# What's the most efficient way to derive negative voltage for a bench power supply?

I’m building a bench power supply based on two LM317 with the following specs:

• Input: 127VAC (from mains)
• Output: adjustable from 1.25VDC to 15VDC
• Current: limit adjustable from 100mA to 1.5A

In the LM317 datasheet from On Semiconductors there is a reference for the “current regulator” circuit that requires a negative voltage (below -1.25V) in the FET (detail #1 in the figure).

Since my power supply is not handling negative voltages I was wondering on using a 24VAC center tapped transformer (127VAC to 12-0-12) ignoring the center tap to power the whole circuit (detail #2 in the figure).

So, my question is how to derive the negative voltage necessary for the FET?

Also, If I decided to use the center tap of the transformer, changing it to a 48VAC one (127VAC to 24-0-24), I’ll use the full rectifier bridge between +24VAC a 0VAC and I’ll try to extract the negative from the -24VAC of the trafo. Since I only need something around -1.5V isn’t that a waste of cooper on the trafo (detail #3 in the figure)?

What is the best approach to power the power supply and also provide the necessary negative voltage for the current regulator circuit?

• 24VAC center tapped (or 12-0-12 in some older nomenclatures) - not +12 and -12 VAC - AC is AC, and does not have polarity, though it does have phase. – Ecnerwal Mar 7 '14 at 2:39
• 2N5640 has a spec'd saturation current of 5mA min. But you need at least 12.5 mA (preferably 14 mA) to get current down to zero. So maybe the 2N5640 is not the right choice. Also, the 2N5640 is nearly impossible to obtain. – markrages Mar 7 '14 at 4:42
• Thanks @Ecnerwal . I'm having a hard time to find the 2N5640 and other alternatives (2N3822, 2N4858 and 2N4861) here in Brazil and also on other international online retailers. Can help me to find an alternative to that? Let's say I can't find that JFET. What are the remaining alternatives? – mlemos Mar 7 '14 at 19:07

A 24VAC transformer will give you as much as 40VDC with a light load. That's pretty marginal for an LM317 with zero output. Way too close for comfort, in my opinion.

Anyway, if you were to use a third LM317 to regulate the input voltage down to +5VDC, you could use that 5V to power a 7660 charge pump to generate about -4.5V for your JFET current sink.

The filtered unregulated voltage likely is too close to the maximum 40V input voltage of the 317 (actually maximum difference between in and out). Maybe you can use 317HV, but an 18V transformer would be better.

Unfortunately the LM7805 is only good for 35V max input, so you'd need a Zener in series with the input of the LM7805 if you wanted to use that part to drop the voltage for the 7660 with a 24VAC transformer.

• +1. The TC7660 should easily supply the dozen or so mA for this circuit. – markrages Mar 7 '14 at 4:41
• Thanks @SpehroPefhany . So, the two points here are: 1) 24VAC in the input may be too close due to all the lost voltage in the regulators, diodes, etc to reach my desired max 15VDC output? 2) Use this TC7660 to generate the negative voltage. Can I use a LM7805 to feed him? – mlemos Mar 7 '14 at 11:43
• See edits clarifying problem. – Spehro Pefhany Mar 7 '14 at 12:20

Best is awfully opinion based and depends on too many criteria we don't know. Simplest (and most efficient) these days would be a little DC-DC converter unit that's cheap and requires no real thought. If the challenge of design is the goal, or a significant part of the goal, you could do that part less simply and probably less cheaply...of course you could also move beyond LM317's too - they have certain nice features but they are neither remotely efficient (they get hot - they are for most intents best considered as a "smart resistor") nor modern.

Given part supply issues mentioned in comment to the question, I wonder if going to a +/- symmetrical supply (pairing with an LM337) might be both easy and potentially useful in a bench power supply...