It’s fair to say that I’ve been an electronics hobbyist for almost two decades, yet during that expanse of time I’ve never owned a “real” lab bench power supply. At home I got away with wall-warts, ATX supplies, and homebrew linear deals, and at work and school I had access to The Good Stuff. Somehow I’ve gotten by, until I recently felt the need to obtain a quality, low-noise lab supply for home use.
The hardest part was deciding what model of supply to get. I knew I wanted a triple supply, which generally has a ~5V rail and a positive/negative tracking 20V -ish pair. My default is to go with HP/Agilent/Keysight. At work we have Agilent/Keysight E3631A triple supplies, which I don’t enjoy the interface of, and are more spendy than I’d like. (Disclosure: I work for Keysight Technologies. Opinions are my own.) Plus I was told by old-timers that the noise performance of the E36xx couldn’t match the older range of models, which were apparently acquired from a company called Harrison. So for a while I was interested in the HP 6236B, which can do 6V@2.5A and +/-20V@0.5A, enough for my purposes. I started trolling eBay to get an idea of how much they were going for. But then things took an unexpected turn.
A friend of mine watched the eevBlog video on the Power Designs 2005A and decided to get one of his own. The 2005A is a different beast, a precision voltage source, but nonetheless I was intrigued by this company that I’d never heard of, with its brushed-aluminum aesthetic and reputation for high quality. Looking around, I found that Power Designs had (they seem to be dead now) produced some triple supplies as well. The PD 340 and 340A have dual rails that go to 32V@1A, and a peculiar third rail that showcases PD’s “Uniply” design: 6V@5A or 15V@2.5A. Aside from the move to plastic knobs, I am not sure if there is a difference between the 340 and 340A.
What is a Uniply and what is it good for? (A colleague said that it sounded like a cheap brand of toilet paper. He also doesn’t like the brushed aluminum, FWIW.) To set the scene, a linear power supply generally consists of a rectifier stage, a big honking capacitor filter, and a pass element. If, say, the supply produced 20V internally in order to support a 15V output, 5V would be ‘burned’ across the pass element. If 5V was requested at the output, 15V would have to be burned. This is not great in terms of efficiency, to say the least. Now enter the Uniply. Let’s say we change up the rectifier/capacitor scheme to make a 10V rail as well as a 20V rail internally. With the right arrangement of diodes and pass elements, a 15V output will use the 20V rail only. A 5V output will use the 10V rail, up to a certain current, whereupon it will make use of the 20V rail as well. This grand idea (US Patent 3699352) allows for power savings at lower output voltages when a wide output range is required. (Contrast this to a common but different strategy of having many switchable secondaries with different winding ratios.)
I made a $50 offer on a ‘parts’ TP340 that seemed to be in decent cosmetic shape, though one of the fault lights was on, and the offer was accepted. On closer inspection, the fault-y channel had a control pot that had somehow gotten bashed in, popping out the plastic housing of the Bourns 3590S. The pot was open for most of the travel; I deemed it unrecoverable after halfheartedly trying to put it back together, so I had to order a new one. On initial power-on, the dual rails seemed to work, and tracked well.
The middle pot’s knob had a bit of lean, but it was still working smoothly. Though it’s a mechanical issue and not an electrical one, I find this the most puzzling aspect of this unit. The holes for the pots are 3/8″, which goes with the diameter of the threaded portion of the pot shaft. However there is a feature at the bottom of the shaft which is 0.405″. What the factory had done was try to balance the 3/8″ anti-rotation washer on the 0.405″ feature, to get the pot shaft perpendicular to the surface of the the front panel, but it seems to be a precarious arrangement; indeed the middle pot had developed a lean; a real-world example of spontaneous symmetry breaking. What the Bourns datasheet recommends is to drill a ‘Z’ sized hole in the front panel which the 0.405″ feature can fit into. I was going to enlarge the holes, but my backup idea turned out to be simpler. A 7/16″ anti-rotation washer fits loosely around the 0.405″ feature and keeps the body of the pot aligned with the front panel. The washer (Grainger 22UJ14) is rather thick but leaves just the right amount of thread on the face side.
