PSU design assistance needed

[juniorhifikit]

Thanks everyone.

Ya know, it's funny... I looked at a lot of bipolar supplies from different products, and used one from a "budget manufacturer" (who will remain nameless) as a model. I swear they screw up their schematics on purpose to keep other manufacturers and people like myself from copying their circuits. Anyway, I had a look at the data sheets for the regulators. I still need to calculate the C for the given ripple, but I re-uploaded my circuit and I think it's more in line. Thanks for taking the time to upload the theory for me. Once digested, I'll hopefully have a better handle on PSU design. I also found this website, which was more dealing with power amps and their supply needs, but it all seemed applicable:

http://sound.westhost.com/power-supplies.htm#intro

Merry Christmas y'all

[mediatechnology]

Heres' the LM317/337 regulator board schematic we sold for the Picos. They are no longer available but it may help you out:

http://www.proaudiodesignforum.com/Pico ... uments.htm

I had forgotten that there were links to Sam Groener's PSU calculator spreadsheet and to the NSC Audio design Handbook.

[juniorhifikit]

Heres' the LM317/337 regulator board schematic we sold for the Picos. They are no longer available but it may help you out:

http://www.proaudiodesignforum.com/Pico ... uments.htm

I had forgotten that there were links to Sam Groener's PSU calculator spreadsheet and to the NSC Audio design Handbook.

Exactly the rosetta stone I was needing - thanks! Couple questions about the Pico PSU design though:

Why the extra diode from regulator output to ground on each leg (D7 & D10)?
Why the resistors across the protection diodes (R3 & R4)?
Why the 2200uF in parallel with the 1uF (I get why the .1uF) Just three different ranges of filtering? Still not clear on this calculation (must read more and drink less spiked eggnog)

[JR.]

Why the extra diode from regulator output to ground on each leg (D7 & D10)?

The extra diodes may be over kill but they are to prevent latch up, during transient start up conditions. With some older regulators when one supply comes up first, the other can latch in a reverse bias and be damaged, or never start properly.

Why the resistors across the protection diodes (R3 & R4)?

Actually the diodes are across the resistors, and probably there to discharge the caps at the adj terminals after turn off, so they don't discharge through the internal circuitry.

Why the 2200uF in parallel with the 1uF (I get why the .1uF) Just three different ranges of filtering? Still not clear on this calculation (must read more and drink less spiked eggnog)

Yup... apparently still an undesirable impedance coverage gap between .1 and 2k... modern low impedance (switcher) caps probably help.

FWIW, this is not a sharp pencil design. This is over designed to be more robust and reliable than any application should require. Not a bad thing.

JR

[juniorhifikit]

John

Thanks. That is a luxury we have in small volume and DIY designs... Go ahead, add that penny diode or extra bypassing cap.
You could remove most of that stuff and the Pico PSU would work just as well.

Roger

Yeah, why not right!?

I'm having trouble locating a single transformer with suitable secondary coils. Without custom ordering something, do you think it's reasonable to expect to find an off-the-shelf transformer with one secondary suitable for +/-18V at around 50VA, and another secondary for the 5V low current stuff? Most transformers I'm finding have two identical secondaries. Getting one coil substantial enough to handle the bipolar part seems to leave the other way too beefy for the 5V leg. There would be a lot of thrown-away voltage.

Or, is it kosher to do the +/-18V for the audio circuitry and the 5V for the logic off the same secondary? Originally I was thinking of something like this:

[JR.]

Please read my post about the TS-1 PS

While that was dealing with a relatively high current 5V supply, powered from a higher voltage winding. If the 5V supply is low current, it doesn't really matter.

JR

[juniorhifikit]

Please read my post about the TS-1 PS

I did read the post, but the jpg is so pixelated that I can't make out much of anything. I'm not in a position to be able to ASSume anything about circuit design

While that was dealing with a relatively high current 5V supply, powered from a higher voltage winding. If the 5V supply is low current, it doesn't really matter.

JR

Does it need to have it's own rectifier, or can the 5V regulator just parallel off the positive leg? Or, could I use an LM338 for the +18V leg (since it can supply "in excess of" 5A) and hang the LM317 for the +5V leg after that?

