DC Power Supplies +-24V
This large state machine provides some high-function control of the $55, 8.3A, commercial, open-frame power supplies, Mean Well PPS-200-24. One supply provides +24V for the NPN output transistors of the speaker amplifiers, & the other provides -24V. Each supply gets a full state machine. This control is so elaborate because the time to check for voltage from a supply isn't right as you apply 120VAC to it; the specs for these supplies say it may take 2.5s for DC to come good, a very long time.
There are other protection circuits: woofer relay putting extra resistance in series with woofer in case voice coil gets too hot, 6 ohms in series with each 24V supply when needed, output diodes on supplies to prevent reverse voltage application, meltwater level, power transistor short, 6 ohm overtemp, variable fan speed on the +-24V supplies (high speed at high audio volume).
These supplies don't have inrush current limiting, so to lower the arcing in the relays, I will do as Bob Cordell says, put in Epcos $.37 NTC thermistors, 5ohm (when cold) 4.5A, B57236S509M54, on each supply. Epcos has a good app note about this. These thermistors run hot as normal operation, reaching way below 1ohm at 4.5A RMS, and require cool-off time (like 45s) before you restart a supply. To make this less time, I intend to cool both thermistors with the fans that cool the supplies. And I intend to sense the thermistor temps & let a restart happen as early as possible, extending the 3.5s wait from state 1000 to 0000 as needed. Furthermore, I intend to start one 24V supply about 1/2s before the other so they don't do a double whammy on the 120VAC electrical system.
Here is a question: if the audio program going to the speaker is a low level, the +-24V supplies will be at low current, and the thermistors will be at about 2ohm. What happens if the audio program suddenly requires 150W from each supply? The supplies need to pull high 120VAC current as the bulk caps discharge, like in the next + or - cycle of AC. Does the 2ohms of an inrush thermistor prevent adequate charging of the bulk caps? Acknowledged, that the 2ohms will do down in 5 milliseconds or so as the thermistor heats, but the peak charging current from 120VAC needs to be about 20A (10A per supply) near the peak voltage of the AC, and a thermistor at 2ohms would drop 20V from the sq.root.of.2 * 120V = 169V peak. In case this turns out to be a problem, provide room in the supply box for relays to short out thermistors at appropriate states.
Fans on +-24V supplies will run full speed except at state 0010, or when DC current draw is high or inrush thermistors are hot.
April 26: all three orders are in, including the $55 switching power supplies for +-24V. With real parts to handle, I modeled the power supply box using stiff paper to represent major assemblies, & decided how things stack together. The auxiliary supplies (-5.15, +7.5, +16 regulated down to +12) take up a lot of room! The open-frame supplies are heavy & have tiny mounting areas, making them susceptible to breaking the fiberglass PCB if the power supply box is dropped. Came up with an alternative way to mount the supplies, cradling them between wooden blocks, and mounting the fan for +-24V cooling on the blocks. The assemblies fit together with 120VAC coming in at the bottom through AC line filter, then conversion to DC happens toward the top of the box. Dangerous AC is at the bottom, less likely to be touched, and quiet DC is at the top. Time to purchase hardware cloth that main shielding is to be made from (the sides of power supply & amplifier).
May 10: Lots of hardware construction has been happening. The supply box has a 1" plywood base, angle-iron right-back edge for rigidity, flanked by 4"-wide, 16 gauge steel panels. 1/2" square dowels make up the rest of the frame. Galvanized hardware cloth is over just about everything, making for good shielding; combined with filtering on all I/O wires, should make for good containment of the SMPS noise. The three aux supplies are in the right back. Below them is the A.C. line filter & input MOV protector. At the back of all that is the convenience A.C. outlet.
The cradled pair of SMPS 24V 8.3A commercial supplies is mounted, with inrush thermistors & A.C. relays at the bottom of that module.
Front-panel controls & ammeters (with shunts for 9A full scale) are mounted.
The 6-ohm power resistors that give soft start (for charging 15,000uF caps) are mounted on an exhaust fan. +-24V wiring is complete through these resistors & on to LC filtering at the back, with pilot lamps & reverse-voltage diodes.
