Initially i was planning on finding / getting a "drop-in replacement" toroidal transformer, ie. similarly sized physically and power-wise, but with less insanely-high secondary voltages. The stock one, according to some "inspection notes" i found online, seems to have had roughly 2x20VAC windings for the analog +/-15V (meaning the voltage regulators had to "burn off" uselessly much power / voltage), 12VAC for the 5V regulators, and a whopping 58VAC winding for the phantom power regulator. That's about a 50% overkill across the board. One can only wonder just what possessed them to spec the transformer quite like that, especially since it must've been a custom order - you'll rarely find retail units with more than two (identical) secondary windings, and when you do, expect to pay a pretty penny for'em.
Either way, after a cursory glance at some prices, 2x12VAC or 2x15VAC 20VA-ish transformer prices seemed to hover in the 15-20$ area. Not too bad, i suppose, but since i didn't have an immediate need for the thing, i was in no rush to grab one. Especially since shipping alone would've been another 10$ or so on top of that.
Fast-forward to last week when, after having finally gotten around to getting some multicoloured alligator clips and making some decent-length cables, i remembered about this unit, and decided to hook it up to the bench supply. After adjusting the outputs to +/-18V for the analog circuitry, and the 6V maximum the third rail could supply (i'm gonna have to tweak / mod that someday) on the old TTi PL330QMT, i switched on the power, and... straight away hit the current-limit on the negative rail (which i had set to about 500mA). A bit of poking around later, showed that the heatsinked 7915 was internally shorted.
A quick rummage through my parts collection didn't provide a replacement, but i did find two 7912's. The solution was clear - move the non-heatsinked 7915 onto the heatsink, slip a 7912 in its place, and swap its partner for a 7812. A few minutes of soldering iron "wizardry" later, the next power-up attempt went (much more) normally. The "Master" and "Locked" clock-indicator LEDs on the front panel lit up as expected, since the clock selector on the back was set to "44.1 internal", and they went off when sliding the switch to the "wordclock in" or "ADAT in", both of which are "slave" modes. We got a live one!
Right then... By now, you should be aware i'm quite the cheapskate (in certain cases, at least), but dare i say it, a relatively ingenious one. A few years back, as i found myself browsing through a few local thrift-stores with a friend who was looking for cheap furniture for the apartment he had recently moved into, i got my hands on a few old, 50Hz-transformer-based "wall warts", for a mere couple bucks a pop. I figured i could probably coax one of those to work in this converter, with the aid of some clever(?) circuit-workarounds and whatnot.
First one i grabbed off the shelf was a 24VAC 16VA one. After a 10-minute struggle cracking the casing open (bastards had plugged up the mounting screw holes), my faint hope of the transformer "miraculously" being a center-tapped one was shattered - well, that's a no-go. Next one, 12VAC 1.67A - NOW we're talkin'!..
Granted, that would provide a potentially too-low resulting DC voltage for the +/-15V regulators to work (datasheet says input needs to be at least 2V higher than the output), but i have no qualms about replacing the remaining 7815/7915 pair with 12V versions. With that said, i took to Google once again, to see what wiser people than i might've come up with, for obtaining two symmetrical output voltages out of a single secondary transformer winding. The search was brief, but thoroughly fruitful, as it provided me with the answer (or at least a suitable-enough one) within seconds.
Two "extra" capacitors? No sweat then, problem solved! Well, at least the opamp supply. The digital supply was a no-brainer (see positive side of the above-mentioned circuit). But what about the phantom power? Ehrm... Yeah - Houston, we've got a boo-boo...
Took me a good hour or two, googling for various voltage-multiplier topologies (without needing to ground one end of the transformer secondary) and simulating a few ideas in LTspice, but it quickly became quite clear that the number of stages needed to get some 55V out of a 12VAC winding would've required too much complexity and too many components, to be reasonable or worth-while. Enter our old friend, the MC34063.
