and 19 Volts at the bridge under a 12 Amp load. What I'm going to try first with this MOT and its 14.7V AC output, is the 35 Amp diode bridge, the LM2940- 12 LDV regulator, and 2 x MJ11015 transistors. Capacitors I'm trying with this will be snap in radial electrolytics 10000 each x 6 capacitors in a bank. If that can manage above 11.40 Volts at 12 Amps i will be happy with that. But closer to 12 Volts for the same load would be better. If it drops under 11.40 Volts with the above, then i will wind a slightly lower wire diameter on the MOT. Last one i built below.

[Ian.M]

While you are gold plating it, what about using a DC to DC converter or an auxiliary transformer or PSU to provide a boosted^* supply for the control circuit? Then you can use a LM723 with a simple NPN Quasi-darlington^# pass transistor so you can regulate right down to 0.5V input-output voltage differential! 

The LM723 makes it easy to implement foldback current limiting and load voltage sensing.  Protect it against sense disconnect by shunting each sense terminal to the corresponding power terminal with a 1N4001. 

*  You only need 5V extra headroom - a 5V 1A USB supply, riding on top of Vunreg would do nicely.

# Using multiple output transistors, with emitter resistors for current sharing, and feeding the driver transistor from the boosted rail

[davelectronic]

Now you really got my mind in overdrive...
I don't know about a dc to dc converter, but its got me thinking. So has an additional transformer for regulation with higher voltage for the control transformer. I'm thinking first which one to try, and a parallel configuration maybe. But how to implement the control transformer, so it regulates only, and all power is taken from the MOT.
I never gold plate btw, its to expensive. But you've sure got me thinking.

[Ian.M]

That's fairly simple.  The boosted rail feeds *everything* except the fan and the collectors of the paralleled output (pass) transistors. I You need 1A so you've got plenty of base drive for the output transistors (forced H_FE of 20)  2A would be better if you are using crappy low gain output transistors.   It will only draw what it needs from the boost rail till the unreg rail collapses to within about 0.5V of the output, at which point the boost rail current will start to rapidly increase due to the driver transistor trying to keep the load voltage up.  It might be a good idea to detect the excessive increase in driver collector current by adding a sense resistor and use it to turn on a PNP to inject current into the LM723 CL pin (via a resistor) and trigger the current limiting.

Run the fan off Vunreg via a LM317 set to 14V to prevent it being over-voltaged.  You can even add a thermistor on the heatsink to vary the fan voltage according to temperature so its not too noisy on light load.

I've just figured out how to get the boost rail essentially for free
No extra mains wiring, no extra SMPSU boards or DC to DC converters, just 4 components (2 if you use a common cathode Schottky, and a single large cap).
 
Use a pair of 2A Schottky diodes off the AC terminals of the bridge to feed their own reservoir capacitor.  10mF will give you a boost rail that stays above 16.5V when loaded with 1A - either use two of the same caps as the main output or something cheaper as it doesn't need the high ripple current rating.  Revised LTspice schematic attached

[davelectronic]

As intersting as it would be to build a control circuit for the regulator, i feel it would start getting to busy under the hood. Compared with different cable diameter it would work out more expensive.

I've took the plunge and ordered the same quality high temperature silicon wire, but in 14AWG instead of the 12AWG that's on it now. The 14AWG wire has an outer diameter of 3.6mm and conductor diameter of 2.1mm, and I've ordered 10 meters of it. I should easily get 18 Volts plus on it with that. But i will aim for 20 Volts AC so i can use components i already have. Cable cost £10.58, if the MOT had a slightly bigger core i wreckon I'd have got away with it.