Fast 4013s and fast RAM / ROM for 1802 Membership Card Lee Hart Feb-Mar 2024 Summary: Lee Hart has produced the 1802 Membership card for about 13 years as of 2024. In that time he's used various aged to modern parts for the CD4013 quad gate, and various RAM chips, and various ROM chips. In this note he measures the propagation delay Tpd for a 4013 (1802 clock flip-flop) of various dates. Decoding of MRD from TPB in some 1802MC's depends on a 4013 flip-flop and the speed of the RAM. In the latest revision of the 1802MC CPU, he provides an option to adjust the timing of RAM / ROM decode. This note also provides guidance for reworking older 1802 MC's and use of RAM and 4013. Additionally there's a note about Lee reviewing UV EPROMs he's bought, for actual speed and current consumption. - Herb Johnson Fast 4013s (topic posted in groups.io cosmacelf; these notes edited with further revisions) Feb 7 2024 #43739 ------------------- I spent some time testing the various 4013's I have. It's clear that they have had die improvements over the years as well (just like the 1802). Today's parts are *significantly* faster than the data sheets indicate. Each chip was tested on a breadboard socket (which adds about 20pF of load) and scope probe (which adds another 25pf): date, brand, part#, datecode, Tpd, Tpd with 100pF, Tpd with 330pf 1975 Motorola MC14013, (7519) 160ns, 210ns, 320ns 1978 RCA CD4013BE, (849), 130ns, 180ns, 280ns 1986 Nat.Semi CD4013BCN, (P8624SK),130ns, 230ns, 370ns 1989 T.I. CD4013BP, (89HCYFK), 110ns, 170ns, 270ns 1995 Goldstar CD4013B, (9534), 90ns, 140ns, 220ns 2019 T.I. CD4013BE, (19B2HQT), 70ns, 110ns, 150ns 2021 T.I. CD4013BE, (21C4CDKe4), 60ns, 80ns, 130ns 2022 T.I. CD4013BE, (22DN9JNe4), 50ns, 70ns, 120ns There is a pretty clear improvement in propagation delay as the years passed, regardless of brand. My guess is that all the manufacturers were improving their processes without changing the data sheets ("typicals" are no longer typical). I measured propagation delay to the 50% of VDD point (2.5v on a 5v supply). Adding capacitance increases the propagation delay, mainly by changing the sharp vertical drop into a slower gradual drop. 330pF roughly doubles the propagation delay. I started making the 1802 Membership Cards way back in 2011. I had lots of old slow RAMs and EPROMs (450 nsec or worse). So until rev.E, I included a 100pF capacitor (C6) on the output of the 4013 to slow down its rising edge, and latch the output port as late as possible to allow for the slow memory access time. After rev.E, I had used up my old memory ICs, and the newer chips I was getting were much faster (50-100 nsec); so C6 was deleted. I also observed that the newer memory ICs used a lot more power; 100-150mA was common! I'm kind of a low-power nut, so I changed the design to only chip-select memory during TPB. This reduced supply current by about 30% during reads, and 75% during writes. What I failed to notice was that regular 4000-series CMOS parts were also getting faster. In rev.L (the latest version), new 4013's had propagation delays that were faster than RAMs and EPROMs! These 4013's tried to latch the output port data too soon (before the memory access time could provide it). Assuming 150 nsec max access time for memory, the 4013 needs to have at least 150 nsec delay. It looks like a capacitor of at least 330pF needs to be added from 4013 pin 1 to GND. With the newest fastest 4013's, 470pF is even better. So, If you have a rev.L Membership Card, add a 470pF capacitor between 4013 pin 1 [OUT4 signal] and GND to avoid any problems. Or (as noted in a previous email), if you don't care about power consumption, you can cut the trace to address decoder U4 pin 13 (74HC139 B input), and tie pin 13 to VDD. This selects RAM and ROM for the entire memory cycle, not just TPB. Then the memory data will be active much earlier, and there's no problem no matter how fast the 4013 is. This fix also lets you use old memories with much longer access times. Going forward to rev.M... I'm undecided on a permanent fix. Add the capacitor, remove TPB from the address decoder, or something else? Thoughts? - Lee Lee Hart, Feb 13 #43788 --------------------- I couldn't decide whether it was better to make the Membership Card work with any memory chips and 4013s (but use more power); or to minimize power consumption when using modern fast-but-high-power memory chips (but risk it not working with a 4013 that is too fast). So I added a jumper option P10, and a place for a slow-down capacitor C9 on the 4013. This CPU PC board will be revision L1. FAST (the default): Speed jumper P10 in FAST position, and C9 is installed. This supports new fast memory chips (150 nSec or better), and minimizes power consumption (by only chip-selecting memory during TPB). C9 delays latching the output port until the end of TPB even with a new fast 4013. SLOW (optional): Speed jumper P10 in SLOW position. No C9 installed. This supports any memory chip (even old slow ones), and any speed 4013 (new or old). But power consumption is about 30% higher; it depends on the specific memory chips and program that's running. Power consumption is also higher in standby, so the supercapacitor has a shorter memory retention time with modern memory chips. (For example, about 1 day with a CY7C199 RAM in SLOW vs. 4 days in FAST) Another change concerns the power-on-jump to 8000h if the IN button is pressed. I increased C6 to 4.7uF to be sure it is longer than the power-on-CLEAR delay. Also, I changed Q1 from the special FJN4303 with its internal base resistors to a standard 2N3906 with an external base resistor R18 (hiding under the 1802 -- no room!) - Lee Private correspondence, Lee Hart Mar 5 2024 -------------------------------- I use 27C256 EPROMs on several of my products. I used to get them from BG Micro, then Rochester Electronics, but they no longer have them at a reasonable price. I ordered some from a China ebay seller. They are "marked" SGS M27C256-10. Most of them worked, so it seemed like a reasonable source. But I've had customers complain about "bad" EPROMs. When they send them back, they work fine for me. Today, I tracked down the problem. The part# says they are 100 nsec access time. *They are not*! I measured ~8 of them, and found the access times vary from 200-300 nsec. So the Chinese are re-marking EPROMs as well (harder, since they are ceramic). (Grump grump) I'll have to sort them, and use the faster ones in the 1802MC, and slower ones in [products] that allow slower parts. Being the suspicious type, I then went back and checked some of the "marked" NMC 27C32BQ-200 EPROMs I got from another Chinese source. Their access time is about 200 nsec, but *they are not CMOS*! The power consumption is the regular old 2732 levels. I use these for the Membership Card Front Panel hex-to-7seg decoder. Luckily, I added a FAST-SLOW jumper to 1802MC rev.L1. Even slow EPROMs [>200ns] are fine in the SLOW position. But it increases supply current, and shortens supercapacitor memory retention time. These two EPROMs and the SLOW position increase supply current from 5mA to 50mA. (Sigh) It probably doesn't matter to anyone but me; but that's life. - Lee Hart