Even 35V seems a little on the high side, but perhaps the transformer design is consistent with 50% over voltage at no load. Something closer to 25% would be more in line with (general) expectations.
Without wishing to derail the thread, does this link work for everyone? I've been unable to access Rod's site the last few days. "could not connect to server..." same on phone and PC
The absolute maximum input voltage for a 7918 is -35V. (ON/Fairchild data sheet.)
I recall +40V is the absolute maximum for the 7824 only.
As others have said you need to find out why the regulator input voltage is -44V (and +44V.)
@warpie Why two bridge rectifiers? There is a reason for that in some circumstances but as the OP said in this case it was not for the reason I'm about to give.
Two independent bridges fed from isolated windings (as drawn earlier) are a better solution when the load currents between the plus and minus rails are heavily imbalanced. A heavy relay or illumination load on the positive rail, returning to ground, are two examples. A single "fullwave bridge" providing split rails are actually two "full-wave center-tapped" rectifiers. When a load current imbalance occurs, DC is setup in the center tap which, if large enough, will saturate the transformer core and increase even-order hum, even-order radiated hum fields and heat. Using two full-wave bridges avoids DC being developed in the windings and core.
I had never given this serious consideration until I found it in the Texar Audio Prism power supply which puts about 50 LEDs on the positive rail. Though not the ideal solution, and counter-intuitive, increasing the load on the negative rail, as a test, reduced the even-order hum by pulling the transformer core back into its linear region. https://www.proaudiodesignforum.com/forum/php/viewtopic.php?p=15586#p15586 The final solution was to use a switcher.
1. The IC´s need 100nF on input and output against oscillation.
Just add a power resistor across the negative rail.
Back last century when I was designing rack mount SKUs for Peavey I/we tried to make every SKU work with the same 1A/16VAC wall wart. This sometimes required gymnastics to generate useful power supplies. I recall one SKU that had a relatively high current 5V digital rail while requiring bipolar +/-V rails. The simple wall wart was not center tapped so it was not easy to generate similar voltage split supplies. The 5V rail loaded down the + rail resulting in imbalanced +/- rails. This was before switching PS were widely used (especially by me). I used an op amp driving a power resistor to load down the - rail to bring the two bipolar rails into balance. IIRC I only realized +/- 8V regulated rails for the audio path but this SKU was intended for use in a -10dBV bedroom recording environment so it was adequate.
Do they need a minimum load ? I've never considered that with linear regulators (unlike switching devices). The internal circuitry uses a small current. And measuring with a voltmeter/ DMM pulls a finite current, albeit very small in power terms.
Yes. The output voltage specification specifies a 5 mA minimum output load for the 78XX-series. This is in addition to the 3.5 mA typical quiescent current measured at the input terminal. They typically will measure within range without a minimum load but its not guaranteed. IIRC ST regulators will go out of regulation with no load. I have a foggy memory that we dealt with an ST regulator having that problem here a decade or two ago.
I've done this before and it worked great. It would be desirable to have less voltage on both inputs. And to use silpads under both regulators before you bolt them to a heatsink.
