48VDC and <= 18VDC out of 48V transformer?

[JJones]

Hi, I have a 48V transformer @ 95VA, and would like to get 48VDC for a fan I have, as well as ~5-18VDC for a small PWM controller I've designed for said fan. I'm thinking that if I got the 48VDC, I could probably simply use a LM317 to drop the 48 to 15 or so.

Right now, I have a rectifier and 2200µF capacitor on the transformer, and am thus getting about 67VDC, so I need to drop about 20V. FYI, my fan is ~0.75A. The PWM controller is basically a 555 circuit driving a MOSFET (IRFZ46 or the like), and I'd expect less than 0.2A draw on that end of things. Speaking of which, does anyone see a problem using the 5-18VDC from the 555 to switch the 48V line with the MOSFET? I'm very experienced with using MOSFETs, so I want to make sure I'm not going to fry anything

 

[Hero999]

You could add a buck switching regulator to get 48V.

A linear (LM317) would do but it's very inefficient, you'll need a large heatsink. The LM317 also isn't capable of dropping 20V at 0.75A so you'll need to add a booster transistor.

There's nothing wrong with dropping 48V to 15V for a PWM controller as it's only low current.

Does the supply need to be regulated?

If not then you could use a 35V transformer which should give you nearly 48V without adding a regulator.

If the answer is yes then you could use a 40V transformer plus a linear regulator which will be much more efficient.

 

[JJones]

Hrmmm, not familiar with buck switching regulators... I found "buck converter" on wikipedia, but have no idea how I'd apply it.

The supply doesn't need to be particularly regulated (±2V isn't going to make much of a difference to a 48V fan); however, I'm trying to build this out of parts I can buy at work - which is why I have the 48V transformer. I seem to recall seeing a transformer with twin 19V secondaries, so I might be able to get 38V... what sort of linear regulator would you recommend in that case?

BTW, as you may have guessed, I basically know just enough about electronics to get myself in trouble... but I learn best by example, so I'm trying to build things, asking for help along the way, and trying to see how various things work (or smoke!)

 

[Hero999]

38V would certainly do, an LM317 regulator could be used but it won't be short circuit proof because the maximum voltage between the input and output will be exceeded.

 

[bailey45]

The LM317 is only good for 40VDC Vin max.

 

[Hero999]

That's a common myth, the LM317 is good for any input voltage providing it is no more than 40V greater than the output voltage.

You could use an LM317 to regulate 130V down to 100V if you like but don't expect any short circuit protection. When the output is shorted, it will try to shut down, only then the input-output differential will exceed the absolute maximum rating of 40V causing it to emit smoke.

 

[bailey45]

When some of the features of a component do not work above a certain voltage I call that exceeding the manufacturers recommended maximums.

Dragons and Elves are a common myth.

 

[JJones]

What if I used a 317 on the first secondary of the 19V transformer, and a 337 (negative complement to the 317) on the second? I'd need to rectify and filter each secondary separately to do this, wouldn't I? I'm not terribly familiar with playing with negative voltages; would I basically connect the (+) of the rectified output as the ground, and the (-) to the 337's input?

BTW, putting a 1A breaker on the output would address the short-circuit issue, wouldn't it?

 

[bailey45]

See Attached

 

[crutschow]

If you have a transformer with dual 19V secondaries you can connect them in series to a bridge rectifier to get about 52V for the 48VDC requirement. At the same time you can also connect a single rectifier from the transformer tap (junction of the two windings) to generate about 25V for the lower voltage requirement for the controller. (See below) Interestingly, this tap connection acts like a full-wave rectifier configuration so both outputs are full-wave rectified to minimize ripple.

You can then use regulators as required to get the voltages you need.


The breaker for short circuits will not be fast enough to protect the regulator from damage.
(The semiconductor is always faster than the fuse).

 

[JJones]

Okay, I'm at work now, here's the exact transformer I have:

Primary: 117V
Sec 1: 18V
Sec 2: 18V
Sec 3: 8.9V

(18+18) * 1.4 ≈ 50.4VDC
Oddly, the first and second secondaries don't seem to share a center tap (poking with an ohmmeter); it seems they're two completely different windings. The thing I particularly like about this transformer though, is that it has a third secondary winding that puts out 8.9V, or about 12.46VDC rectified and filtered, so that takes care of my low-voltage supply.

This transformer also has a slightly smaller form factor, so it will fit in some cases I'd been looking at that were just a wee bit too small for the other toroid

 

[Hero999]

When some of the features of a component do not work above a certain voltage I call that exceeding the manufacturers recommended maximums.

Sorry but I disagree, it's fairly common for components not to be rated to two absolute maximum ratings simultaneously because it exceeds another maximum rating.

For example, a transistor cannot simultaneously take the maximum collector current at the maximum collector-emitter voltage because the maximum power rating would be exceeded.

The the same applies to the LM317 which can't pass 1.5A if the input-output differential exceeds 15V because the maximum power rating will be exceeded. In this case the safe operating area protection is triggered causing the current to be limited.

