Need to find a Vreg for a Bipolar Supply

[bob p]

I'm looking for a way to regulate a bipolar power supply, and I'm not sufficiently familiar with the available SS devices to know what I should be looking for. Here's what I need to do:

I have an amp with a bipolar power supply that delivers +/-85VDC. I need to provide a low voltage bipolar supply at +/- 12VDC to power a few opamps. The amp's current embodiment uses a pair of large series resistors and zener diodes in a brute force effort to create the +/- 12V supply rails. Unfortunately, this has the unfortunate side effect of dissipating lots of watts into a closed / poorly ventilated space. In use, the series resistors are reaching temps of 300*F when the box is open with no top or bottom in place. This trapped heat causes the amp's cap compliment to suffer a premature death. So I'm looking for a more thermally efficient way of delivering the bipolar 12V supply.

Can anyone offer ideas for a more efficient way to deliver the +/- 12VDC supplies, using the +/- 85V supplies as a power source? I'm interested in a regulated / semiconductor type of solution, rather than using resistors, as I need to minimize the heat that gets generated along the way.

TIA.

[Steve Conner]

To achieve a big reduction in the heat, you would need a switching regulator or a second small transformer. Linear regulators still dissipate heat. A bit less than the Zener shunt regulator to be sure, because they only pass as much current as the load requires.

I've used a few approaches. You can add a pass transistor to the Zener shunt regulator. The gain of the transistor allows you to drive a high load current with a small Zener current. You can make your own regulator with a TL431 driving a MOSFET follower. Or you can use a LM317 with a Zener in series with the input pin to drop some voltage. None of these options will survive a short circuit on the 15V rail unless you're careful with the design.

[bob p]

thanks for the input, Steve.

I'm thinking that the three op-amps are going to draw negligible current, which should help to tame the amount of heat that a regulator would generate. Using something in a TO-220 package would allow me to sink the regulator's heat to the chassis, rather than dumping it into the interior compartment as the pair of 7W resistors is doing right now.

I looked at this TI product, which seems appealing. It'll readily take an 85 VDC input and produce a 12V output. The problem is that although it'll work fine on the (+) rail, I have no idea what kind of device normally gets used on the (-) side in a bipolar setup.

http://www.ti.com/lit/ds/symlink/tl783.pdf

I'd appreciate a link to a schematic that demonstrates a suitable approach to the problem if anyone's got one. I'm thinking that I shouldn't have to go about re-inventing the wheel to solve this problem. This has got to be a problem that's already been solved, it's just that I don't have enough familiarity with the SS devices to know what to use.

[g1]

You should be able to use the same tl783 in a different configuration to get the negative side. Searching for "tl783 negative voltage" gives several results, such as this: -VE Voltage Regulator - Linear Regulators - Forum - Linear Regulators - TI E2E Community although the poster doesn't specify how he got it to work in the end, may have been pilot error.

[bob p]

That fellow's setup is a floating ground design.
I have a non-floating ground that is comprised of the power xformer CT going to a pair of 6,600uF PS caps and the amp chassis.
I can't use the 783. Won't work.

[Enzo]

You are focused on using 85v rails to make 12, but as you state, the op amps will draw a tiny amount, so a little 6VA transformer is smaller than a ping pong ball, and would be perfect for a small supply. A little round bridge and a couple caps and a pair of TO92 regulators like 78L12/79L12, and it hardly uses any space or makes much heat. I see Mouser also has 1.1VA transformers. for example 20vCT @ 55ma

[Steve Conner]

I've never seen a high voltage negative IC regulator. :/ You can't use a positive regulator like the TL783 to regulate the negative rail, because the two unregulated rails share a common 0v point.

This circuit is pretty simple and a negative version can be made with a PNP transistor. The regulation is no worse than the original zener arrangement.
Zener Controlled Transistor Voltage Regulators - Electronic Circuits and Diagram-Electronics Projects and Design

I built a high voltage discrete regulator once that regulated +/-110V down to +/-65.
http://scopeboy.com/regulate.gif

[Mick Bailey]

If the current requirement is low then it could be worth experimenting with a series capacitor as a reactive component off the AC side of the 85v rails. You'd need a series cap, series limiting resistor, rectifier, 12v zener and smoothing cap. (one set for each side of the supply). The success or otherwise depends on the current draw requirement and reasonably constant load. The cap doesn't get hot like a dropper resistor, but cap sizing is important and it has to be NP.

[R.G.]

The critical issue is **how much current do you need for the opamps?**

This determines how much power comes out of the +/- 85v if it comes from there, or the size of the capacitive divider if it comes from the AC.

The zener regulators for opamps in many amps with higher supplies are a direct result of three-terminal regulators not living well with more than 40V or so as an input. Zener regulators are particularly bad for supplies like this because the zener must be designed to run at essentially maximum possible power all the time, and the dropping resistors have to be designed to let the maximum possible current through all the time. The tradeoff is using cheap zeners and power resistors and letting the thing heat up as much as it wants. The buyer is paying for the electricity after all.

As possible improvements:
1. If the opamp power is small, a 120Vac:30Vct transformer for up to about 10W is under $10 and so making an entirely new power supply for the opamps is not all that unreasonable, and completely sidesteps the issue of the heat. Small toroids are quiet and efficient.

2. If you're determined to get it out of the +/-85, then the minimum heating that's going to be possible with a non-switching power supply is 85V times the opamp currents. This will almost certainly be much smaller than the heat wasted to keep the dropping resistors and zener going. The only trick is making this work on both + and - sides. Steve's shown one way.

