32V 8A Regulated PSU Schematics

[otherside]

Hey everyone,

Here is a simple implementation of a regulated 32V 8A power supply for use with an LM4780 Amp.

Is this any good at your opinion? Am i missing stuff?

Any critisism is welcome when accompanied with recommendations :-).

[gmikol]

I've been working on designing a supply like this, too. Except mine is 65V/15A, but the same principles should apply. Here's my thoughts...but keep in mind that I haven't had a chance to test them out yet.

-First, I think you've got your resistor values R2/R3 and R5/R6 backwards.

-Second, the resistor ratio you have will only give you +/-16V instead of +/-32 (11.8k/1k)*1.25=16

-Third, I think the value of the upper resistor is too high at 1k. Typically, you would choose the upper resistor so that it draws the minimum load current from the regulator. It looks like worst-case minimum load current for LM317L is 2.5mA. So you want the upper resistor to be R=V/I=1.25/2.5m=500 Ohms. The datasheet shows 240, so that gives a little margin. You'll have to re-calculate your lower resistor based on what you choose for your upper resistor.

-I don't think a darlington is appropriate in this application (MJ11015,11016). This is because it would require R1 and R4 to drop ~1.2V instead of 0.6V. Since all the base current ends up flowing through the regulator anyway, it doesn't matter. Also, you'll be operating the regulator at 120mA in order to generate that Vbe drop. That's very near the limit. What about using a regular LM317/337?

-Personally, I would use LM338 so that the regulator and pass transistor can share the power dissipation more equally. Using the LM338 would require you to build 2 positive supplies and hook them up in series to get your + and - voltages, since there isn't a 5A rated negative regulator in the LM family. There aren't any real drawbacks to this arrangement for a chipamp-type application like yours.

-There's one more option, but it's pretty expensive. And that is to use LT1083 regulator instead. It's rated for 7.5A minimum and 9A typical. You could do away with the current-boosted approach altogether. But they cost ~$13 each.

That's an awful lot of food for thought I've just given you (sorry 'bout that). Good luck with your design!!

--Greg

[Shredly]

For best ripple rejection, you should bypass R2 and R5; the data sheet recommends 10uF tantalum, but for this application I'd use 100uF at least 30V aluminum electrolytic because of the prohibitive cost of 10uF tantalum in higher voltages; there is no need to worry about the ripple rating. The ability to do this, and the enhanced ripple rejection it provides, is one of the great advantages of using adjustable regulators in a supply as opposed to fixed regulators, which are not designed to accomodate this cap, nor for that matter to have their ground-sense pin raised (though I'm sure we've all done it to them ).

gmikol is right on the money about the values of R3 and R6; these should be 240R for the best regulation accuracy, and in fact they need to be right next to the regulator devices, as close as possible; the regulation accuracy depends upon it. Recalculate R2 and R5 to get your desired voltage; I suggest that the correct value is in the neighborhood of 5.6k to 6.8k. I'd recommend a pot as part of R2 and R5, and I'd go with a high-stability trimmer, perhaps a 5k, with the additional 600R to 1.8k as a fixed resistor in series with the pot.

When using both output and R2/R5 bypass capacitors, discharge diodes bypassing R3 and R6, and bypassing the regulator from Vin to Vout, are strongly recommended by the manufacturer, and having blown a couple of 7805s and such by leaving them out early in my adventures, I'd say they have excellent reason to recommend it. Note that this will also protect the transistor from reverse current when the input collapses and C5 and C6 want to discharge to it. 1N400x should be fine; I've never blown one doing this with it, though technically at above 24V, you ought to use something more substantial in its surge rating. I don't build a 3-terminal regulator circuit without them, personally. If you're not sure where the diodes go, let me know, if you can't find the data sheet. I can give you a reference to the page in the data sheet, a link to it, or even draw you a schematic and post it.

[otherside]

Hey guys

First of all thanks for your great input! Appreciate it

The Schematic i posted earlier was just about the concept not the values, however your comments helped a lot on doing a more analytical design.