I wouldn't use an LM317 with a input greater than 40V in an application where short circuit protection is required such as a bench-top power supply. However I wouldn't hesitate using it in an application where short circuit protection isn't required: i.e when an appliance will be permanently connected to its output such as a fan.

You just need to be aware of the absolute maximum rating of components and if you can, design around them. When using the LM317 where the input >40, a 39V zener could be connected from the input to output if there's a concern that capacitors on the output could cause current surges that might cause it to shut down. Current limiting could also be added back in by adding a transistor current regulator in series with the LM317 circuit.

The datasheet also explains that the LM317 is a floating regulator and that it's the input-output voltage differential that's important, not the input voltage.

Okay, I'm at work now, here's the exact transformer I have:

Primary: 117V
Sec 1: 18V
Sec 2: 18V
Sec 3: 8.9V

(18+18) * 1.4 ≈ 50.4VDC
Oddly, the first and second secondaries don't seem to share a center tap (poking with an ohmmeter); it seems they're two completely different windings.

Check the absolute maximum voltage rating of the fan, it will probably be good up to 60VDC.

The thing I particularly like about this transformer though, is that it has a third secondary winding that puts out 8.9V, or about 12.46VDC rectified and filtered, so that takes care of my low-voltage supply.

Great but don't forget the connect the 0V of the 12V supply to the 0V of the 50V supply.

 

[JJones]

Check the absolute maximum voltage rating of the fan, it will probably be good up to 60VDC.

The **broken link removed** has an operating voltage range of 28.0V to 53.0V, but I'd still like to regulate it to 48 if I can.

If I'm understanding things correctly... would I use R1=240Ω and R2=9KΩ with the ~50V input to get a regulated 48V out? Also, since the LM317 would be dissipating ≤2W of energy, it should be fine with its built-in heatsink, correct? I may put a small snap-on heatsink on it as well, but just want to make sure I've got things in the right perspective.

Great but don't forget the connect the 0V of the 12V supply to the 0V of the 50V supply.

I assume I should connect the grounds post-rectification (that is, connect on the DC side of things) right?

 

[Hero999]

The **broken link removed** has an operating voltage range of 28.0V to 53.0V, but I'd still like to regulate it to 48 if I can.

According to the datasheet you linked, it's good for up to 56V.

Is 38V across both secondariness the measured voltage or the voltage rating?

If it's the voltage rating then you're sensible regulating because the voltage will be 10% to 20% higher under light loads.

If it's the measured voltage then a regulator is probably not needed.

If I'm understanding things correctly... would I use R1=240Ω and R2=9KΩ with the ~50V input to get a regulated 48V out?

The input needs to be 3V> than the output to ensure good regulation.

The diodes will drop 1.4V

38√2 = 53.74V - 1.4 = 52.34.

The minimum input voltage for the regulator to regulate properly is:
52.34 - 3 = 49.34V

The maximum allowable ripple on the filter capacitor is:
49.34 - 48 = 1.34V

To calculate the filter capacitor size use the following formula:

[latex]C = \frac{I}{2F \times V_R}[/latex]

Vr is the maximum ripple, I is the current and F is the mains frequency, assuming 50Hz mains frequency:

[latex] C = \frac{I}{2 \times 50 \times 1.34} = 0.0056 = 5600 \mu F[/latex]

I'd recommend using a 6800µF capacitor which needs to be rated to 60V and will be very large.

Also, since the LM317 would be dissipating ≤2W of energy, it should be fine with its built-in heatsink, correct? I may put a small snap-on heatsink on it as well, but just want to make sure I've got things in the right perspective.

No, you need a heatsink.

Look at the datasheet.
https://www.electro-tech-online.com/custompdfs/2009/07/LM117-1.pdf

The maximum temperature rating is 125°C.

The thermal resistance from junction to ambient is 50°C/W, even at 2W the temperature rise would be 100°C which means if the ambient temperature is over 25°C it will overheat.

In this case the power dissipation is:

Voltage across regulator × current through regulator:

52.34 - 48 = 4.34×0.75 = 3.26W

Suppose the maximum ambient is 25°C, the maximum temperature rise is 100°C

The maximum thermal resistance between the IC and ambient is:
100/3.26 = 30.7°C/W

The thermal resistance of the die to the case is 5°C/W (see datasheet).

The maximu thermal resistance of the heat sink plus the thermal resistance of any insulating tab is:
30.7 - 5 = 25.7°C/W.

For example if the insulating tab you want to use has a thermal resistance of 6°C/W then you'll need a heatsink with a thermal resistance of less than 19.7°C/W.

Note that the above calculations assume the regulator is an LM317 in a TO-220 (T package on the datasheet) made by National Semiconductor. Always check the datasheet from the supplier where you've bought your IC from, it might differ slightly.

I assume I should connect the grounds post-rectification (that is, connect on the DC side of things) right?

Yes.

 

[JJones]

According to the datasheet you linked, it's good for up to 56V.

Well the model I have is the EHE version, which for some reason has a maximum rating of only 53V; the lower-performance VHE and SHE fans can tolerate 56V.

Is 38V across both secondariness the measured voltage or the voltage rating?

