Tech Note: Reduced power operation, Aug 15 2012 by Lee Hart update Nov 4 2012 Herb Johnson Refer to the 1802 Membership card schematic for details. Additional information added in []'s by Herb Johnson. Some of these notes may refer to earlier versions of the Membership Card, so use details with caution. - Herb Johnson Power connector --------------- The power connector at P4 has +power on pin 1, ground on pin 4, and a jumper between 2 and 3. To "power off", set the CLEAR switch low to stop the CPU, then unplug the power connector at P4. This automatically disables the LEDs (the biggest power user). [The LED resistors are grounded through the jumper between 2 and 3.] Low Power ROM/RAM ------------------ The original plan was to use true CMOS RAMs, where simply switching to LOAD or CLEAR or stopping the clock (by unplugging the front panel board, for example) was all it took. But these RAMS seem to have become unobtainium. [Modern RAMs and ROMs consume much more power, and more still when selected via CS, etc.] Early CMOS ROM and RAM chips: RCA RAMs were MW or CDM 6116 or 6264. Harris or Intersil HM65xxx numbers are the RAMs, HM66xxx numbers are the ROM/PROM/EPROMs. Or, Toshiba TC55xx. The xx numbers are like the other CMOS chips; 6516=2k bytes, 6564=8k bytes 65256=32k, etc. [More modern CMOS ROMs or RAMs usually have a "C" in their name, and often have a "-L" suffix. Check their data sheets to find the lowest power version of their product line. For instance, the SEC brand KM62256CLP-7L, from the 1990's but still in production. 20uA standby, the lowest in that design series, most are 100uA. But the RCA early CMOS chips consumed 1uA in standby! - Herb] Micropower applications ----------------------- For micropower operation, I planned to use the Membership CPU board by itself, without the front panel. In this case, "+power" is pin 13, "common" is pin 1 or 30, and your on/off switch connects pins 13 to 14. Parts R4 R6 R7 C2 C3 are designed to minimize power when the front panel board is unplugged. When you turn off the power switch, R4 and R6 pull RUN (J1 pin 14) low. C2 discharges through R4. In about 47 msec, /CLEAR goes low and stops the CPU. Meanwhile C3 discharges through R7. In about 100 msec, it pulls pin 12 of U4D low and stops the clock. When you turn power back on (short J1 pins 13-13), D10 recharges C3 immediately, starting the clock. /CLEAR is still held low by C2, providing the power-on-reset. R6 charges C2. In about 10 msec, /CLEAR goes high and the CPU begins execution. Standby mode for ROM/RAM ------------------------ I worked on the Membership board some this weekend. I figured out a circuit that adds a transistor to control chip select for the RAM/EPROM. [Deselected RAM and ROM consume much less power.] old: A15 to U2 /CE new: A15 to source of Q1 (2N7000 n-channel MOSFET) TPB to gate of Q1 /CE to drain of Q1 /DMA-OUT was tied directly to VCC, instead of through a pullup resistor in R5. This freed a resistor, which pulls /CE high when Q1 is off. Operation: U2 /CE is low when TPB is high and A15 is low. TPB is high for 1/8th of the time (1 out of 8 clock cycles during each CPU bus cycle. So the effect of this is to cut memory chip power consumption by 8:1. TPB is also low when the CPU is in RESET, keeping the memory off when not running. 3-volt considerations [Aug 2012] -------------------------------- Basically, most chip manufacturers only test their parts at one supply voltage. On parts of the vintage used in the Membership Card, they tested them at 5v. They work at other voltages; they just weren't explicitly tested there. The Membership Card works fine at lower voltages. You have to slow down the clock as the supply voltage goes down, but there's a trimpot to do that. Power consumption goes down roughly as the square of the voltage. If it draws 10ma at 5v, it draws 2.5ma (1/4th as much) at 2.5v. Since you're at 1/2 the voltage and 1/4 the current, total power consumption is 8 times less. :-) The Membership Card still works at even 1.6v, however, the LEDs don't so you can't see anything happening. Two AA cells will work, but as they get old and their voltage drops to near 1.0v per cell, the LEDs get too dim to be of any use. Thus three AA cells is a more reasonable power source, as you can keep using them and still see the LEDs until they are almost totally dead. I've never seen [an 1802 CPU] that didn't work at 3v, and most work even at 2v. I have more problems with the external logic and memories quitting before the 1802 itself. You do have to reduce the clock speed as the supply voltage drops. The maximum clock frequency change is roughly proportional to supply voltage. It is spec'd at 3.2 MHz at 5v, and 6.4 MHz at 10v for example. At 3.3v, I'd estimate it to be (3.3v/5v) x 3.2 MHz = 2.1 MHz. But it gets nonlinear as the supply voltage goes under 3v. The propagation delays get longer. If you don't reduce the clock speed, the delays get so long that the high byte of the address isn't yet stable before TPA, and when the 1802 is outputting a data byte, it isn't available soon enough for the memory to successfully write it. If you have both 5v and 3.3v supplies, just run VDD at 5v, and VCC at 3.3 so all your interface pins use 3.3v levels. -- Lee Hart Unexpected power consumption in manual LOAD mode ------------------------------------------------ The 1802 has an undocumented "feature" in Load mode. After the first LOAD cycle (first DMA-IN cycle, triggered by pressing the IN button), the 1802 leaves the data bus floating. This allows the memory or I/O chips to put data on the bus. But right after CLEAR, and before the first LOAD pulse, the 1802 *grounds* the data bus! This will result in high power consumption if any of the 8 data input switches are set to fight it. So measure your power consumption in LOAD *after* the first IN pulse. The lowest power will generally be measured with the clock stopped (the RUN line low). But, that assumes the 1802 was stopped at a point where it is not driving the bus low, and thus fighting the memory or I/O logic. - Lee Hart, Nov 5 2012 (and reported previously)