I forgot the 140mm fan at right, I can add it on the outside with no problem.
White LEDs light up a lot of the interior of the box.
Remaining cavities for PC boards are not very big. There is room at bottom, beside line filter, for +-15V regulators. The Atmel CPLD state machines that control application of 120VAC to SMPS will be at the top. The 7-segment display that displays some of the conditions will be on a little PCB seen from the front, beside the +24V ammeter. Other 7-segment conditions will show up on a second display, in the amplifier box.
All in all, space planning has been shown to be worthwhile. The various assemblies in the supply box fit together quite well. They have been time consuming to build.
May 15: printed circuit board design has started, for the board with supply regulators and fan supplies (dependent on thermistor temperatures). This is being built upon a heat sink with four pass transistors, for +-15V (adjustable down to +-6V), 5V, & 12V. The follow-on board will be the Atmel CPLD state machines to control the AC relays for the SMPS +-24V supplies.
June 1, 2015: The PCB with supply regulators & fan supplies is working. It needs an external fan when significant load is drawn from +- 6 to 15V since there is little built-in air ducting to the sink. To facilitate this, there are four binding posts added on the back of the box, two being ground and 12V.
Design of the Atmel CPLD with the state machines, one per 24V SMPS, is proceeding. A significant addition to this board is a DIP switch to give manual controls.
August 2015: All the boards in the power supply are finished, the supply is working. Still need to finish up some steel work. The protection scheme in the two state machines has been shown to be useful, I accidentally wired ground to -24V during midrange amplifier work and the supply shut down, showing a 1 fault. Pretty quickly, I found the wiring error.
Dec 17, 2015: A late change in the power supply box is on the logic board. There was increasing fault 1 (loss of +-24V supply, to the tune of 10% to 20% decrease from 24V) as I have had the amplifier box's 15,000uF caps on the +-24V supplies. My only oscilloscope, the XProtolab tiny $55 oscilloscope, seemed to show a longer-term, seconds-type problem, not a microsecond-type problem. I redesigned the two 339 comparators that look at the 24V supplies, detecting loss of regulation (they also show up sudden charging of big capacitors that suddenly get applied). The elaborated design is on a double-sided PCB, 1.2" x 1.4", stuck above the logic board's bottom area. The 339 remains on the logic board but the added circuits replace the passive parts that had been around the 339. I did the little board without a thorough breadboard trial, hoping that the added circuits would be a solution, and it seems to be effective. Here is the new +-24V sensing solution. For the first 2.7s after +24V comes on, the acceptance limits for +-24V are about 17.8V. Furthermore, the hysteresis band of 339 detection is about 17.8V to 20.2V, a 2.4V spread. After 2.7s, the hysteresis band moves to 21 to 22.4V, a 1.4V spread. This dynamic acceptance band is more forgiving of transients that happen in the first 4.2s after the AC relays apply power to the commercial, Mean Well SMPS, especially when the two relays short out the 6 ohm inline resistors that ease charging of 30,000 uF on each of the busses. An important feature of the added circuitry is trimpots that trim the hysteresis band. The 2.7s timer is a divider on +24V, a zener, and two NPN transistors with hysteresis feedback resistor of 300k, giving a 0 to 12V logic signal. A front-panel switch allows the operator to enable the coarser detection band at any time (any time there are extra load transients going on). An orange LED indicates that the coarser detection band is enabled. The control signal from the two-NPN circuit feeds all four analog switches in a 4066 IC. Two analog switches are used for +24V sensing, and the other two are for -24V sensing. Both of these channels have a pot divider that is pulled lower by analog switch after 2.7s, feeding 500k-.1uF (.05s time constant) transient filter, then a NPN Darlington buffer, then the hysteresis parts for the 339 comparator. The hysteresis parts have an analog switch making hysteresis narrower after 2.7s. Hysteresis feedback has some RC lead (27us, 10k with 2700pF) to give added hysteresis in the microsecond area. A red LED indicates +24V is out of spec, and a green LED is for -24V. This is a lot of circuitry on a little add-on board.
Power supply work now moves on to some labeling.