When i was looking into modding the tube section in the SM Pro Audio TB202 preamp, i had come across this little circuit, employing a 34063 to drive a MOSFET, in a high-voltage boost-converter. This time i wouldn't need an output in the hundreds of volts, but still considerably higher than the 40V max that the switching transistors inside the chip can take. Since a short-circuit on the phantom power on all eight mic inputs would amount up to 112mA or so, the power requirements wouldn't exactly be impractical either - an absolute worst-case maximum requirement of roughly 5W. In other words, not quite what one might call a big deal.
In a couple hours (that is, including several revisions and numerous "tweaks" and "polishes"), the above schematic gained a board layout counterpart as well. I included a mounting hole / pad for the earthing screw to poke through, which i'll also use as one mounting element of the board. On the "overhanging" end of the board, i figured i'd squirt a big blob of RTV silicone - nice and sticky until it cures, after which it has a pretty firm grip, while also being relatively compliant (as opposed to brittle or prone to fractures).
The 12VAC transformer winding comes in to the bottom-right side, while on the top, from left to right, are the 50-something-volt (pending feedback resistor value figuring-out) output for the phantom power regulator, the 16-17V output going to the digital supply 7805's (which will pass through a couple diodes still on the mainboard, to take some stress off the regulators themselves), and the two symmetrical outputs for the analog circuitry (pending measurements to see whether the remaining 15V regulators might need to be replaced by 12V ones). I even went whole-hog and liberally sprinkled some extra filtering inductors on each output, mainly because i could (and had a few in the parts bin, that were unlikely to be used anywhere else, anytime soon).
The board was specifically laid out in order to make it as simple as possible to realize via my now "classic" method of scoring isolation slots between islands on a blank copper-clad board - cheap, simple and fast, with readily-available tools and materials. True, a bit labor-intensive, but with an added bonus at the end, consisting in the satisfaction of having done something with your own hands. Isn't that what DIY is all about, anyway?
Instead of simply eyeballing it and hoping for the best (the phrase "spit and a prayer" comes to mind), i went with a quarter-"scientific" method, because half would be an overstatement: printed the top layer (on some paper, then cut it out), used the ground-off shank of a broken drill-bit to spot all the corners / intersections of the separations between the copper "islands", drew the lines to cut with a pencil, and proceeded to "manual-CNC mill" the cutouts. About half an hour (and a couple minor slip-ups) later, job done. Onwards, to soldering!
Preliminary test stage, after populating the boost-converter side, and we've got a winner. At least with no load, the MC34063 circuit puts out the 58V that the 10K / 220 ohm feedback network sets, drawing about 30mA at 14V input. More than enough voltage headroom for the LM317-based 50V phantom-power regulator needs for proper functioning.
After a couple of little "oopsies" (ie. having installed three of the twelve diodes, the wrong way around - thank you, brainfarts!), i'm glad to report all is well. The two analog supplies put out something in the region of +/-15.6V, so the two 15V regulators definitely need to be swapped out for 12V ones. The third rail, considering it's loaded down a bit by the boost-converter, measures in at 15.1V, but that also supplies the three regulators for the digital circuitry.
Unloaded tests done; time to start hooking things up. I figured i'd start with the rear board only first - No smoke-release, which was good news. Opamp supply rails measure in at 12V and -11.8V - slightly odd, but the reading's stable. The non-heatsinked 7912 measures -12v directly on its output pin, so the "low" one must be the heatsinked one. Either way, i'm not gonna sweat over a mere 200mV. 5V rails measure as nominal, phantom power regulator outputs 49.5V.
Plugging in the connectors from the front section (preamps & indicator LEDs) changed nothing, so that means we have another rousing(?) success here! But fair enough, allow me to rain on my own parade - i'll admit, it looks like what some might call a "haywire abortion", but it IS functional. And once i get my hands on a little CNC mill hopefully sometime next year (ie. 2018), all these prototypes will start looking a bit less... umm... Well, i'll let you fill that one in on your own.
Hello, your work is very good, congratulations for helping us with this pre, I intend to remove this transformer from him and also simplify the circuit. Don't stop, always continue like this, hugs from BR, thank you
ReplyDeleteThank you too, and i'm glad you found this useful!
Delete