Some specifics

Personally, I would use LM338 so that the regulator and pass transistor can share the power dissipation more equally.

I know what you mean and i can completely understand your idea, however it doesn't work exactly like that. To do so you'll have to use two bridges and seperate the two secondary windings, then after the regulation you connect them in series. Maybe i'm wrong? that the case as far as i know. Also there is a response on national about that:

National's Staff Engineer - Applications: It won't work. I have also used positive regulators as "negative" rails but we had to have a completely different secondary winding and rectifiers feeding the input voltage to it. Then, you take the point which is normally the positive output and hook it to system ground. The regulators "GND" point then becomes the negative regulated output. I do not believe this can be made to work with a center tapped transformer with a positive regulator on the negative side.

-There's one more option, but it's pretty expensive. And that is to use LT1083 regulator instead. It's rated for 7.5A minimum and 9A typical. You could do away with the current-boosted approach altogether. But they cost ~$13 each.

Usually most of the stuff in my designs i build with free samples. Unfortunatelly Linear only let you order 2 and not sure what time limit is before order again.. On farnell here in the UK these cost around 11 pounds which is almost 20$.. nooooo way ;-)
(I ordered already the 2 samples just to try them out)

Shredly thanks for the great tip these notes are not in all datasheets, however i found them and read a bit about it.
About the pot, yeap i could use one, however having 10mA (which the max (min) required current for LM337) flowing through a pot i don't like it much, i'll have to spent more on a good quality pot than just a 0.1% smt resistor which is around 50cent. Anyway that's just an implementation issue, as we usually do at work ... put the pads there and if you want you use them ;-), most of the times on the protos we use them ;-).

OK here is a more complete version.

About the current distribution, At the moment i have on the schematic an 1ohm resistor, this will allow 700mA to flow through the regulator and the rest from the power TR. There are some issues here. The chipamp which i'm designing this for doesn't draw steady current so at some level the transistor will be at a state where will turn-on/off all the time, this can only be bad for noise i guess. So i might change that to allow only 150mA through which is the quiet current. Another issue is the heatsink, if i allow only small current through the regulator IC then i don't need to heatsink it whilst all components (SMT) will be close together and will avoid noise etc. So i might be better of throwing all the current in the transistor and place it on another heatsink. Either way the voltage drop is small (3V), even at 10A this is 30W dissipation worse case, which is not too much.

Also about the TIP33/34 power transistors, i just used these as an example i'm not 100% about all their characteristics, something else might be better.

Tell me what you think about the new schematic.

Again guys thanks a lot for your help. I'm quite new in this forum, my knowledge in analog is a bit limited but i rock in digital, programming micros etc... (my job is actually that) so pretty soon i hope i'll publish here some stuff i've done which you might like to add on your designs.

[dangus]

This thread may help, although the VA requirements of the Leach are a little higher: Leach Super Amp Regulated supply

[Shredly]

Yeah, that's the right idea for the diodes.

Interestingly, the resistor for this particular regulator appears to be recommended at 120R; the rest of them all recommend 240R for the 2xx and 3xx, and 120R only for the military grade 1xx part. Silly me, I shoulda read the data sheet for both parts.

I wouldn't worry too much about the transistor turning on and off; it'll run in the active region until you hammer it to the rail, so you'll get a pretty smooth transition as the current increases. Those 1000uF caps out on the output will help with that, too.

I might consider going with a 2N3055/MJ2955 pair rather than the TIP34/33 pair you've chosen if I needed more amperes; but you probably don't so there's no point, and it'll just make it bulkier without adding value. The TIP33/34 handle 10A; 2N3055/MJ2955 handle 15A (not that I'd put 15A through one, or 10A through a TIP33/34 either; call me superstitious, I don't like to push transistors all the freakin way out there to their limits- I guess I just don't like the smell of burning heat sink grease). =:O

BTW, be sure to get at least the TIP33A/34A; they're prolly an extra ten pence or whatnot, big deal, skimp on your next candy bar, or get some Old Peculier instead of the usual Samuel Smith's Nut Brown Ale or whatever.