3. If you're willing to take it off the AC side of the rectifiers, use a capacitive divider, not a capacitive dropper. Analogous to a resistive divider, a capacitive divider can provide a divided-down AC voltage. This may be ticklish to design well, but offers much of the advantages of the capacitive dropper, but directly provides some voltage dropping that's not as tricky to balance with the loading on the rectifiers eventually used.

4. If you're willing to switch, you can probably use an idiot switcher. All non-flyback switching power supplies use a combination of pulse width modulation and filtering to turn a DC voltage into a regulated voltage. You have +/-85, and want +/-12. If you had +/-18V, you could use a three terminal regulator on each side and have regulated 12. So if you were to use a high-side switch feeding an inductor to a capacitor, you could do a fixed duty cycle into the inductor/cap, and the voltage on the cap would approach the duty cycle times the 85V. The idiot switcher would have a voltage output with ripple that would vary with load, but (a) the load doesn't vary much and (b) the three terminals could smooth out that ripple stuff.

Frankly, the low cost/low frustration method is probably the baby-transformer with rectifier/filter/regulators.

[Mick Bailey]

3. If you're willing to take it off the AC side of the rectifiers, use a capacitive divider, not a capacitive dropper. Analogous to a resistive divider, a capacitive divider can provide a divided-down AC voltage. This may be ticklish to design well, but offers much of the advantages of the capacitive dropper, but directly provides some voltage dropping that's not as tricky to balance with the loading on the rectifiers eventually used.

I see your point here. A series dropper isn't so easy to balance with a variable load (though using a zener and setting the voltage slightly high will give a reasonable window for regulation) but the divider can be tailored to give a more efficient circuit.

[J M Fahey]

In a nutshell it's either add an extra smalltransformer *or* passive dropping, which will dissipate exactly the same heat no matter if active or passive, the Math is the same.

RG hit the nail in the head: determine the *exact* current consumption, for which we'll need to see your preamp schematic or at least the full IC complement and any other device sucking current from there (any LEDs/Relays/whatever?)

Then we add a meager couple mA to that resistors can be optimized , reducing heat to bare minimum .... which can be *much* less than current dissipation.

So please answer this doubt.
Thanks.

[J M Fahey]

***************
max current: 10ma Left, 10ma Right, 20mA total.
***************

Ok, then you must dissipate at least (70V * 0.02A) per dropping resistor=1.4W each. Quite bearable.

Now, you want 50% more current "just in case" so it's 30mA per rail, resistor will be 70/.03=2333 ohms= 2K2 or 2K4 (or plain 2K) , use a 5W one per side, although they will dissipate around 2W each.
Will be warm to hot but well within specs.

And no Voltage problems.

The Zeners will dissipate Max. (supposing you unplug the preamp board) .03*15=450mW so 1W Zeners are fine.

I'm quite sure your big amp can tolerate extra 4 or 5W (tops) total dissipation added inside the chassis

Does it have a fan ?

[bob p]

Thanks for the help, Juan. It looks like you derived the same solution that the original designer did, which dissipates the thermal losses into a closed box. Unfortunately, the amp already used a dropping resistor + zener approach, which dumps a lot of heat into the cabinet, which is passively ventilated with very poor ventilation.

The problem is that the original embodiment (mentioned in post 1) retains a LOT of trapped heat inside the chassis, and that trapped heat has had a tremendous deleterious effect on de-rating capacitor life. The amp came with some very expensive Mallory computer grade screw-in caps, which had already been replaced one time. I'm working on the amp now because even the high end replacement caps test to have less than 1/3 of their capacitance rating; in real world listening terms, the amp delivers OK power up to about 30W, but pushed beyond that the amp's power output falls apart. With the bad PSU caps it's performance is nowhere near it's 455Winto 4R rating. Further complicating the issue is the fact that the retained heat has de-rated all of the small caps in the amp to the point that it's frequency response is just awful. It looks fine with a 1k test signal, but when I sweep the amp, or when I test it with a square wave, the LF response is just awful.

I'd really prefer not to continue to use the original designer's method of supplying the low voltage rail using dropping resistors and zeners -- although this design was good enough to prevent thumping, and it made the caps last long enough to get through the warranty period, it's a design that just doesn't work well for the long haul because of the heat issue.

Because space in the amp is tight, tight now I'm leaning toward a series regulator that uses TO-220 transistors bolted to the chassis. The math says that I can cut the heat dissipation down from 7W to about 3W, and the good news is that instead of dumping 7W of heat into trapped air inside of the chassis, I'll be diverting it into the chassis itself, which will hopefully result in less heating of the parts trapped inside of the box.

I really don't like the fact that this amp's design ignores the heat problem. The retained heat turns it into a cap-eater, in an era when the type of PSU caps that the amp uses are both expensive and becoming harder and harder to find.

Thanks for subscribing to the thread. FYI the reason that you had trouble quoting my post is because I had deleted it while I went to double-check my math.

[km6xz]

I recently had a project where the single supply voltage was 60v and needed low current bipolar supply for analog control circuits. Heat was an design issue and of course efficiency. I ended up using a TI LM5009A switching regulator down to 18v and feeding a LM5009 for +15 out and TSP84259 for -15v for a total dissipation of 100mw and about 78% efficiency.
The layout was not terrible critical, the first breadboard worked OK but efficiency improved when laying out a production board. Total BOM was less than $9. I only needed 12-14 ma max and regulation and noise were very good, with a low parts count. The finished board or foot print was smaller than a transformer and heat was insignificant.
Just a thought...