If it's the voltage rating then you're sensible regulating because the voltage will be 10% to 20% higher under light loads.

If it's the measured voltage then a regulator is probably not needed.

That's the rated voltage... unfortunately this transformer doesn't have a VA rating on it, and since it is physically smaller, it might have a slightly lower rated output capacity. I'm trying to contact the manufacturer to find this out.

The input needs to be 3V> than the output to ensure good regulation.

The diodes will drop 1.4V

38√2 = 53.74V - 1.4 = 52.34.

The minimum input voltage for the regulator to regulate properly is:
52.34 - 3 = 49.34V

Well if you add even 5%, the numbers should change as follows, right?

39.9√2 = 56.43V - 1.4 = 55.03

Thus, the maximum allowable ripple on the filter capacitor is:
55.03 - 48 = 7.03V

Vr is the maximum ripple, I is the current and F is the mains frequency, assuming 60Hz mains frequency:

[latex] C = \frac{1}{2 \times 60 \times 7.03} = 0.00119 = 1200 \mu F[/latex]
(assuming a generous current draw of 1A)

I think I can find a 1200 µF 63V capacitor without much trouble, and it'll be much smaller, relatively.

Of course all this means I'll need a bigger heatsink, won't it?

 

[JJones]

Another thought; since I don't need the 48V terribly well-regulated, what if I just used some diodes to drop it? I've got some at work that I think drop ~1.5V apiece. Three in series would drop 4.5V, which should put me at around 47.5-50V, no?

Also, what effect, if any, will high frequency PWM have on the filter caps? I've got my PWM tuned for about 40kHz, so the power draw will be cycling very rapidly. I haven't done the math exactly (plus the duty cycle range will depend on the actual values of the resistors and caps, versus ratings) but I should have a range in duty cycle of ~10-90%.

 

[Hero999]

Well if you add even 5%, the numbers should change as follows, right?

The trouble is the voltage will go down when a load is applied.

Of course all this means I'll need a bigger heatsink, won't it?

Yes, if the input voltage is higher, you'll need a larger heatsink to get rid of more heat.

Another thought; since I don't need the 48V terribly well-regulated, what if I just used some diodes to drop it? I've got some at work that I think drop ~1.5V apiece. Three in series would drop 4.5V, which should put me at around 47.5-50V, no?

Yes, diodes will work but not as well as a voltage regulator, you might even get away with a resistor.

Also, what effect, if any, will high frequency PWM have on the filter caps? I've got my PWM tuned for about 40kHz, so the power draw will be cycling very rapidly. I haven't done the math exactly (plus the duty cycle range will depend on the actual values of the resistors and caps, versus ratings) but I should have a range in duty cycle of ~10-90%.

It shoudn't make much difference.

 

[JJones]

Okay, I think I've solved my problem... instead of playing around with an LM317T, I managed to dig up an LM338K - an adjustable voltage regulator in a TO-3 package, and I have a TO-3 heatsink that's about 1.8" square, and 3/4" high, with 16 fins.

I've poked around and drawn up this PCB layout, and intend to go ahead with it unless anyone can see some glaring flaws: **broken link removed**

Mind you this is just for the power; the 120V in, the transformer, the filter caps and rectifiers, and the regulator. The regulated 48V and unregulated ~13V (I'll squeeze in a little TO-92 12V regulator if need be) go out via the four-pin connector which will be jumpered to the other board that actually handles the PWM. The large empty chunk in the bottom right is where the toroid will sit; to save space I poked the 13V filter cap in the hole in the toroid

In case the four-cap array was a bit confusing (or wrong), my intent is to put the four 3300µF caps in a 2x2 series-parallel matrix. Since I couldn't find any reasonably-sized 2200+µF caps at ≥ 63V (yes, this set of four was smaller than the ones I did find), I figured two 3300µF@50V caps in series will give me an effective 100V rating (yay for safety margins!), but whack me down to 1650µF. Thus I take two pairs of series caps and put them in parallel, and presto, I'm back up to 3300µF, with a 100V rating.

Oh, one last question: any ideas how I can ballpark the current draw for the transformer? I did this math:

55V * 1.0A (pre-regulated voltage, and rounding up on the current draw) = 55W
15V * 0.2A (maximum unregulated voltage, and being generous on the current draw) = 3W
55W + 3W = 58W
Assume a dismal efficiency of 50%... so it would need 116W on the primary
116W / 120V = 0.97A

Am I going about it the correct way? My concern is the trace width for the line voltage; according to some online calculators I've tried, 1A through 1 mil copper only needs a trace width of ~65 mil (I'm using the math for internal traces, as I intend to epoxy the traces carrying 120VAC). This just seems really, really small to me. BTW, my current design has ~80 mil traces on the 120, and ~60 mil traces elsewhere.

 

[Hero999]

You haven't included the silk screen or schematic so it's difficult to know.

You'll also find that PCBs tend to etch more reliably if you avoid 90° bends or T-junctions.

 

[JJones]

If you tick the checkboxes, it will overlay what amounts to a silkscreen. I did it this way so you can just review one part of it at a time, or turn it all on to see the whole thing. I also added a schematic next to it