I've seen people decouple these at the input with 0.1uF disks or CK05s; I generally do myself, it's probably just another superstition. You might try it, though. I'd say it's pretty much gold at this point. Build it, load test it at like 6A or so, and check the noise at the output- my guess is, when you get your 'scope down to like 10mV/div, it'll start to get fuzzy and you might just be able to convince yourself you see some ripple down there if you happen to have had one more brewski than usual that evening. IOW, it should look like a battery.

Have fun.

[otherside]

About the input decoupling, sometimes can create problems depeneding on the values, so i was thinking to put the pads there and test out if i need them.


Any ideas about ac mains filter/protection?

[sawreyrw]

otherside,

If you really want to get +/- 32 volts at 8 amps, you're probably going to have some thermal issues. What is the raw DC voltage?

I would use about 30 ohm from base to emitters. This will reduce the load on the IC's, but you still need to watch the power losses in them. You may need to heatsink them.

I don't think a Darlington would be a bad choice in order to keep the IC currents low. Adjust the base emitter resistors accordingly.

You may need several power transistors in parallel; use emitter resistors to help with current sharing. The reason I am suggesting several parallel devices is because of the safe operating area (SOA).

This design should work, but heat kills so check the power losses and SOA.

As far as a mains filter and protection, you might want to use transzorbs or MOVs across the secondaries of the transformers to protect the diodes from large voltage spikes.

Rick

[otherside]

Hi sawreyrw,

This design was based on a 25V TR which results in ~35V DC.

From one of my previous posts

About the current distribution, At the moment i have on the schematic an 1ohm resistor, this will allow 700mA to flow through the regulator and the rest from the power TR. There are some issues here. The chipamp which i'm designing this for doesn't draw steady current so at some level the transistor will be at a state where will turn-on/off all the time, this can only be bad for noise i guess. So i might change that to allow only 150mA through which is the quiet current. Another issue is the heatsink, if i allow only small current through the regulator IC then i don't need to heatsink it whilst all components (SMT) will be close together and will avoid noise etc. So i might be better of throwing all the current in the transistor and place it on another heatsink. Either way the voltage drop is small (3V), even at 10A this is 30W dissipation worse case, which is not too much.

30W dissipation on the transistor on a relative large heatsink will not be a problem i think. It's well inside the SOA of the tr.

I would use about 30 ohm from base to emitters. This will reduce the load on the IC's, but you still need to watch the power losses in them. You may need to heatsink them.

Could you explain this a bit further? I'm not sure i understand the purpose of the resistor and how will effect the current.


Thanks

[gmikol]

Okay, otherside, I think I understand better what you're trying to accomplish re: using SMT parts and making the TIP33/34 carry the whole load.

The reason I suggested dual positive regulators was because your schematic looked like you had independent secondaries which you tied together in a center-tapped configuration. Reading your later posts, it sounds like you have a center-tapped transformer, which does, in fact, make my suggestion useless. But for 8A load, you may want to consider independent secondaries and full-wave bridge rectification. You'll be trading off higher voltage loss against better transformer utilization. Though at 35VDC at the caps, you don't have a whole lot of margin before you're bumping up against the dropout of the regulator.

But I do think you need to be careful on SOA for the TIP33/34. This isn't the output stage of an amp, so it's not seeing massive voltage swings, so you might think you're OK. Assuming really poor regulation, it will see maximum about 5 volts across it (if your output sags all the way to 30V). That's 8A*5V=40W dissipation. Tjmax is 150degC. That meas your total thermal resistance to ambient needs to be less than 3degC/W. (3*40 + 25 = 145)

Rjc + Rins + Rhs = 3

Rjc=1.56 C/W from the data sheet
Rins=0.5 C/W for the insulator and grease
That leaves ~1 C/W for the heatsink. And that's to just barely stay under Tjmax. A high-line condition could kill the device.

Hope this helps. And when you finally get it built, be sure to let us know how it turned out.

--Greg