I didn’t have a good mental model of this unit before it arrived. Manuals and pictures of the TP340/TP340A are available around the internet from those in the know, but I was still surprised by the design when I opened it up. The back case is held on by the four outermost front screws plus three in the back. The box is mostly empty, with boards on the top and right sides, sheet metal with three transistors on the left, and the bottom open. The power transformer is attached to the back wall, opposite to the heatsink. A mess of wires runs between boards and parts, zip-tied into bundles. The wire used feels like that fancy Teflon-jacketed stuff, silver-plated stranded copper.
I’m not an expert on line voltage safety, but this box seems a little deficient in that regard. My unit has a bit of clear heat-shrink (?) on the power switch, but the terminals on the fuse holder are totally exposed. Line voltage finds its way through two terminals on the front output board, unprotected. The filter cap on the primary is a plain Y5U 1KV type; I replaced it with an X1-rated film cap. I plan to replace the fuse holder whenever I open the box next, since the plastic retaining nut has cracked.
The PCBs are a little odd. They are single-sided, probably laid out by hand, without soldermask. The larger components are attached via hollow turrets which seem totally unnecessary. The big primary filter caps are axial electrolytics, zip-tied to the boards for mechanical strength. Rumor has it that PD swore by manual assembly (likely contributing to their eventual demise) and there is a remarkable amount of rosin splatter on these boards. When I removed the big caps, the smell of the rosin took me back to the cheap Radio Shack solder of my childhood…
About those big caps. There are three 10000uF 25V and two 2200uF 80V that serve as the main filtering caps. A few months ago I acquired a DER EE 5000 LCR meter, so I was able to use it to check out all the electrolytic caps. I found a small one with a high DF that I replaced (C304), but it turns out I was using the LCR completely wrong on the big filter caps. What did I do? Well I failed to realize that big caps have extremely low impedance. If you don’t use a Kelvin connection, which compensates for measuring lead impedance, your results will be wonky. Hence when using grabber cables with the DER EE 5000, I was Doing It Wrong™.
I thought my caps were dead or high ESR and ordered new ones. Curiously, Digi-Key carries the exact same axial caps, the Illinois Capacitor TTA series. (I am not sure if these were the factory original caps though. There is no capacitor date code, and most of the date codes in this unit are 1987.) Something happened to the font over the years; I think the new font looks cheap.
It was in the process of installing the new caps that I realized my error when measuring the old caps. I went back to the old set, measured them with Kelvin clips, and found something strange. While the dissipation factor (DF) for the 2200uF caps is similar between the old and new (~0.05), it’s quite different for the 10000uF caps. The new ones have DF=~0.2 and the old ones have DF=~0.1! Yes, the new caps would appear to be worse than the old. I had already installed the new caps so I will leave them be, but this is one of those things that leaves you feeling unsatisfied.
One oddity on the inside, perhaps best left for a rainy day. The way my box is wired leaves out the fuse in the picture. There ought to be a wire attached to post 4, but there isn’t. Between the patchy image quality on the schematic and the wire bundle scheme, it’s not easy to tell where the wires are supposed to go.
And finally, I replaced all the feet, since two were missing. The new feet are Keystone 721 (Newark 25C3781) rubber feet. They are supposedly designed for #4 screws, but I used 6-32 hardware and nylon locking nuts. The screws could be a little shorter (I used 1/2″) but there seems to be enough internal clearance.
That’s about it for now. At some point I’d like to verify this power supply’s performance more rigorously. I can check stability and accuracy with a 3456A. Ripple, if the specs are to be believed, I am not equipped to measure. Load step, I will have to come up with an electronic load. Anyhow, I now have a usable lab supply and no more excuse for holding up my other projects.