You had mentioned thermal resistance and I see it mentioned in the data sheets, but all I can manage to take away from that is that the bigger metal can TO3 version might dissipate heat better. Is there a big difference between a heat-sinked TO3 package and a heat-sinked TO-220 package in regards to heat and the regulator's max current output (probably also depends on how much voltage it's having to throw away)?

[JR.]

An important first question that comes to mind is how much 5V current do you need? If the unregulated supply is nominally 20v, you will be dissipating 15v x that current in the pass regulator. Not a big deal if drawing 25 mA, but a few hundred mA could be several watts.

So task #1 is to quantify the power.

If the draw is steady, you can also use a power resistor in series to scrub off some of that extra voltage. If you put the resistot in series with the diode before the reservoir cap, you will scrub off even more voltage. Since you obviously have more unregulated voltage than you need, using half wave rectification, and perhaps an undersized reservoir cap, would also give you more ripple voltage, but lower average voltage and less wasted power.

To answer you specific thermal resistance question, the important number to minimize is the total resistance, from junction (silicon) to ambient (air).. There will typically be a number for junction to case, case to heatsink, and heatsink to air. TO-3 while falling out of popular use (due to cost), have the lowest junction to case resistance, and good low case to heatsink resistance. The math is very similar to ohms law, while watts of power dissipated or flowing through the mechanical path is equivalent to the current, and thermal resistance is resistance. Instead of a voltage drop across a resistance, we are calculating a temperature rise, from ambient, up to the silicon junction. You probably want to keep silicon (In plastic devices) less than 150'C. Metal cans can tolerate a little higher temp, but 150'C is still a good target. Note:It gets kind of complex to calculate thermal resistance from heatsinks to ambient. Most high power heat sink designs, build and measure results (or at least we did).

If it looks like the TS-1 approach is useful, I can dig up the values, while i would still be inclined to suggest you dial them in for the transformer you use, transistors you use for switches, and actual circuit current draw. Briefly the resistor driving the base of the first NPN must be adequate to saturate the switch when not turned off (more like peak than average charging current). The three resistors driving the base of the second transistor have the current gain of the second transistor reducing the work they must do to turn off the first, so can be higher impedance. Two of them set the threshold voltage for the lower unregulated supply, and the third resistor is hysteresis (positive feedback) so it turns off the transistor again once nominal voltage is reached, otherwise that transistor doesn't switch off and heats up like a simple pass transistor.

If the 5v current draw is low, there is no need for this complexity. If 5V is really high, you could always use a switcher.

JR

[juniorhifikit]

Since I don't yet know exactly what will be on the 5V leg, and this is kind of a prototype project, I'll just wait and add that part of the PSU later. I do know however that it will be low current draw, as it's only microprocessor & DAC stuff. All the theory and advice was not wasted though, and I thank everyone for pointing me down the right path.

[juniorhifikit]

Thanx.. the trick PS is in the upper left hand corner.. IIRC I gave the 5V 3-terminal regulator it's own roughly 10V unregulated supply so it wouldn't drop out of regulation with ripple. A significant power savings from the nominal 20V unregulated.

One perhaps unobvious part of that circuit, is that I add some hysteresis to the switch threshold comparison, to insure the switch turns off sharply (resistor from raw rectified AC). If you don't add this hysteresis, the switch doesn't turn off completely and overheats, not to mention it doesn't save you much power.

This is close to the efficiency of using a switcher, with far less complexity.

JR

Would anyone happen to have a more hi-res image of JR's Loftec with the cool power supply? I can't make out much from the posted image. I'm basically revisiting my original PSU design and remembered that what I'm thinking of is pretty much what he had suggested. I'm basically trying to efficiently get an extra 5V leg from my +/- 18V supply. I've got about +/- 22 volts after the rectifier. Since my 5V circuit will draw around 1.2 Amps, that's a lot of voltage for the 5V regulator to throw away (as heat). I've also toyed with half-wave rectifying separately for the 5V leg...