Mod to Heath IP-2718 Triple Supply

Most recent update 6-18-2011

Modification of Heath IP-2718 Triple Power Supply
Better regulation and lower ripple and noise on the 20V supplies

Reasons for the mod
I needed a compact, good bench supply for small projects. I found the Heath IP-2718 on eBay at a very good price and bought it. When I hooked it up, I discovered that the 20V supplies were quite noisy with ripple, rectifier peaks, and noise. I was able to download schematics and review the circuitry. What I found was that the two 20V supplies each have a negative sub-supply that provides the regulator's main amplifier differential pair of transistors witn the ability to take the output down to zero Volts. This negative voltage sub-supply only sinks the current from the 62k resistor that is the long-tail emitter resistor for the diff pair -- it has no other use. As shipped, this sub-supply is unregulated, poorly filtered, and noisy.

6-18-2011 -- There are a number of problems with the IP-2718's design. My intent is not to make this a good supply but to make it a better supply than it is stock. A full treatment would invlove changing the base resistors of Q108/208 and making them return to regulated supplies using either zeners or IC regulators; possibly adding a third transistor to the main regulator pair; and re-routing all ground paths for lower noise. That's just too much to do.

In addition, some parts values are problematic and can easily be fixed. In particular, the main filter caps C103/203 can be larger than the 2000uF (2200uF in my unit) stock caps -- newer caps are physically smaller. The bleeder resistors for these caps, R105/205, shown as 1.5k 1W on the schematic, but actually 1k 1W in my unit, are running right at the spec'd power rating or over. They can easily be increased as I'll show below. Negative sub-supply filter caps C105/205 and their bleeders R113/213 have a similar situation and they get replaced, too.

Any improvement in filtering and regulation of this supply would help the 2718 a lot. I dithered a few minutes about using a small PNP transistor with a Zener diode and capacitor in its base for regulation and filtering. The exact output voltage of this sub-supply is relatively unimportant, so there's lots of latitude. But the Zener-regulator uses almost as many parts as a better IC regulator, so I decided to use two LM337-L 100mA Adjustable Voltage Regulator ICs for the mod. Or you can use a 79L05 with resistors adjusted accordingly for 5mA in the resistor string to give roughly -27V of output.

6-18-2011 -- A further mod would be to add a filter cap around ZD108/208, say 10uF, and connect the upper end of R123/223 to ZD108's cathode, giving R123 a regulated source that doesn't change with the main output; concurrently, R124/224 could go up in value to around 33k and be connected from the base of Q108/208 to the negative regulated supply end of R122/222, both lowering noise and I hope improving regulation. This is as yet untried, but when I get a moment, I'll give it a shot.

The mod
Overall, the filtering in the 20V supplies left something to be desired, so I changed filter caps along with the regulator mod. Here's a summary of parts changes:
Remove     C103, 203     2000uF/50V
Install     4700uF/50V axial caps

Remove     C105, 205     10uF/50V
Install     1000uF/50V -- I used radial lead caps but axials work well too

Remove     C108, 208     50uF/50V
Install     10uF/50V -- you can reuse the two you just removed from C105/205

Remove     D110, 210

Remove     R113, 213     1.2k 1W -- these shoulda been 2W anyway!
Install     4.7k 1/2W

Install     regulator boards with LM337-L, 220 ohm, and 4.7k ohm

Install     removed 1.2k 1W (R113/213) in series with R105/205 -- easily done by lifting one end of R105/205 and soldering the other in using the empty PCB hole and joining the bare ends of the resistors

I chose to use small pieces of stripboard five traces wide and five holes deep to hold the regulators and parts. I also chose to put C108 and 208 on the stripboards because I had some low-Z radials, but that doesn't matter -- they can go on the main PCB, especially if you reuse C105/205. The input and output of the 337s go to the empty holes for D110/210.

I drilled a small hole through the PCB trace at the junction of R113/213 and R124/224 for the lead from the 4.7k resistor that comes from the 337's adj. terminal. It all space-wired in OK. I used the free lead from the 4.7k resistor, one from the 220 ohm resistor at the LM337's output trace, and a bare wire from the 337's input trace to serve as the stand-offs to mount the stripboard to the main PCB.

Here's a pictorial view of the mod (it doesn't show the mods to R105/205; those are obvious) -- note that both supplies have exactly the same layout on the PCB, so what works for one works for the other:

How well does it work?
Before the mod, the outputs of the 20V supplies had about 2mVRMS of ripple and noise; now they have less than 200uV. I'll take a 20dB improvement in anything anytime. I didn't check on line and load regulation before the mod, but they're pretty good now -- the load regulation at 16VDC from no load to 0.5A is ~~0.07%~~ 0.1%, while the line regulation for a 10VRMS change from 120VRMS to 110VRMS is ~~0.08%~~ 0.12% -- I didn't give the 2718 enough settling time. I like it when supplies don't have line-related noise -- makes it easier to see how well the circuit you're building is actually doing.

I'm not sure why the line regulation is a little poorer than the load regulation, but both regulation changes may have something to do with the change in zener current through the reference zener ZD104/204 (and the resulting change in the constant-current source Q102/202) with the change in the DC output voltage from the rectifiers -- either through loading or from line voltage change.

Also note that the current metering may show the small currents flowing through the 1.1 ohm sense resistors R109/209 from the negative regualtors, bleeder resistors R113/213, and transistors Q105/205 and Q106/206.

What about those diodes?
Good question. Heath calls D110 and 210 "blocking diodes." I'm not at all sure what they block -- I've seen other regulated bench supplies over the years, and I just don't remember seeing such diodes; I'm guessing they may keep "flyback" currents from flowing where they aren't wanted. You'll note that I didn't remove D103 and 203 -- just an excess of caution, I guess. I'll leave it to you to wonder why Heath didn't do just a little more with the 2718 -- but everything has a cost of some kind.

As a passing note, I sawed away some of the aluminum from the meter retaining bracket to expose more of the back of the meter, and installed a couple white LEDs on the rear panel facing the back of the meter, which is translucent. Makes a very nice illumination for the meter scale. I put the white diodes in parallel and hooked them in series with a 680 ohm resistor, then hooked that assembly up to the terminals of the big filter cap for the 5V supply, which has about 12V across it.