80-Bus News

  

May–June 1984, Volume 3, Issue 3











Page 11 of 51











4 i | | i | i :

11

appears at the output pin Q when R/W goes high. /CAS also acts as an output enable; when it is high, Q switches into the high-impedance state. Because only one bit is stored for each address pattern, 8 chips are needed to service the full data bus. Thus a block of 8 4116s give 16 kbytes and 8 4164s give a full 64 kbytes. This is in contrast to the byte-mapped 6116, where a single chip can be used to provide a block of 2 kbytes.

The other main feature of dynamic RAM is that the charge in the memory cells slowly leaks away and has to be replenished periodically by rewriting. Fach cell has to be refreshed within a time less than 2 mSec for the 4116 and less than 4 mSec for the 4164. This is another feature which has to be provided by the memory board designer. Fortunately, the problem is not as complex as it first appears since cells in the same column can be refreshed in parallel by applying an address to the rows only which is strobed by /RAS. Provided all row addresses are strobed during the stated period, refreshis assured. During this process, /CAS is held high which minimises power consumption and ensures that no data is placed on the data bus during refresh cycles. The refresh process occurs during the second half of the opcode fetch cycle, and the Z80 provides a J-bit refresh counter which makes this part of the memory board design fairly straight-forward. However, when selecting 64k chips a little care has to be taken since, for example, the Texas 4164 requires an 8-bit refresh cycle and therefore presents problems when used with the Z80. Also, some chips, such as the Motorola MCM6664, have an internal refresh counter but require a strobe pulse to be applied to pin 1 to clock the counter on. You should therefore use 7-bit refresh chips which do not utilise pin 1, such as the Mostek MK4564, the Fujitsu MB8264, the NEC uPD4164, etc. The ones I actually used were Motorola MCM6665AP20s.

The 4164 is simpler than the 4116 in that it can operate from a single 5 volt supply, so the –5 and –12 volt supplies will not be required in the new layout. A further advantage of the 4164 is that it actually consumes less power than the 4116; the 4116 typically consumes only 125 mW when running and 17 mW on standby, whereas the correponding figures for the 4116 are 460 and 20 mW respectively. Let us turn now to the circuit modifications required to add a single set of 8 4116s to the RAM A board, and after that we will go into the additional circuitry to achieve paging for the 128k RAM. Incidentally, if your copy of the circuit diagram is an minute as mine, you will probably find the use of a magnifying glass helpful to see the fine details of the circuit.

The first item which must be attended to is the address multiplexing. For the 4116s, A0 to A6 are multiplexed with A7 to A13 by IC20 and IC21, which are 74157 quad 2-to-1 line data selectors. The two remaining lines, A14 and A15, are decoded by IC24 to produce 4 16k decodes – we will utilise these for paging the 128k RAM. For the 4164s, AQ to A7 must be multiplexed with A8 to A15. It does not actually matter which lines are multiplexed together as long as the components of each pair come from the two different groupings, and that A0 to A6 are coupled as a unit to the RAM chips and latched by /RAS so that refreshing is performed correctly. Unfortunately, in the original design, A0 is switched with A7 which is not allowed, so A7 must be disconnected and replaced by A14, which can be picked up from pin 15 of the 74L875 address latch 1025. All the remaining pairs can be left unaltered, which leaves A7 and A15 to be dealt with. As currently configured, the remaining inputs on IC21 toggle between 0 and +5 volts to provide the /CAS signal via three buffers in IC 3 which provide a short delay to allow the address lines to settle. It seemed to me to be more straightforward to use these inputs to toggle between


This is an OCR’d version of the scanned page and likely contains recognition errors.











Page 11 of 51